/* Subroutines used for code generation on IA-32.
Copyright (C) 1988, 1992, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001,
- 2002, 2003, 2004, 2005, 2006, 2007 Free Software Foundation, Inc.
+ 2002, 2003, 2004, 2005, 2006, 2007, 2008
+ Free Software Foundation, Inc.
This file is part of GCC.
GCC is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2, or (at your option)
+the Free Software Foundation; either version 3, or (at your option)
any later version.
GCC is distributed in the hope that it will be useful,
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
-along with GCC; see the file COPYING. If not, write to
-the Free Software Foundation, 51 Franklin Street, Fifth Floor,
-Boston, MA 02110-1301, USA. */
+along with GCC; see the file COPYING3. If not see
+<http://www.gnu.org/licenses/>. */
#include "config.h"
#include "system.h"
#include "insn-codes.h"
#include "insn-attr.h"
#include "flags.h"
+#include "c-common.h"
#include "except.h"
#include "function.h"
#include "recog.h"
#include "cgraph.h"
#include "tree-gimple.h"
#include "dwarf2.h"
+#include "df.h"
#include "tm-constrs.h"
#include "params.h"
+static int x86_builtin_vectorization_cost (bool);
+static rtx legitimize_dllimport_symbol (rtx, bool);
+
#ifndef CHECK_STACK_LIMIT
#define CHECK_STACK_LIMIT (-1)
#endif
{3, 3, 3}, /* cost of storing SSE registers
in SImode, DImode and TImode */
3, /* MMX or SSE register to integer */
+ 0, /* size of l1 cache */
+ 0, /* size of l2 cache */
0, /* size of prefetch block */
0, /* number of parallel prefetches */
2, /* Branch cost */
{{rep_prefix_1_byte, {{-1, rep_prefix_1_byte}}},
{rep_prefix_1_byte, {{-1, rep_prefix_1_byte}}}},
{{rep_prefix_1_byte, {{-1, rep_prefix_1_byte}}},
- {rep_prefix_1_byte, {{-1, rep_prefix_1_byte}}}}
+ {rep_prefix_1_byte, {{-1, rep_prefix_1_byte}}}},
+ 1, /* scalar_stmt_cost. */
+ 1, /* scalar load_cost. */
+ 1, /* scalar_store_cost. */
+ 1, /* vec_stmt_cost. */
+ 1, /* vec_to_scalar_cost. */
+ 1, /* scalar_to_vec_cost. */
+ 1, /* vec_align_load_cost. */
+ 1, /* vec_unalign_load_cost. */
+ 1, /* vec_store_cost. */
+ 1, /* cond_taken_branch_cost. */
+ 1, /* cond_not_taken_branch_cost. */
};
/* Processor costs (relative to an add) */
{4, 8, 16}, /* cost of storing SSE registers
in SImode, DImode and TImode */
3, /* MMX or SSE register to integer */
+ 0, /* size of l1 cache */
+ 0, /* size of l2 cache */
0, /* size of prefetch block */
0, /* number of parallel prefetches */
1, /* Branch cost */
DUMMY_STRINGOP_ALGS},
{{rep_prefix_1_byte, {{-1, rep_prefix_1_byte}}},
DUMMY_STRINGOP_ALGS},
+ 1, /* scalar_stmt_cost. */
+ 1, /* scalar load_cost. */
+ 1, /* scalar_store_cost. */
+ 1, /* vec_stmt_cost. */
+ 1, /* vec_to_scalar_cost. */
+ 1, /* scalar_to_vec_cost. */
+ 1, /* vec_align_load_cost. */
+ 2, /* vec_unalign_load_cost. */
+ 1, /* vec_store_cost. */
+ 3, /* cond_taken_branch_cost. */
+ 1, /* cond_not_taken_branch_cost. */
};
static const
{4, 8, 16}, /* cost of storing SSE registers
in SImode, DImode and TImode */
3, /* MMX or SSE register to integer */
+ 4, /* size of l1 cache. 486 has 8kB cache
+ shared for code and data, so 4kB is
+ not really precise. */
+ 4, /* size of l2 cache */
0, /* size of prefetch block */
0, /* number of parallel prefetches */
1, /* Branch cost */
{{rep_prefix_4_byte, {{-1, rep_prefix_4_byte}}},
DUMMY_STRINGOP_ALGS},
{{rep_prefix_4_byte, {{-1, rep_prefix_4_byte}}},
- DUMMY_STRINGOP_ALGS}
+ DUMMY_STRINGOP_ALGS},
+ 1, /* scalar_stmt_cost. */
+ 1, /* scalar load_cost. */
+ 1, /* scalar_store_cost. */
+ 1, /* vec_stmt_cost. */
+ 1, /* vec_to_scalar_cost. */
+ 1, /* scalar_to_vec_cost. */
+ 1, /* vec_align_load_cost. */
+ 2, /* vec_unalign_load_cost. */
+ 1, /* vec_store_cost. */
+ 3, /* cond_taken_branch_cost. */
+ 1, /* cond_not_taken_branch_cost. */
};
static const
{4, 8, 16}, /* cost of storing SSE registers
in SImode, DImode and TImode */
3, /* MMX or SSE register to integer */
+ 8, /* size of l1 cache. */
+ 8, /* size of l2 cache */
0, /* size of prefetch block */
0, /* number of parallel prefetches */
2, /* Branch cost */
{{libcall, {{256, rep_prefix_4_byte}, {-1, libcall}}},
DUMMY_STRINGOP_ALGS},
{{libcall, {{-1, rep_prefix_4_byte}}},
- DUMMY_STRINGOP_ALGS}
+ DUMMY_STRINGOP_ALGS},
+ 1, /* scalar_stmt_cost. */
+ 1, /* scalar load_cost. */
+ 1, /* scalar_store_cost. */
+ 1, /* vec_stmt_cost. */
+ 1, /* vec_to_scalar_cost. */
+ 1, /* scalar_to_vec_cost. */
+ 1, /* vec_align_load_cost. */
+ 2, /* vec_unalign_load_cost. */
+ 1, /* vec_store_cost. */
+ 3, /* cond_taken_branch_cost. */
+ 1, /* cond_not_taken_branch_cost. */
};
static const
{2, 2, 8}, /* cost of storing SSE registers
in SImode, DImode and TImode */
3, /* MMX or SSE register to integer */
+ 8, /* size of l1 cache. */
+ 256, /* size of l2 cache */
32, /* size of prefetch block */
6, /* number of parallel prefetches */
2, /* Branch cost */
DUMMY_STRINGOP_ALGS},
{{rep_prefix_4_byte, {{1024, unrolled_loop},
{8192, rep_prefix_4_byte}, {-1, libcall}}},
- DUMMY_STRINGOP_ALGS}
+ DUMMY_STRINGOP_ALGS},
+ 1, /* scalar_stmt_cost. */
+ 1, /* scalar load_cost. */
+ 1, /* scalar_store_cost. */
+ 1, /* vec_stmt_cost. */
+ 1, /* vec_to_scalar_cost. */
+ 1, /* scalar_to_vec_cost. */
+ 1, /* vec_align_load_cost. */
+ 2, /* vec_unalign_load_cost. */
+ 1, /* vec_store_cost. */
+ 3, /* cond_taken_branch_cost. */
+ 1, /* cond_not_taken_branch_cost. */
};
static const
{1, 1, 1}, /* cost of storing SSE registers
in SImode, DImode and TImode */
1, /* MMX or SSE register to integer */
+ 64, /* size of l1 cache. */
+ 128, /* size of l2 cache. */
32, /* size of prefetch block */
1, /* number of parallel prefetches */
1, /* Branch cost */
{{libcall, {{256, rep_prefix_4_byte}, {-1, libcall}}},
DUMMY_STRINGOP_ALGS},
{{libcall, {{256, rep_prefix_4_byte}, {-1, libcall}}},
- DUMMY_STRINGOP_ALGS}
+ DUMMY_STRINGOP_ALGS},
+ 1, /* scalar_stmt_cost. */
+ 1, /* scalar load_cost. */
+ 1, /* scalar_store_cost. */
+ 1, /* vec_stmt_cost. */
+ 1, /* vec_to_scalar_cost. */
+ 1, /* scalar_to_vec_cost. */
+ 1, /* vec_align_load_cost. */
+ 2, /* vec_unalign_load_cost. */
+ 1, /* vec_store_cost. */
+ 3, /* cond_taken_branch_cost. */
+ 1, /* cond_not_taken_branch_cost. */
};
static const
{2, 2, 8}, /* cost of storing SSE registers
in SImode, DImode and TImode */
6, /* MMX or SSE register to integer */
+ 32, /* size of l1 cache. */
+ 32, /* size of l2 cache. Some models
+ have integrated l2 cache, but
+ optimizing for k6 is not important
+ enough to worry about that. */
32, /* size of prefetch block */
1, /* number of parallel prefetches */
1, /* Branch cost */
{{libcall, {{256, rep_prefix_4_byte}, {-1, libcall}}},
DUMMY_STRINGOP_ALGS},
{{libcall, {{256, rep_prefix_4_byte}, {-1, libcall}}},
- DUMMY_STRINGOP_ALGS}
+ DUMMY_STRINGOP_ALGS},
+ 1, /* scalar_stmt_cost. */
+ 1, /* scalar load_cost. */
+ 1, /* scalar_store_cost. */
+ 1, /* vec_stmt_cost. */
+ 1, /* vec_to_scalar_cost. */
+ 1, /* scalar_to_vec_cost. */
+ 1, /* vec_align_load_cost. */
+ 2, /* vec_unalign_load_cost. */
+ 1, /* vec_store_cost. */
+ 3, /* cond_taken_branch_cost. */
+ 1, /* cond_not_taken_branch_cost. */
};
static const
{4, 4, 5}, /* cost of storing SSE registers
in SImode, DImode and TImode */
5, /* MMX or SSE register to integer */
+ 64, /* size of l1 cache. */
+ 256, /* size of l2 cache. */
64, /* size of prefetch block */
6, /* number of parallel prefetches */
5, /* Branch cost */
{{libcall, {{2048, rep_prefix_4_byte}, {-1, libcall}}},
DUMMY_STRINGOP_ALGS},
{{libcall, {{2048, rep_prefix_4_byte}, {-1, libcall}}},
- DUMMY_STRINGOP_ALGS}
+ DUMMY_STRINGOP_ALGS},
+ 1, /* scalar_stmt_cost. */
+ 1, /* scalar load_cost. */
+ 1, /* scalar_store_cost. */
+ 1, /* vec_stmt_cost. */
+ 1, /* vec_to_scalar_cost. */
+ 1, /* scalar_to_vec_cost. */
+ 1, /* vec_align_load_cost. */
+ 2, /* vec_unalign_load_cost. */
+ 1, /* vec_store_cost. */
+ 3, /* cond_taken_branch_cost. */
+ 1, /* cond_not_taken_branch_cost. */
};
static const
{4, 4, 5}, /* cost of storing SSE registers
in SImode, DImode and TImode */
5, /* MMX or SSE register to integer */
+ 64, /* size of l1 cache. */
+ 512, /* size of l2 cache. */
64, /* size of prefetch block */
/* New AMD processors never drop prefetches; if they cannot be performed
immediately, they are queued. We set number of simultaneous prefetches
to limit number of prefetches at all, as their execution also takes some
time). */
100, /* number of parallel prefetches */
- 5, /* Branch cost */
+ 3, /* Branch cost */
COSTS_N_INSNS (4), /* cost of FADD and FSUB insns. */
COSTS_N_INSNS (4), /* cost of FMUL instruction. */
COSTS_N_INSNS (19), /* cost of FDIV instruction. */
{libcall, {{16, loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}},
{{libcall, {{8, loop}, {24, unrolled_loop},
{2048, rep_prefix_4_byte}, {-1, libcall}}},
- {libcall, {{48, unrolled_loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}}
+ {libcall, {{48, unrolled_loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}},
+ 4, /* scalar_stmt_cost. */
+ 2, /* scalar load_cost. */
+ 2, /* scalar_store_cost. */
+ 5, /* vec_stmt_cost. */
+ 0, /* vec_to_scalar_cost. */
+ 2, /* scalar_to_vec_cost. */
+ 2, /* vec_align_load_cost. */
+ 3, /* vec_unalign_load_cost. */
+ 3, /* vec_store_cost. */
+ 3, /* cond_taken_branch_cost. */
+ 2, /* cond_not_taken_branch_cost. */
};
struct processor_costs amdfam10_cost = {
1/1 1/1
MOVD reg32, xmmreg Double FADD 3
1/1 1/1 */
+ 64, /* size of l1 cache. */
+ 512, /* size of l2 cache. */
64, /* size of prefetch block */
/* New AMD processors never drop prefetches; if they cannot be performed
immediately, they are queued. We set number of simultaneous prefetches
to limit number of prefetches at all, as their execution also takes some
time). */
100, /* number of parallel prefetches */
- 5, /* Branch cost */
+ 2, /* Branch cost */
COSTS_N_INSNS (4), /* cost of FADD and FSUB insns. */
COSTS_N_INSNS (4), /* cost of FMUL instruction. */
COSTS_N_INSNS (19), /* cost of FDIV instruction. */
{libcall, {{16, loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}},
{{libcall, {{8, loop}, {24, unrolled_loop},
{2048, rep_prefix_4_byte}, {-1, libcall}}},
- {libcall, {{48, unrolled_loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}}
+ {libcall, {{48, unrolled_loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}},
+ 4, /* scalar_stmt_cost. */
+ 2, /* scalar load_cost. */
+ 2, /* scalar_store_cost. */
+ 6, /* vec_stmt_cost. */
+ 0, /* vec_to_scalar_cost. */
+ 2, /* scalar_to_vec_cost. */
+ 2, /* vec_align_load_cost. */
+ 2, /* vec_unalign_load_cost. */
+ 2, /* vec_store_cost. */
+ 2, /* cond_taken_branch_cost. */
+ 1, /* cond_not_taken_branch_cost. */
};
static const
{2, 2, 8}, /* cost of storing SSE registers
in SImode, DImode and TImode */
10, /* MMX or SSE register to integer */
+ 8, /* size of l1 cache. */
+ 256, /* size of l2 cache. */
64, /* size of prefetch block */
6, /* number of parallel prefetches */
2, /* Branch cost */
{{libcall, {{6, loop_1_byte}, {48, loop}, {20480, rep_prefix_4_byte},
{-1, libcall}}},
DUMMY_STRINGOP_ALGS},
+ 1, /* scalar_stmt_cost. */
+ 1, /* scalar load_cost. */
+ 1, /* scalar_store_cost. */
+ 1, /* vec_stmt_cost. */
+ 1, /* vec_to_scalar_cost. */
+ 1, /* scalar_to_vec_cost. */
+ 1, /* vec_align_load_cost. */
+ 2, /* vec_unalign_load_cost. */
+ 1, /* vec_store_cost. */
+ 3, /* cond_taken_branch_cost. */
+ 1, /* cond_not_taken_branch_cost. */
};
static const
{12, 12, 12}, /* cost of storing SSE registers
in SImode, DImode and TImode */
8, /* MMX or SSE register to integer */
+ 8, /* size of l1 cache. */
+ 1024, /* size of l2 cache. */
128, /* size of prefetch block */
8, /* number of parallel prefetches */
1, /* Branch cost */
{{libcall, {{6, loop_1_byte}, {48, loop}, {20480, rep_prefix_4_byte},
{-1, libcall}}},
{libcall, {{24, loop}, {64, unrolled_loop},
- {8192, rep_prefix_8_byte}, {-1, libcall}}}}
+ {8192, rep_prefix_8_byte}, {-1, libcall}}}},
+ 1, /* scalar_stmt_cost. */
+ 1, /* scalar load_cost. */
+ 1, /* scalar_store_cost. */
+ 1, /* vec_stmt_cost. */
+ 1, /* vec_to_scalar_cost. */
+ 1, /* scalar_to_vec_cost. */
+ 1, /* vec_align_load_cost. */
+ 2, /* vec_unalign_load_cost. */
+ 1, /* vec_store_cost. */
+ 3, /* cond_taken_branch_cost. */
+ 1, /* cond_not_taken_branch_cost. */
};
static const
{4, 4, 4}, /* cost of storing SSE registers
in SImode, DImode and TImode */
2, /* MMX or SSE register to integer */
+ 32, /* size of l1 cache. */
+ 2048, /* size of l2 cache. */
128, /* size of prefetch block */
8, /* number of parallel prefetches */
3, /* Branch cost */
{{libcall, {{8, loop}, {15, unrolled_loop},
{2048, rep_prefix_4_byte}, {-1, libcall}}},
{libcall, {{24, loop}, {32, unrolled_loop},
- {8192, rep_prefix_8_byte}, {-1, libcall}}}}
+ {8192, rep_prefix_8_byte}, {-1, libcall}}}},
+ 1, /* scalar_stmt_cost. */
+ 1, /* scalar load_cost. */
+ 1, /* scalar_store_cost. */
+ 1, /* vec_stmt_cost. */
+ 1, /* vec_to_scalar_cost. */
+ 1, /* scalar_to_vec_cost. */
+ 1, /* vec_align_load_cost. */
+ 2, /* vec_unalign_load_cost. */
+ 1, /* vec_store_cost. */
+ 3, /* cond_taken_branch_cost. */
+ 1, /* cond_not_taken_branch_cost. */
};
/* Generic64 should produce code tuned for Nocona and K8. */
{8, 8, 8}, /* cost of storing SSE registers
in SImode, DImode and TImode */
5, /* MMX or SSE register to integer */
+ 32, /* size of l1 cache. */
+ 512, /* size of l2 cache. */
64, /* size of prefetch block */
6, /* number of parallel prefetches */
/* Benchmarks shows large regressions on K8 sixtrack benchmark when this value
{DUMMY_STRINGOP_ALGS,
{libcall, {{32, loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}},
{DUMMY_STRINGOP_ALGS,
- {libcall, {{32, loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}}
+ {libcall, {{32, loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}},
+ 1, /* scalar_stmt_cost. */
+ 1, /* scalar load_cost. */
+ 1, /* scalar_store_cost. */
+ 1, /* vec_stmt_cost. */
+ 1, /* vec_to_scalar_cost. */
+ 1, /* scalar_to_vec_cost. */
+ 1, /* vec_align_load_cost. */
+ 2, /* vec_unalign_load_cost. */
+ 1, /* vec_store_cost. */
+ 3, /* cond_taken_branch_cost. */
+ 1, /* cond_not_taken_branch_cost. */
};
/* Generic32 should produce code tuned for Athlon, PPro, Pentium4, Nocona and K8. */
{8, 8, 8}, /* cost of storing SSE registers
in SImode, DImode and TImode */
5, /* MMX or SSE register to integer */
+ 32, /* size of l1 cache. */
+ 256, /* size of l2 cache. */
64, /* size of prefetch block */
6, /* number of parallel prefetches */
3, /* Branch cost */
DUMMY_STRINGOP_ALGS},
{{libcall, {{32, loop}, {8192, rep_prefix_4_byte}, {-1, libcall}}},
DUMMY_STRINGOP_ALGS},
+ 1, /* scalar_stmt_cost. */
+ 1, /* scalar load_cost. */
+ 1, /* scalar_store_cost. */
+ 1, /* vec_stmt_cost. */
+ 1, /* vec_to_scalar_cost. */
+ 1, /* scalar_to_vec_cost. */
+ 1, /* vec_align_load_cost. */
+ 2, /* vec_unalign_load_cost. */
+ 1, /* vec_store_cost. */
+ 3, /* cond_taken_branch_cost. */
+ 1, /* cond_not_taken_branch_cost. */
};
const struct processor_costs *ix86_cost = &pentium_cost;
#define m_ATHLON (1<<PROCESSOR_ATHLON)
#define m_ATHLON_K8 (m_K8 | m_ATHLON)
#define m_AMDFAM10 (1<<PROCESSOR_AMDFAM10)
-#define m_ATHLON_K8_AMDFAM10 (m_K8 | m_ATHLON | m_AMDFAM10)
+#define m_AMD_MULTIPLE (m_K8 | m_ATHLON | m_AMDFAM10)
#define m_GENERIC32 (1<<PROCESSOR_GENERIC32)
#define m_GENERIC64 (1<<PROCESSOR_GENERIC64)
negatively, so enabling for Generic64 seems like good code size
tradeoff. We can't enable it for 32bit generic because it does not
work well with PPro base chips. */
- m_386 | m_K6_GEODE | m_ATHLON_K8_AMDFAM10 | m_CORE2 | m_GENERIC64,
+ m_386 | m_K6_GEODE | m_AMD_MULTIPLE | m_CORE2 | m_GENERIC64,
/* X86_TUNE_PUSH_MEMORY */
- m_386 | m_K6_GEODE | m_ATHLON_K8_AMDFAM10 | m_PENT4
+ m_386 | m_K6_GEODE | m_AMD_MULTIPLE | m_PENT4
| m_NOCONA | m_CORE2 | m_GENERIC,
/* X86_TUNE_ZERO_EXTEND_WITH_AND */
m_386,
/* X86_TUNE_UNROLL_STRLEN */
- m_486 | m_PENT | m_PPRO | m_ATHLON_K8_AMDFAM10 | m_K6 | m_CORE2 | m_GENERIC,
+ m_486 | m_PENT | m_PPRO | m_AMD_MULTIPLE | m_K6 | m_CORE2 | m_GENERIC,
/* X86_TUNE_DEEP_BRANCH_PREDICTION */
- m_PPRO | m_K6_GEODE | m_ATHLON_K8_AMDFAM10 | m_PENT4
- | m_NOCONA | m_CORE2 | m_GENERIC,
+ m_PPRO | m_K6_GEODE | m_AMD_MULTIPLE | m_PENT4 | m_GENERIC,
/* X86_TUNE_BRANCH_PREDICTION_HINTS: Branch hints were put in P4 based
on simulation result. But after P4 was made, no performance benefit
/* X86_TUNE_DOUBLE_WITH_ADD */
~m_386,
-
+
/* X86_TUNE_USE_SAHF */
m_PPRO | m_K6_GEODE | m_K8 | m_AMDFAM10 | m_PENT4
| m_NOCONA | m_CORE2 | m_GENERIC,
/* X86_TUNE_MOVX: Enable to zero extend integer registers to avoid
partial dependencies. */
- m_ATHLON_K8_AMDFAM10 | m_PPRO | m_PENT4 | m_NOCONA
+ m_AMD_MULTIPLE | m_PPRO | m_PENT4 | m_NOCONA
| m_CORE2 | m_GENERIC | m_GEODE /* m_386 | m_K6 */,
/* X86_TUNE_PARTIAL_REG_STALL: We probably ought to watch for partial
/* X86_TUNE_PARTIAL_FLAG_REG_STALL */
m_CORE2 | m_GENERIC,
-
+
/* X86_TUNE_USE_HIMODE_FIOP */
m_386 | m_486 | m_K6_GEODE,
/* X86_TUNE_USE_SIMODE_FIOP */
- ~(m_PPRO | m_ATHLON_K8_AMDFAM10 | m_PENT | m_CORE2 | m_GENERIC),
+ ~(m_PPRO | m_AMD_MULTIPLE | m_PENT | m_CORE2 | m_GENERIC),
/* X86_TUNE_USE_MOV0 */
m_K6,
-
+
/* X86_TUNE_USE_CLTD */
~(m_PENT | m_K6 | m_CORE2 | m_GENERIC),
~(m_PENT | m_PPRO),
/* X86_TUNE_PROMOTE_QIMODE */
- m_K6_GEODE | m_PENT | m_386 | m_486 | m_ATHLON_K8_AMDFAM10 | m_CORE2
+ m_K6_GEODE | m_PENT | m_386 | m_486 | m_AMD_MULTIPLE | m_CORE2
| m_GENERIC /* | m_PENT4 ? */,
/* X86_TUNE_FAST_PREFIX */
/* X86_TUNE_SINGLE_STRINGOP */
m_386 | m_PENT4 | m_NOCONA,
-
+
/* X86_TUNE_QIMODE_MATH */
~0,
-
+
/* X86_TUNE_HIMODE_MATH: On PPro this flag is meant to avoid partial
register stalls. Just like X86_TUNE_PARTIAL_REG_STALL this option
might be considered for Generic32 if our scheme for avoiding partial
m_PPRO,
/* X86_TUNE_ADD_ESP_4: Enable if add/sub is preferred over 1/2 push/pop. */
- m_ATHLON_K8_AMDFAM10 | m_K6_GEODE | m_PENT4 | m_NOCONA | m_CORE2 | m_GENERIC,
+ m_AMD_MULTIPLE | m_K6_GEODE | m_PENT4 | m_NOCONA | m_CORE2 | m_GENERIC,
/* X86_TUNE_ADD_ESP_8 */
- m_ATHLON_K8_AMDFAM10 | m_PPRO | m_K6_GEODE | m_386
+ m_AMD_MULTIPLE | m_PPRO | m_K6_GEODE | m_386
| m_486 | m_PENT4 | m_NOCONA | m_CORE2 | m_GENERIC,
/* X86_TUNE_SUB_ESP_4 */
- m_ATHLON_K8_AMDFAM10 | m_PPRO | m_PENT4 | m_NOCONA | m_CORE2 | m_GENERIC,
+ m_AMD_MULTIPLE | m_PPRO | m_PENT4 | m_NOCONA | m_CORE2 | m_GENERIC,
/* X86_TUNE_SUB_ESP_8 */
- m_ATHLON_K8_AMDFAM10 | m_PPRO | m_386 | m_486
+ m_AMD_MULTIPLE | m_PPRO | m_386 | m_486
| m_PENT4 | m_NOCONA | m_CORE2 | m_GENERIC,
/* X86_TUNE_INTEGER_DFMODE_MOVES: Enable if integer moves are preferred
for DFmode copies */
- ~(m_ATHLON_K8_AMDFAM10 | m_PENT4 | m_NOCONA | m_PPRO | m_CORE2
+ ~(m_AMD_MULTIPLE | m_PENT4 | m_NOCONA | m_PPRO | m_CORE2
| m_GENERIC | m_GEODE),
/* X86_TUNE_PARTIAL_REG_DEPENDENCY */
- m_ATHLON_K8_AMDFAM10 | m_PENT4 | m_NOCONA | m_CORE2 | m_GENERIC,
+ m_AMD_MULTIPLE | m_PENT4 | m_NOCONA | m_CORE2 | m_GENERIC,
/* X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY: In the Generic model we have a
conflict here in between PPro/Pentium4 based chips that thread 128bit
m_ATHLON_K8,
/* X86_TUNE_SSE_TYPELESS_STORES */
- m_ATHLON_K8_AMDFAM10,
+ m_AMD_MULTIPLE,
/* X86_TUNE_SSE_LOAD0_BY_PXOR */
m_PPRO | m_PENT4 | m_NOCONA,
/* X86_TUNE_MEMORY_MISMATCH_STALL */
- m_ATHLON_K8_AMDFAM10 | m_PENT4 | m_NOCONA | m_CORE2 | m_GENERIC,
+ m_AMD_MULTIPLE | m_PENT4 | m_NOCONA | m_CORE2 | m_GENERIC,
/* X86_TUNE_PROLOGUE_USING_MOVE */
m_ATHLON_K8 | m_PPRO | m_CORE2 | m_GENERIC,
~m_486,
/* X86_TUNE_USE_FFREEP */
- m_ATHLON_K8_AMDFAM10,
+ m_AMD_MULTIPLE,
/* X86_TUNE_INTER_UNIT_MOVES */
- ~(m_ATHLON_K8_AMDFAM10 | m_GENERIC),
+ ~(m_AMD_MULTIPLE | m_GENERIC),
+
+ /* X86_TUNE_INTER_UNIT_CONVERSIONS */
+ ~(m_AMDFAM10),
/* X86_TUNE_FOUR_JUMP_LIMIT: Some CPU cores are not able to predict more
than 4 branch instructions in the 16 byte window. */
- m_PPRO | m_ATHLON_K8_AMDFAM10 | m_PENT4 | m_NOCONA | m_CORE2 | m_GENERIC,
+ m_PPRO | m_AMD_MULTIPLE | m_PENT4 | m_NOCONA | m_CORE2 | m_GENERIC,
/* X86_TUNE_SCHEDULE */
- m_PPRO | m_ATHLON_K8_AMDFAM10 | m_K6_GEODE | m_PENT | m_CORE2 | m_GENERIC,
+ m_PPRO | m_AMD_MULTIPLE | m_K6_GEODE | m_PENT | m_CORE2 | m_GENERIC,
/* X86_TUNE_USE_BT */
- m_ATHLON_K8_AMDFAM10,
+ m_AMD_MULTIPLE,
/* X86_TUNE_USE_INCDEC */
- ~(m_PENT4 | m_NOCONA | m_CORE2 | m_GENERIC),
+ ~(m_PENT4 | m_NOCONA | m_GENERIC),
/* X86_TUNE_PAD_RETURNS */
- m_ATHLON_K8_AMDFAM10 | m_CORE2 | m_GENERIC,
+ m_AMD_MULTIPLE | m_CORE2 | m_GENERIC,
/* X86_TUNE_EXT_80387_CONSTANTS */
m_K6_GEODE | m_ATHLON_K8 | m_PENT4 | m_NOCONA | m_PPRO | m_CORE2 | m_GENERIC,
operand that cannot be represented using a modRM byte. The XOR
replacement is long decoded, so this split helps here as well. */
m_K6,
+
+ /* X86_TUNE_USE_VECTOR_CONVERTS: Prefer vector packed SSE conversion
+ from integer to FP. */
+ m_AMDFAM10,
};
/* Feature tests against the various architecture variations. */
unsigned int ix86_arch_features[X86_ARCH_LAST] = {
- /* X86_ARCH_CMOVE */
- m_PPRO | m_GEODE | m_ATHLON_K8_AMDFAM10 | m_PENT4 | m_NOCONA,
+ /* X86_ARCH_CMOVE: Conditional move was added for pentiumpro. */
+ ~(m_386 | m_486 | m_PENT | m_K6),
/* X86_ARCH_CMPXCHG: Compare and exchange was added for 80486. */
~m_386,
};
static const unsigned int x86_accumulate_outgoing_args
- = m_ATHLON_K8_AMDFAM10 | m_PENT4 | m_NOCONA | m_PPRO | m_CORE2 | m_GENERIC;
+ = m_AMD_MULTIPLE | m_PENT4 | m_NOCONA | m_PPRO | m_CORE2 | m_GENERIC;
static const unsigned int x86_arch_always_fancy_math_387
- = m_PENT | m_PPRO | m_ATHLON_K8_AMDFAM10 | m_PENT4
+ = m_PENT | m_PPRO | m_AMD_MULTIPLE | m_PENT4
| m_NOCONA | m_CORE2 | m_GENERIC;
static enum stringop_alg stringop_alg = no_stringop;
NON_Q_REGS,
/* flags, fpsr, fpcr, frame */
NO_REGS, NO_REGS, NO_REGS, NON_Q_REGS,
- SSE_REGS, SSE_REGS, SSE_REGS, SSE_REGS, SSE_REGS, SSE_REGS,
+ /* SSE registers */
+ SSE_FIRST_REG, SSE_REGS, SSE_REGS, SSE_REGS, SSE_REGS, SSE_REGS,
SSE_REGS, SSE_REGS,
+ /* MMX registers */
MMX_REGS, MMX_REGS, MMX_REGS, MMX_REGS, MMX_REGS, MMX_REGS,
MMX_REGS, MMX_REGS,
+ /* REX registers */
NON_Q_REGS, NON_Q_REGS, NON_Q_REGS, NON_Q_REGS,
NON_Q_REGS, NON_Q_REGS, NON_Q_REGS, NON_Q_REGS,
+ /* SSE REX registers */
SSE_REGS, SSE_REGS, SSE_REGS, SSE_REGS, SSE_REGS, SSE_REGS,
SSE_REGS, SSE_REGS,
};
char internal_label_prefix[16];
int internal_label_prefix_len;
+/* Fence to use after loop using movnt. */
+tree x86_mfence;
+
/* Register class used for passing given 64bit part of the argument.
These represent classes as documented by the PS ABI, with the exception
of SSESF, SSEDF classes, that are basically SSE class, just gcc will
\f
static struct machine_function * ix86_init_machine_status (void);
-static rtx ix86_function_value (tree, tree, bool);
-static int ix86_function_regparm (tree, tree);
+static rtx ix86_function_value (const_tree, const_tree, bool);
+static int ix86_function_regparm (const_tree, const_tree);
static void ix86_compute_frame_layout (struct ix86_frame *);
static bool ix86_expand_vector_init_one_nonzero (bool, enum machine_mode,
rtx, rtx, int);
#define DEFAULT_PCC_STRUCT_RETURN 1
#endif
+/* Bit flags that specify the ISA we are compiling for. */
+int ix86_isa_flags = TARGET_64BIT_DEFAULT | TARGET_SUBTARGET_ISA_DEFAULT;
+
+/* A mask of ix86_isa_flags that includes bit X if X
+ was set or cleared on the command line. */
+static int ix86_isa_flags_explicit;
+
+/* Define a set of ISAs which are available when a given ISA is
+ enabled. MMX and SSE ISAs are handled separately. */
+
+#define OPTION_MASK_ISA_MMX_SET OPTION_MASK_ISA_MMX
+#define OPTION_MASK_ISA_3DNOW_SET \
+ (OPTION_MASK_ISA_3DNOW | OPTION_MASK_ISA_MMX_SET)
+
+#define OPTION_MASK_ISA_SSE_SET OPTION_MASK_ISA_SSE
+#define OPTION_MASK_ISA_SSE2_SET \
+ (OPTION_MASK_ISA_SSE2 | OPTION_MASK_ISA_SSE_SET)
+#define OPTION_MASK_ISA_SSE3_SET \
+ (OPTION_MASK_ISA_SSE3 | OPTION_MASK_ISA_SSE2_SET)
+#define OPTION_MASK_ISA_SSSE3_SET \
+ (OPTION_MASK_ISA_SSSE3 | OPTION_MASK_ISA_SSE3_SET)
+#define OPTION_MASK_ISA_SSE4_1_SET \
+ (OPTION_MASK_ISA_SSE4_1 | OPTION_MASK_ISA_SSSE3_SET)
+#define OPTION_MASK_ISA_SSE4_2_SET \
+ (OPTION_MASK_ISA_SSE4_2 | OPTION_MASK_ISA_SSE4_1_SET)
+
+/* SSE4 includes both SSE4.1 and SSE4.2. -msse4 should be the same
+ as -msse4.2. */
+#define OPTION_MASK_ISA_SSE4_SET OPTION_MASK_ISA_SSE4_2_SET
+
+#define OPTION_MASK_ISA_SSE4A_SET \
+ (OPTION_MASK_ISA_SSE4A | OPTION_MASK_ISA_SSE3_SET)
+#define OPTION_MASK_ISA_SSE5_SET \
+ (OPTION_MASK_ISA_SSE5 | OPTION_MASK_ISA_SSE4A_SET)
+
+/* Define a set of ISAs which aren't available when a given ISA is
+ disabled. MMX and SSE ISAs are handled separately. */
+
+#define OPTION_MASK_ISA_MMX_UNSET \
+ (OPTION_MASK_ISA_MMX | OPTION_MASK_ISA_3DNOW_UNSET)
+#define OPTION_MASK_ISA_3DNOW_UNSET \
+ (OPTION_MASK_ISA_3DNOW | OPTION_MASK_ISA_3DNOW_A_UNSET)
+#define OPTION_MASK_ISA_3DNOW_A_UNSET OPTION_MASK_ISA_3DNOW_A
+
+#define OPTION_MASK_ISA_SSE_UNSET \
+ (OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_SSE2_UNSET)
+#define OPTION_MASK_ISA_SSE2_UNSET \
+ (OPTION_MASK_ISA_SSE2 | OPTION_MASK_ISA_SSE3_UNSET)
+#define OPTION_MASK_ISA_SSE3_UNSET \
+ (OPTION_MASK_ISA_SSE3 \
+ | OPTION_MASK_ISA_SSSE3_UNSET \
+ | OPTION_MASK_ISA_SSE4A_UNSET )
+#define OPTION_MASK_ISA_SSSE3_UNSET \
+ (OPTION_MASK_ISA_SSSE3 | OPTION_MASK_ISA_SSE4_1_UNSET)
+#define OPTION_MASK_ISA_SSE4_1_UNSET \
+ (OPTION_MASK_ISA_SSE4_1 | OPTION_MASK_ISA_SSE4_2_UNSET)
+#define OPTION_MASK_ISA_SSE4_2_UNSET OPTION_MASK_ISA_SSE4_2
+
+/* SSE4 includes both SSE4.1 and SSE4.2. -mno-sse4 should the same
+ as -mno-sse4.1. */
+#define OPTION_MASK_ISA_SSE4_UNSET OPTION_MASK_ISA_SSE4_1_UNSET
+
+#define OPTION_MASK_ISA_SSE4A_UNSET \
+ (OPTION_MASK_ISA_SSE4A | OPTION_MASK_ISA_SSE5_UNSET)
+
+#define OPTION_MASK_ISA_SSE5_UNSET OPTION_MASK_ISA_SSE5
+
+/* Vectorization library interface and handlers. */
+tree (*ix86_veclib_handler)(enum built_in_function, tree, tree) = NULL;
+static tree ix86_veclibabi_svml (enum built_in_function, tree, tree);
+static tree ix86_veclibabi_acml (enum built_in_function, tree, tree);
+
/* Implement TARGET_HANDLE_OPTION. */
static bool
{
switch (code)
{
- case OPT_m3dnow:
- if (!value)
+ case OPT_mmmx:
+ if (value)
+ {
+ ix86_isa_flags |= OPTION_MASK_ISA_MMX_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_MMX_SET;
+ }
+ else
{
- target_flags &= ~MASK_3DNOW_A;
- target_flags_explicit |= MASK_3DNOW_A;
+ ix86_isa_flags &= ~OPTION_MASK_ISA_MMX_UNSET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_MMX_UNSET;
}
return true;
- case OPT_mmmx:
- if (!value)
+ case OPT_m3dnow:
+ if (value)
+ {
+ ix86_isa_flags |= OPTION_MASK_ISA_3DNOW_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_3DNOW_SET;
+ }
+ else
{
- target_flags &= ~(MASK_3DNOW | MASK_3DNOW_A);
- target_flags_explicit |= MASK_3DNOW | MASK_3DNOW_A;
+ ix86_isa_flags &= ~OPTION_MASK_ISA_3DNOW_UNSET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_3DNOW_UNSET;
}
return true;
+ case OPT_m3dnowa:
+ return false;
+
case OPT_msse:
- if (!value)
+ if (value)
{
- target_flags &= ~(MASK_SSE2 | MASK_SSE3 | MASK_SSSE3
- | MASK_SSE4A);
- target_flags_explicit |= (MASK_SSE2 | MASK_SSE3 | MASK_SSSE3
- | MASK_SSE4A);
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE_SET;
+ }
+ else
+ {
+ ix86_isa_flags &= ~OPTION_MASK_ISA_SSE_UNSET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE_UNSET;
}
return true;
case OPT_msse2:
- if (!value)
+ if (value)
+ {
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE2_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE2_SET;
+ }
+ else
{
- target_flags &= ~(MASK_SSE3 | MASK_SSSE3 | MASK_SSE4A);
- target_flags_explicit |= MASK_SSE3 | MASK_SSSE3 | MASK_SSE4A;
+ ix86_isa_flags &= ~OPTION_MASK_ISA_SSE2_UNSET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE2_UNSET;
}
return true;
case OPT_msse3:
- if (!value)
+ if (value)
+ {
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE3_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE3_SET;
+ }
+ else
{
- target_flags &= ~(MASK_SSSE3 | MASK_SSE4A);
- target_flags_explicit |= MASK_SSSE3 | MASK_SSE4A;
+ ix86_isa_flags &= ~OPTION_MASK_ISA_SSE3_UNSET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE3_UNSET;
}
return true;
case OPT_mssse3:
- if (!value)
+ if (value)
+ {
+ ix86_isa_flags |= OPTION_MASK_ISA_SSSE3_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSSE3_SET;
+ }
+ else
+ {
+ ix86_isa_flags &= ~OPTION_MASK_ISA_SSSE3_UNSET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSSE3_UNSET;
+ }
+ return true;
+
+ case OPT_msse4_1:
+ if (value)
+ {
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE4_1_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE4_1_SET;
+ }
+ else
+ {
+ ix86_isa_flags &= ~OPTION_MASK_ISA_SSE4_1_UNSET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE4_1_UNSET;
+ }
+ return true;
+
+ case OPT_msse4_2:
+ if (value)
+ {
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE4_2_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE4_2_SET;
+ }
+ else
+ {
+ ix86_isa_flags &= ~OPTION_MASK_ISA_SSE4_2_UNSET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE4_2_UNSET;
+ }
+ return true;
+
+ case OPT_msse4:
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE4_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE4_SET;
+ return true;
+
+ case OPT_mno_sse4:
+ ix86_isa_flags &= ~OPTION_MASK_ISA_SSE4_UNSET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE4_UNSET;
+ return true;
+
+ case OPT_msse4a:
+ if (value)
+ {
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE4A_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE4A_SET;
+ }
+ else
+ {
+ ix86_isa_flags &= ~OPTION_MASK_ISA_SSE4A_UNSET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE4A_UNSET;
+ }
+ return true;
+
+ case OPT_msse5:
+ if (value)
+ {
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE5_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE5_SET;
+ }
+ else
{
- target_flags &= ~MASK_SSE4A;
- target_flags_explicit |= MASK_SSE4A;
+ ix86_isa_flags &= ~OPTION_MASK_ISA_SSE5_UNSET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE5_UNSET;
}
return true;
{
int i;
int ix86_tune_defaulted = 0;
+ int ix86_arch_specified = 0;
unsigned int ix86_arch_mask, ix86_tune_mask;
/* Comes from final.c -- no real reason to change it. */
static struct ptt
{
const struct processor_costs *cost; /* Processor costs */
- const int target_enable; /* Target flags to enable. */
- const int target_disable; /* Target flags to disable. */
const int align_loop; /* Default alignments. */
const int align_loop_max_skip;
const int align_jump;
}
const processor_target_table[PROCESSOR_max] =
{
- {&i386_cost, 0, 0, 4, 3, 4, 3, 4},
- {&i486_cost, 0, 0, 16, 15, 16, 15, 16},
- {&pentium_cost, 0, 0, 16, 7, 16, 7, 16},
- {&pentiumpro_cost, 0, 0, 16, 15, 16, 7, 16},
- {&geode_cost, 0, 0, 0, 0, 0, 0, 0},
- {&k6_cost, 0, 0, 32, 7, 32, 7, 32},
- {&athlon_cost, 0, 0, 16, 7, 16, 7, 16},
- {&pentium4_cost, 0, 0, 0, 0, 0, 0, 0},
- {&k8_cost, 0, 0, 16, 7, 16, 7, 16},
- {&nocona_cost, 0, 0, 0, 0, 0, 0, 0},
- {&core2_cost, 0, 0, 16, 7, 16, 7, 16},
- {&generic32_cost, 0, 0, 16, 7, 16, 7, 16},
- {&generic64_cost, 0, 0, 16, 7, 16, 7, 16},
- {&amdfam10_cost, 0, 0, 32, 7, 32, 7, 32}
+ {&i386_cost, 4, 3, 4, 3, 4},
+ {&i486_cost, 16, 15, 16, 15, 16},
+ {&pentium_cost, 16, 7, 16, 7, 16},
+ {&pentiumpro_cost, 16, 15, 16, 10, 16},
+ {&geode_cost, 0, 0, 0, 0, 0},
+ {&k6_cost, 32, 7, 32, 7, 32},
+ {&athlon_cost, 16, 7, 16, 7, 16},
+ {&pentium4_cost, 0, 0, 0, 0, 0},
+ {&k8_cost, 16, 7, 16, 7, 16},
+ {&nocona_cost, 0, 0, 0, 0, 0},
+ {&core2_cost, 16, 10, 16, 10, 16},
+ {&generic32_cost, 16, 7, 16, 7, 16},
+ {&generic64_cost, 16, 10, 16, 10, 16},
+ {&amdfam10_cost, 32, 24, 32, 7, 32}
+ };
+
+ static const char *const cpu_names[TARGET_CPU_DEFAULT_max] =
+ {
+ "generic",
+ "i386",
+ "i486",
+ "pentium",
+ "pentium-mmx",
+ "pentiumpro",
+ "pentium2",
+ "pentium3",
+ "pentium4",
+ "pentium-m",
+ "prescott",
+ "nocona",
+ "core2",
+ "geode",
+ "k6",
+ "k6-2",
+ "k6-3",
+ "athlon",
+ "athlon-4",
+ "k8",
+ "amdfam10"
+ };
+
+ enum pta_flags
+ {
+ PTA_SSE = 1 << 0,
+ PTA_SSE2 = 1 << 1,
+ PTA_SSE3 = 1 << 2,
+ PTA_MMX = 1 << 3,
+ PTA_PREFETCH_SSE = 1 << 4,
+ PTA_3DNOW = 1 << 5,
+ PTA_3DNOW_A = 1 << 6,
+ PTA_64BIT = 1 << 7,
+ PTA_SSSE3 = 1 << 8,
+ PTA_CX16 = 1 << 9,
+ PTA_POPCNT = 1 << 10,
+ PTA_ABM = 1 << 11,
+ PTA_SSE4A = 1 << 12,
+ PTA_NO_SAHF = 1 << 13,
+ PTA_SSE4_1 = 1 << 14,
+ PTA_SSE4_2 = 1 << 15,
+ PTA_SSE5 = 1 << 16,
+ PTA_AES = 1 << 17,
+ PTA_PCLMUL = 1 << 18
};
- static const char * const cpu_names[] = TARGET_CPU_DEFAULT_NAMES;
static struct pta
{
const char *const name; /* processor name or nickname. */
const enum processor_type processor;
- const enum pta_flags
- {
- PTA_SSE = 1 << 0,
- PTA_SSE2 = 1 << 1,
- PTA_SSE3 = 1 << 2,
- PTA_MMX = 1 << 3,
- PTA_PREFETCH_SSE = 1 << 4,
- PTA_3DNOW = 1 << 5,
- PTA_3DNOW_A = 1 << 6,
- PTA_64BIT = 1 << 7,
- PTA_SSSE3 = 1 << 8,
- PTA_CX16 = 1 << 9,
- PTA_POPCNT = 1 << 10,
- PTA_ABM = 1 << 11,
- PTA_SSE4A = 1 << 12,
- PTA_NO_SAHF = 1 << 13
- } flags;
+ const unsigned /*enum pta_flags*/ flags;
}
const processor_alias_table[] =
{
{"winchip-c6", PROCESSOR_I486, PTA_MMX},
{"winchip2", PROCESSOR_I486, PTA_MMX | PTA_3DNOW},
{"c3", PROCESSOR_I486, PTA_MMX | PTA_3DNOW},
- {"c3-2", PROCESSOR_PENTIUMPRO, PTA_MMX | PTA_PREFETCH_SSE | PTA_SSE},
+ {"c3-2", PROCESSOR_PENTIUMPRO, PTA_MMX | PTA_SSE},
{"i686", PROCESSOR_PENTIUMPRO, 0},
{"pentiumpro", PROCESSOR_PENTIUMPRO, 0},
{"pentium2", PROCESSOR_PENTIUMPRO, PTA_MMX},
- {"pentium3", PROCESSOR_PENTIUMPRO, PTA_MMX | PTA_SSE | PTA_PREFETCH_SSE},
- {"pentium3m", PROCESSOR_PENTIUMPRO, PTA_MMX | PTA_SSE | PTA_PREFETCH_SSE},
- {"pentium-m", PROCESSOR_PENTIUMPRO, PTA_MMX | PTA_SSE | PTA_PREFETCH_SSE | PTA_SSE2},
- {"pentium4", PROCESSOR_PENTIUM4, PTA_SSE | PTA_SSE2
- | PTA_MMX | PTA_PREFETCH_SSE},
- {"pentium4m", PROCESSOR_PENTIUM4, PTA_SSE | PTA_SSE2
- | PTA_MMX | PTA_PREFETCH_SSE},
- {"prescott", PROCESSOR_NOCONA, PTA_SSE | PTA_SSE2 | PTA_SSE3
- | PTA_MMX | PTA_PREFETCH_SSE},
- {"nocona", PROCESSOR_NOCONA, PTA_SSE | PTA_SSE2 | PTA_SSE3 | PTA_64BIT
- | PTA_MMX | PTA_PREFETCH_SSE
- | PTA_CX16 | PTA_NO_SAHF},
- {"core2", PROCESSOR_CORE2, PTA_SSE | PTA_SSE2 | PTA_SSE3 | PTA_SSSE3
- | PTA_64BIT | PTA_MMX
- | PTA_PREFETCH_SSE | PTA_CX16},
- {"geode", PROCESSOR_GEODE, PTA_MMX | PTA_PREFETCH_SSE | PTA_3DNOW
- | PTA_3DNOW_A},
+ {"pentium3", PROCESSOR_PENTIUMPRO, PTA_MMX | PTA_SSE},
+ {"pentium3m", PROCESSOR_PENTIUMPRO, PTA_MMX | PTA_SSE},
+ {"pentium-m", PROCESSOR_PENTIUMPRO, PTA_MMX | PTA_SSE | PTA_SSE2},
+ {"pentium4", PROCESSOR_PENTIUM4, PTA_MMX |PTA_SSE | PTA_SSE2},
+ {"pentium4m", PROCESSOR_PENTIUM4, PTA_MMX | PTA_SSE | PTA_SSE2},
+ {"prescott", PROCESSOR_NOCONA, PTA_MMX | PTA_SSE | PTA_SSE2 | PTA_SSE3},
+ {"nocona", PROCESSOR_NOCONA, (PTA_64BIT
+ | PTA_MMX | PTA_SSE | PTA_SSE2 | PTA_SSE3
+ | PTA_CX16 | PTA_NO_SAHF)},
+ {"core2", PROCESSOR_CORE2, (PTA_64BIT
+ | PTA_MMX | PTA_SSE | PTA_SSE2 | PTA_SSE3
+ | PTA_SSSE3
+ | PTA_CX16)},
+ {"geode", PROCESSOR_GEODE, (PTA_MMX | PTA_3DNOW | PTA_3DNOW_A
+ |PTA_PREFETCH_SSE)},
{"k6", PROCESSOR_K6, PTA_MMX},
{"k6-2", PROCESSOR_K6, PTA_MMX | PTA_3DNOW},
{"k6-3", PROCESSOR_K6, PTA_MMX | PTA_3DNOW},
- {"athlon", PROCESSOR_ATHLON, PTA_MMX | PTA_PREFETCH_SSE | PTA_3DNOW
- | PTA_3DNOW_A},
- {"athlon-tbird", PROCESSOR_ATHLON, PTA_MMX | PTA_PREFETCH_SSE
- | PTA_3DNOW | PTA_3DNOW_A},
- {"athlon-4", PROCESSOR_ATHLON, PTA_MMX | PTA_PREFETCH_SSE | PTA_3DNOW
- | PTA_3DNOW_A | PTA_SSE},
- {"athlon-xp", PROCESSOR_ATHLON, PTA_MMX | PTA_PREFETCH_SSE | PTA_3DNOW
- | PTA_3DNOW_A | PTA_SSE},
- {"athlon-mp", PROCESSOR_ATHLON, PTA_MMX | PTA_PREFETCH_SSE | PTA_3DNOW
- | PTA_3DNOW_A | PTA_SSE},
- {"x86-64", PROCESSOR_K8, PTA_MMX | PTA_PREFETCH_SSE | PTA_64BIT
- | PTA_SSE | PTA_SSE2 | PTA_NO_SAHF},
- {"k8", PROCESSOR_K8, PTA_MMX | PTA_PREFETCH_SSE | PTA_3DNOW | PTA_64BIT
- | PTA_3DNOW_A | PTA_SSE | PTA_SSE2
- | PTA_NO_SAHF},
- {"opteron", PROCESSOR_K8, PTA_MMX | PTA_PREFETCH_SSE | PTA_3DNOW
- | PTA_64BIT | PTA_3DNOW_A | PTA_SSE
- | PTA_SSE2 | PTA_NO_SAHF},
- {"athlon64", PROCESSOR_K8, PTA_MMX | PTA_PREFETCH_SSE | PTA_3DNOW
- | PTA_64BIT | PTA_3DNOW_A | PTA_SSE
- | PTA_SSE2 | PTA_NO_SAHF},
- {"athlon-fx", PROCESSOR_K8, PTA_MMX | PTA_PREFETCH_SSE | PTA_3DNOW
- | PTA_64BIT | PTA_3DNOW_A | PTA_SSE
- | PTA_SSE2 | PTA_NO_SAHF},
- {"amdfam10", PROCESSOR_AMDFAM10, PTA_MMX | PTA_PREFETCH_SSE | PTA_3DNOW
- | PTA_64BIT | PTA_3DNOW_A | PTA_SSE
- | PTA_SSE2 | PTA_SSE3 | PTA_POPCNT
- | PTA_ABM | PTA_SSE4A | PTA_CX16},
- {"barcelona", PROCESSOR_AMDFAM10, PTA_MMX | PTA_PREFETCH_SSE | PTA_3DNOW
- | PTA_64BIT | PTA_3DNOW_A | PTA_SSE
- | PTA_SSE2 | PTA_SSE3 | PTA_POPCNT
- | PTA_ABM | PTA_SSE4A | PTA_CX16},
+ {"athlon", PROCESSOR_ATHLON, (PTA_MMX | PTA_3DNOW | PTA_3DNOW_A
+ | PTA_PREFETCH_SSE)},
+ {"athlon-tbird", PROCESSOR_ATHLON, (PTA_MMX | PTA_3DNOW | PTA_3DNOW_A
+ | PTA_PREFETCH_SSE)},
+ {"athlon-4", PROCESSOR_ATHLON, (PTA_MMX | PTA_3DNOW | PTA_3DNOW_A
+ | PTA_SSE)},
+ {"athlon-xp", PROCESSOR_ATHLON, (PTA_MMX | PTA_3DNOW | PTA_3DNOW_A
+ | PTA_SSE)},
+ {"athlon-mp", PROCESSOR_ATHLON, (PTA_MMX | PTA_3DNOW | PTA_3DNOW_A
+ | PTA_SSE)},
+ {"x86-64", PROCESSOR_K8, (PTA_64BIT
+ | PTA_MMX | PTA_SSE | PTA_SSE2
+ | PTA_NO_SAHF)},
+ {"k8", PROCESSOR_K8, (PTA_64BIT
+ | PTA_MMX | PTA_3DNOW | PTA_3DNOW_A
+ | PTA_SSE | PTA_SSE2
+ | PTA_NO_SAHF)},
+ {"k8-sse3", PROCESSOR_K8, (PTA_64BIT
+ | PTA_MMX | PTA_3DNOW | PTA_3DNOW_A
+ | PTA_SSE | PTA_SSE2 | PTA_SSE3
+ | PTA_NO_SAHF)},
+ {"opteron", PROCESSOR_K8, (PTA_64BIT
+ | PTA_MMX | PTA_3DNOW | PTA_3DNOW_A
+ | PTA_SSE | PTA_SSE2
+ | PTA_NO_SAHF)},
+ {"opteron-sse3", PROCESSOR_K8, (PTA_64BIT
+ | PTA_MMX | PTA_3DNOW | PTA_3DNOW_A
+ | PTA_SSE | PTA_SSE2 | PTA_SSE3
+ | PTA_NO_SAHF)},
+ {"athlon64", PROCESSOR_K8, (PTA_64BIT
+ | PTA_MMX | PTA_3DNOW | PTA_3DNOW_A
+ | PTA_SSE | PTA_SSE2
+ | PTA_NO_SAHF)},
+ {"athlon64-sse3", PROCESSOR_K8, (PTA_64BIT
+ | PTA_MMX | PTA_3DNOW | PTA_3DNOW_A
+ | PTA_SSE | PTA_SSE2 | PTA_SSE3
+ | PTA_NO_SAHF)},
+ {"athlon-fx", PROCESSOR_K8, (PTA_64BIT
+ | PTA_MMX | PTA_3DNOW | PTA_3DNOW_A
+ | PTA_SSE | PTA_SSE2
+ | PTA_NO_SAHF)},
+ {"amdfam10", PROCESSOR_AMDFAM10, (PTA_64BIT
+ | PTA_MMX | PTA_3DNOW | PTA_3DNOW_A
+ | PTA_SSE | PTA_SSE2 | PTA_SSE3
+ | PTA_SSE4A
+ | PTA_CX16 | PTA_ABM)},
+ {"barcelona", PROCESSOR_AMDFAM10, (PTA_64BIT
+ | PTA_MMX | PTA_3DNOW | PTA_3DNOW_A
+ | PTA_SSE | PTA_SSE2 | PTA_SSE3
+ | PTA_SSE4A
+ | PTA_CX16 | PTA_ABM)},
{"generic32", PROCESSOR_GENERIC32, 0 /* flags are only used for -march switch. */ },
{"generic64", PROCESSOR_GENERIC64, PTA_64BIT /* flags are only used for -march switch. */ },
};
ix86_tune_string = ix86_arch_string;
if (!ix86_tune_string)
{
- ix86_tune_string = cpu_names [TARGET_CPU_DEFAULT];
+ ix86_tune_string = cpu_names[TARGET_CPU_DEFAULT];
ix86_tune_defaulted = 1;
}
if (!ix86_arch_string)
ix86_arch_string = TARGET_64BIT ? "x86-64" : "i386";
+ else
+ ix86_arch_specified = 1;
+
if (!strcmp (ix86_arch_string, "generic"))
error ("generic CPU can be used only for -mtune= switch");
if (!strncmp (ix86_arch_string, "generic", 7))
if ((TARGET_64BIT == 0) != (ix86_cmodel == CM_32))
error ("code model %qs not supported in the %s bit mode",
ix86_cmodel_string, TARGET_64BIT ? "64" : "32");
- if ((TARGET_64BIT != 0) != ((target_flags & MASK_64BIT) != 0))
+ if ((TARGET_64BIT != 0) != ((ix86_isa_flags & OPTION_MASK_ISA_64BIT) != 0))
sorry ("%i-bit mode not compiled in",
- (target_flags & MASK_64BIT) ? 64 : 32);
+ (ix86_isa_flags & OPTION_MASK_ISA_64BIT) ? 64 : 32);
for (i = 0; i < pta_size; i++)
if (! strcmp (ix86_arch_string, processor_alias_table[i].name))
ix86_arch = processor_alias_table[i].processor;
/* Default cpu tuning to the architecture. */
ix86_tune = ix86_arch;
+
+ if (TARGET_64BIT && !(processor_alias_table[i].flags & PTA_64BIT))
+ error ("CPU you selected does not support x86-64 "
+ "instruction set");
+
if (processor_alias_table[i].flags & PTA_MMX
- && !(target_flags_explicit & MASK_MMX))
- target_flags |= MASK_MMX;
+ && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_MMX))
+ ix86_isa_flags |= OPTION_MASK_ISA_MMX;
if (processor_alias_table[i].flags & PTA_3DNOW
- && !(target_flags_explicit & MASK_3DNOW))
- target_flags |= MASK_3DNOW;
+ && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_3DNOW))
+ ix86_isa_flags |= OPTION_MASK_ISA_3DNOW;
if (processor_alias_table[i].flags & PTA_3DNOW_A
- && !(target_flags_explicit & MASK_3DNOW_A))
- target_flags |= MASK_3DNOW_A;
+ && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_3DNOW_A))
+ ix86_isa_flags |= OPTION_MASK_ISA_3DNOW_A;
if (processor_alias_table[i].flags & PTA_SSE
- && !(target_flags_explicit & MASK_SSE))
- target_flags |= MASK_SSE;
+ && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_SSE))
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE;
if (processor_alias_table[i].flags & PTA_SSE2
- && !(target_flags_explicit & MASK_SSE2))
- target_flags |= MASK_SSE2;
+ && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_SSE2))
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE2;
if (processor_alias_table[i].flags & PTA_SSE3
- && !(target_flags_explicit & MASK_SSE3))
- target_flags |= MASK_SSE3;
+ && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_SSE3))
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE3;
if (processor_alias_table[i].flags & PTA_SSSE3
- && !(target_flags_explicit & MASK_SSSE3))
- target_flags |= MASK_SSSE3;
- if (processor_alias_table[i].flags & PTA_PREFETCH_SSE)
- x86_prefetch_sse = true;
+ && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_SSSE3))
+ ix86_isa_flags |= OPTION_MASK_ISA_SSSE3;
+ if (processor_alias_table[i].flags & PTA_SSE4_1
+ && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_SSE4_1))
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE4_1;
+ if (processor_alias_table[i].flags & PTA_SSE4_2
+ && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_SSE4_2))
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE4_2;
+ if (processor_alias_table[i].flags & PTA_SSE4A
+ && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_SSE4A))
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE4A;
+ if (processor_alias_table[i].flags & PTA_SSE5
+ && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_SSE5))
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE5;
+
+ if (processor_alias_table[i].flags & PTA_ABM)
+ x86_abm = true;
if (processor_alias_table[i].flags & PTA_CX16)
x86_cmpxchg16b = true;
- if (processor_alias_table[i].flags & PTA_POPCNT
- && !(target_flags_explicit & MASK_POPCNT))
- target_flags |= MASK_POPCNT;
- if (processor_alias_table[i].flags & PTA_ABM
- && !(target_flags_explicit & MASK_ABM))
- target_flags |= MASK_ABM;
- if (processor_alias_table[i].flags & PTA_SSE4A
- && !(target_flags_explicit & MASK_SSE4A))
- target_flags |= MASK_SSE4A;
+ if (processor_alias_table[i].flags & (PTA_POPCNT | PTA_ABM))
+ x86_popcnt = true;
+ if (processor_alias_table[i].flags & (PTA_PREFETCH_SSE | PTA_SSE))
+ x86_prefetch_sse = true;
if (!(TARGET_64BIT && (processor_alias_table[i].flags & PTA_NO_SAHF)))
x86_sahf = true;
- if (TARGET_64BIT && !(processor_alias_table[i].flags & PTA_64BIT))
- error ("CPU you selected does not support x86-64 "
- "instruction set");
+ if (processor_alias_table[i].flags & PTA_AES)
+ x86_aes = true;
+ if (processor_alias_table[i].flags & PTA_PCLMUL)
+ x86_pclmul = true;
+
break;
}
-mtune (rather than -march) points us to a processor that has them.
However, the VIA C3 gives a SIGILL, so we only do that for i686 and
higher processors. */
- if (TARGET_CMOVE && (processor_alias_table[i].flags & PTA_PREFETCH_SSE))
+ if (TARGET_CMOVE
+ && (processor_alias_table[i].flags & (PTA_PREFETCH_SSE | PTA_SSE)))
x86_prefetch_sse = true;
break;
}
if (i == pta_size)
error ("bad value (%s) for -mtune= switch", ix86_tune_string);
+ /* Enable SSE2 if AES or PCLMUL is enabled. */
+ if ((x86_aes || x86_pclmul)
+ && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_SSE2))
+ {
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE2_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE2_SET;
+ }
+
ix86_tune_mask = 1u << ix86_tune;
for (i = 0; i < X86_TUNE_LAST; ++i)
ix86_tune_features[i] &= ix86_tune_mask;
ix86_cost = &size_cost;
else
ix86_cost = processor_target_table[ix86_tune].cost;
- target_flags |= processor_target_table[ix86_tune].target_enable;
- target_flags &= ~processor_target_table[ix86_tune].target_disable;
/* Arrange to set up i386_stack_locals for all functions. */
init_machine_status = ix86_init_machine_status;
error ("pc%d is not valid precision setting (32, 64 or 80)", i);
}
+ if (TARGET_64BIT)
+ {
+ target_flags |= TARGET_SUBTARGET64_DEFAULT & ~target_flags_explicit;
+
+ /* Enable by default the SSE and MMX builtins. Do allow the user to
+ explicitly disable any of these. In particular, disabling SSE and
+ MMX for kernel code is extremely useful. */
+ if (!ix86_arch_specified)
+ ix86_isa_flags
+ |= ((OPTION_MASK_ISA_SSE2 | OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_MMX
+ | TARGET_SUBTARGET64_ISA_DEFAULT) & ~ix86_isa_flags_explicit);
+
+ if (TARGET_RTD)
+ warning (0, "-mrtd is ignored in 64bit mode");
+ }
+ else
+ {
+ target_flags |= TARGET_SUBTARGET32_DEFAULT & ~target_flags_explicit;
+
+ if (!ix86_arch_specified)
+ ix86_isa_flags
+ |= TARGET_SUBTARGET32_ISA_DEFAULT & ~ix86_isa_flags_explicit;
+
+ /* i386 ABI does not specify red zone. It still makes sense to use it
+ when programmer takes care to stack from being destroyed. */
+ if (!(target_flags_explicit & MASK_NO_RED_ZONE))
+ target_flags |= MASK_NO_RED_ZONE;
+ }
+
/* Keep nonleaf frame pointers. */
if (flag_omit_frame_pointer)
target_flags &= ~MASK_OMIT_LEAF_FRAME_POINTER;
if (!TARGET_80387)
target_flags |= MASK_NO_FANCY_MATH_387;
- /* Turn on SSE3 builtins for -mssse3. */
- if (TARGET_SSSE3)
- target_flags |= MASK_SSE3;
-
- /* Turn on SSE3 builtins for -msse4a. */
- if (TARGET_SSE4A)
- target_flags |= MASK_SSE3;
-
- /* Turn on SSE2 builtins for -msse3. */
- if (TARGET_SSE3)
- target_flags |= MASK_SSE2;
-
- /* Turn on SSE builtins for -msse2. */
- if (TARGET_SSE2)
- target_flags |= MASK_SSE;
-
/* Turn on MMX builtins for -msse. */
if (TARGET_SSE)
{
- target_flags |= MASK_MMX & ~target_flags_explicit;
+ ix86_isa_flags |= OPTION_MASK_ISA_MMX & ~ix86_isa_flags_explicit;
x86_prefetch_sse = true;
}
- /* Turn on MMX builtins for 3Dnow. */
- if (TARGET_3DNOW)
- target_flags |= MASK_MMX;
-
- /* Turn on POPCNT builtins for -mabm. */
- if (TARGET_ABM)
- target_flags |= MASK_POPCNT;
-
- if (TARGET_64BIT)
- {
- if (TARGET_RTD)
- warning (0, "-mrtd is ignored in 64bit mode");
-
- /* Enable by default the SSE and MMX builtins. Do allow the user to
- explicitly disable any of these. In particular, disabling SSE and
- MMX for kernel code is extremely useful. */
- target_flags
- |= ((MASK_SSE2 | MASK_SSE | MASK_MMX | TARGET_SUBTARGET64_DEFAULT)
- & ~target_flags_explicit);
- }
- else
- {
- /* i386 ABI does not specify red zone. It still makes sense to use it
- when programmer takes care to stack from being destroyed. */
- if (!(target_flags_explicit & MASK_NO_RED_ZONE))
- target_flags |= MASK_NO_RED_ZONE;
- }
+ /* Turn on popcnt instruction for -msse4.2 or -mabm. */
+ if (TARGET_SSE4_2 || TARGET_ABM)
+ x86_popcnt = true;
/* Validate -mpreferred-stack-boundary= value, or provide default.
The default of 128 bits is for Pentium III's SSE __m128. We can't
ix86_fpmath = FPMATH_SSE;
}
else
- ix86_fpmath = FPMATH_SSE | FPMATH_387;
+ ix86_fpmath = (enum fpmath_unit) (FPMATH_SSE | FPMATH_387);
}
else
error ("bad value (%s) for -mfpmath= switch", ix86_fpmath_string);
if (!TARGET_80387)
target_flags &= ~MASK_FLOAT_RETURNS;
+ /* Use external vectorized library in vectorizing intrinsics. */
+ if (ix86_veclibabi_string)
+ {
+ if (strcmp (ix86_veclibabi_string, "svml") == 0)
+ ix86_veclib_handler = ix86_veclibabi_svml;
+ else if (strcmp (ix86_veclibabi_string, "acml") == 0)
+ ix86_veclib_handler = ix86_veclibabi_acml;
+ else
+ error ("unknown vectorization library ABI type (%s) for "
+ "-mveclibabi= switch", ix86_veclibabi_string);
+ }
+
if ((x86_accumulate_outgoing_args & ix86_tune_mask)
&& !(target_flags_explicit & MASK_ACCUMULATE_OUTGOING_ARGS)
&& !optimize_size)
target_flags |= MASK_ACCUMULATE_OUTGOING_ARGS;
}
+ /* If stack probes are required, the space used for large function
+ arguments on the stack must also be probed, so enable
+ -maccumulate-outgoing-args so this happens in the prologue. */
+ if (TARGET_STACK_PROBE
+ && !(target_flags & MASK_ACCUMULATE_OUTGOING_ARGS))
+ {
+ if (target_flags_explicit & MASK_ACCUMULATE_OUTGOING_ARGS)
+ warning (0, "stack probing requires -maccumulate-outgoing-args "
+ "for correctness");
+ target_flags |= MASK_ACCUMULATE_OUTGOING_ARGS;
+ }
+
/* For sane SSE instruction set generation we need fcomi instruction.
It is safe to enable all CMOVE instructions. */
if (TARGET_SSE)
ix86_cost->simultaneous_prefetches);
if (!PARAM_SET_P (PARAM_L1_CACHE_LINE_SIZE))
set_param_value ("l1-cache-line-size", ix86_cost->prefetch_block);
+ if (!PARAM_SET_P (PARAM_L1_CACHE_SIZE))
+ set_param_value ("l1-cache-size", ix86_cost->l1_cache_size);
+ if (!PARAM_SET_P (PARAM_L2_CACHE_SIZE))
+ set_param_value ("l2-cache-size", ix86_cost->l2_cache_size);
+
+ /* If using typedef char *va_list, signal that __builtin_va_start (&ap, 0)
+ can be optimized to ap = __builtin_next_arg (0). */
+ if (!TARGET_64BIT || TARGET_64BIT_MS_ABI)
+ targetm.expand_builtin_va_start = NULL;
}
\f
/* Return true if this goes in large data/bss. */
/* We don't split these for medium model. Place them into
default sections and hope for best. */
break;
+ case SECCAT_EMUTLS_VAR:
+ case SECCAT_EMUTLS_TMPL:
+ gcc_unreachable ();
}
if (sname)
{
case SECCAT_DATA_REL_LOCAL:
case SECCAT_DATA_REL_RO:
case SECCAT_DATA_REL_RO_LOCAL:
- prefix = one_only ? ".gnu.linkonce.ld." : ".ldata.";
+ prefix = one_only ? ".ld" : ".ldata";
break;
case SECCAT_BSS:
- prefix = one_only ? ".gnu.linkonce.lb." : ".lbss.";
+ prefix = one_only ? ".lb" : ".lbss";
break;
case SECCAT_RODATA:
case SECCAT_RODATA_MERGE_STR:
case SECCAT_RODATA_MERGE_STR_INIT:
case SECCAT_RODATA_MERGE_CONST:
- prefix = one_only ? ".gnu.linkonce.lr." : ".lrodata.";
+ prefix = one_only ? ".lr" : ".lrodata";
break;
case SECCAT_SRODATA:
case SECCAT_SDATA:
/* We don't split these for medium model. Place them into
default sections and hope for best. */
break;
+ case SECCAT_EMUTLS_VAR:
+ prefix = targetm.emutls.var_section;
+ break;
+ case SECCAT_EMUTLS_TMPL:
+ prefix = targetm.emutls.tmpl_section;
+ break;
}
if (prefix)
{
- const char *name;
- size_t nlen, plen;
+ const char *name, *linkonce;
char *string;
- plen = strlen (prefix);
name = IDENTIFIER_POINTER (DECL_ASSEMBLER_NAME (decl));
name = targetm.strip_name_encoding (name);
- nlen = strlen (name);
-
- string = alloca (nlen + plen + 1);
- memcpy (string, prefix, plen);
- memcpy (string + plen, name, nlen + 1);
-
- DECL_SECTION_NAME (decl) = build_string (nlen + plen, string);
+
+ /* If we're using one_only, then there needs to be a .gnu.linkonce
+ prefix to the section name. */
+ linkonce = one_only ? ".gnu.linkonce" : "";
+
+ string = ACONCAT ((linkonce, prefix, ".", name, NULL));
+
+ DECL_SECTION_NAME (decl) = build_string (strlen (string), string);
return;
}
}
flag_omit_frame_pointer = 2;
flag_pcc_struct_return = 2;
flag_asynchronous_unwind_tables = 2;
+ flag_vect_cost_model = 1;
#ifdef SUBTARGET_OPTIMIZATION_OPTIONS
SUBTARGET_OPTIMIZATION_OPTIONS;
#endif
warning to be generated). */
static int
-ix86_comp_type_attributes (tree type1, tree type2)
+ix86_comp_type_attributes (const_tree type1, const_tree type2)
{
/* Check for mismatch of non-default calling convention. */
const char *const rtdstr = TARGET_RTD ? "cdecl" : "stdcall";
- if (TREE_CODE (type1) != FUNCTION_TYPE)
+ if (TREE_CODE (type1) != FUNCTION_TYPE
+ && TREE_CODE (type1) != METHOD_TYPE)
return 1;
/* Check for mismatched fastcall/regparm types. */
or considering a libcall. */
static int
-ix86_function_regparm (tree type, tree decl)
+ix86_function_regparm (const_tree type, const_tree decl)
{
tree attr;
int regparm = ix86_regparm;
+ static bool error_issued;
+
if (TARGET_64BIT)
return regparm;
attr = lookup_attribute ("regparm", TYPE_ATTRIBUTES (type));
if (attr)
- return TREE_INT_CST_LOW (TREE_VALUE (TREE_VALUE (attr)));
+ {
+ regparm
+ = TREE_INT_CST_LOW (TREE_VALUE (TREE_VALUE (attr)));
+
+ if (decl && TREE_CODE (decl) == FUNCTION_DECL)
+ {
+ /* We can't use regparm(3) for nested functions because
+ these pass static chain pointer in %ecx register. */
+ if (!error_issued && regparm == 3
+ && decl_function_context (decl)
+ && !DECL_NO_STATIC_CHAIN (decl))
+ {
+ error ("nested functions are limited to 2 register parameters");
+ error_issued = true;
+ return 0;
+ }
+ }
+
+ return regparm;
+ }
if (lookup_attribute ("fastcall", TYPE_ATTRIBUTES (type)))
return 2;
/* Use register calling convention for local functions when possible. */
- if (decl && flag_unit_at_a_time && !profile_flag)
+ if (decl && TREE_CODE (decl) == FUNCTION_DECL
+ && flag_unit_at_a_time && !profile_flag)
{
- struct cgraph_local_info *i = cgraph_local_info (decl);
+ /* FIXME: remove this CONST_CAST when cgraph.[ch] is constified. */
+ struct cgraph_local_info *i = cgraph_local_info (CONST_CAST_TREE(decl));
if (i && i->local)
{
int local_regparm, globals = 0, regno;
struct function *f;
/* Make sure no regparm register is taken by a
- global register variable. */
- for (local_regparm = 0; local_regparm < 3; local_regparm++)
- if (global_regs[local_regparm])
+ fixed register variable. */
+ for (local_regparm = 0; local_regparm < REGPARM_MAX; local_regparm++)
+ if (fixed_regs[local_regparm])
break;
/* We can't use regparm(3) for nested functions as these use
static chain pointer in third argument. */
if (local_regparm == 3
- && decl_function_context (decl)
+ && (decl_function_context (decl)
+ || ix86_force_align_arg_pointer)
&& !DECL_NO_STATIC_CHAIN (decl))
local_regparm = 2;
TYPE_ATTRIBUTES (TREE_TYPE (decl)))))
local_regparm = 2;
- /* Each global register variable increases register preassure,
- so the more global reg vars there are, the smaller regparm
- optimization use, unless requested by the user explicitly. */
- for (regno = 0; regno < 6; regno++)
- if (global_regs[regno])
+ /* Each fixed register usage increases register pressure,
+ so less registers should be used for argument passing.
+ This functionality can be overriden by an explicit
+ regparm value. */
+ for (regno = 0; regno <= DI_REG; regno++)
+ if (fixed_regs[regno])
globals++;
+
local_regparm
= globals < local_regparm ? local_regparm - globals : 0;
indirectly or considering a libcall. Otherwise return 0. */
static int
-ix86_function_sseregparm (tree type, tree decl)
+ix86_function_sseregparm (const_tree type, const_tree decl, bool warn)
{
gcc_assert (!TARGET_64BIT);
{
if (!TARGET_SSE)
{
- if (decl)
- error ("Calling %qD with attribute sseregparm without "
- "SSE/SSE2 enabled", decl);
- else
- error ("Calling %qT with attribute sseregparm without "
- "SSE/SSE2 enabled", type);
+ if (warn)
+ {
+ if (decl)
+ error ("Calling %qD with attribute sseregparm without "
+ "SSE/SSE2 enabled", decl);
+ else
+ error ("Calling %qT with attribute sseregparm without "
+ "SSE/SSE2 enabled", type);
+ }
return 0;
}
(and DFmode for SSE2) arguments in SSE registers. */
if (decl && TARGET_SSE_MATH && flag_unit_at_a_time && !profile_flag)
{
- struct cgraph_local_info *i = cgraph_local_info (decl);
+ /* FIXME: remove this CONST_CAST when cgraph.[ch] is constified. */
+ struct cgraph_local_info *i = cgraph_local_info (CONST_CAST_TREE(decl));
if (i && i->local)
return TARGET_SSE2 ? 2 : 1;
}
to correct at this point. This gives false positives for broken
functions that might use uninitialized data that happens to be
allocated in eax, but who cares? */
- return REGNO_REG_SET_P (ENTRY_BLOCK_PTR->il.rtl->global_live_at_end, 0);
-}
-
-/* Return true if TYPE has a variable argument list. */
-
-static bool
-type_has_variadic_args_p (tree type)
-{
- tree t;
-
- for (t = TYPE_ARG_TYPES (type); t; t = TREE_CHAIN (t))
- if (t == void_list_node)
- return false;
- return true;
+ return REGNO_REG_SET_P (df_get_live_out (ENTRY_BLOCK_PTR), 0);
}
/* Value is the number of bytes of arguments automatically
|| lookup_attribute ("fastcall", TYPE_ATTRIBUTES (funtype)))
rtd = 1;
- if (rtd && ! type_has_variadic_args_p (funtype))
+ if (rtd && ! stdarg_p (funtype))
return size;
}
}
/* RAX is used as hidden argument to va_arg functions. */
- if (!TARGET_64BIT_MS_ABI && regno == 0)
+ if (!TARGET_64BIT_MS_ABI && regno == AX_REG)
return true;
if (TARGET_64BIT_MS_ABI)
/* Return if we do not know how to pass TYPE solely in registers. */
static bool
-ix86_must_pass_in_stack (enum machine_mode mode, tree type)
+ix86_must_pass_in_stack (enum machine_mode mode, const_tree type)
{
if (must_pass_in_stack_var_size_or_pad (mode, type))
return true;
rtx libname, /* SYMBOL_REF of library name or 0 */
tree fndecl)
{
+ struct cgraph_local_info *i = fndecl ? cgraph_local_info (fndecl) : NULL;
memset (cum, 0, sizeof (*cum));
/* Set up the number of registers to use for passing arguments. */
cum->mmx_nregs = MMX_REGPARM_MAX;
cum->warn_sse = true;
cum->warn_mmx = true;
- cum->maybe_vaarg = (fntype ? type_has_variadic_args_p (fntype) : !libname);
+
+ /* Because type might mismatch in between caller and callee, we need to
+ use actual type of function for local calls.
+ FIXME: cgraph_analyze can be told to actually record if function uses
+ va_start so for local functions maybe_vaarg can be made aggressive
+ helping K&R code.
+ FIXME: once typesytem is fixed, we won't need this code anymore. */
+ if (i && i->local)
+ fntype = TREE_TYPE (fndecl);
+ cum->maybe_vaarg = (fntype
+ ? (!prototype_p (fntype) || stdarg_p (fntype))
+ : !libname);
if (!TARGET_64BIT)
{
/* If there are variable arguments, then we won't pass anything
in registers in 32-bit mode. */
- if (cum->maybe_vaarg)
+ if (stdarg_p (fntype))
{
cum->nregs = 0;
cum->sse_nregs = 0;
/* Set up the number of SSE registers used for passing SFmode
and DFmode arguments. Warn for mismatching ABI. */
- cum->float_in_sse = ix86_function_sseregparm (fntype, fndecl);
+ cum->float_in_sse = ix86_function_sseregparm (fntype, fndecl, true);
}
}
the middle-end decides to do with these vector types. */
static enum machine_mode
-type_natural_mode (tree type)
+type_natural_mode (const_tree type)
{
enum machine_mode mode = TYPE_MODE (type);
*/
static int
-classify_argument (enum machine_mode mode, tree type,
+classify_argument (enum machine_mode mode, const_tree type,
enum x86_64_reg_class classes[MAX_CLASSES], int bit_offset)
{
HOST_WIDE_INT bytes =
}
/* for V1xx modes, just use the base mode */
- if (VECTOR_MODE_P (mode)
+ if (VECTOR_MODE_P (mode) && mode != V1DImode
&& GET_MODE_SIZE (GET_MODE_INNER (mode)) == bytes)
mode = GET_MODE_INNER (mode);
classes[0] = X86_64_SSE_CLASS;
classes[1] = X86_64_SSEUP_CLASS;
return 2;
+ case V1DImode:
case V2SFmode:
case V2SImode:
case V4HImode:
/* Examine the argument and return set number of register required in each
class. Return 0 iff parameter should be passed in memory. */
static int
-examine_argument (enum machine_mode mode, tree type, int in_return,
+examine_argument (enum machine_mode mode, const_tree type, int in_return,
int *int_nregs, int *sse_nregs)
{
- enum x86_64_reg_class class[MAX_CLASSES];
- int n = classify_argument (mode, type, class, 0);
+ enum x86_64_reg_class regclass[MAX_CLASSES];
+ int n = classify_argument (mode, type, regclass, 0);
*int_nregs = 0;
*sse_nregs = 0;
if (!n)
return 0;
for (n--; n >= 0; n--)
- switch (class[n])
+ switch (regclass[n])
{
case X86_64_INTEGER_CLASS:
case X86_64_INTEGERSI_CLASS:
static rtx
construct_container (enum machine_mode mode, enum machine_mode orig_mode,
- tree type, int in_return, int nintregs, int nsseregs,
+ const_tree type, int in_return, int nintregs, int nsseregs,
const int *intreg, int sse_regno)
{
/* The following variables hold the static issued_error state. */
enum machine_mode tmpmode;
int bytes =
(mode == BLKmode) ? int_size_in_bytes (type) : (int) GET_MODE_SIZE (mode);
- enum x86_64_reg_class class[MAX_CLASSES];
+ enum x86_64_reg_class regclass[MAX_CLASSES];
int n;
int i;
int nexps = 0;
rtx exp[MAX_CLASSES];
rtx ret;
- n = classify_argument (mode, type, class, 0);
+ n = classify_argument (mode, type, regclass, 0);
if (!n)
return NULL;
if (!examine_argument (mode, type, in_return, &needed_intregs,
x87 registers and the user specified -mno-80387. */
if (!TARGET_80387 && in_return)
for (i = 0; i < n; i++)
- if (class[i] == X86_64_X87_CLASS
- || class[i] == X86_64_X87UP_CLASS
- || class[i] == X86_64_COMPLEX_X87_CLASS)
+ if (regclass[i] == X86_64_X87_CLASS
+ || regclass[i] == X86_64_X87UP_CLASS
+ || regclass[i] == X86_64_COMPLEX_X87_CLASS)
{
if (!issued_x87_ret_error)
{
/* First construct simple cases. Avoid SCmode, since we want to use
single register to pass this type. */
if (n == 1 && mode != SCmode)
- switch (class[0])
+ switch (regclass[0])
{
case X86_64_INTEGER_CLASS:
case X86_64_INTEGERSI_CLASS:
default:
gcc_unreachable ();
}
- if (n == 2 && class[0] == X86_64_SSE_CLASS && class[1] == X86_64_SSEUP_CLASS
- && mode != BLKmode)
+ if (n == 2 && regclass[0] == X86_64_SSE_CLASS
+ && regclass[1] == X86_64_SSEUP_CLASS && mode != BLKmode)
return gen_rtx_REG (mode, SSE_REGNO (sse_regno));
if (n == 2
- && class[0] == X86_64_X87_CLASS && class[1] == X86_64_X87UP_CLASS)
+ && regclass[0] == X86_64_X87_CLASS && regclass[1] == X86_64_X87UP_CLASS)
return gen_rtx_REG (XFmode, FIRST_STACK_REG);
- if (n == 2 && class[0] == X86_64_INTEGER_CLASS
- && class[1] == X86_64_INTEGER_CLASS
+ if (n == 2 && regclass[0] == X86_64_INTEGER_CLASS
+ && regclass[1] == X86_64_INTEGER_CLASS
&& (mode == CDImode || mode == TImode || mode == TFmode)
&& intreg[0] + 1 == intreg[1])
return gen_rtx_REG (mode, intreg[0]);
/* Otherwise figure out the entries of the PARALLEL. */
for (i = 0; i < n; i++)
{
- switch (class[i])
+ switch (regclass[i])
{
case X86_64_NO_CLASS:
break;
/* Merge TImodes on aligned occasions here too. */
if (i * 8 + 8 > bytes)
tmpmode = mode_for_size ((bytes - i * 8) * BITS_PER_UNIT, MODE_INT, 0);
- else if (class[i] == X86_64_INTEGERSI_CLASS)
+ else if (regclass[i] == X86_64_INTEGERSI_CLASS)
tmpmode = SImode;
else
tmpmode = DImode;
sse_regno++;
break;
case X86_64_SSE_CLASS:
- if (i < n - 1 && class[i + 1] == X86_64_SSEUP_CLASS)
+ if (i < n - 1 && regclass[i + 1] == X86_64_SSEUP_CLASS)
tmpmode = TImode;
else
tmpmode = DImode;
case V4HImode:
case V2SImode:
case V2SFmode:
+ case V1DImode:
if (!type || !AGGREGATE_TYPE_P (type))
{
cum->mmx_words += words;
int regno = cum->regno;
/* Fastcall allocates the first two DWORD (SImode) or
- smaller arguments to ECX and EDX. */
+ smaller arguments to ECX and EDX if it isn't an
+ aggregate type . */
if (cum->fastcall)
{
- if (mode == BLKmode || mode == DImode)
+ if (mode == BLKmode
+ || mode == DImode
+ || (type && AGGREGATE_TYPE_P (type)))
break;
/* ECX not EAX is the first allocated register. */
- if (regno == 0)
- regno = 2;
+ if (regno == AX_REG)
+ regno = CX_REG;
}
return gen_rtx_REG (mode, regno);
}
case V4HImode:
case V2SImode:
case V2SFmode:
+ case V1DImode:
if (!type || !AGGREGATE_TYPE_P (type))
{
if (!TARGET_MMX && !warnedmmx && cum->warn_mmx)
static rtx
function_arg_ms_64 (CUMULATIVE_ARGS *cum, enum machine_mode mode,
- enum machine_mode orig_mode, int named)
+ enum machine_mode orig_mode, int named,
+ HOST_WIDE_INT bytes)
{
unsigned int regno;
return gen_rtx_PARALLEL (mode, gen_rtvec (2, t1, t2));
}
}
+ /* Handle aggregated types passed in register. */
+ if (orig_mode == BLKmode)
+ {
+ if (bytes > 0 && bytes <= 8)
+ mode = (bytes > 4 ? DImode : SImode);
+ if (mode == BLKmode)
+ mode = DImode;
+ }
return gen_reg_or_parallel (mode, orig_mode, regno);
}
mode = type_natural_mode (type);
if (TARGET_64BIT_MS_ABI)
- return function_arg_ms_64 (cum, mode, omode, named);
+ return function_arg_ms_64 (cum, mode, omode, named, bytes);
else if (TARGET_64BIT)
return function_arg_64 (cum, mode, omode, type);
else
static bool
ix86_pass_by_reference (CUMULATIVE_ARGS *cum ATTRIBUTE_UNUSED,
enum machine_mode mode ATTRIBUTE_UNUSED,
- tree type, bool named ATTRIBUTE_UNUSED)
+ const_tree type, bool named ATTRIBUTE_UNUSED)
{
+ /* See Windows x64 Software Convention. */
if (TARGET_64BIT_MS_ABI)
{
+ int msize = (int) GET_MODE_SIZE (mode);
if (type)
{
/* Arrays are passed by reference. */
{
/* Structs/unions of sizes other than 8, 16, 32, or 64 bits
are passed by reference. */
- int el2 = exact_log2 (int_size_in_bytes (type));
- return !(el2 >= 0 && el2 <= 3);
+ msize = int_size_in_bytes (type);
}
}
/* __m128 is passed by reference. */
- /* ??? How to handle complex? For now treat them as structs,
- and pass them by reference if they're too large. */
- if (GET_MODE_SIZE (mode) > 8)
- return true;
+ switch (msize) {
+ case 1: case 2: case 4: case 8:
+ break;
+ default:
+ return true;
+ }
}
else if (TARGET_64BIT && type && int_size_in_bytes (type) == -1)
return 1;
}
/* Return true when TYPE should be 128bit aligned for 32bit argument passing
- ABI. Only called if TARGET_SSE. */
+ ABI. */
static bool
-contains_128bit_aligned_vector_p (tree type)
+contains_aligned_value_p (tree type)
{
enum machine_mode mode = TYPE_MODE (type);
- if (SSE_REG_MODE_P (mode)
+ if (((TARGET_SSE && SSE_REG_MODE_P (mode)) || mode == TDmode)
&& (!TYPE_USER_ALIGN (type) || TYPE_ALIGN (type) > 128))
return true;
if (TYPE_ALIGN (type) < 128)
for (field = TYPE_FIELDS (type); field; field = TREE_CHAIN (field))
{
if (TREE_CODE (field) == FIELD_DECL
- && contains_128bit_aligned_vector_p (TREE_TYPE (field)))
+ && contains_aligned_value_p (TREE_TYPE (field)))
return true;
}
break;
case ARRAY_TYPE:
/* Just for use if some languages passes arrays by value. */
- if (contains_128bit_aligned_vector_p (TREE_TYPE (type)))
+ if (contains_aligned_value_p (TREE_TYPE (type)))
return true;
break;
align = GET_MODE_ALIGNMENT (mode);
if (align < PARM_BOUNDARY)
align = PARM_BOUNDARY;
- if (!TARGET_64BIT)
+ /* In 32bit, only _Decimal128 is aligned to its natural boundary. */
+ if (!TARGET_64BIT && mode != TDmode)
{
/* i386 ABI defines all arguments to be 4 byte aligned. We have to
make an exception for SSE modes since these require 128bit
The handling here differs from field_alignment. ICC aligns MMX
arguments to 4 byte boundaries, while structure fields are aligned
to 8 byte boundaries. */
- if (!TARGET_SSE)
- align = PARM_BOUNDARY;
- else if (!type)
+ if (!type)
{
- if (!SSE_REG_MODE_P (mode))
+ if (!(TARGET_SSE && SSE_REG_MODE_P (mode)) && mode != TDmode)
align = PARM_BOUNDARY;
}
else
{
- if (!contains_128bit_aligned_vector_p (type))
+ if (!contains_aligned_value_p (type))
align = PARM_BOUNDARY;
}
}
- if (align > 128)
- align = 128;
+ if (align > BIGGEST_ALIGNMENT)
+ align = BIGGEST_ALIGNMENT;
return align;
}
static rtx
function_value_32 (enum machine_mode orig_mode, enum machine_mode mode,
- tree fntype, tree fn)
+ const_tree fntype, const_tree fn)
{
unsigned int regno;
|| (VECTOR_MODE_P (mode) && GET_MODE_SIZE (mode) == 16))
regno = TARGET_SSE ? FIRST_SSE_REG : 0;
- /* Decimal floating point values can go in %eax, unlike other float modes. */
- else if (DECIMAL_FLOAT_MODE_P (mode))
- regno = 0;
-
- /* Most things go in %eax, except (unless -mno-fp-ret-in-387) fp values. */
- else if (!SCALAR_FLOAT_MODE_P (mode) || !TARGET_FLOAT_RETURNS_IN_80387)
- regno = 0;
+ /* Floating point return values in %st(0) (unless -mno-fp-ret-in-387). */
+ else if (X87_FLOAT_MODE_P (mode) && TARGET_FLOAT_RETURNS_IN_80387)
+ regno = FIRST_FLOAT_REG;
+ else
+ /* Most things go in %eax. */
+ regno = AX_REG;
- /* Floating point return values in %st(0), except for local functions when
+ /* Override FP return register with %xmm0 for local functions when
SSE math is enabled or for functions with sseregparm attribute. */
- else
+ if ((fn || fntype) && (mode == SFmode || mode == DFmode))
{
- regno = FIRST_FLOAT_REG;
-
- if ((fn || fntype) && (mode == SFmode || mode == DFmode))
- {
- int sse_level = ix86_function_sseregparm (fntype, fn);
- if ((sse_level >= 1 && mode == SFmode)
- || (sse_level == 2 && mode == DFmode))
- regno = FIRST_SSE_REG;
- }
+ int sse_level = ix86_function_sseregparm (fntype, fn, false);
+ if ((sse_level >= 1 && mode == SFmode)
+ || (sse_level == 2 && mode == DFmode))
+ regno = FIRST_SSE_REG;
}
return gen_rtx_REG (orig_mode, regno);
static rtx
function_value_64 (enum machine_mode orig_mode, enum machine_mode mode,
- tree valtype)
+ const_tree valtype)
{
rtx ret;
case TCmode:
return NULL;
default:
- return gen_rtx_REG (mode, 0);
+ return gen_rtx_REG (mode, AX_REG);
}
}
/* For zero sized structures, construct_container returns NULL, but we
need to keep rest of compiler happy by returning meaningful value. */
if (!ret)
- ret = gen_rtx_REG (orig_mode, 0);
+ ret = gen_rtx_REG (orig_mode, AX_REG);
return ret;
}
static rtx
function_value_ms_64 (enum machine_mode orig_mode, enum machine_mode mode)
{
- unsigned int regno = 0;
+ unsigned int regno = AX_REG;
if (TARGET_SSE)
{
- if (mode == SFmode || mode == DFmode)
- regno = FIRST_SSE_REG;
- else if (VECTOR_MODE_P (mode) || GET_MODE_SIZE (mode) == 16)
- regno = FIRST_SSE_REG;
+ switch (GET_MODE_SIZE (mode))
+ {
+ case 16:
+ if((SCALAR_INT_MODE_P (mode) || VECTOR_MODE_P (mode))
+ && !COMPLEX_MODE_P (mode))
+ regno = FIRST_SSE_REG;
+ break;
+ case 8:
+ case 4:
+ if (mode == SFmode || mode == DFmode)
+ regno = FIRST_SSE_REG;
+ break;
+ default:
+ break;
+ }
}
-
return gen_rtx_REG (orig_mode, regno);
}
static rtx
-ix86_function_value_1 (tree valtype, tree fntype_or_decl,
+ix86_function_value_1 (const_tree valtype, const_tree fntype_or_decl,
enum machine_mode orig_mode, enum machine_mode mode)
{
- tree fn, fntype;
+ const_tree fn, fntype;
fn = NULL_TREE;
if (fntype_or_decl && DECL_P (fntype_or_decl))
}
static rtx
-ix86_function_value (tree valtype, tree fntype_or_decl,
+ix86_function_value (const_tree valtype, const_tree fntype_or_decl,
bool outgoing ATTRIBUTE_UNUSED)
{
enum machine_mode mode, orig_mode;
/* Return true iff type is returned in memory. */
static int
-return_in_memory_32 (tree type, enum machine_mode mode)
+return_in_memory_32 (const_tree type, enum machine_mode mode)
{
HOST_WIDE_INT size;
}
static int
-return_in_memory_64 (tree type, enum machine_mode mode)
+return_in_memory_64 (const_tree type, enum machine_mode mode)
{
int needed_intregs, needed_sseregs;
return !examine_argument (mode, type, 1, &needed_intregs, &needed_sseregs);
}
static int
-return_in_memory_ms_64 (tree type, enum machine_mode mode)
+return_in_memory_ms_64 (const_tree type, enum machine_mode mode)
{
HOST_WIDE_INT size = int_size_in_bytes (type);
- /* __m128 and friends are returned in xmm0. */
- if (size == 16 && VECTOR_MODE_P (mode))
+ /* __m128 is returned in xmm0. */
+ if ((SCALAR_INT_MODE_P (mode) || VECTOR_MODE_P (mode))
+ && !COMPLEX_MODE_P (mode) && (GET_MODE_SIZE (mode) == 16 || size == 16))
return 0;
- /* Otherwise, the size must be exactly in [1248]. */
+ /* Otherwise, the size must be exactly in [1248]. */
return (size != 1 && size != 2 && size != 4 && size != 8);
}
-int
-ix86_return_in_memory (tree type)
+bool
+ix86_return_in_memory (const_tree type, const_tree fntype ATTRIBUTE_UNUSED)
{
- enum machine_mode mode = type_natural_mode (type);
+ const enum machine_mode mode = type_natural_mode (type);
if (TARGET_64BIT_MS_ABI)
return return_in_memory_ms_64 (type, mode);
return return_in_memory_32 (type, mode);
}
-/* When returning SSE vector types, we have a choice of either
- (1) being abi incompatible with a -march switch, or
- (2) generating an error.
- Given no good solution, I think the safest thing is one warning.
- The user won't be able to use -Werror, but....
-
- Choose the STRUCT_VALUE_RTX hook because that's (at present) only
- called in response to actually generating a caller or callee that
- uses such a type. As opposed to RETURN_IN_MEMORY, which is called
- via aggregate_value_p for general type probing from tree-ssa. */
+/* Return false iff TYPE is returned in memory. This version is used
+ on Solaris 10. It is similar to the generic ix86_return_in_memory,
+ but differs notably in that when MMX is available, 8-byte vectors
+ are returned in memory, rather than in MMX registers. */
-static rtx
-ix86_struct_value_rtx (tree type, int incoming ATTRIBUTE_UNUSED)
+bool
+ix86_sol10_return_in_memory (const_tree type, const_tree fntype ATTRIBUTE_UNUSED)
{
- static bool warnedsse, warnedmmx;
+ int size;
+ enum machine_mode mode = type_natural_mode (type);
- if (!TARGET_64BIT && type)
- {
- /* Look at the return type of the function, not the function type. */
- enum machine_mode mode = TYPE_MODE (TREE_TYPE (type));
+ if (TARGET_64BIT)
+ return return_in_memory_64 (type, mode);
- if (!TARGET_SSE && !warnedsse)
- {
- if (mode == TImode
- || (VECTOR_MODE_P (mode) && GET_MODE_SIZE (mode) == 16))
- {
+ if (mode == BLKmode)
+ return 1;
+
+ size = int_size_in_bytes (type);
+
+ if (VECTOR_MODE_P (mode))
+ {
+ /* Return in memory only if MMX registers *are* available. This
+ seems backwards, but it is consistent with the existing
+ Solaris x86 ABI. */
+ if (size == 8)
+ return TARGET_MMX;
+ if (size == 16)
+ return !TARGET_SSE;
+ }
+ else if (mode == TImode)
+ return !TARGET_SSE;
+ else if (mode == XFmode)
+ return 0;
+
+ return size > 12;
+}
+
+bool
+ix86_i386elf_return_in_memory (const_tree type, const_tree fntype ATTRIBUTE_UNUSED)
+{
+ return (TYPE_MODE (type) == BLKmode
+ || (VECTOR_MODE_P (TYPE_MODE (type)) && int_size_in_bytes (type) == 8));
+}
+
+bool
+ix86_i386interix_return_in_memory (const_tree type, const_tree fntype ATTRIBUTE_UNUSED)
+{
+ return (TYPE_MODE (type) == BLKmode
+ || (AGGREGATE_TYPE_P (type) && int_size_in_bytes(type) > 8 ));
+}
+
+/* When returning SSE vector types, we have a choice of either
+ (1) being abi incompatible with a -march switch, or
+ (2) generating an error.
+ Given no good solution, I think the safest thing is one warning.
+ The user won't be able to use -Werror, but....
+
+ Choose the STRUCT_VALUE_RTX hook because that's (at present) only
+ called in response to actually generating a caller or callee that
+ uses such a type. As opposed to TARGET_RETURN_IN_MEMORY, which is called
+ via aggregate_value_p for general type probing from tree-ssa. */
+
+static rtx
+ix86_struct_value_rtx (tree type, int incoming ATTRIBUTE_UNUSED)
+{
+ static bool warnedsse, warnedmmx;
+
+ if (!TARGET_64BIT && type)
+ {
+ /* Look at the return type of the function, not the function type. */
+ enum machine_mode mode = TYPE_MODE (TREE_TYPE (type));
+
+ if (!TARGET_SSE && !warnedsse)
+ {
+ if (mode == TImode
+ || (VECTOR_MODE_P (mode) && GET_MODE_SIZE (mode) == 16))
+ {
warnedsse = true;
warning (0, "SSE vector return without SSE enabled "
"changes the ABI");
rtx label_ref;
rtx tmp_reg;
rtx nsse_reg;
- int set;
+ alias_set_type set;
int i;
if (! cfun->va_list_gpr_size && ! cfun->va_list_fpr_size)
/* Indicate to allocate space on the stack for varargs save area. */
ix86_save_varrargs_registers = 1;
- cfun->stack_alignment_needed = 128;
+ /* We need 16-byte stack alignment to save SSE registers. If user
+ asked for lower preferred_stack_boundary, lets just hope that he knows
+ what he is doing and won't varargs SSE values.
+
+ We also may end up assuming that only 64bit values are stored in SSE
+ register let some floating point program work. */
+ if (ix86_preferred_stack_boundary >= BIGGEST_ALIGNMENT)
+ crtl->stack_alignment_needed = BIGGEST_ALIGNMENT;
save_area = frame_pointer_rtx;
set = get_varargs_alias_set ();
label - 5*eax + nnamed_sse_arguments*5 */
tmp_reg = gen_reg_rtx (Pmode);
nsse_reg = gen_reg_rtx (Pmode);
- emit_insn (gen_zero_extendqidi2 (nsse_reg, gen_rtx_REG (QImode, 0)));
+ emit_insn (gen_zero_extendqidi2 (nsse_reg, gen_rtx_REG (QImode, AX_REG)));
emit_insn (gen_rtx_SET (VOIDmode, tmp_reg,
gen_rtx_MULT (Pmode, nsse_reg,
GEN_INT (4))));
static void
setup_incoming_varargs_ms_64 (CUMULATIVE_ARGS *cum)
{
- int set = get_varargs_alias_set ();
+ alias_set_type set = get_varargs_alias_set ();
int i;
for (i = cum->regno; i < REGPARM_MAX; i++)
{
CUMULATIVE_ARGS next_cum;
tree fntype;
- int stdarg_p;
/* This argument doesn't appear to be used anymore. Which is good,
because the old code here didn't suppress rtl generation. */
return;
fntype = TREE_TYPE (current_function_decl);
- stdarg_p = (TYPE_ARG_TYPES (fntype) != 0
- && (TREE_VALUE (tree_last (TYPE_ARG_TYPES (fntype)))
- != void_type_node));
/* For varargs, we do not want to skip the dummy va_dcl argument.
For stdargs, we do want to skip the last named argument. */
next_cum = *cum;
- if (stdarg_p)
+ if (stdarg_p (fntype))
function_arg_advance (&next_cum, mode, type, 1);
if (TARGET_64BIT_MS_ABI)
/* Implement va_start. */
-void
+static void
ix86_va_start (tree valist, rtx nextarg)
{
HOST_WIDE_INT words, n_gpr, n_fpr;
sav = build3 (COMPONENT_REF, TREE_TYPE (f_sav), valist, f_sav, NULL_TREE);
/* Count number of gp and fp argument registers used. */
- words = current_function_args_info.words;
- n_gpr = current_function_args_info.regno;
- n_fpr = current_function_args_info.sse_regno;
+ words = crtl->args.info.words;
+ n_gpr = crtl->args.info.regno;
+ n_fpr = crtl->args.info.sse_regno;
if (cfun->va_list_gpr_size)
{
type = TREE_TYPE (ovf);
t = make_tree (type, virtual_incoming_args_rtx);
if (words != 0)
- t = build2 (PLUS_EXPR, type, t,
- build_int_cst (type, words * UNITS_PER_WORD));
+ t = build2 (POINTER_PLUS_EXPR, type, t,
+ size_int (words * UNITS_PER_WORD));
t = build2 (GIMPLE_MODIFY_STMT, type, ovf, t);
TREE_SIDE_EFFECTS (t) = 1;
expand_expr (t, const0_rtx, VOIDmode, EXPAND_NORMAL);
if (needed_intregs)
{
/* int_addr = gpr + sav; */
- t = fold_convert (ptr_type_node, gpr);
- t = build2 (PLUS_EXPR, ptr_type_node, sav, t);
+ t = fold_convert (sizetype, gpr);
+ t = build2 (POINTER_PLUS_EXPR, ptr_type_node, sav, t);
t = build2 (GIMPLE_MODIFY_STMT, void_type_node, int_addr, t);
gimplify_and_add (t, pre_p);
}
if (needed_sseregs)
{
/* sse_addr = fpr + sav; */
- t = fold_convert (ptr_type_node, fpr);
- t = build2 (PLUS_EXPR, ptr_type_node, sav, t);
+ t = fold_convert (sizetype, fpr);
+ t = build2 (POINTER_PLUS_EXPR, ptr_type_node, sav, t);
t = build2 (GIMPLE_MODIFY_STMT, void_type_node, sse_addr, t);
gimplify_and_add (t, pre_p);
}
src_offset = REGNO (reg) * 8;
}
src_addr = fold_convert (addr_type, src_addr);
- src_addr = fold_build2 (PLUS_EXPR, addr_type, src_addr,
+ src_addr = fold_build2 (POINTER_PLUS_EXPR, addr_type, src_addr,
size_int (src_offset));
src = build_va_arg_indirect_ref (src_addr);
dest_addr = fold_convert (addr_type, addr);
- dest_addr = fold_build2 (PLUS_EXPR, addr_type, dest_addr,
+ dest_addr = fold_build2 (POINTER_PLUS_EXPR, addr_type, dest_addr,
size_int (INTVAL (XEXP (slot, 1))));
dest = build_va_arg_indirect_ref (dest_addr);
if (FUNCTION_ARG_BOUNDARY (VOIDmode, type) <= 64
|| integer_zerop (TYPE_SIZE (type)))
t = ovf;
- else
+ else
{
HOST_WIDE_INT align = FUNCTION_ARG_BOUNDARY (VOIDmode, type) / 8;
- t = build2 (PLUS_EXPR, TREE_TYPE (ovf), ovf,
- build_int_cst (TREE_TYPE (ovf), align - 1));
+ t = build2 (POINTER_PLUS_EXPR, TREE_TYPE (ovf), ovf,
+ size_int (align - 1));
+ t = fold_convert (sizetype, t);
t = build2 (BIT_AND_EXPR, TREE_TYPE (t), t,
- build_int_cst (TREE_TYPE (t), -align));
+ size_int (-align));
+ t = fold_convert (TREE_TYPE (ovf), t);
}
gimplify_expr (&t, pre_p, NULL, is_gimple_val, fb_rvalue);
t2 = build2 (GIMPLE_MODIFY_STMT, void_type_node, addr, t);
gimplify_and_add (t2, pre_p);
- t = build2 (PLUS_EXPR, TREE_TYPE (t), t,
- build_int_cst (TREE_TYPE (t), rsize * UNITS_PER_WORD));
+ t = build2 (POINTER_PLUS_EXPR, TREE_TYPE (t), t,
+ size_int (rsize * UNITS_PER_WORD));
t = build2 (GIMPLE_MODIFY_STMT, TREE_TYPE (ovf), ovf, t);
gimplify_and_add (t, pre_p);
int
standard_80387_constant_p (rtx x)
{
+ enum machine_mode mode = GET_MODE (x);
+
REAL_VALUE_TYPE r;
- if (GET_CODE (x) != CONST_DOUBLE || !FLOAT_MODE_P (GET_MODE (x)))
+ if (!(X87_FLOAT_MODE_P (mode) && (GET_CODE (x) == CONST_DOUBLE)))
return -1;
- if (x == CONST0_RTX (GET_MODE (x)))
+ if (x == CONST0_RTX (mode))
return 1;
- if (x == CONST1_RTX (GET_MODE (x)))
+ if (x == CONST1_RTX (mode))
return 2;
REAL_VALUE_FROM_CONST_DOUBLE (r, x);
/* For XFmode constants, try to find a special 80387 instruction when
optimizing for size or on those CPUs that benefit from them. */
- if (GET_MODE (x) == XFmode
+ if (mode == XFmode
&& (optimize_size || TARGET_EXT_80387_CONSTANTS))
{
int i;
/* Don't allow more than 32 pop, since that's all we can do
with one instruction. */
- if (current_function_pops_args
- && current_function_args_size >= 32768)
+ if (crtl->args.pops_args
+ && crtl->args.size >= 32768)
return 0;
ix86_compute_frame_layout (&frame);
|| ix86_current_function_calls_tls_descriptor))
return 1;
- if (current_function_profile)
+ if (crtl->profile)
return 1;
return 0;
ASM_OUTPUT_LABEL (asm_out_file, name);
}
- xops[0] = gen_rtx_REG (SImode, regno);
- xops[1] = gen_rtx_MEM (SImode, stack_pointer_rtx);
- output_asm_insn ("mov{l}\t{%1, %0|%0, %1}", xops);
+ xops[0] = gen_rtx_REG (Pmode, regno);
+ xops[1] = gen_rtx_MEM (Pmode, stack_pointer_rtx);
+ if (TARGET_64BIT)
+ output_asm_insn ("mov{q}\t{%1, %0|%0, %1}", xops);
+ else
+ output_asm_insn ("mov{l}\t{%1, %0|%0, %1}", xops);
output_asm_insn ("ret", xops);
}
xops[2] = gen_rtx_LABEL_REF (Pmode, label ? label : gen_label_rtx ());
if (!flag_pic)
- output_asm_insn ("mov{l}\t{%2, %0|%0, %2}", xops);
+ {
+ if (TARGET_64BIT)
+ output_asm_insn ("mov{q}\t{%2, %0|%0, %2}", xops);
+ else
+ output_asm_insn ("mov{l}\t{%2, %0|%0, %2}", xops);
+ }
else
output_asm_insn ("call\t%a2", xops);
CODE_LABEL_NUMBER (XEXP (xops[2], 0)));
if (flag_pic)
- output_asm_insn ("pop{l}\t%0", xops);
+ {
+ if (TARGET_64BIT)
+ output_asm_insn ("pop{q}\t%0", xops);
+ else
+ output_asm_insn ("pop{l}\t%0", xops);
+ }
}
else
{
return "";
if (!flag_pic || TARGET_DEEP_BRANCH_PREDICTION)
- output_asm_insn ("add{l}\t{%1, %0|%0, %1}", xops);
+ {
+ if (TARGET_64BIT)
+ output_asm_insn ("add{q}\t{%1, %0|%0, %1}", xops);
+ else
+ output_asm_insn ("add{l}\t{%1, %0|%0, %1}", xops);
+ }
else
- output_asm_insn ("add{l}\t{%1+[.-%a2], %0|%0, %1+(.-%a2)}", xops);
+ {
+ if (TARGET_64BIT)
+ output_asm_insn ("add{q}\t{%1+[.-%a2], %0|%0, %1+(.-%a2)}", xops);
+ else
+ output_asm_insn ("add{l}\t{%1+[.-%a2], %0|%0, %1+(.-%a2)}", xops);
+ }
return "";
}
static unsigned int
ix86_select_alt_pic_regnum (void)
{
- if (current_function_is_leaf && !current_function_profile
+ if (current_function_is_leaf && !crtl->profile
&& !ix86_current_function_calls_tls_descriptor)
{
int i;
for (i = 2; i >= 0; --i)
- if (!regs_ever_live[i])
+ if (!df_regs_ever_live_p (i))
return i;
}
{
if (pic_offset_table_rtx
&& regno == REAL_PIC_OFFSET_TABLE_REGNUM
- && (regs_ever_live[REAL_PIC_OFFSET_TABLE_REGNUM]
- || current_function_profile
- || current_function_calls_eh_return
- || current_function_uses_const_pool))
+ && (df_regs_ever_live_p (REAL_PIC_OFFSET_TABLE_REGNUM)
+ || crtl->profile
+ || crtl->calls_eh_return
+ || crtl->uses_const_pool))
{
if (ix86_select_alt_pic_regnum () != INVALID_REGNUM)
return 0;
return 1;
}
- if (current_function_calls_eh_return && maybe_eh_return)
+ if (crtl->calls_eh_return && maybe_eh_return)
{
unsigned i;
for (i = 0; ; i++)
&& regno == REGNO (cfun->machine->force_align_arg_pointer))
return 1;
- return (regs_ever_live[regno]
+ return (df_regs_ever_live_p (regno)
&& !call_used_regs[regno]
&& !fixed_regs[regno]
&& (regno != HARD_FRAME_POINTER_REGNUM || !frame_pointer_needed));
frame->nregs = ix86_nsaved_regs ();
total_size = size;
- stack_alignment_needed = cfun->stack_alignment_needed / BITS_PER_UNIT;
- preferred_alignment = cfun->preferred_stack_boundary / BITS_PER_UNIT;
+ stack_alignment_needed = crtl->stack_alignment_needed / BITS_PER_UNIT;
+ preferred_alignment = crtl->preferred_stack_boundary / BITS_PER_UNIT;
/* During reload iteration the amount of registers saved can change.
Recompute the value as needed. Do not recompute when amount of registers
/* Add outgoing arguments area. Can be skipped if we eliminated
all the function calls as dead code.
Skipping is however impossible when function calls alloca. Alloca
- expander assumes that last current_function_outgoing_args_size
+ expander assumes that last crtl->outgoing_args_size
of stack frame are unused. */
if (ACCUMULATE_OUTGOING_ARGS
- && (!current_function_is_leaf || current_function_calls_alloca
+ && (!current_function_is_leaf || cfun->calls_alloca
|| ix86_current_function_calls_tls_descriptor))
{
- offset += current_function_outgoing_args_size;
- frame->outgoing_arguments_size = current_function_outgoing_args_size;
+ offset += crtl->outgoing_args_size;
+ frame->outgoing_arguments_size = crtl->outgoing_args_size;
}
else
frame->outgoing_arguments_size = 0;
/* Align stack boundary. Only needed if we're calling another function
or using alloca. */
- if (!current_function_is_leaf || current_function_calls_alloca
+ if (!current_function_is_leaf || cfun->calls_alloca
|| ix86_current_function_calls_tls_descriptor)
frame->padding2 = ((offset + preferred_alignment - 1)
& -preferred_alignment) - offset;
(long)frame->hard_frame_pointer_offset);
fprintf (stderr, "stack_pointer_offset: %ld\n", (long)frame->stack_pointer_offset);
fprintf (stderr, "current_function_is_leaf: %ld\n", (long)current_function_is_leaf);
- fprintf (stderr, "current_function_calls_alloca: %ld\n", (long)current_function_calls_alloca);
+ fprintf (stderr, "cfun->calls_alloca: %ld\n", (long)cfun->calls_alloca);
fprintf (stderr, "x86_current_function_calls_tls_descriptor: %ld\n", (long)ix86_current_function_calls_tls_descriptor);
#endif
}
ix86_force_align_arg_pointer_string);
return virtual_incoming_args_rtx;
}
- cfun->machine->force_align_arg_pointer = gen_rtx_REG (Pmode, 2);
+ cfun->machine->force_align_arg_pointer = gen_rtx_REG (Pmode, CX_REG);
return copy_to_reg (cfun->machine->force_align_arg_pointer);
}
else
allocate += frame.nregs * UNITS_PER_WORD;
/* When using red zone we may start register saving before allocating
- the stack frame saving one cycle of the prologue. */
- if (TARGET_RED_ZONE && frame.save_regs_using_mov)
+ the stack frame saving one cycle of the prologue. However I will
+ avoid doing this if I am going to have to probe the stack since
+ at least on x86_64 the stack probe can turn into a call that clobbers
+ a red zone location */
+ if (TARGET_RED_ZONE && frame.save_regs_using_mov
+ && (! TARGET_STACK_PROBE || allocate < CHECK_STACK_LIMIT))
ix86_emit_save_regs_using_mov (frame_pointer_needed ? hard_frame_pointer_rtx
: stack_pointer_rtx,
-frame.nregs * UNITS_PER_WORD);
else
{
/* Only valid for Win32. */
- rtx eax = gen_rtx_REG (Pmode, 0);
+ rtx eax = gen_rtx_REG (Pmode, AX_REG);
bool eax_live;
rtx t;
}
}
- if (frame.save_regs_using_mov && !TARGET_RED_ZONE)
+ if (frame.save_regs_using_mov
+ && !(TARGET_RED_ZONE
+ && (! TARGET_STACK_PROBE || allocate < CHECK_STACK_LIMIT)))
{
if (!frame_pointer_needed || !frame.to_allocate)
ix86_emit_save_regs_using_mov (stack_pointer_rtx, frame.to_allocate);
pic_reg_used = false;
if (pic_offset_table_rtx
- && (regs_ever_live[REAL_PIC_OFFSET_TABLE_REGNUM]
- || current_function_profile))
+ && (df_regs_ever_live_p (REAL_PIC_OFFSET_TABLE_REGNUM)
+ || crtl->profile))
{
unsigned int alt_pic_reg_used = ix86_select_alt_pic_regnum ();
if (alt_pic_reg_used != INVALID_REGNUM)
- REGNO (pic_offset_table_rtx) = alt_pic_reg_used;
+ SET_REGNO (pic_offset_table_rtx, alt_pic_reg_used);
pic_reg_used = true;
}
{
if (ix86_cmodel == CM_LARGE_PIC)
{
- rtx tmp_reg = gen_rtx_REG (DImode,
- FIRST_REX_INT_REG + 3 /* R11 */);
+ rtx tmp_reg = gen_rtx_REG (DImode, R11_REG);
rtx label = gen_label_rtx ();
emit_label (label);
LABEL_PRESERVE_P (label) = 1;
gcc_assert (REGNO (pic_offset_table_rtx) != REGNO (tmp_reg));
insn = emit_insn (gen_set_rip_rex64 (pic_offset_table_rtx, label));
- REG_NOTES (insn) = gen_rtx_EXPR_LIST (REG_MAYBE_DEAD, const0_rtx, NULL);
insn = emit_insn (gen_set_got_offset_rex64 (tmp_reg, label));
- REG_NOTES (insn) = gen_rtx_EXPR_LIST (REG_MAYBE_DEAD, const0_rtx, NULL);
insn = emit_insn (gen_adddi3 (pic_offset_table_rtx,
pic_offset_table_rtx, tmp_reg));
}
}
else
insn = emit_insn (gen_set_got (pic_offset_table_rtx));
-
- /* Even with accurate pre-reload life analysis, we can wind up
- deleting all references to the pic register after reload.
- Consider if cross-jumping unifies two sides of a branch
- controlled by a comparison vs the only read from a global.
- In which case, allow the set_got to be deleted, though we're
- too late to do anything about the ebx save in the prologue. */
- REG_NOTES (insn) = gen_rtx_EXPR_LIST (REG_MAYBE_DEAD, const0_rtx, NULL);
}
- /* Prevent function calls from be scheduled before the call to mcount.
+ /* Prevent function calls from being scheduled before the call to mcount.
In the pic_reg_used case, make sure that the got load isn't deleted. */
- if (current_function_profile)
- emit_insn (gen_blockage (pic_reg_used ? pic_offset_table_rtx : const0_rtx));
+ if (crtl->profile)
+ {
+ if (pic_reg_used)
+ emit_insn (gen_prologue_use (pic_offset_table_rtx));
+ emit_insn (gen_blockage ());
+ }
}
/* Emit code to restore saved registers using MOV insns. First register
eh_return: the eax and edx registers are marked as saved, but not
restored along this path. */
offset = frame.nregs;
- if (current_function_calls_eh_return && style != 2)
+ if (crtl->calls_eh_return && style != 2)
offset -= 2;
offset *= -UNITS_PER_WORD;
|| (frame_pointer_needed && TARGET_USE_LEAVE
&& cfun->machine->use_fast_prologue_epilogue
&& frame.nregs == 1)
- || current_function_calls_eh_return)
+ || crtl->calls_eh_return)
{
/* Restore registers. We can use ebp or esp to address the memory
locations. If both are available, default to ebp, since offsets
if (style == 0)
return;
- if (current_function_pops_args && current_function_args_size)
+ if (crtl->args.pops_args && crtl->args.size)
{
- rtx popc = GEN_INT (current_function_pops_args);
+ rtx popc = GEN_INT (crtl->args.pops_args);
/* i386 can only pop 64K bytes. If asked to pop more, pop
return address, do explicit add, and jump indirectly to the
caller. */
- if (current_function_pops_args >= 65536)
+ if (crtl->args.pops_args >= 65536)
{
- rtx ecx = gen_rtx_REG (SImode, 2);
+ rtx ecx = gen_rtx_REG (SImode, CX_REG);
/* There is no "pascal" calling convention in any 64bit ABI. */
gcc_assert (!TARGET_64BIT);
HOST_WIDE_INT size ATTRIBUTE_UNUSED)
{
if (pic_offset_table_rtx)
- REGNO (pic_offset_table_rtx) = REAL_PIC_OFFSET_TABLE_REGNUM;
+ SET_REGNO (pic_offset_table_rtx, REAL_PIC_OFFSET_TABLE_REGNUM);
#if TARGET_MACHO
/* Mach-O doesn't support labels at the end of objects, so if
it looks like we might want one, insert a NOP. */
rtx insn = get_last_insn ();
while (insn
&& NOTE_P (insn)
- && NOTE_LINE_NUMBER (insn) != NOTE_INSN_DELETED_LABEL)
+ && NOTE_KIND (insn) != NOTE_INSN_DELETED_LABEL)
insn = PREV_INSN (insn);
if (insn
&& (LABEL_P (insn)
|| (NOTE_P (insn)
- && NOTE_LINE_NUMBER (insn) == NOTE_INSN_DELETED_LABEL)))
+ && NOTE_KIND (insn) == NOTE_INSN_DELETED_LABEL)))
fputs ("\tnop\n", file);
}
#endif
/* Special case: on K6, [%esi] makes the instruction vector decoded.
Avoid this by transforming to [%esi+0]. */
- if (ix86_tune == PROCESSOR_K6 && !optimize_size
+ if (TARGET_K6 && !optimize_size
&& base_reg && !index_reg && !disp
&& REG_P (base_reg)
&& REGNO_REG_CLASS (REGNO (base_reg)) == SIREG)
if (parts.index && GET_CODE (parts.index) == SUBREG)
parts.index = SUBREG_REG (parts.index);
- /* More complex memory references are better. */
- if (parts.disp && parts.disp != const0_rtx)
- cost--;
- if (parts.seg != SEG_DEFAULT)
- cost--;
-
/* Attempt to minimize number of registers in the address. */
if ((parts.base
&& (!REG_P (parts.base) || REGNO (parts.base) >= FIRST_PSEUDO_REGISTER))
\f
/* Return a unique alias set for the GOT. */
-static HOST_WIDE_INT
+static alias_set_type
ix86_GOT_alias_set (void)
{
- static HOST_WIDE_INT set = -1;
+ static alias_set_type set = -1;
if (set == -1)
set = new_alias_set ();
return set;
legitimize_pic_address (rtx orig, rtx reg)
{
rtx addr = orig;
- rtx new = orig;
+ rtx new_rtx = orig;
rtx base;
#if TARGET_MACHO
#endif
if (TARGET_64BIT && legitimate_pic_address_disp_p (addr))
- new = addr;
+ new_rtx = addr;
else if (TARGET_64BIT
&& ix86_cmodel != CM_SMALL_PIC
&& gotoff_operand (addr, Pmode))
base address (@GOTOFF). */
if (reload_in_progress)
- regs_ever_live[PIC_OFFSET_TABLE_REGNUM] = 1;
+ df_set_regs_ever_live (PIC_OFFSET_TABLE_REGNUM, true);
if (GET_CODE (addr) == CONST)
addr = XEXP (addr, 0);
if (GET_CODE (addr) == PLUS)
{
- new = gen_rtx_UNSPEC (Pmode, gen_rtvec (1, XEXP (addr, 0)),
- UNSPEC_GOTOFF);
- new = gen_rtx_PLUS (Pmode, new, XEXP (addr, 1));
+ new_rtx = gen_rtx_UNSPEC (Pmode, gen_rtvec (1, XEXP (addr, 0)),
+ UNSPEC_GOTOFF);
+ new_rtx = gen_rtx_PLUS (Pmode, new_rtx, XEXP (addr, 1));
}
else
- new = gen_rtx_UNSPEC (Pmode, gen_rtvec (1, addr), UNSPEC_GOTOFF);
- new = gen_rtx_CONST (Pmode, new);
+ new_rtx = gen_rtx_UNSPEC (Pmode, gen_rtvec (1, addr), UNSPEC_GOTOFF);
+ new_rtx = gen_rtx_CONST (Pmode, new_rtx);
if (!reg)
tmpreg = gen_reg_rtx (Pmode);
else
tmpreg = reg;
- emit_move_insn (tmpreg, new);
+ emit_move_insn (tmpreg, new_rtx);
if (reg != 0)
{
- new = expand_simple_binop (Pmode, PLUS, reg, pic_offset_table_rtx,
- tmpreg, 1, OPTAB_DIRECT);
- new = reg;
+ new_rtx = expand_simple_binop (Pmode, PLUS, reg, pic_offset_table_rtx,
+ tmpreg, 1, OPTAB_DIRECT);
+ new_rtx = reg;
}
- else new = gen_rtx_PLUS (Pmode, pic_offset_table_rtx, tmpreg);
+ else new_rtx = gen_rtx_PLUS (Pmode, pic_offset_table_rtx, tmpreg);
}
else if (!TARGET_64BIT && gotoff_operand (addr, Pmode))
{
base address (@GOTOFF). */
if (reload_in_progress)
- regs_ever_live[PIC_OFFSET_TABLE_REGNUM] = 1;
+ df_set_regs_ever_live (PIC_OFFSET_TABLE_REGNUM, true);
if (GET_CODE (addr) == CONST)
addr = XEXP (addr, 0);
if (GET_CODE (addr) == PLUS)
{
- new = gen_rtx_UNSPEC (Pmode, gen_rtvec (1, XEXP (addr, 0)),
- UNSPEC_GOTOFF);
- new = gen_rtx_PLUS (Pmode, new, XEXP (addr, 1));
+ new_rtx = gen_rtx_UNSPEC (Pmode, gen_rtvec (1, XEXP (addr, 0)),
+ UNSPEC_GOTOFF);
+ new_rtx = gen_rtx_PLUS (Pmode, new_rtx, XEXP (addr, 1));
}
else
- new = gen_rtx_UNSPEC (Pmode, gen_rtvec (1, addr), UNSPEC_GOTOFF);
- new = gen_rtx_CONST (Pmode, new);
- new = gen_rtx_PLUS (Pmode, pic_offset_table_rtx, new);
+ new_rtx = gen_rtx_UNSPEC (Pmode, gen_rtvec (1, addr), UNSPEC_GOTOFF);
+ new_rtx = gen_rtx_CONST (Pmode, new_rtx);
+ new_rtx = gen_rtx_PLUS (Pmode, pic_offset_table_rtx, new_rtx);
if (reg != 0)
{
- emit_move_insn (reg, new);
- new = reg;
+ emit_move_insn (reg, new_rtx);
+ new_rtx = reg;
}
}
else if ((GET_CODE (addr) == SYMBOL_REF && SYMBOL_REF_TLS_MODEL (addr) == 0)
see gotoff_operand. */
|| (TARGET_VXWORKS_RTP && GET_CODE (addr) == LABEL_REF))
{
- /* Given that we've already handled dllimport variables separately
- in legitimize_address, and all other variables should satisfy
- legitimate_pic_address_disp_p, we should never arrive here. */
- gcc_assert (!TARGET_64BIT_MS_ABI);
+ if (TARGET_DLLIMPORT_DECL_ATTRIBUTES)
+ {
+ if (GET_CODE (addr) == SYMBOL_REF && SYMBOL_REF_DLLIMPORT_P (addr))
+ return legitimize_dllimport_symbol (addr, true);
+ if (GET_CODE (addr) == CONST && GET_CODE (XEXP (addr, 0)) == PLUS
+ && GET_CODE (XEXP (XEXP (addr, 0), 0)) == SYMBOL_REF
+ && SYMBOL_REF_DLLIMPORT_P (XEXP (XEXP (addr, 0), 0)))
+ {
+ rtx t = legitimize_dllimport_symbol (XEXP (XEXP (addr, 0), 0), true);
+ return gen_rtx_PLUS (Pmode, t, XEXP (XEXP (addr, 0), 1));
+ }
+ }
if (TARGET_64BIT && ix86_cmodel != CM_LARGE_PIC)
{
- new = gen_rtx_UNSPEC (Pmode, gen_rtvec (1, addr), UNSPEC_GOTPCREL);
- new = gen_rtx_CONST (Pmode, new);
- new = gen_const_mem (Pmode, new);
- set_mem_alias_set (new, ix86_GOT_alias_set ());
+ new_rtx = gen_rtx_UNSPEC (Pmode, gen_rtvec (1, addr), UNSPEC_GOTPCREL);
+ new_rtx = gen_rtx_CONST (Pmode, new_rtx);
+ new_rtx = gen_const_mem (Pmode, new_rtx);
+ set_mem_alias_set (new_rtx, ix86_GOT_alias_set ());
if (reg == 0)
reg = gen_reg_rtx (Pmode);
/* Use directly gen_movsi, otherwise the address is loaded
into register for CSE. We don't want to CSE this addresses,
instead we CSE addresses from the GOT table, so skip this. */
- emit_insn (gen_movsi (reg, new));
- new = reg;
+ emit_insn (gen_movsi (reg, new_rtx));
+ new_rtx = reg;
}
else
{
Global Offset Table (@GOT). */
if (reload_in_progress)
- regs_ever_live[PIC_OFFSET_TABLE_REGNUM] = 1;
- new = gen_rtx_UNSPEC (Pmode, gen_rtvec (1, addr), UNSPEC_GOT);
- new = gen_rtx_CONST (Pmode, new);
+ df_set_regs_ever_live (PIC_OFFSET_TABLE_REGNUM, true);
+ new_rtx = gen_rtx_UNSPEC (Pmode, gen_rtvec (1, addr), UNSPEC_GOT);
+ new_rtx = gen_rtx_CONST (Pmode, new_rtx);
if (TARGET_64BIT)
- new = force_reg (Pmode, new);
- new = gen_rtx_PLUS (Pmode, pic_offset_table_rtx, new);
- new = gen_const_mem (Pmode, new);
- set_mem_alias_set (new, ix86_GOT_alias_set ());
+ new_rtx = force_reg (Pmode, new_rtx);
+ new_rtx = gen_rtx_PLUS (Pmode, pic_offset_table_rtx, new_rtx);
+ new_rtx = gen_const_mem (Pmode, new_rtx);
+ set_mem_alias_set (new_rtx, ix86_GOT_alias_set ());
if (reg == 0)
reg = gen_reg_rtx (Pmode);
- emit_move_insn (reg, new);
- new = reg;
+ emit_move_insn (reg, new_rtx);
+ new_rtx = reg;
}
}
else
if (reg)
{
emit_move_insn (reg, addr);
- new = reg;
+ new_rtx = reg;
}
else
- new = force_reg (Pmode, addr);
+ new_rtx = force_reg (Pmode, addr);
}
else if (GET_CODE (addr) == CONST)
{
if (!TARGET_64BIT)
{
if (reload_in_progress)
- regs_ever_live[PIC_OFFSET_TABLE_REGNUM] = 1;
- new = gen_rtx_UNSPEC (Pmode, gen_rtvec (1, op0),
- UNSPEC_GOTOFF);
- new = gen_rtx_PLUS (Pmode, new, op1);
- new = gen_rtx_CONST (Pmode, new);
- new = gen_rtx_PLUS (Pmode, pic_offset_table_rtx, new);
+ df_set_regs_ever_live (PIC_OFFSET_TABLE_REGNUM, true);
+ new_rtx = gen_rtx_UNSPEC (Pmode, gen_rtvec (1, op0),
+ UNSPEC_GOTOFF);
+ new_rtx = gen_rtx_PLUS (Pmode, new_rtx, op1);
+ new_rtx = gen_rtx_CONST (Pmode, new_rtx);
+ new_rtx = gen_rtx_PLUS (Pmode, pic_offset_table_rtx, new_rtx);
if (reg != 0)
{
- emit_move_insn (reg, new);
- new = reg;
+ emit_move_insn (reg, new_rtx);
+ new_rtx = reg;
}
}
else
{
if (!x86_64_immediate_operand (op1, Pmode))
op1 = force_reg (Pmode, op1);
- new = gen_rtx_PLUS (Pmode, force_reg (Pmode, op0), op1);
+ new_rtx = gen_rtx_PLUS (Pmode, force_reg (Pmode, op0), op1);
}
}
}
else
{
base = legitimize_pic_address (XEXP (addr, 0), reg);
- new = legitimize_pic_address (XEXP (addr, 1),
- base == reg ? NULL_RTX : reg);
+ new_rtx = legitimize_pic_address (XEXP (addr, 1),
+ base == reg ? NULL_RTX : reg);
- if (CONST_INT_P (new))
- new = plus_constant (base, INTVAL (new));
+ if (CONST_INT_P (new_rtx))
+ new_rtx = plus_constant (base, INTVAL (new_rtx));
else
{
- if (GET_CODE (new) == PLUS && CONSTANT_P (XEXP (new, 1)))
+ if (GET_CODE (new_rtx) == PLUS && CONSTANT_P (XEXP (new_rtx, 1)))
{
- base = gen_rtx_PLUS (Pmode, base, XEXP (new, 0));
- new = XEXP (new, 1);
+ base = gen_rtx_PLUS (Pmode, base, XEXP (new_rtx, 0));
+ new_rtx = XEXP (new_rtx, 1);
}
- new = gen_rtx_PLUS (Pmode, base, new);
+ new_rtx = gen_rtx_PLUS (Pmode, base, new_rtx);
}
}
}
}
- return new;
+ return new_rtx;
}
\f
/* Load the thread pointer. If TO_REG is true, force it into a register. */
if (TARGET_64BIT && ! TARGET_GNU2_TLS)
{
- rtx rax = gen_rtx_REG (Pmode, 0), insns;
+ rtx rax = gen_rtx_REG (Pmode, AX_REG), insns;
start_sequence ();
emit_call_insn (gen_tls_global_dynamic_64 (rax, x));
insns = get_insns ();
end_sequence ();
+ RTL_CONST_CALL_P (insns) = 1;
emit_libcall_block (insns, dest, rax, x);
}
else if (TARGET_64BIT && TARGET_GNU2_TLS)
if (TARGET_64BIT && ! TARGET_GNU2_TLS)
{
- rtx rax = gen_rtx_REG (Pmode, 0), insns, note;
+ rtx rax = gen_rtx_REG (Pmode, AX_REG), insns, note;
start_sequence ();
emit_call_insn (gen_tls_local_dynamic_base_64 (rax));
note = gen_rtx_EXPR_LIST (VOIDmode, const0_rtx, NULL);
note = gen_rtx_EXPR_LIST (VOIDmode, ix86_tls_get_addr (), note);
+ RTL_CONST_CALL_P (insns) = 1;
emit_libcall_block (insns, base, rax, note);
}
else if (TARGET_64BIT && TARGET_GNU2_TLS)
else if (flag_pic)
{
if (reload_in_progress)
- regs_ever_live[PIC_OFFSET_TABLE_REGNUM] = 1;
+ df_set_regs_ever_live (PIC_OFFSET_TABLE_REGNUM, true);
pic = pic_offset_table_rtx;
type = TARGET_ANY_GNU_TLS ? UNSPEC_GOTNTPOFF : UNSPEC_GOTTPOFF;
}
in.hash = htab_hash_pointer (decl);
in.base.from = decl;
loc = htab_find_slot_with_hash (dllimport_map, &in, in.hash, INSERT);
- h = *loc;
+ h = (struct tree_map *) *loc;
if (h)
return h->to;
- *loc = h = ggc_alloc (sizeof (struct tree_map));
+ *loc = h = GGC_NEW (struct tree_map);
h->hash = in.hash;
h->base.from = decl;
h->to = to = build_decl (VAR_DECL, NULL, ptr_type_node);
name = IDENTIFIER_POINTER (DECL_ASSEMBLER_NAME (decl));
name = targetm.strip_name_encoding (name);
- if (name[0] == FASTCALL_PREFIX)
- {
- name++;
- prefix = "*__imp_";
- }
- else
- prefix = "*__imp__";
-
+ prefix = name[0] == FASTCALL_PREFIX ? "*__imp_": "*__imp__";
namelen = strlen (name);
prefixlen = strlen (prefix);
- imp_name = alloca (namelen + prefixlen + 1);
+ imp_name = (char *) alloca (namelen + prefixlen + 1);
memcpy (imp_name, prefix, prefixlen);
memcpy (imp_name + prefixlen, name, namelen + 1);
set_mem_alias_set (rtl, ix86_GOT_alias_set ());
SET_DECL_RTL (to, rtl);
+ SET_DECL_ASSEMBLER_NAME (to, get_identifier (name));
return to;
}
log = GET_CODE (x) == SYMBOL_REF ? SYMBOL_REF_TLS_MODEL (x) : 0;
if (log)
- return legitimize_tls_address (x, log, false);
+ return legitimize_tls_address (x, (enum tls_model) log, false);
if (GET_CODE (x) == CONST
&& GET_CODE (XEXP (x, 0)) == PLUS
&& GET_CODE (XEXP (XEXP (x, 0), 0)) == SYMBOL_REF
&& (log = SYMBOL_REF_TLS_MODEL (XEXP (XEXP (x, 0), 0))))
{
- rtx t = legitimize_tls_address (XEXP (XEXP (x, 0), 0), log, false);
+ rtx t = legitimize_tls_address (XEXP (XEXP (x, 0), 0),
+ (enum tls_model) log, false);
return gen_rtx_PLUS (Pmode, t, XEXP (XEXP (x, 0), 1));
}
- if (flag_pic && SYMBOLIC_CONST (x))
- return legitimize_pic_address (x, 0);
-
if (TARGET_DLLIMPORT_DECL_ATTRIBUTES)
{
if (GET_CODE (x) == SYMBOL_REF && SYMBOL_REF_DLLIMPORT_P (x))
}
}
+ if (flag_pic && SYMBOLIC_CONST (x))
+ return legitimize_pic_address (x, 0);
+
/* Canonicalize shifts by 0, 1, 2, 3 into multiply */
if (GET_CODE (x) == ASHIFT
&& CONST_INT_P (XEXP (x, 1))
fputs ("@PLTOFF", file);
break;
case UNSPEC_GOTPCREL:
- fputs ("@GOTPCREL(%rip)", file);
+ fputs (ASSEMBLER_DIALECT == ASM_ATT ?
+ "@GOTPCREL(%rip)" : "@GOTPCREL[rip]", file);
break;
case UNSPEC_GOTTPOFF:
/* FIXME: This might be @TPOFF in Sun ld too. */
break;
case UNSPEC_GOTNTPOFF:
if (TARGET_64BIT)
- fputs ("@GOTTPOFF(%rip)", file);
+ fputs (ASSEMBLER_DIALECT == ASM_ATT ?
+ "@GOTTPOFF(%rip)": "@GOTTPOFF[rip]", file);
else
fputs ("@GOTNTPOFF", file);
break;
switch (code)
{
case EQ:
- suffix = "e";
+ switch (mode)
+ {
+ case CCAmode:
+ suffix = "a";
+ break;
+
+ case CCCmode:
+ suffix = "c";
+ break;
+
+ case CCOmode:
+ suffix = "o";
+ break;
+
+ case CCSmode:
+ suffix = "s";
+ break;
+
+ default:
+ suffix = "e";
+ }
break;
case NE:
- suffix = "ne";
+ switch (mode)
+ {
+ case CCAmode:
+ suffix = "na";
+ break;
+
+ case CCCmode:
+ suffix = "nc";
+ break;
+
+ case CCOmode:
+ suffix = "no";
+ break;
+
+ case CCSmode:
+ suffix = "ns";
+ break;
+
+ default:
+ suffix = "ne";
+ }
break;
case GT:
gcc_assert (mode == CCmode || mode == CCNOmode || mode == CCGCmode);
case GTU:
/* ??? Use "nbe" instead of "a" for fcmov lossage on some assemblers.
Those same assemblers have the same but opposite lossage on cmov. */
- gcc_assert (mode == CCmode);
- suffix = fp ? "nbe" : "a";
+ if (mode == CCmode)
+ suffix = fp ? "nbe" : "a";
+ else if (mode == CCCmode)
+ suffix = "b";
+ else
+ gcc_unreachable ();
break;
case LT:
switch (mode)
}
break;
case LTU:
- gcc_assert (mode == CCmode);
+ gcc_assert (mode == CCmode || mode == CCCmode);
suffix = "b";
break;
case GE:
break;
case GEU:
/* ??? As above. */
- gcc_assert (mode == CCmode);
+ gcc_assert (mode == CCmode || mode == CCCmode);
suffix = fp ? "nb" : "ae";
break;
case LE:
suffix = "le";
break;
case LEU:
- gcc_assert (mode == CCmode);
- suffix = "be";
+ /* ??? As above. */
+ if (mode == CCmode)
+ suffix = "be";
+ else if (mode == CCCmode)
+ suffix = fp ? "nb" : "ae";
+ else
+ gcc_unreachable ();
break;
case UNORDERED:
suffix = fp ? "u" : "p";
void
print_reg (rtx x, int code, FILE *file)
{
- gcc_assert (REGNO (x) != ARG_POINTER_REGNUM
- && REGNO (x) != FRAME_POINTER_REGNUM
- && REGNO (x) != FLAGS_REG
- && REGNO (x) != FPSR_REG
- && REGNO (x) != FPCR_REG);
+ gcc_assert (x == pc_rtx
+ || (REGNO (x) != ARG_POINTER_REGNUM
+ && REGNO (x) != FRAME_POINTER_REGNUM
+ && REGNO (x) != FLAGS_REG
+ && REGNO (x) != FPSR_REG
+ && REGNO (x) != FPCR_REG));
- if (ASSEMBLER_DIALECT == ASM_ATT || USER_LABEL_PREFIX[0] == 0)
+ if (ASSEMBLER_DIALECT == ASM_ATT)
putc ('%', file);
+ if (x == pc_rtx)
+ {
+ gcc_assert (TARGET_64BIT);
+ fputs ("rip", file);
+ return;
+ }
+
if (code == 'w' || MMX_REG_P (x))
code = 2;
else if (code == 'b')
X -- don't print any sort of PIC '@' suffix for a symbol.
& -- print some in-use local-dynamic symbol name.
H -- print a memory address offset by 8; used for sse high-parts
+ Y -- print condition for SSE5 com* instruction.
+ + -- print a branch hint as 'cs' or 'ds' prefix
+ ; -- print a semicolon (after prefixes due to bug in older gas).
*/
void
return;
case 2:
+ if (MEM_P (x))
+ {
#ifdef HAVE_GAS_FILDS_FISTS
- putc ('s', file);
+ putc ('s', file);
#endif
+ return;
+ }
+ else
+ putc ('w', file);
return;
case 4:
}
return;
}
+
+ case 'Y':
+ switch (GET_CODE (x))
+ {
+ case NE:
+ fputs ("neq", file);
+ break;
+ case EQ:
+ fputs ("eq", file);
+ break;
+ case GE:
+ case GEU:
+ fputs (INTEGRAL_MODE_P (GET_MODE (x)) ? "ge" : "unlt", file);
+ break;
+ case GT:
+ case GTU:
+ fputs (INTEGRAL_MODE_P (GET_MODE (x)) ? "gt" : "unle", file);
+ break;
+ case LE:
+ case LEU:
+ fputs ("le", file);
+ break;
+ case LT:
+ case LTU:
+ fputs ("lt", file);
+ break;
+ case UNORDERED:
+ fputs ("unord", file);
+ break;
+ case ORDERED:
+ fputs ("ord", file);
+ break;
+ case UNEQ:
+ fputs ("ueq", file);
+ break;
+ case UNGE:
+ fputs ("nlt", file);
+ break;
+ case UNGT:
+ fputs ("nle", file);
+ break;
+ case UNLE:
+ fputs ("ule", file);
+ break;
+ case UNLT:
+ fputs ("ult", file);
+ break;
+ case LTGT:
+ fputs ("une", file);
+ break;
+ default:
+ gcc_unreachable ();
+ }
+ return;
+
+ case ';':
+#if TARGET_MACHO
+ fputs (" ; ", file);
+#else
+ fputc (' ', file);
+#endif
+ return;
+
default:
output_operand_lossage ("invalid operand code '%c'", code);
}
else if (MEM_P (x))
{
- /* No `byte ptr' prefix for call instructions. */
- if (ASSEMBLER_DIALECT == ASM_INTEL && code != 'X' && code != 'P')
+ /* No `byte ptr' prefix for call instructions or BLKmode operands. */
+ if (ASSEMBLER_DIALECT == ASM_INTEL && code != 'X' && code != 'P'
+ && GET_MODE (x) != BLKmode)
{
const char * size;
switch (GET_MODE_SIZE (GET_MODE (x)))
case 4: size = "DWORD"; break;
case 8: size = "QWORD"; break;
case 12: size = "XWORD"; break;
- case 16: size = "XMMWORD"; break;
+ case 16:
+ if (GET_MODE (x) == XFmode)
+ size = "XWORD";
+ else
+ size = "XMMWORD";
+ break;
default:
gcc_unreachable ();
}
if (ASSEMBLER_DIALECT == ASM_ATT)
putc ('$', file);
- fprintf (file, "0x%08lx", l);
+ fprintf (file, "0x%08lx", (long unsigned int) l);
}
/* These float cases don't actually occur as immediate operands. */
break;
case SEG_FS:
case SEG_GS:
- if (USER_LABEL_PREFIX[0] == 0)
+ if (ASSEMBLER_DIALECT == ASM_ATT)
putc ('%', file);
fputs ((parts.seg == SEG_FS ? "fs:" : "gs:"), file);
break;
gcc_unreachable ();
}
+ /* Use one byte shorter RIP relative addressing for 64bit mode. */
+ if (TARGET_64BIT && !base && !index)
+ {
+ rtx symbol = disp;
+
+ if (GET_CODE (disp) == CONST
+ && GET_CODE (XEXP (disp, 0)) == PLUS
+ && CONST_INT_P (XEXP (XEXP (disp, 0), 1)))
+ symbol = XEXP (XEXP (disp, 0), 0);
+
+ if (GET_CODE (symbol) == LABEL_REF
+ || (GET_CODE (symbol) == SYMBOL_REF
+ && SYMBOL_REF_TLS_MODEL (symbol) == 0))
+ base = pc_rtx;
+ }
if (!base && !index)
{
/* Displacement only requires special attention. */
if (CONST_INT_P (disp))
{
if (ASSEMBLER_DIALECT == ASM_INTEL && parts.seg == SEG_DEFAULT)
- {
- if (USER_LABEL_PREFIX[0] == 0)
- putc ('%', file);
- fputs ("ds:", file);
- }
+ fputs ("ds:", file);
fprintf (file, HOST_WIDE_INT_PRINT_DEC, INTVAL (disp));
}
else if (flag_pic)
output_pic_addr_const (file, disp, 0);
else
output_addr_const (file, disp);
-
- /* Use one byte shorter RIP relative addressing for 64bit mode. */
- if (TARGET_64BIT)
- {
- if (GET_CODE (disp) == CONST
- && GET_CODE (XEXP (disp, 0)) == PLUS
- && CONST_INT_P (XEXP (XEXP (disp, 0), 1)))
- disp = XEXP (XEXP (disp, 0), 0);
- if (GET_CODE (disp) == LABEL_REF
- || (GET_CODE (disp) == SYMBOL_REF
- && SYMBOL_REF_TLS_MODEL (disp) == 0))
- fputs ("(%rip)", file);
- }
}
else
{
case UNSPEC_GOTNTPOFF:
output_addr_const (file, op);
if (TARGET_64BIT)
- fputs ("@GOTTPOFF(%rip)", file);
+ fputs (ASSEMBLER_DIALECT == ASM_ATT ?
+ "@GOTTPOFF(%rip)" : "@GOTTPOFF[rip]", file);
else
fputs ("@GOTNTPOFF", file);
break;
rtx stored_mode = assign_386_stack_local (HImode, SLOT_CW_STORED);
rtx new_mode;
- int slot;
+ enum ix86_stack_slot slot;
rtx reg = gen_reg_rtx (HImode);
gcc_assert (STACK_TOP_P (operands[1]));
gcc_assert (MEM_P (operands[0]));
+ gcc_assert (GET_MODE (operands[1]) != TFmode);
if (fisttp)
output_asm_insn ("fisttp%z0\t%0", operands);
{
static char retval[] = ".word\t0xc_df";
int regno = REGNO (operands[opno]);
-
+
gcc_assert (FP_REGNO_P (regno));
retval[9] = '0' + (regno - FIRST_STACK_REG);
/* This predicate should match that for movsi_xor and movdi_xor_rex64. */
if (reload_completed && (!TARGET_USE_MOV0 || optimize_size))
{
- rtx clob = gen_rtx_CLOBBER (VOIDmode, gen_rtx_REG (CCmode, 17));
+ rtx clob = gen_rtx_CLOBBER (VOIDmode, gen_rtx_REG (CCmode, FLAGS_REG));
tmp = gen_rtx_PARALLEL (VOIDmode, gen_rtvec (2, tmp, clob));
}
void
ix86_expand_move (enum machine_mode mode, rtx operands[])
{
- int strict = (reload_in_progress || reload_completed);
rtx op0, op1;
enum tls_model model;
op1 = force_reg (Pmode, op1);
else if (!TARGET_64BIT || !x86_64_movabs_operand (op1, Pmode))
{
- rtx reg = no_new_pseudos ? op0 : NULL_RTX;
+ rtx reg = !can_create_pseudo_p () ? op0 : NULL_RTX;
op1 = legitimize_pic_address (op1, reg);
if (op0 == op1)
return;
/* Force large constants in 64bit compilation into register
to get them CSEed. */
- if (TARGET_64BIT && mode == DImode
+ if (can_create_pseudo_p ()
+ && (mode == DImode) && TARGET_64BIT
&& immediate_operand (op1, mode)
&& !x86_64_zext_immediate_operand (op1, VOIDmode)
&& !register_operand (op0, mode)
- && optimize && !reload_completed && !reload_in_progress)
+ && optimize)
op1 = copy_to_mode_reg (mode, op1);
- if (FLOAT_MODE_P (mode))
+ if (can_create_pseudo_p ()
+ && FLOAT_MODE_P (mode)
+ && GET_CODE (op1) == CONST_DOUBLE)
{
/* If we are loading a floating point constant to a register,
force the value to memory now, since we'll get better code
out the back end. */
- if (strict)
- ;
- else if (GET_CODE (op1) == CONST_DOUBLE)
+ op1 = validize_mem (force_const_mem (mode, op1));
+ if (!register_operand (op0, mode))
{
- op1 = validize_mem (force_const_mem (mode, op1));
- if (!register_operand (op0, mode))
- {
- rtx temp = gen_reg_rtx (mode);
- emit_insn (gen_rtx_SET (VOIDmode, temp, op1));
- emit_move_insn (op0, temp);
- return;
- }
+ rtx temp = gen_reg_rtx (mode);
+ emit_insn (gen_rtx_SET (VOIDmode, temp, op1));
+ emit_move_insn (op0, temp);
+ return;
}
}
}
ix86_expand_vector_move (enum machine_mode mode, rtx operands[])
{
rtx op0 = operands[0], op1 = operands[1];
+ unsigned int align = GET_MODE_ALIGNMENT (mode);
/* Force constants other than zero into memory. We do not know how
the instructions used to build constants modify the upper 64 bits
of the register, once we have that information we may be able
to handle some of them more efficiently. */
- if ((reload_in_progress | reload_completed) == 0
+ if (can_create_pseudo_p ()
&& register_operand (op0, mode)
- && CONSTANT_P (op1)
+ && (CONSTANT_P (op1)
+ || (GET_CODE (op1) == SUBREG
+ && CONSTANT_P (SUBREG_REG (op1))))
&& standard_sse_constant_p (op1) <= 0)
op1 = validize_mem (force_const_mem (mode, op1));
+ /* TDmode values are passed as TImode on the stack. TImode values
+ are moved via xmm registers, and moving them to stack can result in
+ unaligned memory access. Use ix86_expand_vector_move_misalign()
+ if memory operand is not aligned correctly. */
+ if (can_create_pseudo_p ()
+ && (mode == TImode) && !TARGET_64BIT
+ && ((MEM_P (op0) && (MEM_ALIGN (op0) < align))
+ || (MEM_P (op1) && (MEM_ALIGN (op1) < align))))
+ {
+ rtx tmp[2];
+
+ /* ix86_expand_vector_move_misalign() does not like constants ... */
+ if (CONSTANT_P (op1)
+ || (GET_CODE (op1) == SUBREG
+ && CONSTANT_P (SUBREG_REG (op1))))
+ op1 = validize_mem (force_const_mem (mode, op1));
+
+ /* ... nor both arguments in memory. */
+ if (!register_operand (op0, mode)
+ && !register_operand (op1, mode))
+ op1 = force_reg (mode, op1);
+
+ tmp[0] = op0; tmp[1] = op1;
+ ix86_expand_vector_move_misalign (mode, tmp);
+ return;
+ }
+
/* Make operand1 a register if it isn't already. */
- if (!no_new_pseudos
+ if (can_create_pseudo_p ()
&& !register_operand (op0, mode)
&& !register_operand (op1, mode))
{
movlpd mem, reg (gas syntax)
else
movsd mem, reg
-
+
Code generation for unaligned packed loads of single precision data
(x86_sse_unaligned_move_optimal overrides x86_sse_partial_reg_dependency):
if (x86_sse_unaligned_move_optimal)
/* Helper function of ix86_fixup_binary_operands to canonicalize
operand order. Returns true if the operands should be swapped. */
-
+
static bool
ix86_swap_binary_operands_p (enum rtx_code code, enum machine_mode mode,
rtx operands[])
/* Canonicalize operand order. */
if (ix86_swap_binary_operands_p (code, mode, operands))
{
- rtx temp = src1;
+ rtx temp;
+
+ /* It is invalid to swap operands of different modes. */
+ gcc_assert (GET_MODE (src1) == GET_MODE (src2));
+
+ temp = src1;
src1 = src2;
src2 = temp;
}
ix86_expand_vector_extract (false, target, fp_xmm, 0);
}
+/* Not used, but eases macroization of patterns. */
+void
+ix86_expand_convert_uns_sixf_sse (rtx target ATTRIBUTE_UNUSED,
+ rtx input ATTRIBUTE_UNUSED)
+{
+ gcc_unreachable ();
+}
+
/* Convert an unsigned SImode value into a DFmode. Only currently used
for SSE, but applicable anywhere. */
{
REAL_VALUE_TYPE TWO32r;
rtx fp_lo, fp_hi, x;
-
+
fp_lo = gen_reg_rtx (DFmode);
fp_hi = gen_reg_rtx (DFmode);
rtvec v;
switch (mode)
{
+ case SImode:
+ gcc_assert (vect);
+ v = gen_rtvec (4, value, value, value, value);
+ return gen_rtx_CONST_VECTOR (V4SImode, v);
+
+ case DImode:
+ gcc_assert (vect);
+ v = gen_rtvec (2, value, value);
+ return gen_rtx_CONST_VECTOR (V2DImode, v);
+
case SFmode:
if (vect)
v = gen_rtvec (4, value, value, value, value);
}
}
-/* A subroutine of ix86_expand_fp_absneg_operator and copysign expanders.
- Create a mask for the sign bit in MODE for an SSE register. If VECT is
- true, then replicate the mask for all elements of the vector register.
- If INVERT is true, then create a mask excluding the sign bit. */
+/* A subroutine of ix86_expand_fp_absneg_operator, copysign expanders
+ and ix86_expand_int_vcond. Create a mask for the sign bit in MODE
+ for an SSE register. If VECT is true, then replicate the mask for
+ all elements of the vector register. If INVERT is true, then create
+ a mask excluding the sign bit. */
rtx
ix86_build_signbit_mask (enum machine_mode mode, bool vect, bool invert)
{
- enum machine_mode vec_mode;
+ enum machine_mode vec_mode, imode;
HOST_WIDE_INT hi, lo;
int shift = 63;
rtx v;
rtx mask;
/* Find the sign bit, sign extended to 2*HWI. */
- if (mode == SFmode)
- lo = 0x80000000, hi = lo < 0;
- else if (HOST_BITS_PER_WIDE_INT >= 64)
- lo = (HOST_WIDE_INT)1 << shift, hi = -1;
- else
- lo = 0, hi = (HOST_WIDE_INT)1 << (shift - HOST_BITS_PER_WIDE_INT);
+ switch (mode)
+ {
+ case SImode:
+ case SFmode:
+ imode = SImode;
+ vec_mode = (mode == SImode) ? V4SImode : V4SFmode;
+ lo = 0x80000000, hi = lo < 0;
+ break;
+
+ case DImode:
+ case DFmode:
+ imode = DImode;
+ vec_mode = (mode == DImode) ? V2DImode : V2DFmode;
+ if (HOST_BITS_PER_WIDE_INT >= 64)
+ lo = (HOST_WIDE_INT)1 << shift, hi = -1;
+ else
+ lo = 0, hi = (HOST_WIDE_INT)1 << (shift - HOST_BITS_PER_WIDE_INT);
+ break;
+
+ case TImode:
+ case TFmode:
+ imode = TImode;
+ vec_mode = VOIDmode;
+ gcc_assert (HOST_BITS_PER_WIDE_INT >= 64);
+ lo = 0, hi = (HOST_WIDE_INT)1 << shift;
+ break;
+
+ default:
+ gcc_unreachable ();
+ }
if (invert)
lo = ~lo, hi = ~hi;
/* Force this value into the low part of a fp vector constant. */
- mask = immed_double_const (lo, hi, mode == SFmode ? SImode : DImode);
+ mask = immed_double_const (lo, hi, imode);
mask = gen_lowpart (mode, mask);
+ if (vec_mode == VOIDmode)
+ return force_reg (mode, mask);
+
v = ix86_build_const_vector (mode, vect, mask);
- vec_mode = (mode == SFmode) ? V4SFmode : V2DFmode;
return force_reg (vec_mode, v);
}
rtx operands[])
{
rtx mask, set, use, clob, dst, src;
- bool matching_memory;
bool use_sse = false;
bool vector_mode = VECTOR_MODE_P (mode);
enum machine_mode elt_mode = mode;
elt_mode = GET_MODE_INNER (mode);
use_sse = true;
}
+ else if (mode == TFmode)
+ use_sse = true;
else if (TARGET_SSE_MATH)
use_sse = SSE_FLOAT_MODE_P (mode);
dst = operands[0];
src = operands[1];
- /* If the destination is memory, and we don't have matching source
- operands or we're using the x87, do things in registers. */
- matching_memory = false;
- if (MEM_P (dst))
- {
- if (use_sse && rtx_equal_p (dst, src))
- matching_memory = true;
- else
- dst = gen_reg_rtx (mode);
- }
- if (MEM_P (src) && !matching_memory)
- src = force_reg (mode, src);
-
if (vector_mode)
{
set = gen_rtx_fmt_ee (code == NEG ? XOR : AND, mode, src, mask);
else
emit_insn (set);
}
-
- if (dst != operands[0])
- emit_move_insn (operands[0], dst);
}
/* Expand a copysign operation. Special case operand 0 being a constant. */
void
ix86_expand_copysign (rtx operands[])
{
- enum machine_mode mode, vmode;
+ enum machine_mode mode;
rtx dest, op0, op1, mask, nmask;
dest = operands[0];
op1 = operands[2];
mode = GET_MODE (dest);
- vmode = mode == SFmode ? V4SFmode : V2DFmode;
if (GET_CODE (op0) == CONST_DOUBLE)
{
- rtvec v;
+ rtx (*copysign_insn)(rtx, rtx, rtx, rtx);
if (real_isneg (CONST_DOUBLE_REAL_VALUE (op0)))
op0 = simplify_unary_operation (ABS, mode, op0, mode);
- if (op0 == CONST0_RTX (mode))
- op0 = CONST0_RTX (vmode);
- else
- {
- if (mode == SFmode)
- v = gen_rtvec (4, op0, CONST0_RTX (SFmode),
- CONST0_RTX (SFmode), CONST0_RTX (SFmode));
- else
- v = gen_rtvec (2, op0, CONST0_RTX (DFmode));
- op0 = force_reg (vmode, gen_rtx_CONST_VECTOR (vmode, v));
- }
+ if (mode == SFmode || mode == DFmode)
+ {
+ enum machine_mode vmode;
+
+ vmode = mode == SFmode ? V4SFmode : V2DFmode;
+
+ if (op0 == CONST0_RTX (mode))
+ op0 = CONST0_RTX (vmode);
+ else
+ {
+ rtvec v;
+
+ if (mode == SFmode)
+ v = gen_rtvec (4, op0, CONST0_RTX (SFmode),
+ CONST0_RTX (SFmode), CONST0_RTX (SFmode));
+ else
+ v = gen_rtvec (2, op0, CONST0_RTX (DFmode));
+
+ op0 = force_reg (vmode, gen_rtx_CONST_VECTOR (vmode, v));
+ }
+ }
+ else if (op0 != CONST0_RTX (mode))
+ op0 = force_reg (mode, op0);
mask = ix86_build_signbit_mask (mode, 0, 0);
if (mode == SFmode)
- emit_insn (gen_copysignsf3_const (dest, op0, op1, mask));
+ copysign_insn = gen_copysignsf3_const;
+ else if (mode == DFmode)
+ copysign_insn = gen_copysigndf3_const;
else
- emit_insn (gen_copysigndf3_const (dest, op0, op1, mask));
+ copysign_insn = gen_copysigntf3_const;
+
+ emit_insn (copysign_insn (dest, op0, op1, mask));
}
else
{
+ rtx (*copysign_insn)(rtx, rtx, rtx, rtx, rtx, rtx);
+
nmask = ix86_build_signbit_mask (mode, 0, 1);
mask = ix86_build_signbit_mask (mode, 0, 0);
if (mode == SFmode)
- emit_insn (gen_copysignsf3_var (dest, NULL, op0, op1, nmask, mask));
+ copysign_insn = gen_copysignsf3_var;
+ else if (mode == DFmode)
+ copysign_insn = gen_copysigndf3_var;
else
- emit_insn (gen_copysigndf3_var (dest, NULL, op0, op1, nmask, mask));
+ copysign_insn = gen_copysigntf3_var;
+
+ emit_insn (copysign_insn (dest, NULL_RTX, op0, op1, nmask, mask));
}
}
enum machine_mode
ix86_cc_mode (enum rtx_code code, rtx op0, rtx op1)
{
- if (SCALAR_FLOAT_MODE_P (GET_MODE (op0)))
- return ix86_fp_compare_mode (code);
+ enum machine_mode mode = GET_MODE (op0);
+
+ if (SCALAR_FLOAT_MODE_P (mode))
+ {
+ gcc_assert (!DECIMAL_FLOAT_MODE_P (mode));
+ return ix86_fp_compare_mode (code);
+ }
+
switch (code)
{
/* Only zero flag is needed. */
return CCZmode;
/* Codes needing carry flag. */
case GEU: /* CF=0 */
- case GTU: /* CF=0 & ZF=0 */
case LTU: /* CF=1 */
+ /* Detect overflow checks. They need just the carry flag. */
+ if (GET_CODE (op0) == PLUS
+ && rtx_equal_p (op1, XEXP (op0, 0)))
+ return CCCmode;
+ else
+ return CCmode;
+ case GTU: /* CF=0 & ZF=0 */
case LEU: /* CF=1 | ZF=1 */
- return CCmode;
+ /* Detect overflow checks. They need just the carry flag. */
+ if (GET_CODE (op0) == MINUS
+ && rtx_equal_p (op1, XEXP (op0, 0)))
+ return CCCmode;
+ else
+ return CCmode;
/* Codes possibly doable only with sign flag when
comparing against zero. */
case GE: /* SF=OF or SF=0 */
case CCGCmode:
case CCGOCmode:
case CCNOmode:
+ case CCAmode:
+ case CCCmode:
+ case CCOmode:
+ case CCSmode:
case CCZmode:
switch (m2)
{
case CCGCmode:
case CCGOCmode:
case CCNOmode:
+ case CCAmode:
+ case CCCmode:
+ case CCOmode:
+ case CCSmode:
case CCZmode:
return CCmode;
}
/* Try to rearrange the comparison to make it cheaper. */
if (ix86_fp_comparison_cost (code)
> ix86_fp_comparison_cost (swap_condition (code))
- && (REG_P (op1) || !no_new_pseudos))
+ && (REG_P (op1) || can_create_pseudo_p ()))
{
rtx tmp;
tmp = op0, op0 = op1, op1 = tmp;
ix86_fp_comparison_codes (code, &bypass_code, &first_code, &second_code);
/* Do fcomi/sahf based test when profitable. */
- if ((TARGET_CMOVE || TARGET_SAHF)
+ if (ix86_fp_comparison_arithmetics_cost (code) > cost
&& (bypass_code == UNKNOWN || bypass_test)
- && (second_code == UNKNOWN || second_test)
- && ix86_fp_comparison_arithmetics_cost (code) > cost)
+ && (second_code == UNKNOWN || second_test))
{
+ tmp = gen_rtx_COMPARE (fpcmp_mode, op0, op1);
+ tmp = gen_rtx_SET (VOIDmode, gen_rtx_REG (fpcmp_mode, FLAGS_REG),
+ tmp);
if (TARGET_CMOVE)
- {
- tmp = gen_rtx_COMPARE (fpcmp_mode, op0, op1);
- tmp = gen_rtx_SET (VOIDmode, gen_rtx_REG (fpcmp_mode, FLAGS_REG),
- tmp);
- emit_insn (tmp);
- }
+ emit_insn (tmp);
else
{
- tmp = gen_rtx_COMPARE (fpcmp_mode, op0, op1);
- tmp2 = gen_rtx_UNSPEC (HImode, gen_rtvec (1, tmp), UNSPEC_FNSTSW);
+ gcc_assert (TARGET_SAHF);
+
if (!scratch)
scratch = gen_reg_rtx (HImode);
- emit_insn (gen_rtx_SET (VOIDmode, scratch, tmp2));
- emit_insn (gen_x86_sahf_1 (scratch));
+ tmp2 = gen_rtx_CLOBBER (VOIDmode, scratch);
+
+ emit_insn (gen_rtx_PARALLEL (VOIDmode, gen_rtvec (2, tmp, tmp2)));
}
/* The FP codes work out to act like unsigned. */
ix86_compare_emitted = NULL_RTX;
}
else if (SCALAR_FLOAT_MODE_P (GET_MODE (op0)))
- ret = ix86_expand_fp_compare (code, op0, op1, NULL_RTX,
- second_test, bypass_test);
+ {
+ gcc_assert (!DECIMAL_FLOAT_MODE_P (GET_MODE (op0)));
+ ret = ix86_expand_fp_compare (code, op0, op1, NULL_RTX,
+ second_test, bypass_test);
+ }
else
ret = ix86_expand_int_compare (code, op0, op1);
/* Check whether we will use the natural sequence with one jump. If
so, we can expand jump early. Otherwise delay expansion by
creating compound insn to not confuse optimizers. */
- if (bypass_code == UNKNOWN && second_code == UNKNOWN
- && TARGET_CMOVE)
+ if (bypass_code == UNKNOWN && second_code == UNKNOWN)
{
ix86_split_fp_branch (code, ix86_compare_op0, ix86_compare_op1,
gen_rtx_LABEL_REF (VOIDmode, label),
vec = rtvec_alloc (3 + !use_fcomi);
RTVEC_ELT (vec, 0) = tmp;
RTVEC_ELT (vec, 1)
- = gen_rtx_CLOBBER (VOIDmode, gen_rtx_REG (CCFPmode, 18));
+ = gen_rtx_CLOBBER (VOIDmode, gen_rtx_REG (CCFPmode, FPSR_REG));
RTVEC_ELT (vec, 2)
- = gen_rtx_CLOBBER (VOIDmode, gen_rtx_REG (CCFPmode, 17));
+ = gen_rtx_CLOBBER (VOIDmode, gen_rtx_REG (CCFPmode, FLAGS_REG));
if (! use_fcomi)
RTVEC_ELT (vec, 3)
= gen_rtx_CLOBBER (VOIDmode, gen_rtx_SCRATCH (HImode));
/* Otherwise, if we are doing less-than or greater-or-equal-than,
op1 is a constant and the low word is zero, then we can just
- examine the high word. */
+ examine the high word. Similarly for low word -1 and
+ less-or-equal-than or greater-than. */
- if (CONST_INT_P (hi[1]) && lo[1] == const0_rtx)
+ if (CONST_INT_P (hi[1]))
switch (code)
{
case LT: case LTU: case GE: case GEU:
- ix86_compare_op0 = hi[0];
- ix86_compare_op1 = hi[1];
- ix86_expand_branch (code, label);
- return;
+ if (lo[1] == const0_rtx)
+ {
+ ix86_compare_op0 = hi[0];
+ ix86_compare_op1 = hi[1];
+ ix86_expand_branch (code, label);
+ return;
+ }
+ break;
+ case LE: case LEU: case GT: case GTU:
+ if (lo[1] == constm1_rtx)
+ {
+ ix86_compare_op0 = hi[0];
+ ix86_compare_op1 = hi[1];
+ ix86_expand_branch (code, label);
+ return;
+ }
+ break;
default:
break;
}
enum machine_mode mode =
GET_MODE (op0) != VOIDmode ? GET_MODE (op0) : GET_MODE (op1);
- /* Do not handle DImode compares that go through special path. Also we can't
- deal with FP compares yet. This is possible to add. */
+ /* Do not handle DImode compares that go through special path. */
if (mode == (TARGET_64BIT ? TImode : DImode))
return false;
- if (FLOAT_MODE_P (mode))
+
+ if (SCALAR_FLOAT_MODE_P (mode))
{
rtx second_test = NULL, bypass_test = NULL;
rtx compare_op, compare_seq;
- /* Shortcut: following common codes never translate into carry flag compares. */
+ gcc_assert (!DECIMAL_FLOAT_MODE_P (mode));
+
+ /* Shortcut: following common codes never translate
+ into carry flag compares. */
if (code == EQ || code == NE || code == UNEQ || code == LTGT
|| code == ORDERED || code == UNORDERED)
return false;
code = swap_condition (code);
}
- /* Try to expand the comparison and verify that we end up with carry flag
- based comparison. This is fails to be true only when we decide to expand
- comparison using arithmetic that is not too common scenario. */
+ /* Try to expand the comparison and verify that we end up with
+ carry flag based comparison. This fails to be true only when
+ we decide to expand comparison using arithmetic that is not
+ too common scenario. */
start_sequence ();
compare_op = ix86_expand_fp_compare (code, op0, op1, NULL_RTX,
&second_test, &bypass_test);
if (second_test || bypass_test)
return false;
+
if (GET_MODE (XEXP (compare_op, 0)) == CCFPmode
|| GET_MODE (XEXP (compare_op, 0)) == CCFPUmode)
code = ix86_fp_compare_code_to_integer (GET_CODE (compare_op));
else
code = GET_CODE (compare_op);
+
if (code != LTU && code != GEU)
return false;
+
emit_insn (compare_seq);
*pop = compare_op;
return true;
}
+
if (!INTEGRAL_MODE_P (mode))
return false;
+
switch (code)
{
case LTU:
/* Swapping operands may cause constant to appear as first operand. */
if (!nonimmediate_operand (op0, VOIDmode))
{
- if (no_new_pseudos)
+ if (!can_create_pseudo_p ())
return false;
op0 = force_reg (mode, op0);
}
if (diff < 0)
{
+ enum machine_mode cmp_mode = GET_MODE (ix86_compare_op0);
+
HOST_WIDE_INT tmp;
tmp = ct, ct = cf, cf = tmp;
diff = -diff;
- if (FLOAT_MODE_P (GET_MODE (ix86_compare_op0)))
+
+ if (SCALAR_FLOAT_MODE_P (cmp_mode))
{
+ gcc_assert (!DECIMAL_FLOAT_MODE_P (cmp_mode));
+
/* We may be reversing unordered compare to normal compare, that
is not valid in general (we may convert non-trapping condition
to trapping one), however on i386 we currently emit all
{
if (cf == 0)
{
+ enum machine_mode cmp_mode = GET_MODE (ix86_compare_op0);
+
cf = ct;
ct = 0;
- if (FLOAT_MODE_P (GET_MODE (ix86_compare_op0)))
- /* We may be reversing unordered compare to normal compare,
- that is not valid in general (we may convert non-trapping
- condition to trapping one), however on i386 we currently
- emit all comparisons unordered. */
- code = reverse_condition_maybe_unordered (code);
+
+ if (SCALAR_FLOAT_MODE_P (cmp_mode))
+ {
+ gcc_assert (!DECIMAL_FLOAT_MODE_P (cmp_mode));
+
+ /* We may be reversing unordered compare to normal compare,
+ that is not valid in general (we may convert non-trapping
+ condition to trapping one), however on i386 we currently
+ emit all comparisons unordered. */
+ code = reverse_condition_maybe_unordered (code);
+ }
else
{
code = reverse_condition (code);
x = gen_rtx_AND (mode, x, op_false);
emit_insn (gen_rtx_SET (VOIDmode, dest, x));
}
+ else if (TARGET_SSE5)
+ {
+ rtx pcmov = gen_rtx_SET (mode, dest,
+ gen_rtx_IF_THEN_ELSE (mode, cmp,
+ op_true,
+ op_false));
+ emit_insn (pcmov);
+ }
else
{
op_true = force_reg (mode, op_true);
return true;
}
-/* Expand a signed integral vector conditional move. */
+/* Expand a signed/unsigned integral vector conditional move. */
bool
ix86_expand_int_vcond (rtx operands[])
cop0 = operands[4];
cop1 = operands[5];
- /* Canonicalize the comparison to EQ, GT, GTU. */
- switch (code)
+ /* SSE5 supports all of the comparisons on all vector int types. */
+ if (!TARGET_SSE5)
{
- case EQ:
- case GT:
- case GTU:
- break;
+ /* Canonicalize the comparison to EQ, GT, GTU. */
+ switch (code)
+ {
+ case EQ:
+ case GT:
+ case GTU:
+ break;
- case NE:
- case LE:
- case LEU:
- code = reverse_condition (code);
- negate = true;
- break;
+ case NE:
+ case LE:
+ case LEU:
+ code = reverse_condition (code);
+ negate = true;
+ break;
- case GE:
- case GEU:
- code = reverse_condition (code);
- negate = true;
- /* FALLTHRU */
+ case GE:
+ case GEU:
+ code = reverse_condition (code);
+ negate = true;
+ /* FALLTHRU */
- case LT:
- case LTU:
- code = swap_condition (code);
- x = cop0, cop0 = cop1, cop1 = x;
- break;
+ case LT:
+ case LTU:
+ code = swap_condition (code);
+ x = cop0, cop0 = cop1, cop1 = x;
+ break;
- default:
- gcc_unreachable ();
- }
+ default:
+ gcc_unreachable ();
+ }
- /* Unsigned parallel compare is not supported by the hardware. Play some
- tricks to turn this into a signed comparison against 0. */
- if (code == GTU)
- {
- cop0 = force_reg (mode, cop0);
+ /* Only SSE4.1/SSE4.2 supports V2DImode. */
+ if (mode == V2DImode)
+ {
+ switch (code)
+ {
+ case EQ:
+ /* SSE4.1 supports EQ. */
+ if (!TARGET_SSE4_1)
+ return false;
+ break;
- switch (mode)
+ case GT:
+ case GTU:
+ /* SSE4.2 supports GT/GTU. */
+ if (!TARGET_SSE4_2)
+ return false;
+ break;
+
+ default:
+ gcc_unreachable ();
+ }
+ }
+
+ /* Unsigned parallel compare is not supported by the hardware. Play some
+ tricks to turn this into a signed comparison against 0. */
+ if (code == GTU)
{
- case V4SImode:
- {
- rtx t1, t2, mask;
-
- /* Perform a parallel modulo subtraction. */
- t1 = gen_reg_rtx (mode);
- emit_insn (gen_subv4si3 (t1, cop0, cop1));
-
- /* Extract the original sign bit of op0. */
- mask = GEN_INT (-0x80000000);
- mask = gen_rtx_CONST_VECTOR (mode,
- gen_rtvec (4, mask, mask, mask, mask));
- mask = force_reg (mode, mask);
- t2 = gen_reg_rtx (mode);
- emit_insn (gen_andv4si3 (t2, cop0, mask));
-
- /* XOR it back into the result of the subtraction. This results
- in the sign bit set iff we saw unsigned underflow. */
- x = gen_reg_rtx (mode);
- emit_insn (gen_xorv4si3 (x, t1, t2));
-
- code = GT;
- }
- break;
+ cop0 = force_reg (mode, cop0);
+
+ switch (mode)
+ {
+ case V4SImode:
+ case V2DImode:
+ {
+ rtx t1, t2, mask;
+
+ /* Perform a parallel modulo subtraction. */
+ t1 = gen_reg_rtx (mode);
+ emit_insn ((mode == V4SImode
+ ? gen_subv4si3
+ : gen_subv2di3) (t1, cop0, cop1));
+
+ /* Extract the original sign bit of op0. */
+ mask = ix86_build_signbit_mask (GET_MODE_INNER (mode),
+ true, false);
+ t2 = gen_reg_rtx (mode);
+ emit_insn ((mode == V4SImode
+ ? gen_andv4si3
+ : gen_andv2di3) (t2, cop0, mask));
+
+ /* XOR it back into the result of the subtraction. This results
+ in the sign bit set iff we saw unsigned underflow. */
+ x = gen_reg_rtx (mode);
+ emit_insn ((mode == V4SImode
+ ? gen_xorv4si3
+ : gen_xorv2di3) (x, t1, t2));
+
+ code = GT;
+ }
+ break;
- case V16QImode:
- case V8HImode:
- /* Perform a parallel unsigned saturating subtraction. */
- x = gen_reg_rtx (mode);
- emit_insn (gen_rtx_SET (VOIDmode, x,
- gen_rtx_US_MINUS (mode, cop0, cop1)));
+ case V16QImode:
+ case V8HImode:
+ /* Perform a parallel unsigned saturating subtraction. */
+ x = gen_reg_rtx (mode);
+ emit_insn (gen_rtx_SET (VOIDmode, x,
+ gen_rtx_US_MINUS (mode, cop0, cop1)));
- code = EQ;
- negate = !negate;
- break;
+ code = EQ;
+ negate = !negate;
+ break;
- default:
- gcc_unreachable ();
- }
+ default:
+ gcc_unreachable ();
+ }
- cop0 = x;
- cop1 = CONST0_RTX (mode);
+ cop0 = x;
+ cop1 = CONST0_RTX (mode);
+ }
}
x = ix86_expand_sse_cmp (operands[0], code, cop0, cop1,
case V4SImode:
if (high_p)
unpack = gen_vec_interleave_highv4si;
- else
+ else
unpack = gen_vec_interleave_lowv4si;
break;
default:
- gcc_unreachable ();
+ gcc_unreachable ();
}
dest = gen_lowpart (imode, operands[0]);
emit_insn (unpack (dest, operands[1], se));
}
+/* This function performs the same task as ix86_expand_sse_unpack,
+ but with SSE4.1 instructions. */
+
+void
+ix86_expand_sse4_unpack (rtx operands[2], bool unsigned_p, bool high_p)
+{
+ enum machine_mode imode = GET_MODE (operands[1]);
+ rtx (*unpack)(rtx, rtx);
+ rtx src, dest;
+
+ switch (imode)
+ {
+ case V16QImode:
+ if (unsigned_p)
+ unpack = gen_sse4_1_zero_extendv8qiv8hi2;
+ else
+ unpack = gen_sse4_1_extendv8qiv8hi2;
+ break;
+ case V8HImode:
+ if (unsigned_p)
+ unpack = gen_sse4_1_zero_extendv4hiv4si2;
+ else
+ unpack = gen_sse4_1_extendv4hiv4si2;
+ break;
+ case V4SImode:
+ if (unsigned_p)
+ unpack = gen_sse4_1_zero_extendv2siv2di2;
+ else
+ unpack = gen_sse4_1_extendv2siv2di2;
+ break;
+ default:
+ gcc_unreachable ();
+ }
+
+ dest = operands[0];
+ if (high_p)
+ {
+ /* Shift higher 8 bytes to lower 8 bytes. */
+ src = gen_reg_rtx (imode);
+ emit_insn (gen_sse2_lshrti3 (gen_lowpart (TImode, src),
+ gen_lowpart (TImode, operands[1]),
+ GEN_INT (64)));
+ }
+ else
+ src = operands[1];
+
+ emit_insn (unpack (dest, src));
+}
+
+/* This function performs the same task as ix86_expand_sse_unpack,
+ but with sse5 instructions. */
+
+void
+ix86_expand_sse5_unpack (rtx operands[2], bool unsigned_p, bool high_p)
+{
+ enum machine_mode imode = GET_MODE (operands[1]);
+ int pperm_bytes[16];
+ int i;
+ int h = (high_p) ? 8 : 0;
+ int h2;
+ int sign_extend;
+ rtvec v = rtvec_alloc (16);
+ rtvec vs;
+ rtx x, p;
+ rtx op0 = operands[0], op1 = operands[1];
+
+ switch (imode)
+ {
+ case V16QImode:
+ vs = rtvec_alloc (8);
+ h2 = (high_p) ? 8 : 0;
+ for (i = 0; i < 8; i++)
+ {
+ pperm_bytes[2*i+0] = PPERM_SRC | PPERM_SRC2 | i | h;
+ pperm_bytes[2*i+1] = ((unsigned_p)
+ ? PPERM_ZERO
+ : PPERM_SIGN | PPERM_SRC2 | i | h);
+ }
+
+ for (i = 0; i < 16; i++)
+ RTVEC_ELT (v, i) = GEN_INT (pperm_bytes[i]);
+
+ for (i = 0; i < 8; i++)
+ RTVEC_ELT (vs, i) = GEN_INT (i + h2);
+
+ p = gen_rtx_PARALLEL (VOIDmode, vs);
+ x = force_reg (V16QImode, gen_rtx_CONST_VECTOR (V16QImode, v));
+ if (unsigned_p)
+ emit_insn (gen_sse5_pperm_zero_v16qi_v8hi (op0, op1, p, x));
+ else
+ emit_insn (gen_sse5_pperm_sign_v16qi_v8hi (op0, op1, p, x));
+ break;
+
+ case V8HImode:
+ vs = rtvec_alloc (4);
+ h2 = (high_p) ? 4 : 0;
+ for (i = 0; i < 4; i++)
+ {
+ sign_extend = ((unsigned_p)
+ ? PPERM_ZERO
+ : PPERM_SIGN | PPERM_SRC2 | ((2*i) + 1 + h));
+ pperm_bytes[4*i+0] = PPERM_SRC | PPERM_SRC2 | ((2*i) + 0 + h);
+ pperm_bytes[4*i+1] = PPERM_SRC | PPERM_SRC2 | ((2*i) + 1 + h);
+ pperm_bytes[4*i+2] = sign_extend;
+ pperm_bytes[4*i+3] = sign_extend;
+ }
+
+ for (i = 0; i < 16; i++)
+ RTVEC_ELT (v, i) = GEN_INT (pperm_bytes[i]);
+
+ for (i = 0; i < 4; i++)
+ RTVEC_ELT (vs, i) = GEN_INT (i + h2);
+
+ p = gen_rtx_PARALLEL (VOIDmode, vs);
+ x = force_reg (V16QImode, gen_rtx_CONST_VECTOR (V16QImode, v));
+ if (unsigned_p)
+ emit_insn (gen_sse5_pperm_zero_v8hi_v4si (op0, op1, p, x));
+ else
+ emit_insn (gen_sse5_pperm_sign_v8hi_v4si (op0, op1, p, x));
+ break;
+
+ case V4SImode:
+ vs = rtvec_alloc (2);
+ h2 = (high_p) ? 2 : 0;
+ for (i = 0; i < 2; i++)
+ {
+ sign_extend = ((unsigned_p)
+ ? PPERM_ZERO
+ : PPERM_SIGN | PPERM_SRC2 | ((4*i) + 3 + h));
+ pperm_bytes[8*i+0] = PPERM_SRC | PPERM_SRC2 | ((4*i) + 0 + h);
+ pperm_bytes[8*i+1] = PPERM_SRC | PPERM_SRC2 | ((4*i) + 1 + h);
+ pperm_bytes[8*i+2] = PPERM_SRC | PPERM_SRC2 | ((4*i) + 2 + h);
+ pperm_bytes[8*i+3] = PPERM_SRC | PPERM_SRC2 | ((4*i) + 3 + h);
+ pperm_bytes[8*i+4] = sign_extend;
+ pperm_bytes[8*i+5] = sign_extend;
+ pperm_bytes[8*i+6] = sign_extend;
+ pperm_bytes[8*i+7] = sign_extend;
+ }
+
+ for (i = 0; i < 16; i++)
+ RTVEC_ELT (v, i) = GEN_INT (pperm_bytes[i]);
+
+ for (i = 0; i < 2; i++)
+ RTVEC_ELT (vs, i) = GEN_INT (i + h2);
+
+ p = gen_rtx_PARALLEL (VOIDmode, vs);
+ x = force_reg (V16QImode, gen_rtx_CONST_VECTOR (V16QImode, v));
+ if (unsigned_p)
+ emit_insn (gen_sse5_pperm_zero_v4si_v2di (op0, op1, p, x));
+ else
+ emit_insn (gen_sse5_pperm_sign_v4si_v2di (op0, op1, p, x));
+ break;
+
+ default:
+ gcc_unreachable ();
+ }
+
+ return;
+}
+
+/* Pack the high bits from OPERANDS[1] and low bits from OPERANDS[2] into the
+ next narrower integer vector type */
+void
+ix86_expand_sse5_pack (rtx operands[3])
+{
+ enum machine_mode imode = GET_MODE (operands[0]);
+ int pperm_bytes[16];
+ int i;
+ rtvec v = rtvec_alloc (16);
+ rtx x;
+ rtx op0 = operands[0];
+ rtx op1 = operands[1];
+ rtx op2 = operands[2];
+
+ switch (imode)
+ {
+ case V16QImode:
+ for (i = 0; i < 8; i++)
+ {
+ pperm_bytes[i+0] = PPERM_SRC | PPERM_SRC1 | (i*2);
+ pperm_bytes[i+8] = PPERM_SRC | PPERM_SRC2 | (i*2);
+ }
+
+ for (i = 0; i < 16; i++)
+ RTVEC_ELT (v, i) = GEN_INT (pperm_bytes[i]);
+
+ x = force_reg (V16QImode, gen_rtx_CONST_VECTOR (V16QImode, v));
+ emit_insn (gen_sse5_pperm_pack_v8hi_v16qi (op0, op1, op2, x));
+ break;
+
+ case V8HImode:
+ for (i = 0; i < 4; i++)
+ {
+ pperm_bytes[(2*i)+0] = PPERM_SRC | PPERM_SRC1 | ((i*4) + 0);
+ pperm_bytes[(2*i)+1] = PPERM_SRC | PPERM_SRC1 | ((i*4) + 1);
+ pperm_bytes[(2*i)+8] = PPERM_SRC | PPERM_SRC2 | ((i*4) + 0);
+ pperm_bytes[(2*i)+9] = PPERM_SRC | PPERM_SRC2 | ((i*4) + 1);
+ }
+
+ for (i = 0; i < 16; i++)
+ RTVEC_ELT (v, i) = GEN_INT (pperm_bytes[i]);
+
+ x = force_reg (V16QImode, gen_rtx_CONST_VECTOR (V16QImode, v));
+ emit_insn (gen_sse5_pperm_pack_v4si_v8hi (op0, op1, op2, x));
+ break;
+
+ case V4SImode:
+ for (i = 0; i < 2; i++)
+ {
+ pperm_bytes[(4*i)+0] = PPERM_SRC | PPERM_SRC1 | ((i*8) + 0);
+ pperm_bytes[(4*i)+1] = PPERM_SRC | PPERM_SRC1 | ((i*8) + 1);
+ pperm_bytes[(4*i)+2] = PPERM_SRC | PPERM_SRC1 | ((i*8) + 2);
+ pperm_bytes[(4*i)+3] = PPERM_SRC | PPERM_SRC1 | ((i*8) + 3);
+ pperm_bytes[(4*i)+8] = PPERM_SRC | PPERM_SRC2 | ((i*8) + 0);
+ pperm_bytes[(4*i)+9] = PPERM_SRC | PPERM_SRC2 | ((i*8) + 1);
+ pperm_bytes[(4*i)+10] = PPERM_SRC | PPERM_SRC2 | ((i*8) + 2);
+ pperm_bytes[(4*i)+11] = PPERM_SRC | PPERM_SRC2 | ((i*8) + 3);
+ }
+
+ for (i = 0; i < 16; i++)
+ RTVEC_ELT (v, i) = GEN_INT (pperm_bytes[i]);
+
+ x = force_reg (V16QImode, gen_rtx_CONST_VECTOR (V16QImode, v));
+ emit_insn (gen_sse5_pperm_pack_v2di_v4si (op0, op1, op2, x));
+ break;
+
+ default:
+ gcc_unreachable ();
+ }
+
+ return;
+}
+
/* Expand conditional increment or decrement using adb/sbb instructions.
The default case using setcc followed by the conditional move can be
done by generic code. */
return sc;
}
-/* Return mode for the memcpy/memset loop counter. Preffer SImode over DImode
- for constant loop counts. */
+/* Return mode for the memcpy/memset loop counter. Prefer SImode over
+ DImode for constant loop counts. */
static enum machine_mode
counter_mode (rtx count_exp)
The size is rounded down to whole number of chunk size moved at once.
SRCMEM and DESTMEM provide MEMrtx to feed proper aliasing info. */
-
+
static void
expand_set_or_movmem_via_loop (rtx destmem, rtx srcmem,
srcmem = change_address (srcmem, mode, y_addr);
/* When unrolling for chips that reorder memory reads and writes,
- we can save registers by using single temporary.
+ we can save registers by using single temporary.
Also using 4 temporaries is overkill in 32bit mode. */
if (!TARGET_64BIT && 0)
{
emit_label (out_label);
}
-/* Output "rep; mov" instruction.
+/* Output "rep; mov" instruction.
Arguments have same meaning as for previous function */
static void
expand_movmem_via_rep_mov (rtx destmem, rtx srcmem,
destexp, srcexp));
}
-/* Output "rep; stos" instruction.
+/* Output "rep; stos" instruction.
Arguments have same meaning as for previous function */
static void
expand_setmem_via_rep_stos (rtx destmem, rtx destptr, rtx value,
int *dynamic_check)
{
const struct stringop_algs * algs;
+ /* Algorithms using the rep prefix want at least edi and ecx;
+ additionally, memset wants eax and memcpy wants esi. Don't
+ consider such algorithms if the user has appropriated those
+ registers for their own purposes. */
+ bool rep_prefix_usable = !(fixed_regs[CX_REG] || fixed_regs[DI_REG]
+ || (memset
+ ? fixed_regs[AX_REG] : fixed_regs[SI_REG]));
+
+#define ALG_USABLE_P(alg) (rep_prefix_usable \
+ || (alg != rep_prefix_1_byte \
+ && alg != rep_prefix_4_byte \
+ && alg != rep_prefix_8_byte))
*dynamic_check = -1;
if (memset)
algs = &ix86_cost->memset[TARGET_64BIT != 0];
else
algs = &ix86_cost->memcpy[TARGET_64BIT != 0];
- if (stringop_alg != no_stringop)
+ if (stringop_alg != no_stringop && ALG_USABLE_P (stringop_alg))
return stringop_alg;
/* rep; movq or rep; movl is the smallest variant. */
else if (optimize_size)
{
if (!count || (count & 3))
- return rep_prefix_1_byte;
+ return rep_prefix_usable ? rep_prefix_1_byte : loop_1_byte;
else
- return rep_prefix_4_byte;
+ return rep_prefix_usable ? rep_prefix_4_byte : loop;
}
/* Very tiny blocks are best handled via the loop, REP is expensive to setup.
*/
enum stringop_alg alg = libcall;
for (i = 0; i < NAX_STRINGOP_ALGS; i++)
{
- gcc_assert (algs->size[i].max);
+ /* We get here if the algorithms that were not libcall-based
+ were rep-prefix based and we are unable to use rep prefixes
+ based on global register usage. Break out of the loop and
+ use the heuristic below. */
+ if (algs->size[i].max == 0)
+ break;
if (algs->size[i].max >= expected_size || algs->size[i].max == -1)
{
- if (algs->size[i].alg != libcall)
- alg = algs->size[i].alg;
+ enum stringop_alg candidate = algs->size[i].alg;
+
+ if (candidate != libcall && ALG_USABLE_P (candidate))
+ alg = candidate;
/* Honor TARGET_INLINE_ALL_STRINGOPS by picking
- last non-libcall inline algorithm. */
+ last non-libcall inline algorithm. */
if (TARGET_INLINE_ALL_STRINGOPS)
{
/* When the current size is best to be copied by a libcall,
- but we are still forced to inline, run the heuristic bellow
+ but we are still forced to inline, run the heuristic below
that will pick code for medium sized blocks. */
if (alg != libcall)
return alg;
break;
}
- else
- return algs->size[i].alg;
+ else if (ALG_USABLE_P (candidate))
+ return candidate;
}
}
- gcc_assert (TARGET_INLINE_ALL_STRINGOPS);
+ gcc_assert (TARGET_INLINE_ALL_STRINGOPS || !rep_prefix_usable);
}
/* When asked to inline the call anyway, try to pick meaningful choice.
We look for maximal size of block that is faster to copy by hand and
take blocks of at most of that size guessing that average size will
- be roughly half of the block.
+ be roughly half of the block.
If this turns out to be bad, we might simply specify the preferred
choice in ix86_costs. */
if ((TARGET_INLINE_ALL_STRINGOPS || TARGET_INLINE_STRINGOPS_DYNAMICALLY)
- && algs->unknown_size == libcall)
+ && (algs->unknown_size == libcall || !ALG_USABLE_P (algs->unknown_size)))
{
int max = -1;
enum stringop_alg alg;
int i;
+ bool any_alg_usable_p = true;
for (i = 0; i < NAX_STRINGOP_ALGS; i++)
- if (algs->size[i].alg != libcall && algs->size[i].alg)
- max = algs->size[i].max;
+ {
+ enum stringop_alg candidate = algs->size[i].alg;
+ any_alg_usable_p = any_alg_usable_p && ALG_USABLE_P (candidate);
+
+ if (candidate != libcall && candidate
+ && ALG_USABLE_P (candidate))
+ max = algs->size[i].max;
+ }
+ /* If there aren't any usable algorithms, then recursing on
+ smaller sizes isn't going to find anything. Just return the
+ simple byte-at-a-time copy loop. */
+ if (!any_alg_usable_p)
+ {
+ /* Pick something reasonable. */
+ if (TARGET_INLINE_STRINGOPS_DYNAMICALLY)
+ *dynamic_check = 128;
+ return loop_1_byte;
+ }
if (max == -1)
max = 4096;
alg = decide_alg (count, max / 2, memset, dynamic_check);
*dynamic_check = max;
return alg;
}
- return algs->unknown_size;
+ return ALG_USABLE_P (algs->unknown_size) ? algs->unknown_size : libcall;
+#undef ALG_USABLE_P
}
/* Decide on alignment. We know that the operand is already aligned to ALIGN
}
/* Expand string move (memcpy) operation. Use i386 string operations when
- profitable. expand_clrmem contains similar code. The code depends upon
+ profitable. expand_setmem contains similar code. The code depends upon
architecture, block size and alignment, but always has the same
overall structure:
4) Epilogue: code copying tail of the block that is too small to be
handled by main body (or up to size guarded by prologue guard). */
-
+
int
ix86_expand_movmem (rtx dst, rtx src, rtx count_exp, rtx align_exp,
rtx expected_align_exp, rtx expected_size_exp)
if (CONST_INT_P (expected_size_exp) && count == 0)
expected_size = INTVAL (expected_size_exp);
+ /* Make sure we don't need to care about overflow later on. */
+ if (count > ((unsigned HOST_WIDE_INT) 1 << 30))
+ return 0;
+
/* Step 0: Decide on preferred algorithm, desired alignment and
size of chunks to be copied by main loop. */
/* Alignment code needs count to be in register. */
if (CONST_INT_P (count_exp) && desired_align > align)
- {
- enum machine_mode mode = SImode;
- if (TARGET_64BIT && (count & ~0xffffffff))
- mode = DImode;
- count_exp = force_reg (mode, count_exp);
- }
+ count_exp = force_reg (counter_mode (count_exp), count_exp);
gcc_assert (desired_align >= 1 && align >= 1);
/* Ensure that alignment prologue won't copy past end of block. */
Make sure it is power of 2. */
epilogue_size_needed = smallest_pow2_greater_than (epilogue_size_needed);
- label = gen_label_rtx ();
- emit_cmp_and_jump_insns (count_exp,
- GEN_INT (epilogue_size_needed),
- LTU, 0, counter_mode (count_exp), 1, label);
- if (GET_CODE (count_exp) == CONST_INT)
- ;
- else if (expected_size == -1 || expected_size < epilogue_size_needed)
- predict_jump (REG_BR_PROB_BASE * 60 / 100);
+ if (CONST_INT_P (count_exp))
+ {
+ if (UINTVAL (count_exp) < (unsigned HOST_WIDE_INT)epilogue_size_needed)
+ goto epilogue;
+ }
else
- predict_jump (REG_BR_PROB_BASE * 20 / 100);
+ {
+ label = gen_label_rtx ();
+ emit_cmp_and_jump_insns (count_exp,
+ GEN_INT (epilogue_size_needed),
+ LTU, 0, counter_mode (count_exp), 1, label);
+ if (expected_size == -1 || expected_size < epilogue_size_needed)
+ predict_jump (REG_BR_PROB_BASE * 60 / 100);
+ else
+ predict_jump (REG_BR_PROB_BASE * 20 / 100);
+ }
}
+
/* Emit code to decide on runtime whether library call or inline should be
used. */
if (dynamic_check != -1)
{
- rtx hot_label = gen_label_rtx ();
- jump_around_label = gen_label_rtx ();
- emit_cmp_and_jump_insns (count_exp, GEN_INT (dynamic_check - 1),
- LEU, 0, GET_MODE (count_exp), 1, hot_label);
- predict_jump (REG_BR_PROB_BASE * 90 / 100);
- emit_block_move_via_libcall (dst, src, count_exp, false);
- emit_jump (jump_around_label);
- emit_label (hot_label);
+ if (CONST_INT_P (count_exp))
+ {
+ if (UINTVAL (count_exp) >= (unsigned HOST_WIDE_INT)dynamic_check)
+ {
+ emit_block_move_via_libcall (dst, src, count_exp, false);
+ count_exp = const0_rtx;
+ goto epilogue;
+ }
+ }
+ else
+ {
+ rtx hot_label = gen_label_rtx ();
+ jump_around_label = gen_label_rtx ();
+ emit_cmp_and_jump_insns (count_exp, GEN_INT (dynamic_check - 1),
+ LEU, 0, GET_MODE (count_exp), 1, hot_label);
+ predict_jump (REG_BR_PROB_BASE * 90 / 100);
+ emit_block_move_via_libcall (dst, src, count_exp, false);
+ emit_jump (jump_around_label);
+ emit_label (hot_label);
+ }
}
/* Step 2: Alignment prologue. */
}
/* Step 4: Epilogue to copy the remaining bytes. */
-
+ epilogue:
if (label)
{
/* When the main loop is done, COUNT_EXP might hold original count,
while we want to copy only COUNT_EXP & SIZE_NEEDED bytes.
Epilogue code will actually copy COUNT_EXP & EPILOGUE_SIZE_NEEDED
bytes. Compensate if needed. */
-
+
if (size_needed < epilogue_size_needed)
{
tmp =
if (CONST_INT_P (expected_size_exp) && count == 0)
expected_size = INTVAL (expected_size_exp);
+ /* Make sure we don't need to care about overflow later on. */
+ if (count > ((unsigned HOST_WIDE_INT) 1 << 30))
+ return 0;
+
/* Step 0: Decide on preferred algorithm, desired alignment and
size of chunks to be copied by main loop. */
mode = DImode;
count_exp = force_reg (mode, count_exp);
}
- /* Do the cheap promotion to allow better CSE across the
+ /* Do the cheap promotion to allow better CSE across the
main loop and epilogue (ie one load of the big constant in the
front of all code. */
if (CONST_INT_P (val_exp))
/* Avoid branch in fixing the byte. */
tmpreg = gen_lowpart (QImode, tmpreg);
emit_insn (gen_addqi3_cc (tmpreg, tmpreg, tmpreg));
- cmp = gen_rtx_LTU (Pmode, gen_rtx_REG (CCmode, 17), const0_rtx);
+ cmp = gen_rtx_LTU (Pmode, gen_rtx_REG (CCmode, FLAGS_REG), const0_rtx);
if (TARGET_64BIT)
emit_insn (gen_subdi3_carry_rex64 (out, out, GEN_INT (3), cmp));
else
else
{
rtx unspec;
+
+ /* Can't use this if the user has appropriated eax, ecx, or edi. */
+ if (fixed_regs[AX_REG] || fixed_regs[CX_REG] || fixed_regs[DI_REG])
+ return false;
+
scratch2 = gen_reg_rtx (Pmode);
scratch3 = gen_reg_rtx (Pmode);
scratch4 = force_reg (Pmode, constm1_rtx);
if (TARGET_64BIT && INTVAL (callarg2) >= 0)
{
- rtx al = gen_rtx_REG (QImode, 0);
+ rtx al = gen_rtx_REG (QImode, AX_REG);
emit_move_insn (al, callarg2);
use_reg (&use, al);
}
{
struct machine_function *f;
- f = ggc_alloc_cleared (sizeof (struct machine_function));
+ f = GGC_CNEW (struct machine_function);
f->use_fast_prologue_epilogue_nregs = -1;
f->tls_descriptor_call_expanded_p = 0;
gcc_assert (n < MAX_386_STACK_LOCALS);
+ /* Virtual slot is valid only before vregs are instantiated. */
+ gcc_assert ((n == SLOT_VIRTUAL) == !virtuals_instantiated);
+
for (s = ix86_stack_locals; s; s = s->next)
if (s->mode == mode && s->n == n)
return copy_rtx (s->rtl);
static int
ia32_multipass_dfa_lookahead (void)
{
- if (ix86_tune == PROCESSOR_PENTIUM)
- return 2;
+ switch (ix86_tune)
+ {
+ case PROCESSOR_PENTIUM:
+ return 2;
- if (ix86_tune == PROCESSOR_PENTIUMPRO
- || ix86_tune == PROCESSOR_K6)
- return 1;
+ case PROCESSOR_PENTIUMPRO:
+ case PROCESSOR_K6:
+ return 1;
- else
- return 0;
+ default:
+ return 0;
+ }
}
\f
int
ix86_constant_alignment (tree exp, int align)
{
- if (TREE_CODE (exp) == REAL_CST)
+ if (TREE_CODE (exp) == REAL_CST || TREE_CODE (exp) == VECTOR_CST
+ || TREE_CODE (exp) == INTEGER_CST)
{
if (TYPE_MODE (TREE_TYPE (exp)) == DFmode && align < 64)
return 64;
IX86_BUILTIN_RCPPS,
IX86_BUILTIN_RCPSS,
IX86_BUILTIN_RSQRTPS,
+ IX86_BUILTIN_RSQRTPS_NR,
IX86_BUILTIN_RSQRTSS,
+ IX86_BUILTIN_RSQRTF,
IX86_BUILTIN_SQRTPS,
+ IX86_BUILTIN_SQRTPS_NR,
IX86_BUILTIN_SQRTSS,
IX86_BUILTIN_UNPCKHPS,
IX86_BUILTIN_MFENCE,
IX86_BUILTIN_LFENCE,
- /* Prescott New Instructions. */
+ /* SSE3. */
IX86_BUILTIN_ADDSUBPS,
IX86_BUILTIN_HADDPS,
IX86_BUILTIN_HSUBPS,
IX86_BUILTIN_INSERTQI,
IX86_BUILTIN_INSERTQ,
+ /* SSE4.1. */
+ IX86_BUILTIN_BLENDPD,
+ IX86_BUILTIN_BLENDPS,
+ IX86_BUILTIN_BLENDVPD,
+ IX86_BUILTIN_BLENDVPS,
+ IX86_BUILTIN_PBLENDVB128,
+ IX86_BUILTIN_PBLENDW128,
+
+ IX86_BUILTIN_DPPD,
+ IX86_BUILTIN_DPPS,
+
+ IX86_BUILTIN_INSERTPS128,
+
+ IX86_BUILTIN_MOVNTDQA,
+ IX86_BUILTIN_MPSADBW128,
+ IX86_BUILTIN_PACKUSDW128,
+ IX86_BUILTIN_PCMPEQQ,
+ IX86_BUILTIN_PHMINPOSUW128,
+
+ IX86_BUILTIN_PMAXSB128,
+ IX86_BUILTIN_PMAXSD128,
+ IX86_BUILTIN_PMAXUD128,
+ IX86_BUILTIN_PMAXUW128,
+
+ IX86_BUILTIN_PMINSB128,
+ IX86_BUILTIN_PMINSD128,
+ IX86_BUILTIN_PMINUD128,
+ IX86_BUILTIN_PMINUW128,
+
+ IX86_BUILTIN_PMOVSXBW128,
+ IX86_BUILTIN_PMOVSXBD128,
+ IX86_BUILTIN_PMOVSXBQ128,
+ IX86_BUILTIN_PMOVSXWD128,
+ IX86_BUILTIN_PMOVSXWQ128,
+ IX86_BUILTIN_PMOVSXDQ128,
+
+ IX86_BUILTIN_PMOVZXBW128,
+ IX86_BUILTIN_PMOVZXBD128,
+ IX86_BUILTIN_PMOVZXBQ128,
+ IX86_BUILTIN_PMOVZXWD128,
+ IX86_BUILTIN_PMOVZXWQ128,
+ IX86_BUILTIN_PMOVZXDQ128,
+
+ IX86_BUILTIN_PMULDQ128,
+ IX86_BUILTIN_PMULLD128,
+
+ IX86_BUILTIN_ROUNDPD,
+ IX86_BUILTIN_ROUNDPS,
+ IX86_BUILTIN_ROUNDSD,
+ IX86_BUILTIN_ROUNDSS,
+
+ IX86_BUILTIN_PTESTZ,
+ IX86_BUILTIN_PTESTC,
+ IX86_BUILTIN_PTESTNZC,
+
IX86_BUILTIN_VEC_INIT_V2SI,
IX86_BUILTIN_VEC_INIT_V4HI,
IX86_BUILTIN_VEC_INIT_V8QI,
IX86_BUILTIN_VEC_EXT_V8HI,
IX86_BUILTIN_VEC_EXT_V2SI,
IX86_BUILTIN_VEC_EXT_V4HI,
+ IX86_BUILTIN_VEC_EXT_V16QI,
+ IX86_BUILTIN_VEC_SET_V2DI,
+ IX86_BUILTIN_VEC_SET_V4SF,
+ IX86_BUILTIN_VEC_SET_V4SI,
IX86_BUILTIN_VEC_SET_V8HI,
IX86_BUILTIN_VEC_SET_V4HI,
+ IX86_BUILTIN_VEC_SET_V16QI,
+
+ IX86_BUILTIN_VEC_PACK_SFIX,
+
+ /* SSE4.2. */
+ IX86_BUILTIN_CRC32QI,
+ IX86_BUILTIN_CRC32HI,
+ IX86_BUILTIN_CRC32SI,
+ IX86_BUILTIN_CRC32DI,
+
+ IX86_BUILTIN_PCMPESTRI128,
+ IX86_BUILTIN_PCMPESTRM128,
+ IX86_BUILTIN_PCMPESTRA128,
+ IX86_BUILTIN_PCMPESTRC128,
+ IX86_BUILTIN_PCMPESTRO128,
+ IX86_BUILTIN_PCMPESTRS128,
+ IX86_BUILTIN_PCMPESTRZ128,
+ IX86_BUILTIN_PCMPISTRI128,
+ IX86_BUILTIN_PCMPISTRM128,
+ IX86_BUILTIN_PCMPISTRA128,
+ IX86_BUILTIN_PCMPISTRC128,
+ IX86_BUILTIN_PCMPISTRO128,
+ IX86_BUILTIN_PCMPISTRS128,
+ IX86_BUILTIN_PCMPISTRZ128,
+
+ IX86_BUILTIN_PCMPGTQ,
+
+ /* AES instructions */
+ IX86_BUILTIN_AESENC128,
+ IX86_BUILTIN_AESENCLAST128,
+ IX86_BUILTIN_AESDEC128,
+ IX86_BUILTIN_AESDECLAST128,
+ IX86_BUILTIN_AESIMC128,
+ IX86_BUILTIN_AESKEYGENASSIST128,
+
+ /* PCLMUL instruction */
+ IX86_BUILTIN_PCLMULQDQ128,
+
+ /* TFmode support builtins. */
+ IX86_BUILTIN_INFQ,
+ IX86_BUILTIN_FABSQ,
+ IX86_BUILTIN_COPYSIGNQ,
+
+ /* SSE5 instructions */
+ IX86_BUILTIN_FMADDSS,
+ IX86_BUILTIN_FMADDSD,
+ IX86_BUILTIN_FMADDPS,
+ IX86_BUILTIN_FMADDPD,
+ IX86_BUILTIN_FMSUBSS,
+ IX86_BUILTIN_FMSUBSD,
+ IX86_BUILTIN_FMSUBPS,
+ IX86_BUILTIN_FMSUBPD,
+ IX86_BUILTIN_FNMADDSS,
+ IX86_BUILTIN_FNMADDSD,
+ IX86_BUILTIN_FNMADDPS,
+ IX86_BUILTIN_FNMADDPD,
+ IX86_BUILTIN_FNMSUBSS,
+ IX86_BUILTIN_FNMSUBSD,
+ IX86_BUILTIN_FNMSUBPS,
+ IX86_BUILTIN_FNMSUBPD,
+ IX86_BUILTIN_PCMOV_V2DI,
+ IX86_BUILTIN_PCMOV_V4SI,
+ IX86_BUILTIN_PCMOV_V8HI,
+ IX86_BUILTIN_PCMOV_V16QI,
+ IX86_BUILTIN_PCMOV_V4SF,
+ IX86_BUILTIN_PCMOV_V2DF,
+ IX86_BUILTIN_PPERM,
+ IX86_BUILTIN_PERMPS,
+ IX86_BUILTIN_PERMPD,
+ IX86_BUILTIN_PMACSSWW,
+ IX86_BUILTIN_PMACSWW,
+ IX86_BUILTIN_PMACSSWD,
+ IX86_BUILTIN_PMACSWD,
+ IX86_BUILTIN_PMACSSDD,
+ IX86_BUILTIN_PMACSDD,
+ IX86_BUILTIN_PMACSSDQL,
+ IX86_BUILTIN_PMACSSDQH,
+ IX86_BUILTIN_PMACSDQL,
+ IX86_BUILTIN_PMACSDQH,
+ IX86_BUILTIN_PMADCSSWD,
+ IX86_BUILTIN_PMADCSWD,
+ IX86_BUILTIN_PHADDBW,
+ IX86_BUILTIN_PHADDBD,
+ IX86_BUILTIN_PHADDBQ,
+ IX86_BUILTIN_PHADDWD,
+ IX86_BUILTIN_PHADDWQ,
+ IX86_BUILTIN_PHADDDQ,
+ IX86_BUILTIN_PHADDUBW,
+ IX86_BUILTIN_PHADDUBD,
+ IX86_BUILTIN_PHADDUBQ,
+ IX86_BUILTIN_PHADDUWD,
+ IX86_BUILTIN_PHADDUWQ,
+ IX86_BUILTIN_PHADDUDQ,
+ IX86_BUILTIN_PHSUBBW,
+ IX86_BUILTIN_PHSUBWD,
+ IX86_BUILTIN_PHSUBDQ,
+ IX86_BUILTIN_PROTB,
+ IX86_BUILTIN_PROTW,
+ IX86_BUILTIN_PROTD,
+ IX86_BUILTIN_PROTQ,
+ IX86_BUILTIN_PROTB_IMM,
+ IX86_BUILTIN_PROTW_IMM,
+ IX86_BUILTIN_PROTD_IMM,
+ IX86_BUILTIN_PROTQ_IMM,
+ IX86_BUILTIN_PSHLB,
+ IX86_BUILTIN_PSHLW,
+ IX86_BUILTIN_PSHLD,
+ IX86_BUILTIN_PSHLQ,
+ IX86_BUILTIN_PSHAB,
+ IX86_BUILTIN_PSHAW,
+ IX86_BUILTIN_PSHAD,
+ IX86_BUILTIN_PSHAQ,
+ IX86_BUILTIN_FRCZSS,
+ IX86_BUILTIN_FRCZSD,
+ IX86_BUILTIN_FRCZPS,
+ IX86_BUILTIN_FRCZPD,
+ IX86_BUILTIN_CVTPH2PS,
+ IX86_BUILTIN_CVTPS2PH,
+
+ IX86_BUILTIN_COMEQSS,
+ IX86_BUILTIN_COMNESS,
+ IX86_BUILTIN_COMLTSS,
+ IX86_BUILTIN_COMLESS,
+ IX86_BUILTIN_COMGTSS,
+ IX86_BUILTIN_COMGESS,
+ IX86_BUILTIN_COMUEQSS,
+ IX86_BUILTIN_COMUNESS,
+ IX86_BUILTIN_COMULTSS,
+ IX86_BUILTIN_COMULESS,
+ IX86_BUILTIN_COMUGTSS,
+ IX86_BUILTIN_COMUGESS,
+ IX86_BUILTIN_COMORDSS,
+ IX86_BUILTIN_COMUNORDSS,
+ IX86_BUILTIN_COMFALSESS,
+ IX86_BUILTIN_COMTRUESS,
+
+ IX86_BUILTIN_COMEQSD,
+ IX86_BUILTIN_COMNESD,
+ IX86_BUILTIN_COMLTSD,
+ IX86_BUILTIN_COMLESD,
+ IX86_BUILTIN_COMGTSD,
+ IX86_BUILTIN_COMGESD,
+ IX86_BUILTIN_COMUEQSD,
+ IX86_BUILTIN_COMUNESD,
+ IX86_BUILTIN_COMULTSD,
+ IX86_BUILTIN_COMULESD,
+ IX86_BUILTIN_COMUGTSD,
+ IX86_BUILTIN_COMUGESD,
+ IX86_BUILTIN_COMORDSD,
+ IX86_BUILTIN_COMUNORDSD,
+ IX86_BUILTIN_COMFALSESD,
+ IX86_BUILTIN_COMTRUESD,
+
+ IX86_BUILTIN_COMEQPS,
+ IX86_BUILTIN_COMNEPS,
+ IX86_BUILTIN_COMLTPS,
+ IX86_BUILTIN_COMLEPS,
+ IX86_BUILTIN_COMGTPS,
+ IX86_BUILTIN_COMGEPS,
+ IX86_BUILTIN_COMUEQPS,
+ IX86_BUILTIN_COMUNEPS,
+ IX86_BUILTIN_COMULTPS,
+ IX86_BUILTIN_COMULEPS,
+ IX86_BUILTIN_COMUGTPS,
+ IX86_BUILTIN_COMUGEPS,
+ IX86_BUILTIN_COMORDPS,
+ IX86_BUILTIN_COMUNORDPS,
+ IX86_BUILTIN_COMFALSEPS,
+ IX86_BUILTIN_COMTRUEPS,
+
+ IX86_BUILTIN_COMEQPD,
+ IX86_BUILTIN_COMNEPD,
+ IX86_BUILTIN_COMLTPD,
+ IX86_BUILTIN_COMLEPD,
+ IX86_BUILTIN_COMGTPD,
+ IX86_BUILTIN_COMGEPD,
+ IX86_BUILTIN_COMUEQPD,
+ IX86_BUILTIN_COMUNEPD,
+ IX86_BUILTIN_COMULTPD,
+ IX86_BUILTIN_COMULEPD,
+ IX86_BUILTIN_COMUGTPD,
+ IX86_BUILTIN_COMUGEPD,
+ IX86_BUILTIN_COMORDPD,
+ IX86_BUILTIN_COMUNORDPD,
+ IX86_BUILTIN_COMFALSEPD,
+ IX86_BUILTIN_COMTRUEPD,
+
+ IX86_BUILTIN_PCOMEQUB,
+ IX86_BUILTIN_PCOMNEUB,
+ IX86_BUILTIN_PCOMLTUB,
+ IX86_BUILTIN_PCOMLEUB,
+ IX86_BUILTIN_PCOMGTUB,
+ IX86_BUILTIN_PCOMGEUB,
+ IX86_BUILTIN_PCOMFALSEUB,
+ IX86_BUILTIN_PCOMTRUEUB,
+ IX86_BUILTIN_PCOMEQUW,
+ IX86_BUILTIN_PCOMNEUW,
+ IX86_BUILTIN_PCOMLTUW,
+ IX86_BUILTIN_PCOMLEUW,
+ IX86_BUILTIN_PCOMGTUW,
+ IX86_BUILTIN_PCOMGEUW,
+ IX86_BUILTIN_PCOMFALSEUW,
+ IX86_BUILTIN_PCOMTRUEUW,
+ IX86_BUILTIN_PCOMEQUD,
+ IX86_BUILTIN_PCOMNEUD,
+ IX86_BUILTIN_PCOMLTUD,
+ IX86_BUILTIN_PCOMLEUD,
+ IX86_BUILTIN_PCOMGTUD,
+ IX86_BUILTIN_PCOMGEUD,
+ IX86_BUILTIN_PCOMFALSEUD,
+ IX86_BUILTIN_PCOMTRUEUD,
+ IX86_BUILTIN_PCOMEQUQ,
+ IX86_BUILTIN_PCOMNEUQ,
+ IX86_BUILTIN_PCOMLTUQ,
+ IX86_BUILTIN_PCOMLEUQ,
+ IX86_BUILTIN_PCOMGTUQ,
+ IX86_BUILTIN_PCOMGEUQ,
+ IX86_BUILTIN_PCOMFALSEUQ,
+ IX86_BUILTIN_PCOMTRUEUQ,
+
+ IX86_BUILTIN_PCOMEQB,
+ IX86_BUILTIN_PCOMNEB,
+ IX86_BUILTIN_PCOMLTB,
+ IX86_BUILTIN_PCOMLEB,
+ IX86_BUILTIN_PCOMGTB,
+ IX86_BUILTIN_PCOMGEB,
+ IX86_BUILTIN_PCOMFALSEB,
+ IX86_BUILTIN_PCOMTRUEB,
+ IX86_BUILTIN_PCOMEQW,
+ IX86_BUILTIN_PCOMNEW,
+ IX86_BUILTIN_PCOMLTW,
+ IX86_BUILTIN_PCOMLEW,
+ IX86_BUILTIN_PCOMGTW,
+ IX86_BUILTIN_PCOMGEW,
+ IX86_BUILTIN_PCOMFALSEW,
+ IX86_BUILTIN_PCOMTRUEW,
+ IX86_BUILTIN_PCOMEQD,
+ IX86_BUILTIN_PCOMNED,
+ IX86_BUILTIN_PCOMLTD,
+ IX86_BUILTIN_PCOMLED,
+ IX86_BUILTIN_PCOMGTD,
+ IX86_BUILTIN_PCOMGED,
+ IX86_BUILTIN_PCOMFALSED,
+ IX86_BUILTIN_PCOMTRUED,
+ IX86_BUILTIN_PCOMEQQ,
+ IX86_BUILTIN_PCOMNEQ,
+ IX86_BUILTIN_PCOMLTQ,
+ IX86_BUILTIN_PCOMLEQ,
+ IX86_BUILTIN_PCOMGTQ,
+ IX86_BUILTIN_PCOMGEQ,
+ IX86_BUILTIN_PCOMFALSEQ,
+ IX86_BUILTIN_PCOMTRUEQ,
IX86_BUILTIN_MAX
};
/* Table for the ix86 builtin decls. */
static GTY(()) tree ix86_builtins[(int) IX86_BUILTIN_MAX];
-/* Add a ix86 target builtin function with CODE, NAME and TYPE. Do so,
+/* Add an ix86 target builtin function with CODE, NAME and TYPE. Do so,
* if the target_flags include one of MASK. Stores the function decl
* in the ix86_builtins array.
* Returns the function decl or NULL_TREE, if the builtin was not added. */
{
tree decl = NULL_TREE;
- if (mask & target_flags
- && (!(mask & MASK_64BIT) || TARGET_64BIT))
+ if (mask & ix86_isa_flags
+ && (!(mask & OPTION_MASK_ISA_64BIT) || TARGET_64BIT))
{
decl = add_builtin_function (name, type, code, BUILT_IN_MD,
NULL, NULL_TREE);
const char *const name;
const enum ix86_builtins code;
const enum rtx_code comparison;
- const unsigned int flag;
+ const int flag;
};
static const struct builtin_description bdesc_comi[] =
{
- { MASK_SSE, CODE_FOR_sse_comi, "__builtin_ia32_comieq", IX86_BUILTIN_COMIEQSS, UNEQ, 0 },
- { MASK_SSE, CODE_FOR_sse_comi, "__builtin_ia32_comilt", IX86_BUILTIN_COMILTSS, UNLT, 0 },
- { MASK_SSE, CODE_FOR_sse_comi, "__builtin_ia32_comile", IX86_BUILTIN_COMILESS, UNLE, 0 },
- { MASK_SSE, CODE_FOR_sse_comi, "__builtin_ia32_comigt", IX86_BUILTIN_COMIGTSS, GT, 0 },
- { MASK_SSE, CODE_FOR_sse_comi, "__builtin_ia32_comige", IX86_BUILTIN_COMIGESS, GE, 0 },
- { MASK_SSE, CODE_FOR_sse_comi, "__builtin_ia32_comineq", IX86_BUILTIN_COMINEQSS, LTGT, 0 },
- { MASK_SSE, CODE_FOR_sse_ucomi, "__builtin_ia32_ucomieq", IX86_BUILTIN_UCOMIEQSS, UNEQ, 0 },
- { MASK_SSE, CODE_FOR_sse_ucomi, "__builtin_ia32_ucomilt", IX86_BUILTIN_UCOMILTSS, UNLT, 0 },
- { MASK_SSE, CODE_FOR_sse_ucomi, "__builtin_ia32_ucomile", IX86_BUILTIN_UCOMILESS, UNLE, 0 },
- { MASK_SSE, CODE_FOR_sse_ucomi, "__builtin_ia32_ucomigt", IX86_BUILTIN_UCOMIGTSS, GT, 0 },
- { MASK_SSE, CODE_FOR_sse_ucomi, "__builtin_ia32_ucomige", IX86_BUILTIN_UCOMIGESS, GE, 0 },
- { MASK_SSE, CODE_FOR_sse_ucomi, "__builtin_ia32_ucomineq", IX86_BUILTIN_UCOMINEQSS, LTGT, 0 },
- { MASK_SSE2, CODE_FOR_sse2_comi, "__builtin_ia32_comisdeq", IX86_BUILTIN_COMIEQSD, UNEQ, 0 },
- { MASK_SSE2, CODE_FOR_sse2_comi, "__builtin_ia32_comisdlt", IX86_BUILTIN_COMILTSD, UNLT, 0 },
- { MASK_SSE2, CODE_FOR_sse2_comi, "__builtin_ia32_comisdle", IX86_BUILTIN_COMILESD, UNLE, 0 },
- { MASK_SSE2, CODE_FOR_sse2_comi, "__builtin_ia32_comisdgt", IX86_BUILTIN_COMIGTSD, GT, 0 },
- { MASK_SSE2, CODE_FOR_sse2_comi, "__builtin_ia32_comisdge", IX86_BUILTIN_COMIGESD, GE, 0 },
- { MASK_SSE2, CODE_FOR_sse2_comi, "__builtin_ia32_comisdneq", IX86_BUILTIN_COMINEQSD, LTGT, 0 },
- { MASK_SSE2, CODE_FOR_sse2_ucomi, "__builtin_ia32_ucomisdeq", IX86_BUILTIN_UCOMIEQSD, UNEQ, 0 },
- { MASK_SSE2, CODE_FOR_sse2_ucomi, "__builtin_ia32_ucomisdlt", IX86_BUILTIN_UCOMILTSD, UNLT, 0 },
- { MASK_SSE2, CODE_FOR_sse2_ucomi, "__builtin_ia32_ucomisdle", IX86_BUILTIN_UCOMILESD, UNLE, 0 },
- { MASK_SSE2, CODE_FOR_sse2_ucomi, "__builtin_ia32_ucomisdgt", IX86_BUILTIN_UCOMIGTSD, GT, 0 },
- { MASK_SSE2, CODE_FOR_sse2_ucomi, "__builtin_ia32_ucomisdge", IX86_BUILTIN_UCOMIGESD, GE, 0 },
- { MASK_SSE2, CODE_FOR_sse2_ucomi, "__builtin_ia32_ucomisdneq", IX86_BUILTIN_UCOMINEQSD, LTGT, 0 },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_comi, "__builtin_ia32_comieq", IX86_BUILTIN_COMIEQSS, UNEQ, 0 },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_comi, "__builtin_ia32_comilt", IX86_BUILTIN_COMILTSS, UNLT, 0 },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_comi, "__builtin_ia32_comile", IX86_BUILTIN_COMILESS, UNLE, 0 },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_comi, "__builtin_ia32_comigt", IX86_BUILTIN_COMIGTSS, GT, 0 },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_comi, "__builtin_ia32_comige", IX86_BUILTIN_COMIGESS, GE, 0 },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_comi, "__builtin_ia32_comineq", IX86_BUILTIN_COMINEQSS, LTGT, 0 },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_ucomi, "__builtin_ia32_ucomieq", IX86_BUILTIN_UCOMIEQSS, UNEQ, 0 },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_ucomi, "__builtin_ia32_ucomilt", IX86_BUILTIN_UCOMILTSS, UNLT, 0 },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_ucomi, "__builtin_ia32_ucomile", IX86_BUILTIN_UCOMILESS, UNLE, 0 },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_ucomi, "__builtin_ia32_ucomigt", IX86_BUILTIN_UCOMIGTSS, GT, 0 },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_ucomi, "__builtin_ia32_ucomige", IX86_BUILTIN_UCOMIGESS, GE, 0 },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_ucomi, "__builtin_ia32_ucomineq", IX86_BUILTIN_UCOMINEQSS, LTGT, 0 },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_comi, "__builtin_ia32_comisdeq", IX86_BUILTIN_COMIEQSD, UNEQ, 0 },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_comi, "__builtin_ia32_comisdlt", IX86_BUILTIN_COMILTSD, UNLT, 0 },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_comi, "__builtin_ia32_comisdle", IX86_BUILTIN_COMILESD, UNLE, 0 },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_comi, "__builtin_ia32_comisdgt", IX86_BUILTIN_COMIGTSD, GT, 0 },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_comi, "__builtin_ia32_comisdge", IX86_BUILTIN_COMIGESD, GE, 0 },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_comi, "__builtin_ia32_comisdneq", IX86_BUILTIN_COMINEQSD, LTGT, 0 },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_ucomi, "__builtin_ia32_ucomisdeq", IX86_BUILTIN_UCOMIEQSD, UNEQ, 0 },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_ucomi, "__builtin_ia32_ucomisdlt", IX86_BUILTIN_UCOMILTSD, UNLT, 0 },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_ucomi, "__builtin_ia32_ucomisdle", IX86_BUILTIN_UCOMILESD, UNLE, 0 },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_ucomi, "__builtin_ia32_ucomisdgt", IX86_BUILTIN_UCOMIGTSD, GT, 0 },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_ucomi, "__builtin_ia32_ucomisdge", IX86_BUILTIN_UCOMIGESD, GE, 0 },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_ucomi, "__builtin_ia32_ucomisdneq", IX86_BUILTIN_UCOMINEQSD, LTGT, 0 },
};
-static const struct builtin_description bdesc_2arg[] =
+static const struct builtin_description bdesc_pcmpestr[] =
{
- /* SSE */
- { MASK_SSE, CODE_FOR_addv4sf3, "__builtin_ia32_addps", IX86_BUILTIN_ADDPS, 0, 0 },
- { MASK_SSE, CODE_FOR_subv4sf3, "__builtin_ia32_subps", IX86_BUILTIN_SUBPS, 0, 0 },
- { MASK_SSE, CODE_FOR_mulv4sf3, "__builtin_ia32_mulps", IX86_BUILTIN_MULPS, 0, 0 },
- { MASK_SSE, CODE_FOR_divv4sf3, "__builtin_ia32_divps", IX86_BUILTIN_DIVPS, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_vmaddv4sf3, "__builtin_ia32_addss", IX86_BUILTIN_ADDSS, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_vmsubv4sf3, "__builtin_ia32_subss", IX86_BUILTIN_SUBSS, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_vmmulv4sf3, "__builtin_ia32_mulss", IX86_BUILTIN_MULSS, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_vmdivv4sf3, "__builtin_ia32_divss", IX86_BUILTIN_DIVSS, 0, 0 },
-
- { MASK_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpeqps", IX86_BUILTIN_CMPEQPS, EQ, 0 },
- { MASK_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpltps", IX86_BUILTIN_CMPLTPS, LT, 0 },
- { MASK_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpleps", IX86_BUILTIN_CMPLEPS, LE, 0 },
- { MASK_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpgtps", IX86_BUILTIN_CMPGTPS, LT,
- BUILTIN_DESC_SWAP_OPERANDS },
- { MASK_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpgeps", IX86_BUILTIN_CMPGEPS, LE,
- BUILTIN_DESC_SWAP_OPERANDS },
- { MASK_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpunordps", IX86_BUILTIN_CMPUNORDPS, UNORDERED, 0 },
- { MASK_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpneqps", IX86_BUILTIN_CMPNEQPS, NE, 0 },
- { MASK_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpnltps", IX86_BUILTIN_CMPNLTPS, UNGE, 0 },
- { MASK_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpnleps", IX86_BUILTIN_CMPNLEPS, UNGT, 0 },
- { MASK_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpngtps", IX86_BUILTIN_CMPNGTPS, UNGE,
- BUILTIN_DESC_SWAP_OPERANDS },
- { MASK_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpngeps", IX86_BUILTIN_CMPNGEPS, UNGT,
- BUILTIN_DESC_SWAP_OPERANDS },
- { MASK_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpordps", IX86_BUILTIN_CMPORDPS, ORDERED, 0 },
- { MASK_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpeqss", IX86_BUILTIN_CMPEQSS, EQ, 0 },
- { MASK_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpltss", IX86_BUILTIN_CMPLTSS, LT, 0 },
- { MASK_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpless", IX86_BUILTIN_CMPLESS, LE, 0 },
- { MASK_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpunordss", IX86_BUILTIN_CMPUNORDSS, UNORDERED, 0 },
- { MASK_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpneqss", IX86_BUILTIN_CMPNEQSS, NE, 0 },
- { MASK_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpnltss", IX86_BUILTIN_CMPNLTSS, UNGE, 0 },
- { MASK_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpnless", IX86_BUILTIN_CMPNLESS, UNGT, 0 },
- { MASK_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpngtss", IX86_BUILTIN_CMPNGTSS, UNGE,
- BUILTIN_DESC_SWAP_OPERANDS },
- { MASK_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpngess", IX86_BUILTIN_CMPNGESS, UNGT,
- BUILTIN_DESC_SWAP_OPERANDS },
- { MASK_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpordss", IX86_BUILTIN_CMPORDSS, UNORDERED, 0 },
-
- { MASK_SSE, CODE_FOR_sminv4sf3, "__builtin_ia32_minps", IX86_BUILTIN_MINPS, 0, 0 },
- { MASK_SSE, CODE_FOR_smaxv4sf3, "__builtin_ia32_maxps", IX86_BUILTIN_MAXPS, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_vmsminv4sf3, "__builtin_ia32_minss", IX86_BUILTIN_MINSS, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_vmsmaxv4sf3, "__builtin_ia32_maxss", IX86_BUILTIN_MAXSS, 0, 0 },
-
- { MASK_SSE, CODE_FOR_andv4sf3, "__builtin_ia32_andps", IX86_BUILTIN_ANDPS, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_nandv4sf3, "__builtin_ia32_andnps", IX86_BUILTIN_ANDNPS, 0, 0 },
- { MASK_SSE, CODE_FOR_iorv4sf3, "__builtin_ia32_orps", IX86_BUILTIN_ORPS, 0, 0 },
- { MASK_SSE, CODE_FOR_xorv4sf3, "__builtin_ia32_xorps", IX86_BUILTIN_XORPS, 0, 0 },
-
- { MASK_SSE, CODE_FOR_sse_movss, "__builtin_ia32_movss", IX86_BUILTIN_MOVSS, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_movhlps, "__builtin_ia32_movhlps", IX86_BUILTIN_MOVHLPS, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_movlhps, "__builtin_ia32_movlhps", IX86_BUILTIN_MOVLHPS, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_unpckhps, "__builtin_ia32_unpckhps", IX86_BUILTIN_UNPCKHPS, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_unpcklps, "__builtin_ia32_unpcklps", IX86_BUILTIN_UNPCKLPS, 0, 0 },
-
- /* MMX */
- { MASK_MMX, CODE_FOR_mmx_addv8qi3, "__builtin_ia32_paddb", IX86_BUILTIN_PADDB, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_addv4hi3, "__builtin_ia32_paddw", IX86_BUILTIN_PADDW, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_addv2si3, "__builtin_ia32_paddd", IX86_BUILTIN_PADDD, 0, 0 },
- { MASK_SSE2, CODE_FOR_mmx_adddi3, "__builtin_ia32_paddq", IX86_BUILTIN_PADDQ, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_subv8qi3, "__builtin_ia32_psubb", IX86_BUILTIN_PSUBB, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_subv4hi3, "__builtin_ia32_psubw", IX86_BUILTIN_PSUBW, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_subv2si3, "__builtin_ia32_psubd", IX86_BUILTIN_PSUBD, 0, 0 },
- { MASK_SSE2, CODE_FOR_mmx_subdi3, "__builtin_ia32_psubq", IX86_BUILTIN_PSUBQ, 0, 0 },
-
- { MASK_MMX, CODE_FOR_mmx_ssaddv8qi3, "__builtin_ia32_paddsb", IX86_BUILTIN_PADDSB, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_ssaddv4hi3, "__builtin_ia32_paddsw", IX86_BUILTIN_PADDSW, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_sssubv8qi3, "__builtin_ia32_psubsb", IX86_BUILTIN_PSUBSB, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_sssubv4hi3, "__builtin_ia32_psubsw", IX86_BUILTIN_PSUBSW, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_usaddv8qi3, "__builtin_ia32_paddusb", IX86_BUILTIN_PADDUSB, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_usaddv4hi3, "__builtin_ia32_paddusw", IX86_BUILTIN_PADDUSW, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_ussubv8qi3, "__builtin_ia32_psubusb", IX86_BUILTIN_PSUBUSB, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_ussubv4hi3, "__builtin_ia32_psubusw", IX86_BUILTIN_PSUBUSW, 0, 0 },
-
- { MASK_MMX, CODE_FOR_mmx_mulv4hi3, "__builtin_ia32_pmullw", IX86_BUILTIN_PMULLW, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_smulv4hi3_highpart, "__builtin_ia32_pmulhw", IX86_BUILTIN_PMULHW, 0, 0 },
- { MASK_SSE | MASK_3DNOW_A, CODE_FOR_mmx_umulv4hi3_highpart, "__builtin_ia32_pmulhuw", IX86_BUILTIN_PMULHUW, 0, 0 },
-
- { MASK_MMX, CODE_FOR_mmx_andv2si3, "__builtin_ia32_pand", IX86_BUILTIN_PAND, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_nandv2si3, "__builtin_ia32_pandn", IX86_BUILTIN_PANDN, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_iorv2si3, "__builtin_ia32_por", IX86_BUILTIN_POR, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_xorv2si3, "__builtin_ia32_pxor", IX86_BUILTIN_PXOR, 0, 0 },
-
- { MASK_SSE | MASK_3DNOW_A, CODE_FOR_mmx_uavgv8qi3, "__builtin_ia32_pavgb", IX86_BUILTIN_PAVGB, 0, 0 },
- { MASK_SSE | MASK_3DNOW_A, CODE_FOR_mmx_uavgv4hi3, "__builtin_ia32_pavgw", IX86_BUILTIN_PAVGW, 0, 0 },
-
- { MASK_MMX, CODE_FOR_mmx_eqv8qi3, "__builtin_ia32_pcmpeqb", IX86_BUILTIN_PCMPEQB, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_eqv4hi3, "__builtin_ia32_pcmpeqw", IX86_BUILTIN_PCMPEQW, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_eqv2si3, "__builtin_ia32_pcmpeqd", IX86_BUILTIN_PCMPEQD, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_gtv8qi3, "__builtin_ia32_pcmpgtb", IX86_BUILTIN_PCMPGTB, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_gtv4hi3, "__builtin_ia32_pcmpgtw", IX86_BUILTIN_PCMPGTW, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_gtv2si3, "__builtin_ia32_pcmpgtd", IX86_BUILTIN_PCMPGTD, 0, 0 },
-
- { MASK_SSE | MASK_3DNOW_A, CODE_FOR_mmx_umaxv8qi3, "__builtin_ia32_pmaxub", IX86_BUILTIN_PMAXUB, 0, 0 },
- { MASK_SSE | MASK_3DNOW_A, CODE_FOR_mmx_smaxv4hi3, "__builtin_ia32_pmaxsw", IX86_BUILTIN_PMAXSW, 0, 0 },
- { MASK_SSE | MASK_3DNOW_A, CODE_FOR_mmx_uminv8qi3, "__builtin_ia32_pminub", IX86_BUILTIN_PMINUB, 0, 0 },
- { MASK_SSE | MASK_3DNOW_A, CODE_FOR_mmx_sminv4hi3, "__builtin_ia32_pminsw", IX86_BUILTIN_PMINSW, 0, 0 },
-
- { MASK_MMX, CODE_FOR_mmx_punpckhbw, "__builtin_ia32_punpckhbw", IX86_BUILTIN_PUNPCKHBW, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_punpckhwd, "__builtin_ia32_punpckhwd", IX86_BUILTIN_PUNPCKHWD, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_punpckhdq, "__builtin_ia32_punpckhdq", IX86_BUILTIN_PUNPCKHDQ, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_punpcklbw, "__builtin_ia32_punpcklbw", IX86_BUILTIN_PUNPCKLBW, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_punpcklwd, "__builtin_ia32_punpcklwd", IX86_BUILTIN_PUNPCKLWD, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_punpckldq, "__builtin_ia32_punpckldq", IX86_BUILTIN_PUNPCKLDQ, 0, 0 },
-
- /* Special. */
- { MASK_MMX, CODE_FOR_mmx_packsswb, 0, IX86_BUILTIN_PACKSSWB, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_packssdw, 0, IX86_BUILTIN_PACKSSDW, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_packuswb, 0, IX86_BUILTIN_PACKUSWB, 0, 0 },
-
- { MASK_SSE, CODE_FOR_sse_cvtpi2ps, 0, IX86_BUILTIN_CVTPI2PS, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_cvtsi2ss, 0, IX86_BUILTIN_CVTSI2SS, 0, 0 },
- { MASK_SSE | MASK_64BIT, CODE_FOR_sse_cvtsi2ssq, 0, IX86_BUILTIN_CVTSI642SS, 0, 0 },
-
- { MASK_MMX, CODE_FOR_mmx_ashlv4hi3, 0, IX86_BUILTIN_PSLLW, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_ashlv4hi3, 0, IX86_BUILTIN_PSLLWI, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_ashlv2si3, 0, IX86_BUILTIN_PSLLD, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_ashlv2si3, 0, IX86_BUILTIN_PSLLDI, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_ashldi3, 0, IX86_BUILTIN_PSLLQ, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_ashldi3, 0, IX86_BUILTIN_PSLLQI, 0, 0 },
-
- { MASK_MMX, CODE_FOR_mmx_lshrv4hi3, 0, IX86_BUILTIN_PSRLW, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_lshrv4hi3, 0, IX86_BUILTIN_PSRLWI, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_lshrv2si3, 0, IX86_BUILTIN_PSRLD, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_lshrv2si3, 0, IX86_BUILTIN_PSRLDI, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_lshrdi3, 0, IX86_BUILTIN_PSRLQ, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_lshrdi3, 0, IX86_BUILTIN_PSRLQI, 0, 0 },
-
- { MASK_MMX, CODE_FOR_mmx_ashrv4hi3, 0, IX86_BUILTIN_PSRAW, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_ashrv4hi3, 0, IX86_BUILTIN_PSRAWI, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_ashrv2si3, 0, IX86_BUILTIN_PSRAD, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_ashrv2si3, 0, IX86_BUILTIN_PSRADI, 0, 0 },
-
- { MASK_SSE | MASK_3DNOW_A, CODE_FOR_mmx_psadbw, 0, IX86_BUILTIN_PSADBW, 0, 0 },
- { MASK_MMX, CODE_FOR_mmx_pmaddwd, 0, IX86_BUILTIN_PMADDWD, 0, 0 },
+ /* SSE4.2 */
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_pcmpestr, "__builtin_ia32_pcmpestri128", IX86_BUILTIN_PCMPESTRI128, UNKNOWN, 0 },
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_pcmpestr, "__builtin_ia32_pcmpestrm128", IX86_BUILTIN_PCMPESTRM128, UNKNOWN, 0 },
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_pcmpestr, "__builtin_ia32_pcmpestria128", IX86_BUILTIN_PCMPESTRA128, UNKNOWN, (int) CCAmode },
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_pcmpestr, "__builtin_ia32_pcmpestric128", IX86_BUILTIN_PCMPESTRC128, UNKNOWN, (int) CCCmode },
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_pcmpestr, "__builtin_ia32_pcmpestrio128", IX86_BUILTIN_PCMPESTRO128, UNKNOWN, (int) CCOmode },
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_pcmpestr, "__builtin_ia32_pcmpestris128", IX86_BUILTIN_PCMPESTRS128, UNKNOWN, (int) CCSmode },
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_pcmpestr, "__builtin_ia32_pcmpestriz128", IX86_BUILTIN_PCMPESTRZ128, UNKNOWN, (int) CCZmode },
+};
- /* SSE2 */
- { MASK_SSE2, CODE_FOR_addv2df3, "__builtin_ia32_addpd", IX86_BUILTIN_ADDPD, 0, 0 },
- { MASK_SSE2, CODE_FOR_subv2df3, "__builtin_ia32_subpd", IX86_BUILTIN_SUBPD, 0, 0 },
- { MASK_SSE2, CODE_FOR_mulv2df3, "__builtin_ia32_mulpd", IX86_BUILTIN_MULPD, 0, 0 },
- { MASK_SSE2, CODE_FOR_divv2df3, "__builtin_ia32_divpd", IX86_BUILTIN_DIVPD, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_vmaddv2df3, "__builtin_ia32_addsd", IX86_BUILTIN_ADDSD, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_vmsubv2df3, "__builtin_ia32_subsd", IX86_BUILTIN_SUBSD, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_vmmulv2df3, "__builtin_ia32_mulsd", IX86_BUILTIN_MULSD, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_vmdivv2df3, "__builtin_ia32_divsd", IX86_BUILTIN_DIVSD, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpeqpd", IX86_BUILTIN_CMPEQPD, EQ, 0 },
- { MASK_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpltpd", IX86_BUILTIN_CMPLTPD, LT, 0 },
- { MASK_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmplepd", IX86_BUILTIN_CMPLEPD, LE, 0 },
- { MASK_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpgtpd", IX86_BUILTIN_CMPGTPD, LT,
- BUILTIN_DESC_SWAP_OPERANDS },
- { MASK_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpgepd", IX86_BUILTIN_CMPGEPD, LE,
- BUILTIN_DESC_SWAP_OPERANDS },
- { MASK_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpunordpd", IX86_BUILTIN_CMPUNORDPD, UNORDERED, 0 },
- { MASK_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpneqpd", IX86_BUILTIN_CMPNEQPD, NE, 0 },
- { MASK_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpnltpd", IX86_BUILTIN_CMPNLTPD, UNGE, 0 },
- { MASK_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpnlepd", IX86_BUILTIN_CMPNLEPD, UNGT, 0 },
- { MASK_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpngtpd", IX86_BUILTIN_CMPNGTPD, UNGE,
- BUILTIN_DESC_SWAP_OPERANDS },
- { MASK_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpngepd", IX86_BUILTIN_CMPNGEPD, UNGT,
- BUILTIN_DESC_SWAP_OPERANDS },
- { MASK_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpordpd", IX86_BUILTIN_CMPORDPD, ORDERED, 0 },
- { MASK_SSE2, CODE_FOR_sse2_vmmaskcmpv2df3, "__builtin_ia32_cmpeqsd", IX86_BUILTIN_CMPEQSD, EQ, 0 },
- { MASK_SSE2, CODE_FOR_sse2_vmmaskcmpv2df3, "__builtin_ia32_cmpltsd", IX86_BUILTIN_CMPLTSD, LT, 0 },
- { MASK_SSE2, CODE_FOR_sse2_vmmaskcmpv2df3, "__builtin_ia32_cmplesd", IX86_BUILTIN_CMPLESD, LE, 0 },
- { MASK_SSE2, CODE_FOR_sse2_vmmaskcmpv2df3, "__builtin_ia32_cmpunordsd", IX86_BUILTIN_CMPUNORDSD, UNORDERED, 0 },
- { MASK_SSE2, CODE_FOR_sse2_vmmaskcmpv2df3, "__builtin_ia32_cmpneqsd", IX86_BUILTIN_CMPNEQSD, NE, 0 },
- { MASK_SSE2, CODE_FOR_sse2_vmmaskcmpv2df3, "__builtin_ia32_cmpnltsd", IX86_BUILTIN_CMPNLTSD, UNGE, 0 },
- { MASK_SSE2, CODE_FOR_sse2_vmmaskcmpv2df3, "__builtin_ia32_cmpnlesd", IX86_BUILTIN_CMPNLESD, UNGT, 0 },
- { MASK_SSE2, CODE_FOR_sse2_vmmaskcmpv2df3, "__builtin_ia32_cmpordsd", IX86_BUILTIN_CMPORDSD, ORDERED, 0 },
-
- { MASK_SSE2, CODE_FOR_sminv2df3, "__builtin_ia32_minpd", IX86_BUILTIN_MINPD, 0, 0 },
- { MASK_SSE2, CODE_FOR_smaxv2df3, "__builtin_ia32_maxpd", IX86_BUILTIN_MAXPD, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_vmsminv2df3, "__builtin_ia32_minsd", IX86_BUILTIN_MINSD, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_vmsmaxv2df3, "__builtin_ia32_maxsd", IX86_BUILTIN_MAXSD, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_andv2df3, "__builtin_ia32_andpd", IX86_BUILTIN_ANDPD, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_nandv2df3, "__builtin_ia32_andnpd", IX86_BUILTIN_ANDNPD, 0, 0 },
- { MASK_SSE2, CODE_FOR_iorv2df3, "__builtin_ia32_orpd", IX86_BUILTIN_ORPD, 0, 0 },
- { MASK_SSE2, CODE_FOR_xorv2df3, "__builtin_ia32_xorpd", IX86_BUILTIN_XORPD, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_sse2_movsd, "__builtin_ia32_movsd", IX86_BUILTIN_MOVSD, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_unpckhpd, "__builtin_ia32_unpckhpd", IX86_BUILTIN_UNPCKHPD, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_unpcklpd, "__builtin_ia32_unpcklpd", IX86_BUILTIN_UNPCKLPD, 0, 0 },
+static const struct builtin_description bdesc_pcmpistr[] =
+{
+ /* SSE4.2 */
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_pcmpistr, "__builtin_ia32_pcmpistri128", IX86_BUILTIN_PCMPISTRI128, UNKNOWN, 0 },
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_pcmpistr, "__builtin_ia32_pcmpistrm128", IX86_BUILTIN_PCMPISTRM128, UNKNOWN, 0 },
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_pcmpistr, "__builtin_ia32_pcmpistria128", IX86_BUILTIN_PCMPISTRA128, UNKNOWN, (int) CCAmode },
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_pcmpistr, "__builtin_ia32_pcmpistric128", IX86_BUILTIN_PCMPISTRC128, UNKNOWN, (int) CCCmode },
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_pcmpistr, "__builtin_ia32_pcmpistrio128", IX86_BUILTIN_PCMPISTRO128, UNKNOWN, (int) CCOmode },
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_pcmpistr, "__builtin_ia32_pcmpistris128", IX86_BUILTIN_PCMPISTRS128, UNKNOWN, (int) CCSmode },
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_pcmpistr, "__builtin_ia32_pcmpistriz128", IX86_BUILTIN_PCMPISTRZ128, UNKNOWN, (int) CCZmode },
+};
- /* SSE2 MMX */
- { MASK_SSE2, CODE_FOR_addv16qi3, "__builtin_ia32_paddb128", IX86_BUILTIN_PADDB128, 0, 0 },
- { MASK_SSE2, CODE_FOR_addv8hi3, "__builtin_ia32_paddw128", IX86_BUILTIN_PADDW128, 0, 0 },
- { MASK_SSE2, CODE_FOR_addv4si3, "__builtin_ia32_paddd128", IX86_BUILTIN_PADDD128, 0, 0 },
- { MASK_SSE2, CODE_FOR_addv2di3, "__builtin_ia32_paddq128", IX86_BUILTIN_PADDQ128, 0, 0 },
- { MASK_SSE2, CODE_FOR_subv16qi3, "__builtin_ia32_psubb128", IX86_BUILTIN_PSUBB128, 0, 0 },
- { MASK_SSE2, CODE_FOR_subv8hi3, "__builtin_ia32_psubw128", IX86_BUILTIN_PSUBW128, 0, 0 },
- { MASK_SSE2, CODE_FOR_subv4si3, "__builtin_ia32_psubd128", IX86_BUILTIN_PSUBD128, 0, 0 },
- { MASK_SSE2, CODE_FOR_subv2di3, "__builtin_ia32_psubq128", IX86_BUILTIN_PSUBQ128, 0, 0 },
-
- { MASK_MMX, CODE_FOR_sse2_ssaddv16qi3, "__builtin_ia32_paddsb128", IX86_BUILTIN_PADDSB128, 0, 0 },
- { MASK_MMX, CODE_FOR_sse2_ssaddv8hi3, "__builtin_ia32_paddsw128", IX86_BUILTIN_PADDSW128, 0, 0 },
- { MASK_MMX, CODE_FOR_sse2_sssubv16qi3, "__builtin_ia32_psubsb128", IX86_BUILTIN_PSUBSB128, 0, 0 },
- { MASK_MMX, CODE_FOR_sse2_sssubv8hi3, "__builtin_ia32_psubsw128", IX86_BUILTIN_PSUBSW128, 0, 0 },
- { MASK_MMX, CODE_FOR_sse2_usaddv16qi3, "__builtin_ia32_paddusb128", IX86_BUILTIN_PADDUSB128, 0, 0 },
- { MASK_MMX, CODE_FOR_sse2_usaddv8hi3, "__builtin_ia32_paddusw128", IX86_BUILTIN_PADDUSW128, 0, 0 },
- { MASK_MMX, CODE_FOR_sse2_ussubv16qi3, "__builtin_ia32_psubusb128", IX86_BUILTIN_PSUBUSB128, 0, 0 },
- { MASK_MMX, CODE_FOR_sse2_ussubv8hi3, "__builtin_ia32_psubusw128", IX86_BUILTIN_PSUBUSW128, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_mulv8hi3, "__builtin_ia32_pmullw128", IX86_BUILTIN_PMULLW128, 0, 0 },
- { MASK_SSE2, CODE_FOR_smulv8hi3_highpart, "__builtin_ia32_pmulhw128", IX86_BUILTIN_PMULHW128, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_andv2di3, "__builtin_ia32_pand128", IX86_BUILTIN_PAND128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_nandv2di3, "__builtin_ia32_pandn128", IX86_BUILTIN_PANDN128, 0, 0 },
- { MASK_SSE2, CODE_FOR_iorv2di3, "__builtin_ia32_por128", IX86_BUILTIN_POR128, 0, 0 },
- { MASK_SSE2, CODE_FOR_xorv2di3, "__builtin_ia32_pxor128", IX86_BUILTIN_PXOR128, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_sse2_uavgv16qi3, "__builtin_ia32_pavgb128", IX86_BUILTIN_PAVGB128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_uavgv8hi3, "__builtin_ia32_pavgw128", IX86_BUILTIN_PAVGW128, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_sse2_eqv16qi3, "__builtin_ia32_pcmpeqb128", IX86_BUILTIN_PCMPEQB128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_eqv8hi3, "__builtin_ia32_pcmpeqw128", IX86_BUILTIN_PCMPEQW128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_eqv4si3, "__builtin_ia32_pcmpeqd128", IX86_BUILTIN_PCMPEQD128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_gtv16qi3, "__builtin_ia32_pcmpgtb128", IX86_BUILTIN_PCMPGTB128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_gtv8hi3, "__builtin_ia32_pcmpgtw128", IX86_BUILTIN_PCMPGTW128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_gtv4si3, "__builtin_ia32_pcmpgtd128", IX86_BUILTIN_PCMPGTD128, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_umaxv16qi3, "__builtin_ia32_pmaxub128", IX86_BUILTIN_PMAXUB128, 0, 0 },
- { MASK_SSE2, CODE_FOR_smaxv8hi3, "__builtin_ia32_pmaxsw128", IX86_BUILTIN_PMAXSW128, 0, 0 },
- { MASK_SSE2, CODE_FOR_uminv16qi3, "__builtin_ia32_pminub128", IX86_BUILTIN_PMINUB128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sminv8hi3, "__builtin_ia32_pminsw128", IX86_BUILTIN_PMINSW128, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_sse2_punpckhbw, "__builtin_ia32_punpckhbw128", IX86_BUILTIN_PUNPCKHBW128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_punpckhwd, "__builtin_ia32_punpckhwd128", IX86_BUILTIN_PUNPCKHWD128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_punpckhdq, "__builtin_ia32_punpckhdq128", IX86_BUILTIN_PUNPCKHDQ128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_punpckhqdq, "__builtin_ia32_punpckhqdq128", IX86_BUILTIN_PUNPCKHQDQ128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_punpcklbw, "__builtin_ia32_punpcklbw128", IX86_BUILTIN_PUNPCKLBW128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_punpcklwd, "__builtin_ia32_punpcklwd128", IX86_BUILTIN_PUNPCKLWD128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_punpckldq, "__builtin_ia32_punpckldq128", IX86_BUILTIN_PUNPCKLDQ128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_punpcklqdq, "__builtin_ia32_punpcklqdq128", IX86_BUILTIN_PUNPCKLQDQ128, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_sse2_packsswb, "__builtin_ia32_packsswb128", IX86_BUILTIN_PACKSSWB128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_packssdw, "__builtin_ia32_packssdw128", IX86_BUILTIN_PACKSSDW128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_packuswb, "__builtin_ia32_packuswb128", IX86_BUILTIN_PACKUSWB128, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_umulv8hi3_highpart, "__builtin_ia32_pmulhuw128", IX86_BUILTIN_PMULHUW128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_psadbw, 0, IX86_BUILTIN_PSADBW128, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_sse2_umulsidi3, 0, IX86_BUILTIN_PMULUDQ, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_umulv2siv2di3, 0, IX86_BUILTIN_PMULUDQ128, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_ashlv8hi3, 0, IX86_BUILTIN_PSLLWI128, 0, 0 },
- { MASK_SSE2, CODE_FOR_ashlv4si3, 0, IX86_BUILTIN_PSLLDI128, 0, 0 },
- { MASK_SSE2, CODE_FOR_ashlv2di3, 0, IX86_BUILTIN_PSLLQI128, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_lshrv8hi3, 0, IX86_BUILTIN_PSRLWI128, 0, 0 },
- { MASK_SSE2, CODE_FOR_lshrv4si3, 0, IX86_BUILTIN_PSRLDI128, 0, 0 },
- { MASK_SSE2, CODE_FOR_lshrv2di3, 0, IX86_BUILTIN_PSRLQI128, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_ashrv8hi3, 0, IX86_BUILTIN_PSRAWI128, 0, 0 },
- { MASK_SSE2, CODE_FOR_ashrv4si3, 0, IX86_BUILTIN_PSRADI128, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_sse2_pmaddwd, 0, IX86_BUILTIN_PMADDWD128, 0, 0 },
-
- { MASK_SSE2, CODE_FOR_sse2_cvtsi2sd, 0, IX86_BUILTIN_CVTSI2SD, 0, 0 },
- { MASK_SSE2 | MASK_64BIT, CODE_FOR_sse2_cvtsi2sdq, 0, IX86_BUILTIN_CVTSI642SD, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_cvtsd2ss, 0, IX86_BUILTIN_CVTSD2SS, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_cvtss2sd, 0, IX86_BUILTIN_CVTSS2SD, 0, 0 },
-
- /* SSE3 MMX */
- { MASK_SSE3, CODE_FOR_sse3_addsubv4sf3, "__builtin_ia32_addsubps", IX86_BUILTIN_ADDSUBPS, 0, 0 },
- { MASK_SSE3, CODE_FOR_sse3_addsubv2df3, "__builtin_ia32_addsubpd", IX86_BUILTIN_ADDSUBPD, 0, 0 },
- { MASK_SSE3, CODE_FOR_sse3_haddv4sf3, "__builtin_ia32_haddps", IX86_BUILTIN_HADDPS, 0, 0 },
- { MASK_SSE3, CODE_FOR_sse3_haddv2df3, "__builtin_ia32_haddpd", IX86_BUILTIN_HADDPD, 0, 0 },
- { MASK_SSE3, CODE_FOR_sse3_hsubv4sf3, "__builtin_ia32_hsubps", IX86_BUILTIN_HSUBPS, 0, 0 },
- { MASK_SSE3, CODE_FOR_sse3_hsubv2df3, "__builtin_ia32_hsubpd", IX86_BUILTIN_HSUBPD, 0, 0 },
+/* Special builtin types */
+enum ix86_special_builtin_type
+{
+ SPECIAL_FTYPE_UNKNOWN,
+ VOID_FTYPE_VOID,
+ V16QI_FTYPE_PCCHAR,
+ V4SF_FTYPE_PCFLOAT,
+ V2DF_FTYPE_PCDOUBLE,
+ V4SF_FTYPE_V4SF_PCV2SF,
+ V2DF_FTYPE_V2DF_PCDOUBLE,
+ V2DI_FTYPE_PV2DI,
+ VOID_FTYPE_PV2SF_V4SF,
+ VOID_FTYPE_PV2DI_V2DI,
+ VOID_FTYPE_PCHAR_V16QI,
+ VOID_FTYPE_PFLOAT_V4SF,
+ VOID_FTYPE_PDOUBLE_V2DF,
+ VOID_FTYPE_PDI_DI,
+ VOID_FTYPE_PINT_INT
+};
- /* SSSE3 */
- { MASK_SSSE3, CODE_FOR_ssse3_phaddwv8hi3, "__builtin_ia32_phaddw128", IX86_BUILTIN_PHADDW128, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_phaddwv4hi3, "__builtin_ia32_phaddw", IX86_BUILTIN_PHADDW, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_phadddv4si3, "__builtin_ia32_phaddd128", IX86_BUILTIN_PHADDD128, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_phadddv2si3, "__builtin_ia32_phaddd", IX86_BUILTIN_PHADDD, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_phaddswv8hi3, "__builtin_ia32_phaddsw128", IX86_BUILTIN_PHADDSW128, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_phaddswv4hi3, "__builtin_ia32_phaddsw", IX86_BUILTIN_PHADDSW, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_phsubwv8hi3, "__builtin_ia32_phsubw128", IX86_BUILTIN_PHSUBW128, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_phsubwv4hi3, "__builtin_ia32_phsubw", IX86_BUILTIN_PHSUBW, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_phsubdv4si3, "__builtin_ia32_phsubd128", IX86_BUILTIN_PHSUBD128, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_phsubdv2si3, "__builtin_ia32_phsubd", IX86_BUILTIN_PHSUBD, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_phsubswv8hi3, "__builtin_ia32_phsubsw128", IX86_BUILTIN_PHSUBSW128, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_phsubswv4hi3, "__builtin_ia32_phsubsw", IX86_BUILTIN_PHSUBSW, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_pmaddubswv8hi3, "__builtin_ia32_pmaddubsw128", IX86_BUILTIN_PMADDUBSW128, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_pmaddubswv4hi3, "__builtin_ia32_pmaddubsw", IX86_BUILTIN_PMADDUBSW, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_pmulhrswv8hi3, "__builtin_ia32_pmulhrsw128", IX86_BUILTIN_PMULHRSW128, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_pmulhrswv4hi3, "__builtin_ia32_pmulhrsw", IX86_BUILTIN_PMULHRSW, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_pshufbv16qi3, "__builtin_ia32_pshufb128", IX86_BUILTIN_PSHUFB128, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_pshufbv8qi3, "__builtin_ia32_pshufb", IX86_BUILTIN_PSHUFB, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_psignv16qi3, "__builtin_ia32_psignb128", IX86_BUILTIN_PSIGNB128, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_psignv8qi3, "__builtin_ia32_psignb", IX86_BUILTIN_PSIGNB, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_psignv8hi3, "__builtin_ia32_psignw128", IX86_BUILTIN_PSIGNW128, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_psignv4hi3, "__builtin_ia32_psignw", IX86_BUILTIN_PSIGNW, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_psignv4si3, "__builtin_ia32_psignd128", IX86_BUILTIN_PSIGND128, 0, 0 },
- { MASK_SSSE3, CODE_FOR_ssse3_psignv2si3, "__builtin_ia32_psignd", IX86_BUILTIN_PSIGND, 0, 0 }
+/* Builtin types */
+enum ix86_builtin_type
+{
+ FTYPE_UNKNOWN,
+ FLOAT128_FTYPE_FLOAT128,
+ FLOAT_FTYPE_FLOAT,
+ FLOAT128_FTYPE_FLOAT128_FLOAT128,
+ INT_FTYPE_V2DI_V2DI_PTEST,
+ INT64_FTYPE_V4SF,
+ INT64_FTYPE_V2DF,
+ INT_FTYPE_V16QI,
+ INT_FTYPE_V8QI,
+ INT_FTYPE_V4SF,
+ INT_FTYPE_V2DF,
+ V16QI_FTYPE_V16QI,
+ V8HI_FTYPE_V8HI,
+ V8HI_FTYPE_V16QI,
+ V8QI_FTYPE_V8QI,
+ V4SI_FTYPE_V4SI,
+ V4SI_FTYPE_V16QI,
+ V4SI_FTYPE_V8HI,
+ V4SI_FTYPE_V4SF,
+ V4SI_FTYPE_V2DF,
+ V4HI_FTYPE_V4HI,
+ V4SF_FTYPE_V4SF,
+ V4SF_FTYPE_V4SF_VEC_MERGE,
+ V4SF_FTYPE_V4SI,
+ V4SF_FTYPE_V2DF,
+ V2DI_FTYPE_V2DI,
+ V2DI_FTYPE_V16QI,
+ V2DI_FTYPE_V8HI,
+ V2DI_FTYPE_V4SI,
+ V2DF_FTYPE_V2DF,
+ V2DF_FTYPE_V2DF_VEC_MERGE,
+ V2DF_FTYPE_V4SI,
+ V2DF_FTYPE_V4SF,
+ V2DF_FTYPE_V2SI,
+ V2SI_FTYPE_V2SI,
+ V2SI_FTYPE_V4SF,
+ V2SI_FTYPE_V2SF,
+ V2SI_FTYPE_V2DF,
+ V2SF_FTYPE_V2SF,
+ V2SF_FTYPE_V2SI,
+ V16QI_FTYPE_V16QI_V16QI,
+ V16QI_FTYPE_V8HI_V8HI,
+ V8QI_FTYPE_V8QI_V8QI,
+ V8QI_FTYPE_V4HI_V4HI,
+ V8HI_FTYPE_V8HI_V8HI,
+ V8HI_FTYPE_V8HI_V8HI_COUNT,
+ V8HI_FTYPE_V16QI_V16QI,
+ V8HI_FTYPE_V4SI_V4SI,
+ V8HI_FTYPE_V8HI_SI_COUNT,
+ V4SI_FTYPE_V4SI_V4SI,
+ V4SI_FTYPE_V4SI_V4SI_COUNT,
+ V4SI_FTYPE_V8HI_V8HI,
+ V4SI_FTYPE_V4SF_V4SF,
+ V4SI_FTYPE_V2DF_V2DF,
+ V4SI_FTYPE_V4SI_SI_COUNT,
+ V4HI_FTYPE_V4HI_V4HI,
+ V4HI_FTYPE_V4HI_V4HI_COUNT,
+ V4HI_FTYPE_V8QI_V8QI,
+ V4HI_FTYPE_V2SI_V2SI,
+ V4HI_FTYPE_V4HI_SI_COUNT,
+ V4SF_FTYPE_V4SF_V4SF,
+ V4SF_FTYPE_V4SF_V4SF_SWAP,
+ V4SF_FTYPE_V4SF_V2SI,
+ V4SF_FTYPE_V4SF_V2DF,
+ V4SF_FTYPE_V4SF_DI,
+ V4SF_FTYPE_V4SF_SI,
+ V2DI_FTYPE_V2DI_V2DI,
+ V2DI_FTYPE_V2DI_V2DI_COUNT,
+ V2DI_FTYPE_V16QI_V16QI,
+ V2DI_FTYPE_V4SI_V4SI,
+ V2DI_FTYPE_V2DI_V16QI,
+ V2DI_FTYPE_V2DF_V2DF,
+ V2DI_FTYPE_V2DI_SI_COUNT,
+ V2SI_FTYPE_V2SI_V2SI,
+ V2SI_FTYPE_V2SI_V2SI_COUNT,
+ V2SI_FTYPE_V4HI_V4HI,
+ V2SI_FTYPE_V2SF_V2SF,
+ V2SI_FTYPE_V2SI_SI_COUNT,
+ V2DF_FTYPE_V2DF_V2DF,
+ V2DF_FTYPE_V2DF_V2DF_SWAP,
+ V2DF_FTYPE_V2DF_V4SF,
+ V2DF_FTYPE_V2DF_DI,
+ V2DF_FTYPE_V2DF_SI,
+ V2SF_FTYPE_V2SF_V2SF,
+ V1DI_FTYPE_V1DI_V1DI,
+ V1DI_FTYPE_V1DI_V1DI_COUNT,
+ V1DI_FTYPE_V8QI_V8QI,
+ V1DI_FTYPE_V2SI_V2SI,
+ V1DI_FTYPE_V1DI_SI_COUNT,
+ UINT64_FTYPE_UINT64_UINT64,
+ UINT_FTYPE_UINT_UINT,
+ UINT_FTYPE_UINT_USHORT,
+ UINT_FTYPE_UINT_UCHAR,
+ V8HI_FTYPE_V8HI_INT,
+ V4SI_FTYPE_V4SI_INT,
+ V4HI_FTYPE_V4HI_INT,
+ V4SF_FTYPE_V4SF_INT,
+ V2DI_FTYPE_V2DI_INT,
+ V2DI2TI_FTYPE_V2DI_INT,
+ V2DF_FTYPE_V2DF_INT,
+ V16QI_FTYPE_V16QI_V16QI_V16QI,
+ V4SF_FTYPE_V4SF_V4SF_V4SF,
+ V2DF_FTYPE_V2DF_V2DF_V2DF,
+ V16QI_FTYPE_V16QI_V16QI_INT,
+ V8HI_FTYPE_V8HI_V8HI_INT,
+ V4SI_FTYPE_V4SI_V4SI_INT,
+ V4SF_FTYPE_V4SF_V4SF_INT,
+ V2DI_FTYPE_V2DI_V2DI_INT,
+ V2DI2TI_FTYPE_V2DI_V2DI_INT,
+ V1DI2DI_FTYPE_V1DI_V1DI_INT,
+ V2DF_FTYPE_V2DF_V2DF_INT,
+ V2DI_FTYPE_V2DI_UINT_UINT,
+ V2DI_FTYPE_V2DI_V2DI_UINT_UINT
};
-static const struct builtin_description bdesc_1arg[] =
+/* Special builtins with variable number of arguments. */
+static const struct builtin_description bdesc_special_args[] =
{
- { MASK_SSE | MASK_3DNOW_A, CODE_FOR_mmx_pmovmskb, 0, IX86_BUILTIN_PMOVMSKB, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_movmskps, 0, IX86_BUILTIN_MOVMSKPS, 0, 0 },
+ /* MMX */
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_emms, "__builtin_ia32_emms", IX86_BUILTIN_EMMS, UNKNOWN, (int) VOID_FTYPE_VOID },
+
+ /* 3DNow! */
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_femms, "__builtin_ia32_femms", IX86_BUILTIN_FEMMS, UNKNOWN, (int) VOID_FTYPE_VOID },
- { MASK_SSE, CODE_FOR_sqrtv4sf2, 0, IX86_BUILTIN_SQRTPS, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_rsqrtv4sf2, 0, IX86_BUILTIN_RSQRTPS, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_rcpv4sf2, 0, IX86_BUILTIN_RCPPS, 0, 0 },
+ /* SSE */
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_movups, "__builtin_ia32_storeups", IX86_BUILTIN_STOREUPS, UNKNOWN, (int) VOID_FTYPE_PFLOAT_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_movntv4sf, "__builtin_ia32_movntps", IX86_BUILTIN_MOVNTPS, UNKNOWN, (int) VOID_FTYPE_PFLOAT_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_movups, "__builtin_ia32_loadups", IX86_BUILTIN_LOADUPS, UNKNOWN, (int) V4SF_FTYPE_PCFLOAT },
+
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_loadhps_exp, "__builtin_ia32_loadhps", IX86_BUILTIN_LOADHPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_PCV2SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_loadlps_exp, "__builtin_ia32_loadlps", IX86_BUILTIN_LOADLPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_PCV2SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_storehps, "__builtin_ia32_storehps", IX86_BUILTIN_STOREHPS, UNKNOWN, (int) VOID_FTYPE_PV2SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_storelps, "__builtin_ia32_storelps", IX86_BUILTIN_STORELPS, UNKNOWN, (int) VOID_FTYPE_PV2SF_V4SF },
- { MASK_SSE, CODE_FOR_sse_cvtps2pi, 0, IX86_BUILTIN_CVTPS2PI, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_cvtss2si, 0, IX86_BUILTIN_CVTSS2SI, 0, 0 },
- { MASK_SSE | MASK_64BIT, CODE_FOR_sse_cvtss2siq, 0, IX86_BUILTIN_CVTSS2SI64, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_cvttps2pi, 0, IX86_BUILTIN_CVTTPS2PI, 0, 0 },
- { MASK_SSE, CODE_FOR_sse_cvttss2si, 0, IX86_BUILTIN_CVTTSS2SI, 0, 0 },
- { MASK_SSE | MASK_64BIT, CODE_FOR_sse_cvttss2siq, 0, IX86_BUILTIN_CVTTSS2SI64, 0, 0 },
+ /* SSE or 3DNow!A */
+ { OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_3DNOW_A, CODE_FOR_sse_sfence, "__builtin_ia32_sfence", IX86_BUILTIN_SFENCE, UNKNOWN, (int) VOID_FTYPE_VOID },
+ { OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_3DNOW_A, CODE_FOR_sse_movntdi, "__builtin_ia32_movntq", IX86_BUILTIN_MOVNTQ, UNKNOWN, (int) VOID_FTYPE_PDI_DI },
+
+ /* SSE2 */
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_lfence, "__builtin_ia32_lfence", IX86_BUILTIN_LFENCE, UNKNOWN, (int) VOID_FTYPE_VOID },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_mfence, 0, IX86_BUILTIN_MFENCE, UNKNOWN, (int) VOID_FTYPE_VOID },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_movupd, "__builtin_ia32_storeupd", IX86_BUILTIN_STOREUPD, UNKNOWN, (int) VOID_FTYPE_PDOUBLE_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_movdqu, "__builtin_ia32_storedqu", IX86_BUILTIN_STOREDQU, UNKNOWN, (int) VOID_FTYPE_PCHAR_V16QI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_movntv2df, "__builtin_ia32_movntpd", IX86_BUILTIN_MOVNTPD, UNKNOWN, (int) VOID_FTYPE_PDOUBLE_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_movntv2di, "__builtin_ia32_movntdq", IX86_BUILTIN_MOVNTDQ, UNKNOWN, (int) VOID_FTYPE_PV2DI_V2DI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_movntsi, "__builtin_ia32_movnti", IX86_BUILTIN_MOVNTI, UNKNOWN, (int) VOID_FTYPE_PINT_INT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_movupd, "__builtin_ia32_loadupd", IX86_BUILTIN_LOADUPD, UNKNOWN, (int) V2DF_FTYPE_PCDOUBLE },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_movdqu, "__builtin_ia32_loaddqu", IX86_BUILTIN_LOADDQU, UNKNOWN, (int) V16QI_FTYPE_PCCHAR },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_loadhpd_exp, "__builtin_ia32_loadhpd", IX86_BUILTIN_LOADHPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_PCDOUBLE },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_loadlpd_exp, "__builtin_ia32_loadlpd", IX86_BUILTIN_LOADLPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_PCDOUBLE },
- { MASK_SSE2, CODE_FOR_sse2_pmovmskb, 0, IX86_BUILTIN_PMOVMSKB128, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_movmskpd, 0, IX86_BUILTIN_MOVMSKPD, 0, 0 },
+ /* SSE3 */
+ { OPTION_MASK_ISA_SSE3, CODE_FOR_sse3_lddqu, "__builtin_ia32_lddqu", IX86_BUILTIN_LDDQU, UNKNOWN, (int) V16QI_FTYPE_PCCHAR },
- { MASK_SSE2, CODE_FOR_sqrtv2df2, 0, IX86_BUILTIN_SQRTPD, 0, 0 },
+ /* SSE4.1 */
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_movntdqa, "__builtin_ia32_movntdqa", IX86_BUILTIN_MOVNTDQA, UNKNOWN, (int) V2DI_FTYPE_PV2DI },
- { MASK_SSE2, CODE_FOR_sse2_cvtdq2pd, 0, IX86_BUILTIN_CVTDQ2PD, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_cvtdq2ps, 0, IX86_BUILTIN_CVTDQ2PS, 0, 0 },
+ /* SSE4A */
+ { OPTION_MASK_ISA_SSE4A, CODE_FOR_sse4a_vmmovntv2df, "__builtin_ia32_movntsd", IX86_BUILTIN_MOVNTSD, UNKNOWN, (int) VOID_FTYPE_PDOUBLE_V2DF },
+ { OPTION_MASK_ISA_SSE4A, CODE_FOR_sse4a_vmmovntv4sf, "__builtin_ia32_movntss", IX86_BUILTIN_MOVNTSS, UNKNOWN, (int) VOID_FTYPE_PFLOAT_V4SF },
+};
- { MASK_SSE2, CODE_FOR_sse2_cvtpd2dq, 0, IX86_BUILTIN_CVTPD2DQ, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_cvtpd2pi, 0, IX86_BUILTIN_CVTPD2PI, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_cvtpd2ps, 0, IX86_BUILTIN_CVTPD2PS, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_cvttpd2dq, 0, IX86_BUILTIN_CVTTPD2DQ, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_cvttpd2pi, 0, IX86_BUILTIN_CVTTPD2PI, 0, 0 },
+/* Builtins with variable number of arguments. */
+static const struct builtin_description bdesc_args[] =
+{
+ /* MMX */
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_addv8qi3, "__builtin_ia32_paddb", IX86_BUILTIN_PADDB, UNKNOWN, (int) V8QI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_addv4hi3, "__builtin_ia32_paddw", IX86_BUILTIN_PADDW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_addv2si3, "__builtin_ia32_paddd", IX86_BUILTIN_PADDD, UNKNOWN, (int) V2SI_FTYPE_V2SI_V2SI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_subv8qi3, "__builtin_ia32_psubb", IX86_BUILTIN_PSUBB, UNKNOWN, (int) V8QI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_subv4hi3, "__builtin_ia32_psubw", IX86_BUILTIN_PSUBW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_subv2si3, "__builtin_ia32_psubd", IX86_BUILTIN_PSUBD, UNKNOWN, (int) V2SI_FTYPE_V2SI_V2SI },
+
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_ssaddv8qi3, "__builtin_ia32_paddsb", IX86_BUILTIN_PADDSB, UNKNOWN, (int) V8QI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_ssaddv4hi3, "__builtin_ia32_paddsw", IX86_BUILTIN_PADDSW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_sssubv8qi3, "__builtin_ia32_psubsb", IX86_BUILTIN_PSUBSB, UNKNOWN, (int) V8QI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_sssubv4hi3, "__builtin_ia32_psubsw", IX86_BUILTIN_PSUBSW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_usaddv8qi3, "__builtin_ia32_paddusb", IX86_BUILTIN_PADDUSB, UNKNOWN, (int) V8QI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_usaddv4hi3, "__builtin_ia32_paddusw", IX86_BUILTIN_PADDUSW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_ussubv8qi3, "__builtin_ia32_psubusb", IX86_BUILTIN_PSUBUSB, UNKNOWN, (int) V8QI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_ussubv4hi3, "__builtin_ia32_psubusw", IX86_BUILTIN_PSUBUSW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_mulv4hi3, "__builtin_ia32_pmullw", IX86_BUILTIN_PMULLW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_smulv4hi3_highpart, "__builtin_ia32_pmulhw", IX86_BUILTIN_PMULHW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_andv2si3, "__builtin_ia32_pand", IX86_BUILTIN_PAND, UNKNOWN, (int) V2SI_FTYPE_V2SI_V2SI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_nandv2si3, "__builtin_ia32_pandn", IX86_BUILTIN_PANDN, UNKNOWN, (int) V2SI_FTYPE_V2SI_V2SI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_iorv2si3, "__builtin_ia32_por", IX86_BUILTIN_POR, UNKNOWN, (int) V2SI_FTYPE_V2SI_V2SI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_xorv2si3, "__builtin_ia32_pxor", IX86_BUILTIN_PXOR, UNKNOWN, (int) V2SI_FTYPE_V2SI_V2SI },
+
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_eqv8qi3, "__builtin_ia32_pcmpeqb", IX86_BUILTIN_PCMPEQB, UNKNOWN, (int) V8QI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_eqv4hi3, "__builtin_ia32_pcmpeqw", IX86_BUILTIN_PCMPEQW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_eqv2si3, "__builtin_ia32_pcmpeqd", IX86_BUILTIN_PCMPEQD, UNKNOWN, (int) V2SI_FTYPE_V2SI_V2SI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_gtv8qi3, "__builtin_ia32_pcmpgtb", IX86_BUILTIN_PCMPGTB, UNKNOWN, (int) V8QI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_gtv4hi3, "__builtin_ia32_pcmpgtw", IX86_BUILTIN_PCMPGTW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_gtv2si3, "__builtin_ia32_pcmpgtd", IX86_BUILTIN_PCMPGTD, UNKNOWN, (int) V2SI_FTYPE_V2SI_V2SI },
+
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_punpckhbw, "__builtin_ia32_punpckhbw", IX86_BUILTIN_PUNPCKHBW, UNKNOWN, (int) V8QI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_punpckhwd, "__builtin_ia32_punpckhwd", IX86_BUILTIN_PUNPCKHWD, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_punpckhdq, "__builtin_ia32_punpckhdq", IX86_BUILTIN_PUNPCKHDQ, UNKNOWN, (int) V2SI_FTYPE_V2SI_V2SI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_punpcklbw, "__builtin_ia32_punpcklbw", IX86_BUILTIN_PUNPCKLBW, UNKNOWN, (int) V8QI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_punpcklwd, "__builtin_ia32_punpcklwd", IX86_BUILTIN_PUNPCKLWD, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI},
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_punpckldq, "__builtin_ia32_punpckldq", IX86_BUILTIN_PUNPCKLDQ, UNKNOWN, (int) V2SI_FTYPE_V2SI_V2SI},
+
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_packsswb, "__builtin_ia32_packsswb", IX86_BUILTIN_PACKSSWB, UNKNOWN, (int) V8QI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_packssdw, "__builtin_ia32_packssdw", IX86_BUILTIN_PACKSSDW, UNKNOWN, (int) V4HI_FTYPE_V2SI_V2SI },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_packuswb, "__builtin_ia32_packuswb", IX86_BUILTIN_PACKUSWB, UNKNOWN, (int) V8QI_FTYPE_V4HI_V4HI },
+
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_pmaddwd, "__builtin_ia32_pmaddwd", IX86_BUILTIN_PMADDWD, UNKNOWN, (int) V2SI_FTYPE_V4HI_V4HI },
+
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_ashlv4hi3, "__builtin_ia32_psllwi", IX86_BUILTIN_PSLLWI, UNKNOWN, (int) V4HI_FTYPE_V4HI_SI_COUNT },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_ashlv2si3, "__builtin_ia32_pslldi", IX86_BUILTIN_PSLLDI, UNKNOWN, (int) V2SI_FTYPE_V2SI_SI_COUNT },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_ashlv1di3, "__builtin_ia32_psllqi", IX86_BUILTIN_PSLLQI, UNKNOWN, (int) V1DI_FTYPE_V1DI_SI_COUNT },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_ashlv4hi3, "__builtin_ia32_psllw", IX86_BUILTIN_PSLLW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI_COUNT },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_ashlv2si3, "__builtin_ia32_pslld", IX86_BUILTIN_PSLLD, UNKNOWN, (int) V2SI_FTYPE_V2SI_V2SI_COUNT },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_ashlv1di3, "__builtin_ia32_psllq", IX86_BUILTIN_PSLLQ, UNKNOWN, (int) V1DI_FTYPE_V1DI_V1DI_COUNT },
+
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_lshrv4hi3, "__builtin_ia32_psrlwi", IX86_BUILTIN_PSRLWI, UNKNOWN, (int) V4HI_FTYPE_V4HI_SI_COUNT },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_lshrv2si3, "__builtin_ia32_psrldi", IX86_BUILTIN_PSRLDI, UNKNOWN, (int) V2SI_FTYPE_V2SI_SI_COUNT },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_lshrv1di3, "__builtin_ia32_psrlqi", IX86_BUILTIN_PSRLQI, UNKNOWN, (int) V1DI_FTYPE_V1DI_SI_COUNT },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_lshrv4hi3, "__builtin_ia32_psrlw", IX86_BUILTIN_PSRLW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI_COUNT },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_lshrv2si3, "__builtin_ia32_psrld", IX86_BUILTIN_PSRLD, UNKNOWN, (int) V2SI_FTYPE_V2SI_V2SI_COUNT },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_lshrv1di3, "__builtin_ia32_psrlq", IX86_BUILTIN_PSRLQ, UNKNOWN, (int) V1DI_FTYPE_V1DI_V1DI_COUNT },
+
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_ashrv4hi3, "__builtin_ia32_psrawi", IX86_BUILTIN_PSRAWI, UNKNOWN, (int) V4HI_FTYPE_V4HI_SI_COUNT },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_ashrv2si3, "__builtin_ia32_psradi", IX86_BUILTIN_PSRADI, UNKNOWN, (int) V2SI_FTYPE_V2SI_SI_COUNT },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_ashrv4hi3, "__builtin_ia32_psraw", IX86_BUILTIN_PSRAW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI_COUNT },
+ { OPTION_MASK_ISA_MMX, CODE_FOR_mmx_ashrv2si3, "__builtin_ia32_psrad", IX86_BUILTIN_PSRAD, UNKNOWN, (int) V2SI_FTYPE_V2SI_V2SI_COUNT },
+
+ /* 3DNow! */
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_pf2id, "__builtin_ia32_pf2id", IX86_BUILTIN_PF2ID, UNKNOWN, (int) V2SI_FTYPE_V2SF },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_floatv2si2, "__builtin_ia32_pi2fd", IX86_BUILTIN_PI2FD, UNKNOWN, (int) V2SF_FTYPE_V2SI },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_rcpv2sf2, "__builtin_ia32_pfrcp", IX86_BUILTIN_PFRCP, UNKNOWN, (int) V2SF_FTYPE_V2SF },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_rsqrtv2sf2, "__builtin_ia32_pfrsqrt", IX86_BUILTIN_PFRSQRT, UNKNOWN, (int) V2SF_FTYPE_V2SF },
+
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_uavgv8qi3, "__builtin_ia32_pavgusb", IX86_BUILTIN_PAVGUSB, UNKNOWN, (int) V8QI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_haddv2sf3, "__builtin_ia32_pfacc", IX86_BUILTIN_PFACC, UNKNOWN, (int) V2SF_FTYPE_V2SF_V2SF },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_addv2sf3, "__builtin_ia32_pfadd", IX86_BUILTIN_PFADD, UNKNOWN, (int) V2SF_FTYPE_V2SF_V2SF },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_eqv2sf3, "__builtin_ia32_pfcmpeq", IX86_BUILTIN_PFCMPEQ, UNKNOWN, (int) V2SI_FTYPE_V2SF_V2SF },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_gev2sf3, "__builtin_ia32_pfcmpge", IX86_BUILTIN_PFCMPGE, UNKNOWN, (int) V2SI_FTYPE_V2SF_V2SF },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_gtv2sf3, "__builtin_ia32_pfcmpgt", IX86_BUILTIN_PFCMPGT, UNKNOWN, (int) V2SI_FTYPE_V2SF_V2SF },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_smaxv2sf3, "__builtin_ia32_pfmax", IX86_BUILTIN_PFMAX, UNKNOWN, (int) V2SF_FTYPE_V2SF_V2SF },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_sminv2sf3, "__builtin_ia32_pfmin", IX86_BUILTIN_PFMIN, UNKNOWN, (int) V2SF_FTYPE_V2SF_V2SF },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_mulv2sf3, "__builtin_ia32_pfmul", IX86_BUILTIN_PFMUL, UNKNOWN, (int) V2SF_FTYPE_V2SF_V2SF },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_rcpit1v2sf3, "__builtin_ia32_pfrcpit1", IX86_BUILTIN_PFRCPIT1, UNKNOWN, (int) V2SF_FTYPE_V2SF_V2SF },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_rcpit2v2sf3, "__builtin_ia32_pfrcpit2", IX86_BUILTIN_PFRCPIT2, UNKNOWN, (int) V2SF_FTYPE_V2SF_V2SF },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_rsqit1v2sf3, "__builtin_ia32_pfrsqit1", IX86_BUILTIN_PFRSQIT1, UNKNOWN, (int) V2SF_FTYPE_V2SF_V2SF },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_subv2sf3, "__builtin_ia32_pfsub", IX86_BUILTIN_PFSUB, UNKNOWN, (int) V2SF_FTYPE_V2SF_V2SF },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_subrv2sf3, "__builtin_ia32_pfsubr", IX86_BUILTIN_PFSUBR, UNKNOWN, (int) V2SF_FTYPE_V2SF_V2SF },
+ { OPTION_MASK_ISA_3DNOW, CODE_FOR_mmx_pmulhrwv4hi3, "__builtin_ia32_pmulhrw", IX86_BUILTIN_PMULHRW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+
+ /* 3DNow!A */
+ { OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_pf2iw, "__builtin_ia32_pf2iw", IX86_BUILTIN_PF2IW, UNKNOWN, (int) V2SI_FTYPE_V2SF },
+ { OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_pi2fw, "__builtin_ia32_pi2fw", IX86_BUILTIN_PI2FW, UNKNOWN, (int) V2SF_FTYPE_V2SI },
+ { OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_pswapdv2si2, "__builtin_ia32_pswapdsi", IX86_BUILTIN_PSWAPDSI, UNKNOWN, (int) V2SI_FTYPE_V2SI },
+ { OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_pswapdv2sf2, "__builtin_ia32_pswapdsf", IX86_BUILTIN_PSWAPDSF, UNKNOWN, (int) V2SF_FTYPE_V2SF },
+ { OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_hsubv2sf3, "__builtin_ia32_pfnacc", IX86_BUILTIN_PFNACC, UNKNOWN, (int) V2SF_FTYPE_V2SF_V2SF },
+ { OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_addsubv2sf3, "__builtin_ia32_pfpnacc", IX86_BUILTIN_PFPNACC, UNKNOWN, (int) V2SF_FTYPE_V2SF_V2SF },
- { MASK_SSE2, CODE_FOR_sse2_cvtpi2pd, 0, IX86_BUILTIN_CVTPI2PD, 0, 0 },
+ /* SSE */
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_movmskps, "__builtin_ia32_movmskps", IX86_BUILTIN_MOVMSKPS, UNKNOWN, (int) INT_FTYPE_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_sqrtv4sf2, "__builtin_ia32_sqrtps", IX86_BUILTIN_SQRTPS, UNKNOWN, (int) V4SF_FTYPE_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sqrtv4sf2, "__builtin_ia32_sqrtps_nr", IX86_BUILTIN_SQRTPS_NR, UNKNOWN, (int) V4SF_FTYPE_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_rsqrtv4sf2, "__builtin_ia32_rsqrtps", IX86_BUILTIN_RSQRTPS, UNKNOWN, (int) V4SF_FTYPE_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_rsqrtv4sf2, "__builtin_ia32_rsqrtps_nr", IX86_BUILTIN_RSQRTPS_NR, UNKNOWN, (int) V4SF_FTYPE_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_rcpv4sf2, "__builtin_ia32_rcpps", IX86_BUILTIN_RCPPS, UNKNOWN, (int) V4SF_FTYPE_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_cvtps2pi, "__builtin_ia32_cvtps2pi", IX86_BUILTIN_CVTPS2PI, UNKNOWN, (int) V2SI_FTYPE_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_cvtss2si, "__builtin_ia32_cvtss2si", IX86_BUILTIN_CVTSS2SI, UNKNOWN, (int) INT_FTYPE_V4SF },
+ { OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_64BIT, CODE_FOR_sse_cvtss2siq, "__builtin_ia32_cvtss2si64", IX86_BUILTIN_CVTSS2SI64, UNKNOWN, (int) INT64_FTYPE_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_cvttps2pi, "__builtin_ia32_cvttps2pi", IX86_BUILTIN_CVTTPS2PI, UNKNOWN, (int) V2SI_FTYPE_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_cvttss2si, "__builtin_ia32_cvttss2si", IX86_BUILTIN_CVTTSS2SI, UNKNOWN, (int) INT_FTYPE_V4SF },
+ { OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_64BIT, CODE_FOR_sse_cvttss2siq, "__builtin_ia32_cvttss2si64", IX86_BUILTIN_CVTTSS2SI64, UNKNOWN, (int) INT64_FTYPE_V4SF },
+
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_shufps, "__builtin_ia32_shufps", IX86_BUILTIN_SHUFPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF_INT },
+
+ { OPTION_MASK_ISA_SSE, CODE_FOR_addv4sf3, "__builtin_ia32_addps", IX86_BUILTIN_ADDPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_subv4sf3, "__builtin_ia32_subps", IX86_BUILTIN_SUBPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_mulv4sf3, "__builtin_ia32_mulps", IX86_BUILTIN_MULPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_divv4sf3, "__builtin_ia32_divps", IX86_BUILTIN_DIVPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmaddv4sf3, "__builtin_ia32_addss", IX86_BUILTIN_ADDSS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmsubv4sf3, "__builtin_ia32_subss", IX86_BUILTIN_SUBSS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmmulv4sf3, "__builtin_ia32_mulss", IX86_BUILTIN_MULSS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmdivv4sf3, "__builtin_ia32_divss", IX86_BUILTIN_DIVSS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpeqps", IX86_BUILTIN_CMPEQPS, EQ, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpltps", IX86_BUILTIN_CMPLTPS, LT, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpleps", IX86_BUILTIN_CMPLEPS, LE, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpgtps", IX86_BUILTIN_CMPGTPS, LT, (int) V4SF_FTYPE_V4SF_V4SF_SWAP },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpgeps", IX86_BUILTIN_CMPGEPS, LE, (int) V4SF_FTYPE_V4SF_V4SF_SWAP },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpunordps", IX86_BUILTIN_CMPUNORDPS, UNORDERED, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpneqps", IX86_BUILTIN_CMPNEQPS, NE, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpnltps", IX86_BUILTIN_CMPNLTPS, UNGE, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpnleps", IX86_BUILTIN_CMPNLEPS, UNGT, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpngtps", IX86_BUILTIN_CMPNGTPS, UNGE, (int) V4SF_FTYPE_V4SF_V4SF_SWAP },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpngeps", IX86_BUILTIN_CMPNGEPS, UNGT, (int) V4SF_FTYPE_V4SF_V4SF_SWAP},
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_maskcmpv4sf3, "__builtin_ia32_cmpordps", IX86_BUILTIN_CMPORDPS, ORDERED, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpeqss", IX86_BUILTIN_CMPEQSS, EQ, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpltss", IX86_BUILTIN_CMPLTSS, LT, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpless", IX86_BUILTIN_CMPLESS, LE, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpunordss", IX86_BUILTIN_CMPUNORDSS, UNORDERED, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpneqss", IX86_BUILTIN_CMPNEQSS, NE, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpnltss", IX86_BUILTIN_CMPNLTSS, UNGE, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpnless", IX86_BUILTIN_CMPNLESS, UNGT, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpngtss", IX86_BUILTIN_CMPNGTSS, UNGE, (int) V4SF_FTYPE_V4SF_V4SF_SWAP },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpngess", IX86_BUILTIN_CMPNGESS, UNGT, (int) V4SF_FTYPE_V4SF_V4SF_SWAP },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmmaskcmpv4sf3, "__builtin_ia32_cmpordss", IX86_BUILTIN_CMPORDSS, ORDERED, (int) V4SF_FTYPE_V4SF_V4SF },
+
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sminv4sf3, "__builtin_ia32_minps", IX86_BUILTIN_MINPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_smaxv4sf3, "__builtin_ia32_maxps", IX86_BUILTIN_MAXPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmsminv4sf3, "__builtin_ia32_minss", IX86_BUILTIN_MINSS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmsmaxv4sf3, "__builtin_ia32_maxss", IX86_BUILTIN_MAXSS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+
+ { OPTION_MASK_ISA_SSE, CODE_FOR_andv4sf3, "__builtin_ia32_andps", IX86_BUILTIN_ANDPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_nandv4sf3, "__builtin_ia32_andnps", IX86_BUILTIN_ANDNPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_iorv4sf3, "__builtin_ia32_orps", IX86_BUILTIN_ORPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_xorv4sf3, "__builtin_ia32_xorps", IX86_BUILTIN_XORPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_movss, "__builtin_ia32_movss", IX86_BUILTIN_MOVSS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_movhlps_exp, "__builtin_ia32_movhlps", IX86_BUILTIN_MOVHLPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_movlhps_exp, "__builtin_ia32_movlhps", IX86_BUILTIN_MOVLHPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_unpckhps, "__builtin_ia32_unpckhps", IX86_BUILTIN_UNPCKHPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_unpcklps, "__builtin_ia32_unpcklps", IX86_BUILTIN_UNPCKLPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_cvtpi2ps, "__builtin_ia32_cvtpi2ps", IX86_BUILTIN_CVTPI2PS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V2SI },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_cvtsi2ss, "__builtin_ia32_cvtsi2ss", IX86_BUILTIN_CVTSI2SS, UNKNOWN, (int) V4SF_FTYPE_V4SF_SI },
+ { OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_64BIT, CODE_FOR_sse_cvtsi2ssq, "__builtin_ia32_cvtsi642ss", IX86_BUILTIN_CVTSI642SS, UNKNOWN, V4SF_FTYPE_V4SF_DI },
+
+ { OPTION_MASK_ISA_SSE, CODE_FOR_rsqrtsf2, "__builtin_ia32_rsqrtf", IX86_BUILTIN_RSQRTF, UNKNOWN, (int) FLOAT_FTYPE_FLOAT },
+
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmsqrtv4sf2, "__builtin_ia32_sqrtss", IX86_BUILTIN_SQRTSS, UNKNOWN, (int) V4SF_FTYPE_V4SF_VEC_MERGE },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmrsqrtv4sf2, "__builtin_ia32_rsqrtss", IX86_BUILTIN_RSQRTSS, UNKNOWN, (int) V4SF_FTYPE_V4SF_VEC_MERGE },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmrcpv4sf2, "__builtin_ia32_rcpss", IX86_BUILTIN_RCPSS, UNKNOWN, (int) V4SF_FTYPE_V4SF_VEC_MERGE },
+
+ /* SSE MMX or 3Dnow!A */
+ { OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_uavgv8qi3, "__builtin_ia32_pavgb", IX86_BUILTIN_PAVGB, UNKNOWN, (int) V8QI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_uavgv4hi3, "__builtin_ia32_pavgw", IX86_BUILTIN_PAVGW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_umulv4hi3_highpart, "__builtin_ia32_pmulhuw", IX86_BUILTIN_PMULHUW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+
+ { OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_umaxv8qi3, "__builtin_ia32_pmaxub", IX86_BUILTIN_PMAXUB, UNKNOWN, (int) V8QI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_smaxv4hi3, "__builtin_ia32_pmaxsw", IX86_BUILTIN_PMAXSW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_uminv8qi3, "__builtin_ia32_pminub", IX86_BUILTIN_PMINUB, UNKNOWN, (int) V8QI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_sminv4hi3, "__builtin_ia32_pminsw", IX86_BUILTIN_PMINSW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+
+ { OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_psadbw, "__builtin_ia32_psadbw", IX86_BUILTIN_PSADBW, UNKNOWN, (int) V1DI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_pmovmskb, "__builtin_ia32_pmovmskb", IX86_BUILTIN_PMOVMSKB, UNKNOWN, (int) INT_FTYPE_V8QI },
+
+ { OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_pshufw, "__builtin_ia32_pshufw", IX86_BUILTIN_PSHUFW, UNKNOWN, (int) V4HI_FTYPE_V4HI_INT },
- { MASK_SSE2, CODE_FOR_sse2_cvtsd2si, 0, IX86_BUILTIN_CVTSD2SI, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_cvttsd2si, 0, IX86_BUILTIN_CVTTSD2SI, 0, 0 },
- { MASK_SSE2 | MASK_64BIT, CODE_FOR_sse2_cvtsd2siq, 0, IX86_BUILTIN_CVTSD2SI64, 0, 0 },
- { MASK_SSE2 | MASK_64BIT, CODE_FOR_sse2_cvttsd2siq, 0, IX86_BUILTIN_CVTTSD2SI64, 0, 0 },
+ /* SSE2 */
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_shufpd, "__builtin_ia32_shufpd", IX86_BUILTIN_SHUFPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF_INT },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_movmskpd, "__builtin_ia32_movmskpd", IX86_BUILTIN_MOVMSKPD, UNKNOWN, (int) INT_FTYPE_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_pmovmskb, "__builtin_ia32_pmovmskb128", IX86_BUILTIN_PMOVMSKB128, UNKNOWN, (int) INT_FTYPE_V16QI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sqrtv2df2, "__builtin_ia32_sqrtpd", IX86_BUILTIN_SQRTPD, UNKNOWN, (int) V2DF_FTYPE_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_cvtdq2pd, "__builtin_ia32_cvtdq2pd", IX86_BUILTIN_CVTDQ2PD, UNKNOWN, (int) V2DF_FTYPE_V4SI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_cvtdq2ps, "__builtin_ia32_cvtdq2ps", IX86_BUILTIN_CVTDQ2PS, UNKNOWN, (int) V4SF_FTYPE_V4SI },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_cvtpd2dq, "__builtin_ia32_cvtpd2dq", IX86_BUILTIN_CVTPD2DQ, UNKNOWN, (int) V4SI_FTYPE_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_cvtpd2pi, "__builtin_ia32_cvtpd2pi", IX86_BUILTIN_CVTPD2PI, UNKNOWN, (int) V2SI_FTYPE_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_cvtpd2ps, "__builtin_ia32_cvtpd2ps", IX86_BUILTIN_CVTPD2PS, UNKNOWN, (int) V4SF_FTYPE_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_cvttpd2dq, "__builtin_ia32_cvttpd2dq", IX86_BUILTIN_CVTTPD2DQ, UNKNOWN, (int) V4SI_FTYPE_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_cvttpd2pi, "__builtin_ia32_cvttpd2pi", IX86_BUILTIN_CVTTPD2PI, UNKNOWN, (int) V2SI_FTYPE_V2DF },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_cvtpi2pd, "__builtin_ia32_cvtpi2pd", IX86_BUILTIN_CVTPI2PD, UNKNOWN, (int) V2DF_FTYPE_V2SI },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_cvtsd2si, "__builtin_ia32_cvtsd2si", IX86_BUILTIN_CVTSD2SI, UNKNOWN, (int) INT_FTYPE_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_cvttsd2si, "__builtin_ia32_cvttsd2si", IX86_BUILTIN_CVTTSD2SI, UNKNOWN, (int) INT_FTYPE_V2DF },
+ { OPTION_MASK_ISA_SSE2 | OPTION_MASK_ISA_64BIT, CODE_FOR_sse2_cvtsd2siq, "__builtin_ia32_cvtsd2si64", IX86_BUILTIN_CVTSD2SI64, UNKNOWN, (int) INT64_FTYPE_V2DF },
+ { OPTION_MASK_ISA_SSE2 | OPTION_MASK_ISA_64BIT, CODE_FOR_sse2_cvttsd2siq, "__builtin_ia32_cvttsd2si64", IX86_BUILTIN_CVTTSD2SI64, UNKNOWN, (int) INT64_FTYPE_V2DF },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_cvtps2dq, "__builtin_ia32_cvtps2dq", IX86_BUILTIN_CVTPS2DQ, UNKNOWN, (int) V4SI_FTYPE_V4SF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_cvtps2pd, "__builtin_ia32_cvtps2pd", IX86_BUILTIN_CVTPS2PD, UNKNOWN, (int) V2DF_FTYPE_V4SF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_cvttps2dq, "__builtin_ia32_cvttps2dq", IX86_BUILTIN_CVTTPS2DQ, UNKNOWN, (int) V4SI_FTYPE_V4SF },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_addv2df3, "__builtin_ia32_addpd", IX86_BUILTIN_ADDPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_subv2df3, "__builtin_ia32_subpd", IX86_BUILTIN_SUBPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_mulv2df3, "__builtin_ia32_mulpd", IX86_BUILTIN_MULPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_divv2df3, "__builtin_ia32_divpd", IX86_BUILTIN_DIVPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_vmaddv2df3, "__builtin_ia32_addsd", IX86_BUILTIN_ADDSD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_vmsubv2df3, "__builtin_ia32_subsd", IX86_BUILTIN_SUBSD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_vmmulv2df3, "__builtin_ia32_mulsd", IX86_BUILTIN_MULSD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_vmdivv2df3, "__builtin_ia32_divsd", IX86_BUILTIN_DIVSD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpeqpd", IX86_BUILTIN_CMPEQPD, EQ, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpltpd", IX86_BUILTIN_CMPLTPD, LT, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmplepd", IX86_BUILTIN_CMPLEPD, LE, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpgtpd", IX86_BUILTIN_CMPGTPD, LT, (int) V2DF_FTYPE_V2DF_V2DF_SWAP },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpgepd", IX86_BUILTIN_CMPGEPD, LE, (int) V2DF_FTYPE_V2DF_V2DF_SWAP},
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpunordpd", IX86_BUILTIN_CMPUNORDPD, UNORDERED, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpneqpd", IX86_BUILTIN_CMPNEQPD, NE, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpnltpd", IX86_BUILTIN_CMPNLTPD, UNGE, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpnlepd", IX86_BUILTIN_CMPNLEPD, UNGT, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpngtpd", IX86_BUILTIN_CMPNGTPD, UNGE, (int) V2DF_FTYPE_V2DF_V2DF_SWAP },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpngepd", IX86_BUILTIN_CMPNGEPD, UNGT, (int) V2DF_FTYPE_V2DF_V2DF_SWAP },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_maskcmpv2df3, "__builtin_ia32_cmpordpd", IX86_BUILTIN_CMPORDPD, ORDERED, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_vmmaskcmpv2df3, "__builtin_ia32_cmpeqsd", IX86_BUILTIN_CMPEQSD, EQ, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_vmmaskcmpv2df3, "__builtin_ia32_cmpltsd", IX86_BUILTIN_CMPLTSD, LT, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_vmmaskcmpv2df3, "__builtin_ia32_cmplesd", IX86_BUILTIN_CMPLESD, LE, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_vmmaskcmpv2df3, "__builtin_ia32_cmpunordsd", IX86_BUILTIN_CMPUNORDSD, UNORDERED, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_vmmaskcmpv2df3, "__builtin_ia32_cmpneqsd", IX86_BUILTIN_CMPNEQSD, NE, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_vmmaskcmpv2df3, "__builtin_ia32_cmpnltsd", IX86_BUILTIN_CMPNLTSD, UNGE, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_vmmaskcmpv2df3, "__builtin_ia32_cmpnlesd", IX86_BUILTIN_CMPNLESD, UNGT, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_vmmaskcmpv2df3, "__builtin_ia32_cmpordsd", IX86_BUILTIN_CMPORDSD, ORDERED, (int) V2DF_FTYPE_V2DF_V2DF },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sminv2df3, "__builtin_ia32_minpd", IX86_BUILTIN_MINPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_smaxv2df3, "__builtin_ia32_maxpd", IX86_BUILTIN_MAXPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_vmsminv2df3, "__builtin_ia32_minsd", IX86_BUILTIN_MINSD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_vmsmaxv2df3, "__builtin_ia32_maxsd", IX86_BUILTIN_MAXSD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_andv2df3, "__builtin_ia32_andpd", IX86_BUILTIN_ANDPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_nandv2df3, "__builtin_ia32_andnpd", IX86_BUILTIN_ANDNPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_iorv2df3, "__builtin_ia32_orpd", IX86_BUILTIN_ORPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_xorv2df3, "__builtin_ia32_xorpd", IX86_BUILTIN_XORPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_movsd, "__builtin_ia32_movsd", IX86_BUILTIN_MOVSD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_unpckhpd_exp, "__builtin_ia32_unpckhpd", IX86_BUILTIN_UNPCKHPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_unpcklpd_exp, "__builtin_ia32_unpcklpd", IX86_BUILTIN_UNPCKLPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_vec_pack_sfix_v2df, "__builtin_ia32_vec_pack_sfix", IX86_BUILTIN_VEC_PACK_SFIX, UNKNOWN, (int) V4SI_FTYPE_V2DF_V2DF },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_addv16qi3, "__builtin_ia32_paddb128", IX86_BUILTIN_PADDB128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_addv8hi3, "__builtin_ia32_paddw128", IX86_BUILTIN_PADDW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_addv4si3, "__builtin_ia32_paddd128", IX86_BUILTIN_PADDD128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_addv2di3, "__builtin_ia32_paddq128", IX86_BUILTIN_PADDQ128, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_subv16qi3, "__builtin_ia32_psubb128", IX86_BUILTIN_PSUBB128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_subv8hi3, "__builtin_ia32_psubw128", IX86_BUILTIN_PSUBW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_subv4si3, "__builtin_ia32_psubd128", IX86_BUILTIN_PSUBD128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_subv2di3, "__builtin_ia32_psubq128", IX86_BUILTIN_PSUBQ128, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_ssaddv16qi3, "__builtin_ia32_paddsb128", IX86_BUILTIN_PADDSB128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_ssaddv8hi3, "__builtin_ia32_paddsw128", IX86_BUILTIN_PADDSW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_sssubv16qi3, "__builtin_ia32_psubsb128", IX86_BUILTIN_PSUBSB128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_sssubv8hi3, "__builtin_ia32_psubsw128", IX86_BUILTIN_PSUBSW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_usaddv16qi3, "__builtin_ia32_paddusb128", IX86_BUILTIN_PADDUSB128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_usaddv8hi3, "__builtin_ia32_paddusw128", IX86_BUILTIN_PADDUSW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_ussubv16qi3, "__builtin_ia32_psubusb128", IX86_BUILTIN_PSUBUSB128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_ussubv8hi3, "__builtin_ia32_psubusw128", IX86_BUILTIN_PSUBUSW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_mulv8hi3, "__builtin_ia32_pmullw128", IX86_BUILTIN_PMULLW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_smulv8hi3_highpart, "__builtin_ia32_pmulhw128", IX86_BUILTIN_PMULHW128, UNKNOWN,(int) V8HI_FTYPE_V8HI_V8HI },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_andv2di3, "__builtin_ia32_pand128", IX86_BUILTIN_PAND128, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_nandv2di3, "__builtin_ia32_pandn128", IX86_BUILTIN_PANDN128, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_iorv2di3, "__builtin_ia32_por128", IX86_BUILTIN_POR128, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_xorv2di3, "__builtin_ia32_pxor128", IX86_BUILTIN_PXOR128, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_uavgv16qi3, "__builtin_ia32_pavgb128", IX86_BUILTIN_PAVGB128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_uavgv8hi3, "__builtin_ia32_pavgw128", IX86_BUILTIN_PAVGW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_eqv16qi3, "__builtin_ia32_pcmpeqb128", IX86_BUILTIN_PCMPEQB128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_eqv8hi3, "__builtin_ia32_pcmpeqw128", IX86_BUILTIN_PCMPEQW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_eqv4si3, "__builtin_ia32_pcmpeqd128", IX86_BUILTIN_PCMPEQD128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_gtv16qi3, "__builtin_ia32_pcmpgtb128", IX86_BUILTIN_PCMPGTB128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_gtv8hi3, "__builtin_ia32_pcmpgtw128", IX86_BUILTIN_PCMPGTW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_gtv4si3, "__builtin_ia32_pcmpgtd128", IX86_BUILTIN_PCMPGTD128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_umaxv16qi3, "__builtin_ia32_pmaxub128", IX86_BUILTIN_PMAXUB128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_smaxv8hi3, "__builtin_ia32_pmaxsw128", IX86_BUILTIN_PMAXSW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_uminv16qi3, "__builtin_ia32_pminub128", IX86_BUILTIN_PMINUB128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sminv8hi3, "__builtin_ia32_pminsw128", IX86_BUILTIN_PMINSW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_punpckhbw, "__builtin_ia32_punpckhbw128", IX86_BUILTIN_PUNPCKHBW128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_punpckhwd, "__builtin_ia32_punpckhwd128", IX86_BUILTIN_PUNPCKHWD128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_punpckhdq, "__builtin_ia32_punpckhdq128", IX86_BUILTIN_PUNPCKHDQ128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_punpckhqdq, "__builtin_ia32_punpckhqdq128", IX86_BUILTIN_PUNPCKHQDQ128, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_punpcklbw, "__builtin_ia32_punpcklbw128", IX86_BUILTIN_PUNPCKLBW128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_punpcklwd, "__builtin_ia32_punpcklwd128", IX86_BUILTIN_PUNPCKLWD128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_punpckldq, "__builtin_ia32_punpckldq128", IX86_BUILTIN_PUNPCKLDQ128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_punpcklqdq, "__builtin_ia32_punpcklqdq128", IX86_BUILTIN_PUNPCKLQDQ128, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_packsswb, "__builtin_ia32_packsswb128", IX86_BUILTIN_PACKSSWB128, UNKNOWN, (int) V16QI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_packssdw, "__builtin_ia32_packssdw128", IX86_BUILTIN_PACKSSDW128, UNKNOWN, (int) V8HI_FTYPE_V4SI_V4SI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_packuswb, "__builtin_ia32_packuswb128", IX86_BUILTIN_PACKUSWB128, UNKNOWN, (int) V16QI_FTYPE_V8HI_V8HI },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_umulv8hi3_highpart, "__builtin_ia32_pmulhuw128", IX86_BUILTIN_PMULHUW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_psadbw, "__builtin_ia32_psadbw128", IX86_BUILTIN_PSADBW128, UNKNOWN, (int) V2DI_FTYPE_V16QI_V16QI },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_umulv1siv1di3, "__builtin_ia32_pmuludq", IX86_BUILTIN_PMULUDQ, UNKNOWN, (int) V1DI_FTYPE_V2SI_V2SI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_umulv2siv2di3, "__builtin_ia32_pmuludq128", IX86_BUILTIN_PMULUDQ128, UNKNOWN, (int) V2DI_FTYPE_V4SI_V4SI },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_pmaddwd, "__builtin_ia32_pmaddwd128", IX86_BUILTIN_PMADDWD128, UNKNOWN, (int) V4SI_FTYPE_V8HI_V8HI },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_cvtsi2sd, "__builtin_ia32_cvtsi2sd", IX86_BUILTIN_CVTSI2SD, UNKNOWN, (int) V2DF_FTYPE_V2DF_SI },
+ { OPTION_MASK_ISA_SSE2 | OPTION_MASK_ISA_64BIT, CODE_FOR_sse2_cvtsi2sdq, "__builtin_ia32_cvtsi642sd", IX86_BUILTIN_CVTSI642SD, UNKNOWN, (int) V2DF_FTYPE_V2DF_DI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_cvtsd2ss, "__builtin_ia32_cvtsd2ss", IX86_BUILTIN_CVTSD2SS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V2DF },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_cvtss2sd, "__builtin_ia32_cvtss2sd", IX86_BUILTIN_CVTSS2SD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V4SF },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_ashlti3, "__builtin_ia32_pslldqi128", IX86_BUILTIN_PSLLDQI128, UNKNOWN, (int) V2DI2TI_FTYPE_V2DI_INT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_ashlv8hi3, "__builtin_ia32_psllwi128", IX86_BUILTIN_PSLLWI128, UNKNOWN, (int) V8HI_FTYPE_V8HI_SI_COUNT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_ashlv4si3, "__builtin_ia32_pslldi128", IX86_BUILTIN_PSLLDI128, UNKNOWN, (int) V4SI_FTYPE_V4SI_SI_COUNT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_ashlv2di3, "__builtin_ia32_psllqi128", IX86_BUILTIN_PSLLQI128, UNKNOWN, (int) V2DI_FTYPE_V2DI_SI_COUNT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_ashlv8hi3, "__builtin_ia32_psllw128", IX86_BUILTIN_PSLLW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI_COUNT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_ashlv4si3, "__builtin_ia32_pslld128", IX86_BUILTIN_PSLLD128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI_COUNT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_ashlv2di3, "__builtin_ia32_psllq128", IX86_BUILTIN_PSLLQ128, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI_COUNT },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_lshrti3, "__builtin_ia32_psrldqi128", IX86_BUILTIN_PSRLDQI128, UNKNOWN, (int) V2DI2TI_FTYPE_V2DI_INT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_lshrv8hi3, "__builtin_ia32_psrlwi128", IX86_BUILTIN_PSRLWI128, UNKNOWN, (int) V8HI_FTYPE_V8HI_SI_COUNT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_lshrv4si3, "__builtin_ia32_psrldi128", IX86_BUILTIN_PSRLDI128, UNKNOWN, (int) V4SI_FTYPE_V4SI_SI_COUNT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_lshrv2di3, "__builtin_ia32_psrlqi128", IX86_BUILTIN_PSRLQI128, UNKNOWN, (int) V2DI_FTYPE_V2DI_SI_COUNT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_lshrv8hi3, "__builtin_ia32_psrlw128", IX86_BUILTIN_PSRLW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI_COUNT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_lshrv4si3, "__builtin_ia32_psrld128", IX86_BUILTIN_PSRLD128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI_COUNT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_lshrv2di3, "__builtin_ia32_psrlq128", IX86_BUILTIN_PSRLQ128, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI_COUNT },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_ashrv8hi3, "__builtin_ia32_psrawi128", IX86_BUILTIN_PSRAWI128, UNKNOWN, (int) V8HI_FTYPE_V8HI_SI_COUNT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_ashrv4si3, "__builtin_ia32_psradi128", IX86_BUILTIN_PSRADI128, UNKNOWN, (int) V4SI_FTYPE_V4SI_SI_COUNT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_ashrv8hi3, "__builtin_ia32_psraw128", IX86_BUILTIN_PSRAW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI_COUNT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_ashrv4si3, "__builtin_ia32_psrad128", IX86_BUILTIN_PSRAD128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI_COUNT },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_pshufd, "__builtin_ia32_pshufd", IX86_BUILTIN_PSHUFD, UNKNOWN, (int) V4SI_FTYPE_V4SI_INT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_pshuflw, "__builtin_ia32_pshuflw", IX86_BUILTIN_PSHUFLW, UNKNOWN, (int) V8HI_FTYPE_V8HI_INT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_pshufhw, "__builtin_ia32_pshufhw", IX86_BUILTIN_PSHUFHW, UNKNOWN, (int) V8HI_FTYPE_V8HI_INT },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_sse2_vmsqrtv2df2, "__builtin_ia32_sqrtsd", IX86_BUILTIN_SQRTSD, UNKNOWN, (int) V2DF_FTYPE_V2DF_VEC_MERGE },
- { MASK_SSE2, CODE_FOR_sse2_cvtps2dq, 0, IX86_BUILTIN_CVTPS2DQ, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_cvtps2pd, 0, IX86_BUILTIN_CVTPS2PD, 0, 0 },
- { MASK_SSE2, CODE_FOR_sse2_cvttps2dq, 0, IX86_BUILTIN_CVTTPS2DQ, 0, 0 },
+ /* SSE2 MMX */
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_mmx_addv1di3, "__builtin_ia32_paddq", IX86_BUILTIN_PADDQ, UNKNOWN, (int) V1DI_FTYPE_V1DI_V1DI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_mmx_subv1di3, "__builtin_ia32_psubq", IX86_BUILTIN_PSUBQ, UNKNOWN, (int) V1DI_FTYPE_V1DI_V1DI },
/* SSE3 */
- { MASK_SSE3, CODE_FOR_sse3_movshdup, "__builtin_ia32_movshdup", IX86_BUILTIN_MOVSHDUP, 0, 0 },
- { MASK_SSE3, CODE_FOR_sse3_movsldup, "__builtin_ia32_movsldup", IX86_BUILTIN_MOVSLDUP, 0, 0 },
+ { OPTION_MASK_ISA_SSE3, CODE_FOR_sse3_movshdup, "__builtin_ia32_movshdup", IX86_BUILTIN_MOVSHDUP, UNKNOWN, (int) V4SF_FTYPE_V4SF},
+ { OPTION_MASK_ISA_SSE3, CODE_FOR_sse3_movsldup, "__builtin_ia32_movsldup", IX86_BUILTIN_MOVSLDUP, UNKNOWN, (int) V4SF_FTYPE_V4SF },
+
+ { OPTION_MASK_ISA_SSE3, CODE_FOR_sse3_addsubv4sf3, "__builtin_ia32_addsubps", IX86_BUILTIN_ADDSUBPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE3, CODE_FOR_sse3_addsubv2df3, "__builtin_ia32_addsubpd", IX86_BUILTIN_ADDSUBPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE3, CODE_FOR_sse3_haddv4sf3, "__builtin_ia32_haddps", IX86_BUILTIN_HADDPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE3, CODE_FOR_sse3_haddv2df3, "__builtin_ia32_haddpd", IX86_BUILTIN_HADDPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE3, CODE_FOR_sse3_hsubv4sf3, "__builtin_ia32_hsubps", IX86_BUILTIN_HSUBPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE3, CODE_FOR_sse3_hsubv2df3, "__builtin_ia32_hsubpd", IX86_BUILTIN_HSUBPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF },
/* SSSE3 */
- { MASK_SSSE3, CODE_FOR_absv16qi2, "__builtin_ia32_pabsb128", IX86_BUILTIN_PABSB128, 0, 0 },
- { MASK_SSSE3, CODE_FOR_absv8qi2, "__builtin_ia32_pabsb", IX86_BUILTIN_PABSB, 0, 0 },
- { MASK_SSSE3, CODE_FOR_absv8hi2, "__builtin_ia32_pabsw128", IX86_BUILTIN_PABSW128, 0, 0 },
- { MASK_SSSE3, CODE_FOR_absv4hi2, "__builtin_ia32_pabsw", IX86_BUILTIN_PABSW, 0, 0 },
- { MASK_SSSE3, CODE_FOR_absv4si2, "__builtin_ia32_pabsd128", IX86_BUILTIN_PABSD128, 0, 0 },
- { MASK_SSSE3, CODE_FOR_absv2si2, "__builtin_ia32_pabsd", IX86_BUILTIN_PABSD, 0, 0 },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_absv16qi2, "__builtin_ia32_pabsb128", IX86_BUILTIN_PABSB128, UNKNOWN, (int) V16QI_FTYPE_V16QI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_absv8qi2, "__builtin_ia32_pabsb", IX86_BUILTIN_PABSB, UNKNOWN, (int) V8QI_FTYPE_V8QI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_absv8hi2, "__builtin_ia32_pabsw128", IX86_BUILTIN_PABSW128, UNKNOWN, (int) V8HI_FTYPE_V8HI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_absv4hi2, "__builtin_ia32_pabsw", IX86_BUILTIN_PABSW, UNKNOWN, (int) V4HI_FTYPE_V4HI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_absv4si2, "__builtin_ia32_pabsd128", IX86_BUILTIN_PABSD128, UNKNOWN, (int) V4SI_FTYPE_V4SI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_absv2si2, "__builtin_ia32_pabsd", IX86_BUILTIN_PABSD, UNKNOWN, (int) V2SI_FTYPE_V2SI },
+
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_phaddwv8hi3, "__builtin_ia32_phaddw128", IX86_BUILTIN_PHADDW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_phaddwv4hi3, "__builtin_ia32_phaddw", IX86_BUILTIN_PHADDW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_phadddv4si3, "__builtin_ia32_phaddd128", IX86_BUILTIN_PHADDD128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_phadddv2si3, "__builtin_ia32_phaddd", IX86_BUILTIN_PHADDD, UNKNOWN, (int) V2SI_FTYPE_V2SI_V2SI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_phaddswv8hi3, "__builtin_ia32_phaddsw128", IX86_BUILTIN_PHADDSW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_phaddswv4hi3, "__builtin_ia32_phaddsw", IX86_BUILTIN_PHADDSW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_phsubwv8hi3, "__builtin_ia32_phsubw128", IX86_BUILTIN_PHSUBW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_phsubwv4hi3, "__builtin_ia32_phsubw", IX86_BUILTIN_PHSUBW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_phsubdv4si3, "__builtin_ia32_phsubd128", IX86_BUILTIN_PHSUBD128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_phsubdv2si3, "__builtin_ia32_phsubd", IX86_BUILTIN_PHSUBD, UNKNOWN, (int) V2SI_FTYPE_V2SI_V2SI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_phsubswv8hi3, "__builtin_ia32_phsubsw128", IX86_BUILTIN_PHSUBSW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_phsubswv4hi3, "__builtin_ia32_phsubsw", IX86_BUILTIN_PHSUBSW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_pmaddubsw128, "__builtin_ia32_pmaddubsw128", IX86_BUILTIN_PMADDUBSW128, UNKNOWN, (int) V8HI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_pmaddubsw, "__builtin_ia32_pmaddubsw", IX86_BUILTIN_PMADDUBSW, UNKNOWN, (int) V4HI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_pmulhrswv8hi3, "__builtin_ia32_pmulhrsw128", IX86_BUILTIN_PMULHRSW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_pmulhrswv4hi3, "__builtin_ia32_pmulhrsw", IX86_BUILTIN_PMULHRSW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_pshufbv16qi3, "__builtin_ia32_pshufb128", IX86_BUILTIN_PSHUFB128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_pshufbv8qi3, "__builtin_ia32_pshufb", IX86_BUILTIN_PSHUFB, UNKNOWN, (int) V8QI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_psignv16qi3, "__builtin_ia32_psignb128", IX86_BUILTIN_PSIGNB128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_psignv8qi3, "__builtin_ia32_psignb", IX86_BUILTIN_PSIGNB, UNKNOWN, (int) V8QI_FTYPE_V8QI_V8QI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_psignv8hi3, "__builtin_ia32_psignw128", IX86_BUILTIN_PSIGNW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_psignv4hi3, "__builtin_ia32_psignw", IX86_BUILTIN_PSIGNW, UNKNOWN, (int) V4HI_FTYPE_V4HI_V4HI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_psignv4si3, "__builtin_ia32_psignd128", IX86_BUILTIN_PSIGND128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_psignv2si3, "__builtin_ia32_psignd", IX86_BUILTIN_PSIGND, UNKNOWN, (int) V2SI_FTYPE_V2SI_V2SI },
+
+ /* SSSE3. */
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_palignrti, "__builtin_ia32_palignr128", IX86_BUILTIN_PALIGNR128, UNKNOWN, (int) V2DI2TI_FTYPE_V2DI_V2DI_INT },
+ { OPTION_MASK_ISA_SSSE3, CODE_FOR_ssse3_palignrdi, "__builtin_ia32_palignr", IX86_BUILTIN_PALIGNR, UNKNOWN, (int) V1DI2DI_FTYPE_V1DI_V1DI_INT },
+
+ /* SSE4.1 */
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_blendpd, "__builtin_ia32_blendpd", IX86_BUILTIN_BLENDPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF_INT },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_blendps, "__builtin_ia32_blendps", IX86_BUILTIN_BLENDPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF_INT },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_blendvpd, "__builtin_ia32_blendvpd", IX86_BUILTIN_BLENDVPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF_V2DF },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_blendvps, "__builtin_ia32_blendvps", IX86_BUILTIN_BLENDVPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF_V4SF },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_dppd, "__builtin_ia32_dppd", IX86_BUILTIN_DPPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF_INT },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_dpps, "__builtin_ia32_dpps", IX86_BUILTIN_DPPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF_INT },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_insertps, "__builtin_ia32_insertps128", IX86_BUILTIN_INSERTPS128, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF_INT },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_mpsadbw, "__builtin_ia32_mpsadbw128", IX86_BUILTIN_MPSADBW128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI_INT },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_pblendvb, "__builtin_ia32_pblendvb128", IX86_BUILTIN_PBLENDVB128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_pblendw, "__builtin_ia32_pblendw128", IX86_BUILTIN_PBLENDW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI_INT },
+
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_extendv8qiv8hi2, "__builtin_ia32_pmovsxbw128", IX86_BUILTIN_PMOVSXBW128, UNKNOWN, (int) V8HI_FTYPE_V16QI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_extendv4qiv4si2, "__builtin_ia32_pmovsxbd128", IX86_BUILTIN_PMOVSXBD128, UNKNOWN, (int) V4SI_FTYPE_V16QI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_extendv2qiv2di2, "__builtin_ia32_pmovsxbq128", IX86_BUILTIN_PMOVSXBQ128, UNKNOWN, (int) V2DI_FTYPE_V16QI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_extendv4hiv4si2, "__builtin_ia32_pmovsxwd128", IX86_BUILTIN_PMOVSXWD128, UNKNOWN, (int) V4SI_FTYPE_V8HI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_extendv2hiv2di2, "__builtin_ia32_pmovsxwq128", IX86_BUILTIN_PMOVSXWQ128, UNKNOWN, (int) V2DI_FTYPE_V8HI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_extendv2siv2di2, "__builtin_ia32_pmovsxdq128", IX86_BUILTIN_PMOVSXDQ128, UNKNOWN, (int) V2DI_FTYPE_V4SI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_zero_extendv8qiv8hi2, "__builtin_ia32_pmovzxbw128", IX86_BUILTIN_PMOVZXBW128, UNKNOWN, (int) V8HI_FTYPE_V16QI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_zero_extendv4qiv4si2, "__builtin_ia32_pmovzxbd128", IX86_BUILTIN_PMOVZXBD128, UNKNOWN, (int) V4SI_FTYPE_V16QI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_zero_extendv2qiv2di2, "__builtin_ia32_pmovzxbq128", IX86_BUILTIN_PMOVZXBQ128, UNKNOWN, (int) V2DI_FTYPE_V16QI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_zero_extendv4hiv4si2, "__builtin_ia32_pmovzxwd128", IX86_BUILTIN_PMOVZXWD128, UNKNOWN, (int) V4SI_FTYPE_V8HI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_zero_extendv2hiv2di2, "__builtin_ia32_pmovzxwq128", IX86_BUILTIN_PMOVZXWQ128, UNKNOWN, (int) V2DI_FTYPE_V8HI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_zero_extendv2siv2di2, "__builtin_ia32_pmovzxdq128", IX86_BUILTIN_PMOVZXDQ128, UNKNOWN, (int) V2DI_FTYPE_V4SI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_phminposuw, "__builtin_ia32_phminposuw128", IX86_BUILTIN_PHMINPOSUW128, UNKNOWN, (int) V8HI_FTYPE_V8HI },
+
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_packusdw, "__builtin_ia32_packusdw128", IX86_BUILTIN_PACKUSDW128, UNKNOWN, (int) V8HI_FTYPE_V4SI_V4SI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_eqv2di3, "__builtin_ia32_pcmpeqq", IX86_BUILTIN_PCMPEQQ, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_smaxv16qi3, "__builtin_ia32_pmaxsb128", IX86_BUILTIN_PMAXSB128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_smaxv4si3, "__builtin_ia32_pmaxsd128", IX86_BUILTIN_PMAXSD128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_umaxv4si3, "__builtin_ia32_pmaxud128", IX86_BUILTIN_PMAXUD128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_umaxv8hi3, "__builtin_ia32_pmaxuw128", IX86_BUILTIN_PMAXUW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sminv16qi3, "__builtin_ia32_pminsb128", IX86_BUILTIN_PMINSB128, UNKNOWN, (int) V16QI_FTYPE_V16QI_V16QI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sminv4si3, "__builtin_ia32_pminsd128", IX86_BUILTIN_PMINSD128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_uminv4si3, "__builtin_ia32_pminud128", IX86_BUILTIN_PMINUD128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_uminv8hi3, "__builtin_ia32_pminuw128", IX86_BUILTIN_PMINUW128, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_sse4_1_mulv2siv2di3, "__builtin_ia32_pmuldq128", IX86_BUILTIN_PMULDQ128, UNKNOWN, (int) V2DI_FTYPE_V4SI_V4SI },
+ { OPTION_MASK_ISA_SSE4_1, CODE_FOR_mulv4si3, "__builtin_ia32_pmulld128", IX86_BUILTIN_PMULLD128, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI },
+
+ /* SSE4.1 and SSE5 */
+ { OPTION_MASK_ISA_ROUND, CODE_FOR_sse4_1_roundpd, "__builtin_ia32_roundpd", IX86_BUILTIN_ROUNDPD, UNKNOWN, (int) V2DF_FTYPE_V2DF_INT },
+ { OPTION_MASK_ISA_ROUND, CODE_FOR_sse4_1_roundps, "__builtin_ia32_roundps", IX86_BUILTIN_ROUNDPS, UNKNOWN, (int) V4SF_FTYPE_V4SF_INT },
+ { OPTION_MASK_ISA_ROUND, CODE_FOR_sse4_1_roundsd, "__builtin_ia32_roundsd", IX86_BUILTIN_ROUNDSD, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF_INT },
+ { OPTION_MASK_ISA_ROUND, CODE_FOR_sse4_1_roundss, "__builtin_ia32_roundss", IX86_BUILTIN_ROUNDSS, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF_INT },
+
+ { OPTION_MASK_ISA_ROUND, CODE_FOR_sse4_1_ptest, "__builtin_ia32_ptestz128", IX86_BUILTIN_PTESTZ, EQ, (int) INT_FTYPE_V2DI_V2DI_PTEST },
+ { OPTION_MASK_ISA_ROUND, CODE_FOR_sse4_1_ptest, "__builtin_ia32_ptestc128", IX86_BUILTIN_PTESTC, LTU, (int) INT_FTYPE_V2DI_V2DI_PTEST },
+ { OPTION_MASK_ISA_ROUND, CODE_FOR_sse4_1_ptest, "__builtin_ia32_ptestnzc128", IX86_BUILTIN_PTESTNZC, GTU, (int) INT_FTYPE_V2DI_V2DI_PTEST },
+
+ /* SSE4.2 */
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_gtv2di3, "__builtin_ia32_pcmpgtq", IX86_BUILTIN_PCMPGTQ, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI },
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_crc32qi, "__builtin_ia32_crc32qi", IX86_BUILTIN_CRC32QI, UNKNOWN, (int) UINT_FTYPE_UINT_UCHAR },
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_crc32hi, "__builtin_ia32_crc32hi", IX86_BUILTIN_CRC32HI, UNKNOWN, (int) UINT_FTYPE_UINT_USHORT },
+ { OPTION_MASK_ISA_SSE4_2, CODE_FOR_sse4_2_crc32si, "__builtin_ia32_crc32si", IX86_BUILTIN_CRC32SI, UNKNOWN, (int) UINT_FTYPE_UINT_UINT },
+ { OPTION_MASK_ISA_SSE4_2 | OPTION_MASK_ISA_64BIT, CODE_FOR_sse4_2_crc32di, "__builtin_ia32_crc32di", IX86_BUILTIN_CRC32DI, UNKNOWN, (int) UINT64_FTYPE_UINT64_UINT64 },
+
+ /* SSE4A */
+ { OPTION_MASK_ISA_SSE4A, CODE_FOR_sse4a_extrqi, "__builtin_ia32_extrqi", IX86_BUILTIN_EXTRQI, UNKNOWN, (int) V2DI_FTYPE_V2DI_UINT_UINT },
+ { OPTION_MASK_ISA_SSE4A, CODE_FOR_sse4a_extrq, "__builtin_ia32_extrq", IX86_BUILTIN_EXTRQ, UNKNOWN, (int) V2DI_FTYPE_V2DI_V16QI },
+ { OPTION_MASK_ISA_SSE4A, CODE_FOR_sse4a_insertqi, "__builtin_ia32_insertqi", IX86_BUILTIN_INSERTQI, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI_UINT_UINT },
+ { OPTION_MASK_ISA_SSE4A, CODE_FOR_sse4a_insertq, "__builtin_ia32_insertq", IX86_BUILTIN_INSERTQ, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI },
+
+ /* AES */
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_aeskeygenassist, 0, IX86_BUILTIN_AESKEYGENASSIST128, UNKNOWN, (int) V2DI_FTYPE_V2DI_INT },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_aesimc, 0, IX86_BUILTIN_AESIMC128, UNKNOWN, (int) V2DI_FTYPE_V2DI },
+
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_aesenc, 0, IX86_BUILTIN_AESENC128, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_aesenclast, 0, IX86_BUILTIN_AESENCLAST128, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_aesdec, 0, IX86_BUILTIN_AESDEC128, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI },
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_aesdeclast, 0, IX86_BUILTIN_AESDECLAST128, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI },
+
+ /* PCLMUL */
+ { OPTION_MASK_ISA_SSE2, CODE_FOR_pclmulqdq, 0, IX86_BUILTIN_PCLMULQDQ128, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI_INT },
+
+ /* 64bit */
+ { OPTION_MASK_ISA_64BIT, CODE_FOR_abstf2, 0, IX86_BUILTIN_FABSQ, UNKNOWN, (int) FLOAT128_FTYPE_FLOAT128 },
+ { OPTION_MASK_ISA_64BIT, CODE_FOR_copysigntf3, 0, IX86_BUILTIN_COPYSIGNQ, UNKNOWN, (int) FLOAT128_FTYPE_FLOAT128_FLOAT128 },
+};
+
+/* SSE5 */
+enum multi_arg_type {
+ MULTI_ARG_UNKNOWN,
+ MULTI_ARG_3_SF,
+ MULTI_ARG_3_DF,
+ MULTI_ARG_3_DI,
+ MULTI_ARG_3_SI,
+ MULTI_ARG_3_SI_DI,
+ MULTI_ARG_3_HI,
+ MULTI_ARG_3_HI_SI,
+ MULTI_ARG_3_QI,
+ MULTI_ARG_3_PERMPS,
+ MULTI_ARG_3_PERMPD,
+ MULTI_ARG_2_SF,
+ MULTI_ARG_2_DF,
+ MULTI_ARG_2_DI,
+ MULTI_ARG_2_SI,
+ MULTI_ARG_2_HI,
+ MULTI_ARG_2_QI,
+ MULTI_ARG_2_DI_IMM,
+ MULTI_ARG_2_SI_IMM,
+ MULTI_ARG_2_HI_IMM,
+ MULTI_ARG_2_QI_IMM,
+ MULTI_ARG_2_SF_CMP,
+ MULTI_ARG_2_DF_CMP,
+ MULTI_ARG_2_DI_CMP,
+ MULTI_ARG_2_SI_CMP,
+ MULTI_ARG_2_HI_CMP,
+ MULTI_ARG_2_QI_CMP,
+ MULTI_ARG_2_DI_TF,
+ MULTI_ARG_2_SI_TF,
+ MULTI_ARG_2_HI_TF,
+ MULTI_ARG_2_QI_TF,
+ MULTI_ARG_2_SF_TF,
+ MULTI_ARG_2_DF_TF,
+ MULTI_ARG_1_SF,
+ MULTI_ARG_1_DF,
+ MULTI_ARG_1_DI,
+ MULTI_ARG_1_SI,
+ MULTI_ARG_1_HI,
+ MULTI_ARG_1_QI,
+ MULTI_ARG_1_SI_DI,
+ MULTI_ARG_1_HI_DI,
+ MULTI_ARG_1_HI_SI,
+ MULTI_ARG_1_QI_DI,
+ MULTI_ARG_1_QI_SI,
+ MULTI_ARG_1_QI_HI,
+ MULTI_ARG_1_PH2PS,
+ MULTI_ARG_1_PS2PH
+};
+
+static const struct builtin_description bdesc_multi_arg[] =
+{
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5i_vmfmaddv4sf4, "__builtin_ia32_fmaddss", IX86_BUILTIN_FMADDSS, 0, (int)MULTI_ARG_3_SF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5i_vmfmaddv2df4, "__builtin_ia32_fmaddsd", IX86_BUILTIN_FMADDSD, 0, (int)MULTI_ARG_3_DF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5i_fmaddv4sf4, "__builtin_ia32_fmaddps", IX86_BUILTIN_FMADDPS, 0, (int)MULTI_ARG_3_SF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5i_fmaddv2df4, "__builtin_ia32_fmaddpd", IX86_BUILTIN_FMADDPD, 0, (int)MULTI_ARG_3_DF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5i_vmfmsubv4sf4, "__builtin_ia32_fmsubss", IX86_BUILTIN_FMSUBSS, 0, (int)MULTI_ARG_3_SF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5i_vmfmsubv2df4, "__builtin_ia32_fmsubsd", IX86_BUILTIN_FMSUBSD, 0, (int)MULTI_ARG_3_DF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5i_fmsubv4sf4, "__builtin_ia32_fmsubps", IX86_BUILTIN_FMSUBPS, 0, (int)MULTI_ARG_3_SF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5i_fmsubv2df4, "__builtin_ia32_fmsubpd", IX86_BUILTIN_FMSUBPD, 0, (int)MULTI_ARG_3_DF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5i_vmfnmaddv4sf4, "__builtin_ia32_fnmaddss", IX86_BUILTIN_FNMADDSS, 0, (int)MULTI_ARG_3_SF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5i_vmfnmaddv2df4, "__builtin_ia32_fnmaddsd", IX86_BUILTIN_FNMADDSD, 0, (int)MULTI_ARG_3_DF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5i_fnmaddv4sf4, "__builtin_ia32_fnmaddps", IX86_BUILTIN_FNMADDPS, 0, (int)MULTI_ARG_3_SF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5i_fnmaddv2df4, "__builtin_ia32_fnmaddpd", IX86_BUILTIN_FNMADDPD, 0, (int)MULTI_ARG_3_DF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5i_vmfnmsubv4sf4, "__builtin_ia32_fnmsubss", IX86_BUILTIN_FNMSUBSS, 0, (int)MULTI_ARG_3_SF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5i_vmfnmsubv2df4, "__builtin_ia32_fnmsubsd", IX86_BUILTIN_FNMSUBSD, 0, (int)MULTI_ARG_3_DF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5i_fnmsubv4sf4, "__builtin_ia32_fnmsubps", IX86_BUILTIN_FNMSUBPS, 0, (int)MULTI_ARG_3_SF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5i_fnmsubv2df4, "__builtin_ia32_fnmsubpd", IX86_BUILTIN_FNMSUBPD, 0, (int)MULTI_ARG_3_DF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcmov_v2di, "__builtin_ia32_pcmov", IX86_BUILTIN_PCMOV_V2DI, 0, (int)MULTI_ARG_3_DI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcmov_v2di, "__builtin_ia32_pcmov_v2di", IX86_BUILTIN_PCMOV_V2DI, 0, (int)MULTI_ARG_3_DI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcmov_v4si, "__builtin_ia32_pcmov_v4si", IX86_BUILTIN_PCMOV_V4SI, 0, (int)MULTI_ARG_3_SI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcmov_v8hi, "__builtin_ia32_pcmov_v8hi", IX86_BUILTIN_PCMOV_V8HI, 0, (int)MULTI_ARG_3_HI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcmov_v16qi, "__builtin_ia32_pcmov_v16qi",IX86_BUILTIN_PCMOV_V16QI,0, (int)MULTI_ARG_3_QI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcmov_v2df, "__builtin_ia32_pcmov_v2df", IX86_BUILTIN_PCMOV_V2DF, 0, (int)MULTI_ARG_3_DF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcmov_v4sf, "__builtin_ia32_pcmov_v4sf", IX86_BUILTIN_PCMOV_V4SF, 0, (int)MULTI_ARG_3_SF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pperm, "__builtin_ia32_pperm", IX86_BUILTIN_PPERM, 0, (int)MULTI_ARG_3_QI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_permv4sf, "__builtin_ia32_permps", IX86_BUILTIN_PERMPS, 0, (int)MULTI_ARG_3_PERMPS },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_permv2df, "__builtin_ia32_permpd", IX86_BUILTIN_PERMPD, 0, (int)MULTI_ARG_3_PERMPD },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pmacssww, "__builtin_ia32_pmacssww", IX86_BUILTIN_PMACSSWW, 0, (int)MULTI_ARG_3_HI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pmacsww, "__builtin_ia32_pmacsww", IX86_BUILTIN_PMACSWW, 0, (int)MULTI_ARG_3_HI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pmacsswd, "__builtin_ia32_pmacsswd", IX86_BUILTIN_PMACSSWD, 0, (int)MULTI_ARG_3_HI_SI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pmacswd, "__builtin_ia32_pmacswd", IX86_BUILTIN_PMACSWD, 0, (int)MULTI_ARG_3_HI_SI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pmacssdd, "__builtin_ia32_pmacssdd", IX86_BUILTIN_PMACSSDD, 0, (int)MULTI_ARG_3_SI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pmacsdd, "__builtin_ia32_pmacsdd", IX86_BUILTIN_PMACSDD, 0, (int)MULTI_ARG_3_SI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pmacssdql, "__builtin_ia32_pmacssdql", IX86_BUILTIN_PMACSSDQL, 0, (int)MULTI_ARG_3_SI_DI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pmacssdqh, "__builtin_ia32_pmacssdqh", IX86_BUILTIN_PMACSSDQH, 0, (int)MULTI_ARG_3_SI_DI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pmacsdql, "__builtin_ia32_pmacsdql", IX86_BUILTIN_PMACSDQL, 0, (int)MULTI_ARG_3_SI_DI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pmacsdqh, "__builtin_ia32_pmacsdqh", IX86_BUILTIN_PMACSDQH, 0, (int)MULTI_ARG_3_SI_DI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pmadcsswd, "__builtin_ia32_pmadcsswd", IX86_BUILTIN_PMADCSSWD, 0, (int)MULTI_ARG_3_HI_SI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pmadcswd, "__builtin_ia32_pmadcswd", IX86_BUILTIN_PMADCSWD, 0, (int)MULTI_ARG_3_HI_SI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vrotlv2di3, "__builtin_ia32_protq", IX86_BUILTIN_PROTQ, 0, (int)MULTI_ARG_2_DI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vrotlv4si3, "__builtin_ia32_protd", IX86_BUILTIN_PROTD, 0, (int)MULTI_ARG_2_SI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vrotlv8hi3, "__builtin_ia32_protw", IX86_BUILTIN_PROTW, 0, (int)MULTI_ARG_2_HI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vrotlv16qi3, "__builtin_ia32_protb", IX86_BUILTIN_PROTB, 0, (int)MULTI_ARG_2_QI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_rotlv2di3, "__builtin_ia32_protqi", IX86_BUILTIN_PROTQ_IMM, 0, (int)MULTI_ARG_2_DI_IMM },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_rotlv4si3, "__builtin_ia32_protdi", IX86_BUILTIN_PROTD_IMM, 0, (int)MULTI_ARG_2_SI_IMM },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_rotlv8hi3, "__builtin_ia32_protwi", IX86_BUILTIN_PROTW_IMM, 0, (int)MULTI_ARG_2_HI_IMM },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_rotlv16qi3, "__builtin_ia32_protbi", IX86_BUILTIN_PROTB_IMM, 0, (int)MULTI_ARG_2_QI_IMM },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_ashlv2di3, "__builtin_ia32_pshaq", IX86_BUILTIN_PSHAQ, 0, (int)MULTI_ARG_2_DI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_ashlv4si3, "__builtin_ia32_pshad", IX86_BUILTIN_PSHAD, 0, (int)MULTI_ARG_2_SI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_ashlv8hi3, "__builtin_ia32_pshaw", IX86_BUILTIN_PSHAW, 0, (int)MULTI_ARG_2_HI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_ashlv16qi3, "__builtin_ia32_pshab", IX86_BUILTIN_PSHAB, 0, (int)MULTI_ARG_2_QI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_lshlv2di3, "__builtin_ia32_pshlq", IX86_BUILTIN_PSHLQ, 0, (int)MULTI_ARG_2_DI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_lshlv4si3, "__builtin_ia32_pshld", IX86_BUILTIN_PSHLD, 0, (int)MULTI_ARG_2_SI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_lshlv8hi3, "__builtin_ia32_pshlw", IX86_BUILTIN_PSHLW, 0, (int)MULTI_ARG_2_HI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_lshlv16qi3, "__builtin_ia32_pshlb", IX86_BUILTIN_PSHLB, 0, (int)MULTI_ARG_2_QI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmfrczv4sf2, "__builtin_ia32_frczss", IX86_BUILTIN_FRCZSS, 0, (int)MULTI_ARG_2_SF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmfrczv2df2, "__builtin_ia32_frczsd", IX86_BUILTIN_FRCZSD, 0, (int)MULTI_ARG_2_DF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_frczv4sf2, "__builtin_ia32_frczps", IX86_BUILTIN_FRCZPS, 0, (int)MULTI_ARG_1_SF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_frczv2df2, "__builtin_ia32_frczpd", IX86_BUILTIN_FRCZPD, 0, (int)MULTI_ARG_1_DF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_cvtph2ps, "__builtin_ia32_cvtph2ps", IX86_BUILTIN_CVTPH2PS, 0, (int)MULTI_ARG_1_PH2PS },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_cvtps2ph, "__builtin_ia32_cvtps2ph", IX86_BUILTIN_CVTPS2PH, 0, (int)MULTI_ARG_1_PS2PH },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_phaddbw, "__builtin_ia32_phaddbw", IX86_BUILTIN_PHADDBW, 0, (int)MULTI_ARG_1_QI_HI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_phaddbd, "__builtin_ia32_phaddbd", IX86_BUILTIN_PHADDBD, 0, (int)MULTI_ARG_1_QI_SI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_phaddbq, "__builtin_ia32_phaddbq", IX86_BUILTIN_PHADDBQ, 0, (int)MULTI_ARG_1_QI_DI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_phaddwd, "__builtin_ia32_phaddwd", IX86_BUILTIN_PHADDWD, 0, (int)MULTI_ARG_1_HI_SI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_phaddwq, "__builtin_ia32_phaddwq", IX86_BUILTIN_PHADDWQ, 0, (int)MULTI_ARG_1_HI_DI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_phadddq, "__builtin_ia32_phadddq", IX86_BUILTIN_PHADDDQ, 0, (int)MULTI_ARG_1_SI_DI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_phaddubw, "__builtin_ia32_phaddubw", IX86_BUILTIN_PHADDUBW, 0, (int)MULTI_ARG_1_QI_HI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_phaddubd, "__builtin_ia32_phaddubd", IX86_BUILTIN_PHADDUBD, 0, (int)MULTI_ARG_1_QI_SI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_phaddubq, "__builtin_ia32_phaddubq", IX86_BUILTIN_PHADDUBQ, 0, (int)MULTI_ARG_1_QI_DI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_phadduwd, "__builtin_ia32_phadduwd", IX86_BUILTIN_PHADDUWD, 0, (int)MULTI_ARG_1_HI_SI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_phadduwq, "__builtin_ia32_phadduwq", IX86_BUILTIN_PHADDUWQ, 0, (int)MULTI_ARG_1_HI_DI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_phaddudq, "__builtin_ia32_phaddudq", IX86_BUILTIN_PHADDUDQ, 0, (int)MULTI_ARG_1_SI_DI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_phsubbw, "__builtin_ia32_phsubbw", IX86_BUILTIN_PHSUBBW, 0, (int)MULTI_ARG_1_QI_HI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_phsubwd, "__builtin_ia32_phsubwd", IX86_BUILTIN_PHSUBWD, 0, (int)MULTI_ARG_1_HI_SI },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_phsubdq, "__builtin_ia32_phsubdq", IX86_BUILTIN_PHSUBDQ, 0, (int)MULTI_ARG_1_SI_DI },
+
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv4sf3, "__builtin_ia32_comeqss", IX86_BUILTIN_COMEQSS, EQ, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv4sf3, "__builtin_ia32_comness", IX86_BUILTIN_COMNESS, NE, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv4sf3, "__builtin_ia32_comneqss", IX86_BUILTIN_COMNESS, NE, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv4sf3, "__builtin_ia32_comltss", IX86_BUILTIN_COMLTSS, LT, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv4sf3, "__builtin_ia32_comless", IX86_BUILTIN_COMLESS, LE, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv4sf3, "__builtin_ia32_comgtss", IX86_BUILTIN_COMGTSS, GT, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv4sf3, "__builtin_ia32_comgess", IX86_BUILTIN_COMGESS, GE, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv4sf3, "__builtin_ia32_comueqss", IX86_BUILTIN_COMUEQSS, UNEQ, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv4sf3, "__builtin_ia32_comuness", IX86_BUILTIN_COMUNESS, LTGT, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv4sf3, "__builtin_ia32_comuneqss", IX86_BUILTIN_COMUNESS, LTGT, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv4sf3, "__builtin_ia32_comunltss", IX86_BUILTIN_COMULTSS, UNLT, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv4sf3, "__builtin_ia32_comunless", IX86_BUILTIN_COMULESS, UNLE, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv4sf3, "__builtin_ia32_comungtss", IX86_BUILTIN_COMUGTSS, UNGT, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv4sf3, "__builtin_ia32_comungess", IX86_BUILTIN_COMUGESS, UNGE, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv4sf3, "__builtin_ia32_comordss", IX86_BUILTIN_COMORDSS, ORDERED, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv4sf3, "__builtin_ia32_comunordss", IX86_BUILTIN_COMUNORDSS, UNORDERED, (int)MULTI_ARG_2_SF_CMP },
+
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv2df3, "__builtin_ia32_comeqsd", IX86_BUILTIN_COMEQSD, EQ, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv2df3, "__builtin_ia32_comnesd", IX86_BUILTIN_COMNESD, NE, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv2df3, "__builtin_ia32_comneqsd", IX86_BUILTIN_COMNESD, NE, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv2df3, "__builtin_ia32_comltsd", IX86_BUILTIN_COMLTSD, LT, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv2df3, "__builtin_ia32_comlesd", IX86_BUILTIN_COMLESD, LE, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv2df3, "__builtin_ia32_comgtsd", IX86_BUILTIN_COMGTSD, GT, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv2df3, "__builtin_ia32_comgesd", IX86_BUILTIN_COMGESD, GE, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv2df3, "__builtin_ia32_comueqsd", IX86_BUILTIN_COMUEQSD, UNEQ, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv2df3, "__builtin_ia32_comunesd", IX86_BUILTIN_COMUNESD, LTGT, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv2df3, "__builtin_ia32_comuneqsd", IX86_BUILTIN_COMUNESD, LTGT, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv2df3, "__builtin_ia32_comunltsd", IX86_BUILTIN_COMULTSD, UNLT, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv2df3, "__builtin_ia32_comunlesd", IX86_BUILTIN_COMULESD, UNLE, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv2df3, "__builtin_ia32_comungtsd", IX86_BUILTIN_COMUGTSD, UNGT, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv2df3, "__builtin_ia32_comungesd", IX86_BUILTIN_COMUGESD, UNGE, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv2df3, "__builtin_ia32_comordsd", IX86_BUILTIN_COMORDSD, ORDERED, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_vmmaskcmpv2df3, "__builtin_ia32_comunordsd", IX86_BUILTIN_COMUNORDSD, UNORDERED, (int)MULTI_ARG_2_DF_CMP },
+
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4sf3, "__builtin_ia32_comeqps", IX86_BUILTIN_COMEQPS, EQ, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4sf3, "__builtin_ia32_comneps", IX86_BUILTIN_COMNEPS, NE, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4sf3, "__builtin_ia32_comneqps", IX86_BUILTIN_COMNEPS, NE, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4sf3, "__builtin_ia32_comltps", IX86_BUILTIN_COMLTPS, LT, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4sf3, "__builtin_ia32_comleps", IX86_BUILTIN_COMLEPS, LE, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4sf3, "__builtin_ia32_comgtps", IX86_BUILTIN_COMGTPS, GT, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4sf3, "__builtin_ia32_comgeps", IX86_BUILTIN_COMGEPS, GE, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4sf3, "__builtin_ia32_comueqps", IX86_BUILTIN_COMUEQPS, UNEQ, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4sf3, "__builtin_ia32_comuneps", IX86_BUILTIN_COMUNEPS, LTGT, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4sf3, "__builtin_ia32_comuneqps", IX86_BUILTIN_COMUNEPS, LTGT, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4sf3, "__builtin_ia32_comunltps", IX86_BUILTIN_COMULTPS, UNLT, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4sf3, "__builtin_ia32_comunleps", IX86_BUILTIN_COMULEPS, UNLE, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4sf3, "__builtin_ia32_comungtps", IX86_BUILTIN_COMUGTPS, UNGT, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4sf3, "__builtin_ia32_comungeps", IX86_BUILTIN_COMUGEPS, UNGE, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4sf3, "__builtin_ia32_comordps", IX86_BUILTIN_COMORDPS, ORDERED, (int)MULTI_ARG_2_SF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4sf3, "__builtin_ia32_comunordps", IX86_BUILTIN_COMUNORDPS, UNORDERED, (int)MULTI_ARG_2_SF_CMP },
+
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2df3, "__builtin_ia32_comeqpd", IX86_BUILTIN_COMEQPD, EQ, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2df3, "__builtin_ia32_comnepd", IX86_BUILTIN_COMNEPD, NE, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2df3, "__builtin_ia32_comneqpd", IX86_BUILTIN_COMNEPD, NE, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2df3, "__builtin_ia32_comltpd", IX86_BUILTIN_COMLTPD, LT, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2df3, "__builtin_ia32_comlepd", IX86_BUILTIN_COMLEPD, LE, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2df3, "__builtin_ia32_comgtpd", IX86_BUILTIN_COMGTPD, GT, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2df3, "__builtin_ia32_comgepd", IX86_BUILTIN_COMGEPD, GE, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2df3, "__builtin_ia32_comueqpd", IX86_BUILTIN_COMUEQPD, UNEQ, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2df3, "__builtin_ia32_comunepd", IX86_BUILTIN_COMUNEPD, LTGT, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2df3, "__builtin_ia32_comuneqpd", IX86_BUILTIN_COMUNEPD, LTGT, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2df3, "__builtin_ia32_comunltpd", IX86_BUILTIN_COMULTPD, UNLT, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2df3, "__builtin_ia32_comunlepd", IX86_BUILTIN_COMULEPD, UNLE, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2df3, "__builtin_ia32_comungtpd", IX86_BUILTIN_COMUGTPD, UNGT, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2df3, "__builtin_ia32_comungepd", IX86_BUILTIN_COMUGEPD, UNGE, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2df3, "__builtin_ia32_comordpd", IX86_BUILTIN_COMORDPD, ORDERED, (int)MULTI_ARG_2_DF_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2df3, "__builtin_ia32_comunordpd", IX86_BUILTIN_COMUNORDPD, UNORDERED, (int)MULTI_ARG_2_DF_CMP },
+
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv16qi3, "__builtin_ia32_pcomeqb", IX86_BUILTIN_PCOMEQB, EQ, (int)MULTI_ARG_2_QI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv16qi3, "__builtin_ia32_pcomneb", IX86_BUILTIN_PCOMNEB, NE, (int)MULTI_ARG_2_QI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv16qi3, "__builtin_ia32_pcomneqb", IX86_BUILTIN_PCOMNEB, NE, (int)MULTI_ARG_2_QI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv16qi3, "__builtin_ia32_pcomltb", IX86_BUILTIN_PCOMLTB, LT, (int)MULTI_ARG_2_QI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv16qi3, "__builtin_ia32_pcomleb", IX86_BUILTIN_PCOMLEB, LE, (int)MULTI_ARG_2_QI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv16qi3, "__builtin_ia32_pcomgtb", IX86_BUILTIN_PCOMGTB, GT, (int)MULTI_ARG_2_QI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv16qi3, "__builtin_ia32_pcomgeb", IX86_BUILTIN_PCOMGEB, GE, (int)MULTI_ARG_2_QI_CMP },
+
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv8hi3, "__builtin_ia32_pcomeqw", IX86_BUILTIN_PCOMEQW, EQ, (int)MULTI_ARG_2_HI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv8hi3, "__builtin_ia32_pcomnew", IX86_BUILTIN_PCOMNEW, NE, (int)MULTI_ARG_2_HI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv8hi3, "__builtin_ia32_pcomneqw", IX86_BUILTIN_PCOMNEW, NE, (int)MULTI_ARG_2_HI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv8hi3, "__builtin_ia32_pcomltw", IX86_BUILTIN_PCOMLTW, LT, (int)MULTI_ARG_2_HI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv8hi3, "__builtin_ia32_pcomlew", IX86_BUILTIN_PCOMLEW, LE, (int)MULTI_ARG_2_HI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv8hi3, "__builtin_ia32_pcomgtw", IX86_BUILTIN_PCOMGTW, GT, (int)MULTI_ARG_2_HI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv8hi3, "__builtin_ia32_pcomgew", IX86_BUILTIN_PCOMGEW, GE, (int)MULTI_ARG_2_HI_CMP },
+
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4si3, "__builtin_ia32_pcomeqd", IX86_BUILTIN_PCOMEQD, EQ, (int)MULTI_ARG_2_SI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4si3, "__builtin_ia32_pcomned", IX86_BUILTIN_PCOMNED, NE, (int)MULTI_ARG_2_SI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4si3, "__builtin_ia32_pcomneqd", IX86_BUILTIN_PCOMNED, NE, (int)MULTI_ARG_2_SI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4si3, "__builtin_ia32_pcomltd", IX86_BUILTIN_PCOMLTD, LT, (int)MULTI_ARG_2_SI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4si3, "__builtin_ia32_pcomled", IX86_BUILTIN_PCOMLED, LE, (int)MULTI_ARG_2_SI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4si3, "__builtin_ia32_pcomgtd", IX86_BUILTIN_PCOMGTD, GT, (int)MULTI_ARG_2_SI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv4si3, "__builtin_ia32_pcomged", IX86_BUILTIN_PCOMGED, GE, (int)MULTI_ARG_2_SI_CMP },
+
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2di3, "__builtin_ia32_pcomeqq", IX86_BUILTIN_PCOMEQQ, EQ, (int)MULTI_ARG_2_DI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2di3, "__builtin_ia32_pcomneq", IX86_BUILTIN_PCOMNEQ, NE, (int)MULTI_ARG_2_DI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2di3, "__builtin_ia32_pcomneqq", IX86_BUILTIN_PCOMNEQ, NE, (int)MULTI_ARG_2_DI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2di3, "__builtin_ia32_pcomltq", IX86_BUILTIN_PCOMLTQ, LT, (int)MULTI_ARG_2_DI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2di3, "__builtin_ia32_pcomleq", IX86_BUILTIN_PCOMLEQ, LE, (int)MULTI_ARG_2_DI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2di3, "__builtin_ia32_pcomgtq", IX86_BUILTIN_PCOMGTQ, GT, (int)MULTI_ARG_2_DI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmpv2di3, "__builtin_ia32_pcomgeq", IX86_BUILTIN_PCOMGEQ, GE, (int)MULTI_ARG_2_DI_CMP },
+
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_uns2v16qi3,"__builtin_ia32_pcomequb", IX86_BUILTIN_PCOMEQUB, EQ, (int)MULTI_ARG_2_QI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_uns2v16qi3,"__builtin_ia32_pcomneub", IX86_BUILTIN_PCOMNEUB, NE, (int)MULTI_ARG_2_QI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_uns2v16qi3,"__builtin_ia32_pcomnequb", IX86_BUILTIN_PCOMNEUB, NE, (int)MULTI_ARG_2_QI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_unsv16qi3, "__builtin_ia32_pcomltub", IX86_BUILTIN_PCOMLTUB, LTU, (int)MULTI_ARG_2_QI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_unsv16qi3, "__builtin_ia32_pcomleub", IX86_BUILTIN_PCOMLEUB, LEU, (int)MULTI_ARG_2_QI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_unsv16qi3, "__builtin_ia32_pcomgtub", IX86_BUILTIN_PCOMGTUB, GTU, (int)MULTI_ARG_2_QI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_unsv16qi3, "__builtin_ia32_pcomgeub", IX86_BUILTIN_PCOMGEUB, GEU, (int)MULTI_ARG_2_QI_CMP },
+
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_uns2v8hi3, "__builtin_ia32_pcomequw", IX86_BUILTIN_PCOMEQUW, EQ, (int)MULTI_ARG_2_HI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_uns2v8hi3, "__builtin_ia32_pcomneuw", IX86_BUILTIN_PCOMNEUW, NE, (int)MULTI_ARG_2_HI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_uns2v8hi3, "__builtin_ia32_pcomnequw", IX86_BUILTIN_PCOMNEUW, NE, (int)MULTI_ARG_2_HI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_unsv8hi3, "__builtin_ia32_pcomltuw", IX86_BUILTIN_PCOMLTUW, LTU, (int)MULTI_ARG_2_HI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_unsv8hi3, "__builtin_ia32_pcomleuw", IX86_BUILTIN_PCOMLEUW, LEU, (int)MULTI_ARG_2_HI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_unsv8hi3, "__builtin_ia32_pcomgtuw", IX86_BUILTIN_PCOMGTUW, GTU, (int)MULTI_ARG_2_HI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_unsv8hi3, "__builtin_ia32_pcomgeuw", IX86_BUILTIN_PCOMGEUW, GEU, (int)MULTI_ARG_2_HI_CMP },
+
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_uns2v4si3, "__builtin_ia32_pcomequd", IX86_BUILTIN_PCOMEQUD, EQ, (int)MULTI_ARG_2_SI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_uns2v4si3, "__builtin_ia32_pcomneud", IX86_BUILTIN_PCOMNEUD, NE, (int)MULTI_ARG_2_SI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_uns2v4si3, "__builtin_ia32_pcomnequd", IX86_BUILTIN_PCOMNEUD, NE, (int)MULTI_ARG_2_SI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_unsv4si3, "__builtin_ia32_pcomltud", IX86_BUILTIN_PCOMLTUD, LTU, (int)MULTI_ARG_2_SI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_unsv4si3, "__builtin_ia32_pcomleud", IX86_BUILTIN_PCOMLEUD, LEU, (int)MULTI_ARG_2_SI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_unsv4si3, "__builtin_ia32_pcomgtud", IX86_BUILTIN_PCOMGTUD, GTU, (int)MULTI_ARG_2_SI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_unsv4si3, "__builtin_ia32_pcomgeud", IX86_BUILTIN_PCOMGEUD, GEU, (int)MULTI_ARG_2_SI_CMP },
+
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_uns2v2di3, "__builtin_ia32_pcomequq", IX86_BUILTIN_PCOMEQUQ, EQ, (int)MULTI_ARG_2_DI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_uns2v2di3, "__builtin_ia32_pcomneuq", IX86_BUILTIN_PCOMNEUQ, NE, (int)MULTI_ARG_2_DI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_uns2v2di3, "__builtin_ia32_pcomnequq", IX86_BUILTIN_PCOMNEUQ, NE, (int)MULTI_ARG_2_DI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_unsv2di3, "__builtin_ia32_pcomltuq", IX86_BUILTIN_PCOMLTUQ, LTU, (int)MULTI_ARG_2_DI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_unsv2di3, "__builtin_ia32_pcomleuq", IX86_BUILTIN_PCOMLEUQ, LEU, (int)MULTI_ARG_2_DI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_unsv2di3, "__builtin_ia32_pcomgtuq", IX86_BUILTIN_PCOMGTUQ, GTU, (int)MULTI_ARG_2_DI_CMP },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_maskcmp_unsv2di3, "__builtin_ia32_pcomgeuq", IX86_BUILTIN_PCOMGEUQ, GEU, (int)MULTI_ARG_2_DI_CMP },
+
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_com_tfv4sf3, "__builtin_ia32_comfalsess", IX86_BUILTIN_COMFALSESS, COM_FALSE_S, (int)MULTI_ARG_2_SF_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_com_tfv4sf3, "__builtin_ia32_comtruess", IX86_BUILTIN_COMTRUESS, COM_TRUE_S, (int)MULTI_ARG_2_SF_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_com_tfv4sf3, "__builtin_ia32_comfalseps", IX86_BUILTIN_COMFALSEPS, COM_FALSE_P, (int)MULTI_ARG_2_SF_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_com_tfv4sf3, "__builtin_ia32_comtrueps", IX86_BUILTIN_COMTRUEPS, COM_TRUE_P, (int)MULTI_ARG_2_SF_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_com_tfv2df3, "__builtin_ia32_comfalsesd", IX86_BUILTIN_COMFALSESD, COM_FALSE_S, (int)MULTI_ARG_2_DF_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_com_tfv2df3, "__builtin_ia32_comtruesd", IX86_BUILTIN_COMTRUESD, COM_TRUE_S, (int)MULTI_ARG_2_DF_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_com_tfv2df3, "__builtin_ia32_comfalsepd", IX86_BUILTIN_COMFALSEPD, COM_FALSE_P, (int)MULTI_ARG_2_DF_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_com_tfv2df3, "__builtin_ia32_comtruepd", IX86_BUILTIN_COMTRUEPD, COM_TRUE_P, (int)MULTI_ARG_2_DF_TF },
+
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv16qi3, "__builtin_ia32_pcomfalseb", IX86_BUILTIN_PCOMFALSEB, PCOM_FALSE, (int)MULTI_ARG_2_QI_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv8hi3, "__builtin_ia32_pcomfalsew", IX86_BUILTIN_PCOMFALSEW, PCOM_FALSE, (int)MULTI_ARG_2_HI_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv4si3, "__builtin_ia32_pcomfalsed", IX86_BUILTIN_PCOMFALSED, PCOM_FALSE, (int)MULTI_ARG_2_SI_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv2di3, "__builtin_ia32_pcomfalseq", IX86_BUILTIN_PCOMFALSEQ, PCOM_FALSE, (int)MULTI_ARG_2_DI_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv16qi3, "__builtin_ia32_pcomfalseub",IX86_BUILTIN_PCOMFALSEUB,PCOM_FALSE, (int)MULTI_ARG_2_QI_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv8hi3, "__builtin_ia32_pcomfalseuw",IX86_BUILTIN_PCOMFALSEUW,PCOM_FALSE, (int)MULTI_ARG_2_HI_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv4si3, "__builtin_ia32_pcomfalseud",IX86_BUILTIN_PCOMFALSEUD,PCOM_FALSE, (int)MULTI_ARG_2_SI_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv2di3, "__builtin_ia32_pcomfalseuq",IX86_BUILTIN_PCOMFALSEUQ,PCOM_FALSE, (int)MULTI_ARG_2_DI_TF },
+
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv16qi3, "__builtin_ia32_pcomtrueb", IX86_BUILTIN_PCOMTRUEB, PCOM_TRUE, (int)MULTI_ARG_2_QI_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv8hi3, "__builtin_ia32_pcomtruew", IX86_BUILTIN_PCOMTRUEW, PCOM_TRUE, (int)MULTI_ARG_2_HI_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv4si3, "__builtin_ia32_pcomtrued", IX86_BUILTIN_PCOMTRUED, PCOM_TRUE, (int)MULTI_ARG_2_SI_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv2di3, "__builtin_ia32_pcomtrueq", IX86_BUILTIN_PCOMTRUEQ, PCOM_TRUE, (int)MULTI_ARG_2_DI_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv16qi3, "__builtin_ia32_pcomtrueub", IX86_BUILTIN_PCOMTRUEUB, PCOM_TRUE, (int)MULTI_ARG_2_QI_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv8hi3, "__builtin_ia32_pcomtrueuw", IX86_BUILTIN_PCOMTRUEUW, PCOM_TRUE, (int)MULTI_ARG_2_HI_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv4si3, "__builtin_ia32_pcomtrueud", IX86_BUILTIN_PCOMTRUEUD, PCOM_TRUE, (int)MULTI_ARG_2_SI_TF },
+ { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv2di3, "__builtin_ia32_pcomtrueuq", IX86_BUILTIN_PCOMTRUEUQ, PCOM_TRUE, (int)MULTI_ARG_2_DI_TF },
};
/* Set up all the MMX/SSE builtins. This is not called if TARGET_MMX
tree V16QI_type_node = build_vector_type_for_mode (char_type_node, V16QImode);
tree V2SI_type_node = build_vector_type_for_mode (intSI_type_node, V2SImode);
+ tree V1DI_type_node
+ = build_vector_type_for_mode (long_long_integer_type_node, V1DImode);
tree V2SF_type_node = build_vector_type_for_mode (float_type_node, V2SFmode);
tree V2DI_type_node
= build_vector_type_for_mode (long_long_integer_type_node, V2DImode);
tree V8HI_type_node = build_vector_type_for_mode (intHI_type_node, V8HImode);
tree pchar_type_node = build_pointer_type (char_type_node);
- tree pcchar_type_node = build_pointer_type (
- build_type_variant (char_type_node, 1, 0));
+ tree pcchar_type_node
+ = build_pointer_type (build_type_variant (char_type_node, 1, 0));
tree pfloat_type_node = build_pointer_type (float_type_node);
- tree pcfloat_type_node = build_pointer_type (
- build_type_variant (float_type_node, 1, 0));
- tree pv2si_type_node = build_pointer_type (V2SI_type_node);
+ tree pcfloat_type_node
+ = build_pointer_type (build_type_variant (float_type_node, 1, 0));
+ tree pv2sf_type_node = build_pointer_type (V2SF_type_node);
+ tree pcv2sf_type_node
+ = build_pointer_type (build_type_variant (V2SF_type_node, 1, 0));
tree pv2di_type_node = build_pointer_type (V2DI_type_node);
tree pdi_type_node = build_pointer_type (long_long_unsigned_type_node);
tree v4hi_ftype_v4hi_int
= build_function_type_list (V4HI_type_node,
V4HI_type_node, integer_type_node, NULL_TREE);
- tree v4hi_ftype_v4hi_di
- = build_function_type_list (V4HI_type_node,
- V4HI_type_node, long_long_unsigned_type_node,
- NULL_TREE);
- tree v2si_ftype_v2si_di
+ tree v2si_ftype_v2si_int
= build_function_type_list (V2SI_type_node,
- V2SI_type_node, long_long_unsigned_type_node,
- NULL_TREE);
+ V2SI_type_node, integer_type_node, NULL_TREE);
+ tree v1di_ftype_v1di_int
+ = build_function_type_list (V1DI_type_node,
+ V1DI_type_node, integer_type_node, NULL_TREE);
+
tree void_ftype_void
= build_function_type (void_type_node, void_list_node);
tree void_ftype_unsigned
pchar_type_node, NULL_TREE);
tree v4sf_ftype_pcfloat
= build_function_type_list (V4SF_type_node, pcfloat_type_node, NULL_TREE);
- /* @@@ the type is bogus */
- tree v4sf_ftype_v4sf_pv2si
+ tree v4sf_ftype_v4sf_pcv2sf
= build_function_type_list (V4SF_type_node,
- V4SF_type_node, pv2si_type_node, NULL_TREE);
- tree void_ftype_pv2si_v4sf
+ V4SF_type_node, pcv2sf_type_node, NULL_TREE);
+ tree void_ftype_pv2sf_v4sf
= build_function_type_list (void_type_node,
- pv2si_type_node, V4SF_type_node, NULL_TREE);
+ pv2sf_type_node, V4SF_type_node, NULL_TREE);
tree void_ftype_pfloat_v4sf
= build_function_type_list (void_type_node,
pfloat_type_node, V4SF_type_node, NULL_TREE);
tree v2si_ftype_v2si_v2si
= build_function_type_list (V2SI_type_node,
V2SI_type_node, V2SI_type_node, NULL_TREE);
- tree di_ftype_di_di
- = build_function_type_list (long_long_unsigned_type_node,
- long_long_unsigned_type_node,
- long_long_unsigned_type_node, NULL_TREE);
-
- tree di_ftype_di_di_int
- = build_function_type_list (long_long_unsigned_type_node,
- long_long_unsigned_type_node,
- long_long_unsigned_type_node,
+ tree v1di_ftype_v1di_v1di
+ = build_function_type_list (V1DI_type_node,
+ V1DI_type_node, V1DI_type_node, NULL_TREE);
+ tree v1di_ftype_v1di_v1di_int
+ = build_function_type_list (V1DI_type_node,
+ V1DI_type_node, V1DI_type_node,
integer_type_node, NULL_TREE);
-
tree v2si_ftype_v2sf
= build_function_type_list (V2SI_type_node, V2SF_type_node, NULL_TREE);
tree v2sf_ftype_v2si
= build_function_type_list (V2DF_type_node, V4SI_type_node, NULL_TREE);
tree v4si_ftype_v2df
= build_function_type_list (V4SI_type_node, V2DF_type_node, NULL_TREE);
+ tree v4si_ftype_v2df_v2df
+ = build_function_type_list (V4SI_type_node,
+ V2DF_type_node, V2DF_type_node, NULL_TREE);
tree v2si_ftype_v2df
= build_function_type_list (V2SI_type_node, V2DF_type_node, NULL_TREE);
tree v4sf_ftype_v2df
tree v4si_ftype_v8hi_v8hi
= build_function_type_list (V4SI_type_node,
V8HI_type_node, V8HI_type_node, NULL_TREE);
- tree di_ftype_v8qi_v8qi
- = build_function_type_list (long_long_unsigned_type_node,
+ tree v1di_ftype_v8qi_v8qi
+ = build_function_type_list (V1DI_type_node,
V8QI_type_node, V8QI_type_node, NULL_TREE);
- tree di_ftype_v2si_v2si
- = build_function_type_list (long_long_unsigned_type_node,
+ tree v1di_ftype_v2si_v2si
+ = build_function_type_list (V1DI_type_node,
V2SI_type_node, V2SI_type_node, NULL_TREE);
tree v2di_ftype_v16qi_v16qi
= build_function_type_list (V2DI_type_node,
tree v2di_ftype_v2di_v16qi
= build_function_type_list (V2DI_type_node, V2DI_type_node, V16QI_type_node,
NULL_TREE);
+ tree v2df_ftype_v2df_v2df_v2df
+ = build_function_type_list (V2DF_type_node,
+ V2DF_type_node, V2DF_type_node,
+ V2DF_type_node, NULL_TREE);
+ tree v4sf_ftype_v4sf_v4sf_v4sf
+ = build_function_type_list (V4SF_type_node,
+ V4SF_type_node, V4SF_type_node,
+ V4SF_type_node, NULL_TREE);
+ tree v8hi_ftype_v16qi
+ = build_function_type_list (V8HI_type_node, V16QI_type_node,
+ NULL_TREE);
+ tree v4si_ftype_v16qi
+ = build_function_type_list (V4SI_type_node, V16QI_type_node,
+ NULL_TREE);
+ tree v2di_ftype_v16qi
+ = build_function_type_list (V2DI_type_node, V16QI_type_node,
+ NULL_TREE);
+ tree v4si_ftype_v8hi
+ = build_function_type_list (V4SI_type_node, V8HI_type_node,
+ NULL_TREE);
+ tree v2di_ftype_v8hi
+ = build_function_type_list (V2DI_type_node, V8HI_type_node,
+ NULL_TREE);
+ tree v2di_ftype_v4si
+ = build_function_type_list (V2DI_type_node, V4SI_type_node,
+ NULL_TREE);
+ tree v2di_ftype_pv2di
+ = build_function_type_list (V2DI_type_node, pv2di_type_node,
+ NULL_TREE);
+ tree v16qi_ftype_v16qi_v16qi_int
+ = build_function_type_list (V16QI_type_node, V16QI_type_node,
+ V16QI_type_node, integer_type_node,
+ NULL_TREE);
+ tree v16qi_ftype_v16qi_v16qi_v16qi
+ = build_function_type_list (V16QI_type_node, V16QI_type_node,
+ V16QI_type_node, V16QI_type_node,
+ NULL_TREE);
+ tree v8hi_ftype_v8hi_v8hi_int
+ = build_function_type_list (V8HI_type_node, V8HI_type_node,
+ V8HI_type_node, integer_type_node,
+ NULL_TREE);
+ tree v4si_ftype_v4si_v4si_int
+ = build_function_type_list (V4SI_type_node, V4SI_type_node,
+ V4SI_type_node, integer_type_node,
+ NULL_TREE);
+ tree int_ftype_v2di_v2di
+ = build_function_type_list (integer_type_node,
+ V2DI_type_node, V2DI_type_node,
+ NULL_TREE);
+ tree int_ftype_v16qi_int_v16qi_int_int
+ = build_function_type_list (integer_type_node,
+ V16QI_type_node,
+ integer_type_node,
+ V16QI_type_node,
+ integer_type_node,
+ integer_type_node,
+ NULL_TREE);
+ tree v16qi_ftype_v16qi_int_v16qi_int_int
+ = build_function_type_list (V16QI_type_node,
+ V16QI_type_node,
+ integer_type_node,
+ V16QI_type_node,
+ integer_type_node,
+ integer_type_node,
+ NULL_TREE);
+ tree int_ftype_v16qi_v16qi_int
+ = build_function_type_list (integer_type_node,
+ V16QI_type_node,
+ V16QI_type_node,
+ integer_type_node,
+ NULL_TREE);
+
+ /* SSE5 instructions */
+ tree v2di_ftype_v2di_v2di_v2di
+ = build_function_type_list (V2DI_type_node,
+ V2DI_type_node,
+ V2DI_type_node,
+ V2DI_type_node,
+ NULL_TREE);
+
+ tree v4si_ftype_v4si_v4si_v4si
+ = build_function_type_list (V4SI_type_node,
+ V4SI_type_node,
+ V4SI_type_node,
+ V4SI_type_node,
+ NULL_TREE);
+
+ tree v4si_ftype_v4si_v4si_v2di
+ = build_function_type_list (V4SI_type_node,
+ V4SI_type_node,
+ V4SI_type_node,
+ V2DI_type_node,
+ NULL_TREE);
+
+ tree v8hi_ftype_v8hi_v8hi_v8hi
+ = build_function_type_list (V8HI_type_node,
+ V8HI_type_node,
+ V8HI_type_node,
+ V8HI_type_node,
+ NULL_TREE);
+
+ tree v8hi_ftype_v8hi_v8hi_v4si
+ = build_function_type_list (V8HI_type_node,
+ V8HI_type_node,
+ V8HI_type_node,
+ V4SI_type_node,
+ NULL_TREE);
+
+ tree v2df_ftype_v2df_v2df_v16qi
+ = build_function_type_list (V2DF_type_node,
+ V2DF_type_node,
+ V2DF_type_node,
+ V16QI_type_node,
+ NULL_TREE);
+
+ tree v4sf_ftype_v4sf_v4sf_v16qi
+ = build_function_type_list (V4SF_type_node,
+ V4SF_type_node,
+ V4SF_type_node,
+ V16QI_type_node,
+ NULL_TREE);
+
+ tree v2di_ftype_v2di_si
+ = build_function_type_list (V2DI_type_node,
+ V2DI_type_node,
+ integer_type_node,
+ NULL_TREE);
+
+ tree v4si_ftype_v4si_si
+ = build_function_type_list (V4SI_type_node,
+ V4SI_type_node,
+ integer_type_node,
+ NULL_TREE);
+
+ tree v8hi_ftype_v8hi_si
+ = build_function_type_list (V8HI_type_node,
+ V8HI_type_node,
+ integer_type_node,
+ NULL_TREE);
+
+ tree v16qi_ftype_v16qi_si
+ = build_function_type_list (V16QI_type_node,
+ V16QI_type_node,
+ integer_type_node,
+ NULL_TREE);
+ tree v4sf_ftype_v4hi
+ = build_function_type_list (V4SF_type_node,
+ V4HI_type_node,
+ NULL_TREE);
+
+ tree v4hi_ftype_v4sf
+ = build_function_type_list (V4HI_type_node,
+ V4SF_type_node,
+ NULL_TREE);
+
+ tree v2di_ftype_v2di
+ = build_function_type_list (V2DI_type_node, V2DI_type_node, NULL_TREE);
+
+ tree v16qi_ftype_v8hi_v8hi
+ = build_function_type_list (V16QI_type_node,
+ V8HI_type_node, V8HI_type_node,
+ NULL_TREE);
+ tree v8hi_ftype_v4si_v4si
+ = build_function_type_list (V8HI_type_node,
+ V4SI_type_node, V4SI_type_node,
+ NULL_TREE);
+ tree v8hi_ftype_v16qi_v16qi
+ = build_function_type_list (V8HI_type_node,
+ V16QI_type_node, V16QI_type_node,
+ NULL_TREE);
+ tree v4hi_ftype_v8qi_v8qi
+ = build_function_type_list (V4HI_type_node,
+ V8QI_type_node, V8QI_type_node,
+ NULL_TREE);
+ tree unsigned_ftype_unsigned_uchar
+ = build_function_type_list (unsigned_type_node,
+ unsigned_type_node,
+ unsigned_char_type_node,
+ NULL_TREE);
+ tree unsigned_ftype_unsigned_ushort
+ = build_function_type_list (unsigned_type_node,
+ unsigned_type_node,
+ short_unsigned_type_node,
+ NULL_TREE);
+ tree unsigned_ftype_unsigned_unsigned
+ = build_function_type_list (unsigned_type_node,
+ unsigned_type_node,
+ unsigned_type_node,
+ NULL_TREE);
+ tree uint64_ftype_uint64_uint64
+ = build_function_type_list (long_long_unsigned_type_node,
+ long_long_unsigned_type_node,
+ long_long_unsigned_type_node,
+ NULL_TREE);
+ tree float_ftype_float
+ = build_function_type_list (float_type_node,
+ float_type_node,
+ NULL_TREE);
- tree float80_type;
- tree float128_type;
tree ftype;
/* The __float80 type. */
else
{
/* The __float80 type. */
- float80_type = make_node (REAL_TYPE);
- TYPE_PRECISION (float80_type) = 80;
- layout_type (float80_type);
- (*lang_hooks.types.register_builtin_type) (float80_type, "__float80");
+ tree float80_type_node = make_node (REAL_TYPE);
+
+ TYPE_PRECISION (float80_type_node) = 80;
+ layout_type (float80_type_node);
+ (*lang_hooks.types.register_builtin_type) (float80_type_node,
+ "__float80");
}
if (TARGET_64BIT)
{
- float128_type = make_node (REAL_TYPE);
- TYPE_PRECISION (float128_type) = 128;
- layout_type (float128_type);
- (*lang_hooks.types.register_builtin_type) (float128_type, "__float128");
+ tree float128_type_node = make_node (REAL_TYPE);
+
+ TYPE_PRECISION (float128_type_node) = 128;
+ layout_type (float128_type_node);
+ (*lang_hooks.types.register_builtin_type) (float128_type_node,
+ "__float128");
+
+ /* TFmode support builtins. */
+ ftype = build_function_type (float128_type_node,
+ void_list_node);
+ def_builtin (OPTION_MASK_ISA_64BIT, "__builtin_infq", ftype, IX86_BUILTIN_INFQ);
+
+ ftype = build_function_type_list (float128_type_node,
+ float128_type_node,
+ NULL_TREE);
+ def_builtin_const (OPTION_MASK_ISA_64BIT, "__builtin_fabsq", ftype, IX86_BUILTIN_FABSQ);
+
+ ftype = build_function_type_list (float128_type_node,
+ float128_type_node,
+ float128_type_node,
+ NULL_TREE);
+ def_builtin_const (OPTION_MASK_ISA_64BIT, "__builtin_copysignq", ftype, IX86_BUILTIN_COPYSIGNQ);
}
- /* Add all builtins that are more or less simple operations on two
- operands. */
- for (i = 0, d = bdesc_2arg; i < ARRAY_SIZE (bdesc_2arg); i++, d++)
+ /* Add all special builtins with variable number of operands. */
+ for (i = 0, d = bdesc_special_args;
+ i < ARRAY_SIZE (bdesc_special_args);
+ i++, d++)
{
- /* Use one of the operands; the target can have a different mode for
- mask-generating compares. */
- enum machine_mode mode;
tree type;
if (d->name == 0)
continue;
- mode = insn_data[d->icode].operand[1].mode;
- switch (mode)
+ switch ((enum ix86_special_builtin_type) d->flag)
{
- case V16QImode:
- type = v16qi_ftype_v16qi_v16qi;
+ case VOID_FTYPE_VOID:
+ type = void_ftype_void;
break;
- case V8HImode:
- type = v8hi_ftype_v8hi_v8hi;
+ case V16QI_FTYPE_PCCHAR:
+ type = v16qi_ftype_pcchar;
break;
- case V4SImode:
- type = v4si_ftype_v4si_v4si;
+ case V4SF_FTYPE_PCFLOAT:
+ type = v4sf_ftype_pcfloat;
break;
- case V2DImode:
- type = v2di_ftype_v2di_v2di;
+ case V2DI_FTYPE_PV2DI:
+ type = v2di_ftype_pv2di;
break;
- case V2DFmode:
- type = v2df_ftype_v2df_v2df;
+ case V2DF_FTYPE_PCDOUBLE:
+ type = v2df_ftype_pcdouble;
break;
- case V4SFmode:
- type = v4sf_ftype_v4sf_v4sf;
+ case V4SF_FTYPE_V4SF_PCV2SF:
+ type = v4sf_ftype_v4sf_pcv2sf;
break;
- case V8QImode:
- type = v8qi_ftype_v8qi_v8qi;
+ case V2DF_FTYPE_V2DF_PCDOUBLE:
+ type = v2df_ftype_v2df_pcdouble;
break;
- case V4HImode:
- type = v4hi_ftype_v4hi_v4hi;
+ case VOID_FTYPE_PV2SF_V4SF:
+ type = void_ftype_pv2sf_v4sf;
break;
- case V2SImode:
- type = v2si_ftype_v2si_v2si;
+ case VOID_FTYPE_PV2DI_V2DI:
+ type = void_ftype_pv2di_v2di;
break;
- case DImode:
- type = di_ftype_di_di;
+ case VOID_FTYPE_PCHAR_V16QI:
+ type = void_ftype_pchar_v16qi;
+ break;
+ case VOID_FTYPE_PFLOAT_V4SF:
+ type = void_ftype_pfloat_v4sf;
+ break;
+ case VOID_FTYPE_PDOUBLE_V2DF:
+ type = void_ftype_pdouble_v2df;
+ break;
+ case VOID_FTYPE_PDI_DI:
+ type = void_ftype_pdi_di;
+ break;
+ case VOID_FTYPE_PINT_INT:
+ type = void_ftype_pint_int;
break;
-
default:
gcc_unreachable ();
}
- /* Override for comparisons. */
- if (d->icode == CODE_FOR_sse_maskcmpv4sf3
- || d->icode == CODE_FOR_sse_vmmaskcmpv4sf3)
- type = v4si_ftype_v4sf_v4sf;
-
- if (d->icode == CODE_FOR_sse2_maskcmpv2df3
- || d->icode == CODE_FOR_sse2_vmmaskcmpv2df3)
- type = v2di_ftype_v2df_v2df;
-
def_builtin (d->mask, d->name, type, d->code);
}
- /* Add all builtins that are more or less simple operations on 1 operand. */
- for (i = 0, d = bdesc_1arg; i < ARRAY_SIZE (bdesc_1arg); i++, d++)
+ /* Add all builtins with variable number of operands. */
+ for (i = 0, d = bdesc_args;
+ i < ARRAY_SIZE (bdesc_args);
+ i++, d++)
{
- enum machine_mode mode;
tree type;
if (d->name == 0)
continue;
- mode = insn_data[d->icode].operand[1].mode;
- switch (mode)
+ switch ((enum ix86_builtin_type) d->flag)
{
- case V16QImode:
+ case FLOAT_FTYPE_FLOAT:
+ type = float_ftype_float;
+ break;
+ case INT_FTYPE_V2DI_V2DI_PTEST:
+ type = int_ftype_v2di_v2di;
+ break;
+ case INT64_FTYPE_V4SF:
+ type = int64_ftype_v4sf;
+ break;
+ case INT64_FTYPE_V2DF:
+ type = int64_ftype_v2df;
+ break;
+ case INT_FTYPE_V16QI:
+ type = int_ftype_v16qi;
+ break;
+ case INT_FTYPE_V8QI:
+ type = int_ftype_v8qi;
+ break;
+ case INT_FTYPE_V4SF:
+ type = int_ftype_v4sf;
+ break;
+ case INT_FTYPE_V2DF:
+ type = int_ftype_v2df;
+ break;
+ case V16QI_FTYPE_V16QI:
type = v16qi_ftype_v16qi;
break;
- case V8HImode:
+ case V8HI_FTYPE_V8HI:
type = v8hi_ftype_v8hi;
break;
- case V4SImode:
+ case V8HI_FTYPE_V16QI:
+ type = v8hi_ftype_v16qi;
+ break;
+ case V8QI_FTYPE_V8QI:
+ type = v8qi_ftype_v8qi;
+ break;
+ case V4SI_FTYPE_V4SI:
type = v4si_ftype_v4si;
break;
- case V2DFmode:
- type = v2df_ftype_v2df;
+ case V4SI_FTYPE_V16QI:
+ type = v4si_ftype_v16qi;
break;
- case V4SFmode:
- type = v4sf_ftype_v4sf;
+ case V4SI_FTYPE_V8HI:
+ type = v4si_ftype_v8hi;
break;
- case V8QImode:
- type = v8qi_ftype_v8qi;
+ case V4SI_FTYPE_V4SF:
+ type = v4si_ftype_v4sf;
break;
- case V4HImode:
+ case V4SI_FTYPE_V2DF:
+ type = v4si_ftype_v2df;
+ break;
+ case V4HI_FTYPE_V4HI:
type = v4hi_ftype_v4hi;
break;
- case V2SImode:
+ case V4SF_FTYPE_V4SF:
+ case V4SF_FTYPE_V4SF_VEC_MERGE:
+ type = v4sf_ftype_v4sf;
+ break;
+ case V4SF_FTYPE_V4SI:
+ type = v4sf_ftype_v4si;
+ break;
+ case V4SF_FTYPE_V2DF:
+ type = v4sf_ftype_v2df;
+ break;
+ case V2DI_FTYPE_V2DI:
+ type = v2di_ftype_v2di;
+ break;
+ case V2DI_FTYPE_V16QI:
+ type = v2di_ftype_v16qi;
+ break;
+ case V2DI_FTYPE_V8HI:
+ type = v2di_ftype_v8hi;
+ break;
+ case V2DI_FTYPE_V4SI:
+ type = v2di_ftype_v4si;
+ break;
+ case V2SI_FTYPE_V2SI:
type = v2si_ftype_v2si;
break;
-
- default:
- abort ();
- }
-
- def_builtin (d->mask, d->name, type, d->code);
- }
-
- /* Add the remaining MMX insns with somewhat more complicated types. */
- def_builtin (MASK_MMX, "__builtin_ia32_emms", void_ftype_void, IX86_BUILTIN_EMMS);
- def_builtin (MASK_MMX, "__builtin_ia32_psllw", v4hi_ftype_v4hi_di, IX86_BUILTIN_PSLLW);
- def_builtin (MASK_MMX, "__builtin_ia32_pslld", v2si_ftype_v2si_di, IX86_BUILTIN_PSLLD);
- def_builtin (MASK_MMX, "__builtin_ia32_psllq", di_ftype_di_di, IX86_BUILTIN_PSLLQ);
-
- def_builtin (MASK_MMX, "__builtin_ia32_psrlw", v4hi_ftype_v4hi_di, IX86_BUILTIN_PSRLW);
- def_builtin (MASK_MMX, "__builtin_ia32_psrld", v2si_ftype_v2si_di, IX86_BUILTIN_PSRLD);
- def_builtin (MASK_MMX, "__builtin_ia32_psrlq", di_ftype_di_di, IX86_BUILTIN_PSRLQ);
-
- def_builtin (MASK_MMX, "__builtin_ia32_psraw", v4hi_ftype_v4hi_di, IX86_BUILTIN_PSRAW);
- def_builtin (MASK_MMX, "__builtin_ia32_psrad", v2si_ftype_v2si_di, IX86_BUILTIN_PSRAD);
-
- def_builtin (MASK_SSE | MASK_3DNOW_A, "__builtin_ia32_pshufw", v4hi_ftype_v4hi_int, IX86_BUILTIN_PSHUFW);
- def_builtin (MASK_MMX, "__builtin_ia32_pmaddwd", v2si_ftype_v4hi_v4hi, IX86_BUILTIN_PMADDWD);
-
- /* comi/ucomi insns. */
- for (i = 0, d = bdesc_comi; i < ARRAY_SIZE (bdesc_comi); i++, d++)
- if (d->mask == MASK_SSE2)
- def_builtin (d->mask, d->name, int_ftype_v2df_v2df, d->code);
- else
- def_builtin (d->mask, d->name, int_ftype_v4sf_v4sf, d->code);
-
- def_builtin (MASK_MMX, "__builtin_ia32_packsswb", v8qi_ftype_v4hi_v4hi, IX86_BUILTIN_PACKSSWB);
- def_builtin (MASK_MMX, "__builtin_ia32_packssdw", v4hi_ftype_v2si_v2si, IX86_BUILTIN_PACKSSDW);
- def_builtin (MASK_MMX, "__builtin_ia32_packuswb", v8qi_ftype_v4hi_v4hi, IX86_BUILTIN_PACKUSWB);
-
- def_builtin (MASK_SSE, "__builtin_ia32_ldmxcsr", void_ftype_unsigned, IX86_BUILTIN_LDMXCSR);
- def_builtin (MASK_SSE, "__builtin_ia32_stmxcsr", unsigned_ftype_void, IX86_BUILTIN_STMXCSR);
- def_builtin_const (MASK_SSE, "__builtin_ia32_cvtpi2ps", v4sf_ftype_v4sf_v2si, IX86_BUILTIN_CVTPI2PS);
- def_builtin_const (MASK_SSE, "__builtin_ia32_cvtps2pi", v2si_ftype_v4sf, IX86_BUILTIN_CVTPS2PI);
- def_builtin_const (MASK_SSE, "__builtin_ia32_cvtsi2ss", v4sf_ftype_v4sf_int, IX86_BUILTIN_CVTSI2SS);
- def_builtin_const (MASK_SSE | MASK_64BIT, "__builtin_ia32_cvtsi642ss", v4sf_ftype_v4sf_int64, IX86_BUILTIN_CVTSI642SS);
- def_builtin_const (MASK_SSE, "__builtin_ia32_cvtss2si", int_ftype_v4sf, IX86_BUILTIN_CVTSS2SI);
- def_builtin_const (MASK_SSE | MASK_64BIT, "__builtin_ia32_cvtss2si64", int64_ftype_v4sf, IX86_BUILTIN_CVTSS2SI64);
- def_builtin_const (MASK_SSE, "__builtin_ia32_cvttps2pi", v2si_ftype_v4sf, IX86_BUILTIN_CVTTPS2PI);
- def_builtin_const (MASK_SSE, "__builtin_ia32_cvttss2si", int_ftype_v4sf, IX86_BUILTIN_CVTTSS2SI);
- def_builtin_const (MASK_SSE | MASK_64BIT, "__builtin_ia32_cvttss2si64", int64_ftype_v4sf, IX86_BUILTIN_CVTTSS2SI64);
-
- def_builtin (MASK_SSE | MASK_3DNOW_A, "__builtin_ia32_maskmovq", void_ftype_v8qi_v8qi_pchar, IX86_BUILTIN_MASKMOVQ);
-
- def_builtin (MASK_SSE, "__builtin_ia32_loadups", v4sf_ftype_pcfloat, IX86_BUILTIN_LOADUPS);
- def_builtin (MASK_SSE, "__builtin_ia32_storeups", void_ftype_pfloat_v4sf, IX86_BUILTIN_STOREUPS);
-
- def_builtin (MASK_SSE, "__builtin_ia32_loadhps", v4sf_ftype_v4sf_pv2si, IX86_BUILTIN_LOADHPS);
- def_builtin (MASK_SSE, "__builtin_ia32_loadlps", v4sf_ftype_v4sf_pv2si, IX86_BUILTIN_LOADLPS);
- def_builtin (MASK_SSE, "__builtin_ia32_storehps", void_ftype_pv2si_v4sf, IX86_BUILTIN_STOREHPS);
- def_builtin (MASK_SSE, "__builtin_ia32_storelps", void_ftype_pv2si_v4sf, IX86_BUILTIN_STORELPS);
-
- def_builtin (MASK_SSE, "__builtin_ia32_movmskps", int_ftype_v4sf, IX86_BUILTIN_MOVMSKPS);
- def_builtin (MASK_SSE | MASK_3DNOW_A, "__builtin_ia32_pmovmskb", int_ftype_v8qi, IX86_BUILTIN_PMOVMSKB);
- def_builtin (MASK_SSE, "__builtin_ia32_movntps", void_ftype_pfloat_v4sf, IX86_BUILTIN_MOVNTPS);
- def_builtin (MASK_SSE | MASK_3DNOW_A, "__builtin_ia32_movntq", void_ftype_pdi_di, IX86_BUILTIN_MOVNTQ);
-
- def_builtin (MASK_SSE | MASK_3DNOW_A, "__builtin_ia32_sfence", void_ftype_void, IX86_BUILTIN_SFENCE);
-
- def_builtin (MASK_SSE | MASK_3DNOW_A, "__builtin_ia32_psadbw", di_ftype_v8qi_v8qi, IX86_BUILTIN_PSADBW);
-
- def_builtin (MASK_SSE, "__builtin_ia32_rcpps", v4sf_ftype_v4sf, IX86_BUILTIN_RCPPS);
- def_builtin (MASK_SSE, "__builtin_ia32_rcpss", v4sf_ftype_v4sf, IX86_BUILTIN_RCPSS);
- def_builtin (MASK_SSE, "__builtin_ia32_rsqrtps", v4sf_ftype_v4sf, IX86_BUILTIN_RSQRTPS);
- def_builtin (MASK_SSE, "__builtin_ia32_rsqrtss", v4sf_ftype_v4sf, IX86_BUILTIN_RSQRTSS);
- def_builtin_const (MASK_SSE, "__builtin_ia32_sqrtps", v4sf_ftype_v4sf, IX86_BUILTIN_SQRTPS);
- def_builtin_const (MASK_SSE, "__builtin_ia32_sqrtss", v4sf_ftype_v4sf, IX86_BUILTIN_SQRTSS);
-
- def_builtin (MASK_SSE, "__builtin_ia32_shufps", v4sf_ftype_v4sf_v4sf_int, IX86_BUILTIN_SHUFPS);
-
- /* Original 3DNow! */
- def_builtin (MASK_3DNOW, "__builtin_ia32_femms", void_ftype_void, IX86_BUILTIN_FEMMS);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pavgusb", v8qi_ftype_v8qi_v8qi, IX86_BUILTIN_PAVGUSB);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pf2id", v2si_ftype_v2sf, IX86_BUILTIN_PF2ID);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pfacc", v2sf_ftype_v2sf_v2sf, IX86_BUILTIN_PFACC);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pfadd", v2sf_ftype_v2sf_v2sf, IX86_BUILTIN_PFADD);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pfcmpeq", v2si_ftype_v2sf_v2sf, IX86_BUILTIN_PFCMPEQ);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pfcmpge", v2si_ftype_v2sf_v2sf, IX86_BUILTIN_PFCMPGE);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pfcmpgt", v2si_ftype_v2sf_v2sf, IX86_BUILTIN_PFCMPGT);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pfmax", v2sf_ftype_v2sf_v2sf, IX86_BUILTIN_PFMAX);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pfmin", v2sf_ftype_v2sf_v2sf, IX86_BUILTIN_PFMIN);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pfmul", v2sf_ftype_v2sf_v2sf, IX86_BUILTIN_PFMUL);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pfrcp", v2sf_ftype_v2sf, IX86_BUILTIN_PFRCP);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pfrcpit1", v2sf_ftype_v2sf_v2sf, IX86_BUILTIN_PFRCPIT1);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pfrcpit2", v2sf_ftype_v2sf_v2sf, IX86_BUILTIN_PFRCPIT2);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pfrsqrt", v2sf_ftype_v2sf, IX86_BUILTIN_PFRSQRT);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pfrsqit1", v2sf_ftype_v2sf_v2sf, IX86_BUILTIN_PFRSQIT1);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pfsub", v2sf_ftype_v2sf_v2sf, IX86_BUILTIN_PFSUB);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pfsubr", v2sf_ftype_v2sf_v2sf, IX86_BUILTIN_PFSUBR);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pi2fd", v2sf_ftype_v2si, IX86_BUILTIN_PI2FD);
- def_builtin (MASK_3DNOW, "__builtin_ia32_pmulhrw", v4hi_ftype_v4hi_v4hi, IX86_BUILTIN_PMULHRW);
-
- /* 3DNow! extension as used in the Athlon CPU. */
- def_builtin (MASK_3DNOW_A, "__builtin_ia32_pf2iw", v2si_ftype_v2sf, IX86_BUILTIN_PF2IW);
- def_builtin (MASK_3DNOW_A, "__builtin_ia32_pfnacc", v2sf_ftype_v2sf_v2sf, IX86_BUILTIN_PFNACC);
- def_builtin (MASK_3DNOW_A, "__builtin_ia32_pfpnacc", v2sf_ftype_v2sf_v2sf, IX86_BUILTIN_PFPNACC);
- def_builtin (MASK_3DNOW_A, "__builtin_ia32_pi2fw", v2sf_ftype_v2si, IX86_BUILTIN_PI2FW);
- def_builtin (MASK_3DNOW_A, "__builtin_ia32_pswapdsf", v2sf_ftype_v2sf, IX86_BUILTIN_PSWAPDSF);
- def_builtin (MASK_3DNOW_A, "__builtin_ia32_pswapdsi", v2si_ftype_v2si, IX86_BUILTIN_PSWAPDSI);
-
- /* SSE2 */
- def_builtin (MASK_SSE2, "__builtin_ia32_maskmovdqu", void_ftype_v16qi_v16qi_pchar, IX86_BUILTIN_MASKMOVDQU);
-
- def_builtin (MASK_SSE2, "__builtin_ia32_loadupd", v2df_ftype_pcdouble, IX86_BUILTIN_LOADUPD);
- def_builtin (MASK_SSE2, "__builtin_ia32_storeupd", void_ftype_pdouble_v2df, IX86_BUILTIN_STOREUPD);
-
- def_builtin (MASK_SSE2, "__builtin_ia32_loadhpd", v2df_ftype_v2df_pcdouble, IX86_BUILTIN_LOADHPD);
- def_builtin (MASK_SSE2, "__builtin_ia32_loadlpd", v2df_ftype_v2df_pcdouble, IX86_BUILTIN_LOADLPD);
-
- def_builtin (MASK_SSE2, "__builtin_ia32_movmskpd", int_ftype_v2df, IX86_BUILTIN_MOVMSKPD);
- def_builtin (MASK_SSE2, "__builtin_ia32_pmovmskb128", int_ftype_v16qi, IX86_BUILTIN_PMOVMSKB128);
- def_builtin (MASK_SSE2, "__builtin_ia32_movnti", void_ftype_pint_int, IX86_BUILTIN_MOVNTI);
- def_builtin (MASK_SSE2, "__builtin_ia32_movntpd", void_ftype_pdouble_v2df, IX86_BUILTIN_MOVNTPD);
- def_builtin (MASK_SSE2, "__builtin_ia32_movntdq", void_ftype_pv2di_v2di, IX86_BUILTIN_MOVNTDQ);
-
- def_builtin (MASK_SSE2, "__builtin_ia32_pshufd", v4si_ftype_v4si_int, IX86_BUILTIN_PSHUFD);
- def_builtin (MASK_SSE2, "__builtin_ia32_pshuflw", v8hi_ftype_v8hi_int, IX86_BUILTIN_PSHUFLW);
- def_builtin (MASK_SSE2, "__builtin_ia32_pshufhw", v8hi_ftype_v8hi_int, IX86_BUILTIN_PSHUFHW);
- def_builtin (MASK_SSE2, "__builtin_ia32_psadbw128", v2di_ftype_v16qi_v16qi, IX86_BUILTIN_PSADBW128);
-
- def_builtin_const (MASK_SSE2, "__builtin_ia32_sqrtpd", v2df_ftype_v2df, IX86_BUILTIN_SQRTPD);
- def_builtin_const (MASK_SSE2, "__builtin_ia32_sqrtsd", v2df_ftype_v2df, IX86_BUILTIN_SQRTSD);
-
- def_builtin (MASK_SSE2, "__builtin_ia32_shufpd", v2df_ftype_v2df_v2df_int, IX86_BUILTIN_SHUFPD);
-
- def_builtin_const (MASK_SSE2, "__builtin_ia32_cvtdq2pd", v2df_ftype_v4si, IX86_BUILTIN_CVTDQ2PD);
- def_builtin_const (MASK_SSE2, "__builtin_ia32_cvtdq2ps", v4sf_ftype_v4si, IX86_BUILTIN_CVTDQ2PS);
-
- def_builtin_const (MASK_SSE2, "__builtin_ia32_cvtpd2dq", v4si_ftype_v2df, IX86_BUILTIN_CVTPD2DQ);
- def_builtin_const (MASK_SSE2, "__builtin_ia32_cvtpd2pi", v2si_ftype_v2df, IX86_BUILTIN_CVTPD2PI);
- def_builtin_const (MASK_SSE2, "__builtin_ia32_cvtpd2ps", v4sf_ftype_v2df, IX86_BUILTIN_CVTPD2PS);
- def_builtin_const (MASK_SSE2, "__builtin_ia32_cvttpd2dq", v4si_ftype_v2df, IX86_BUILTIN_CVTTPD2DQ);
- def_builtin_const (MASK_SSE2, "__builtin_ia32_cvttpd2pi", v2si_ftype_v2df, IX86_BUILTIN_CVTTPD2PI);
-
- def_builtin_const (MASK_SSE2, "__builtin_ia32_cvtpi2pd", v2df_ftype_v2si, IX86_BUILTIN_CVTPI2PD);
-
- def_builtin_const (MASK_SSE2, "__builtin_ia32_cvtsd2si", int_ftype_v2df, IX86_BUILTIN_CVTSD2SI);
- def_builtin_const (MASK_SSE2, "__builtin_ia32_cvttsd2si", int_ftype_v2df, IX86_BUILTIN_CVTTSD2SI);
- def_builtin_const (MASK_SSE2 | MASK_64BIT, "__builtin_ia32_cvtsd2si64", int64_ftype_v2df, IX86_BUILTIN_CVTSD2SI64);
- def_builtin_const (MASK_SSE2 | MASK_64BIT, "__builtin_ia32_cvttsd2si64", int64_ftype_v2df, IX86_BUILTIN_CVTTSD2SI64);
-
- def_builtin_const (MASK_SSE2, "__builtin_ia32_cvtps2dq", v4si_ftype_v4sf, IX86_BUILTIN_CVTPS2DQ);
- def_builtin_const (MASK_SSE2, "__builtin_ia32_cvtps2pd", v2df_ftype_v4sf, IX86_BUILTIN_CVTPS2PD);
- def_builtin_const (MASK_SSE2, "__builtin_ia32_cvttps2dq", v4si_ftype_v4sf, IX86_BUILTIN_CVTTPS2DQ);
-
- def_builtin_const (MASK_SSE2, "__builtin_ia32_cvtsi2sd", v2df_ftype_v2df_int, IX86_BUILTIN_CVTSI2SD);
- def_builtin_const (MASK_SSE2 | MASK_64BIT, "__builtin_ia32_cvtsi642sd", v2df_ftype_v2df_int64, IX86_BUILTIN_CVTSI642SD);
- def_builtin_const (MASK_SSE2, "__builtin_ia32_cvtsd2ss", v4sf_ftype_v4sf_v2df, IX86_BUILTIN_CVTSD2SS);
- def_builtin_const (MASK_SSE2, "__builtin_ia32_cvtss2sd", v2df_ftype_v2df_v4sf, IX86_BUILTIN_CVTSS2SD);
-
- def_builtin (MASK_SSE2, "__builtin_ia32_clflush", void_ftype_pcvoid, IX86_BUILTIN_CLFLUSH);
- def_builtin (MASK_SSE2, "__builtin_ia32_lfence", void_ftype_void, IX86_BUILTIN_LFENCE);
- def_builtin (MASK_SSE2, "__builtin_ia32_mfence", void_ftype_void, IX86_BUILTIN_MFENCE);
-
- def_builtin (MASK_SSE2, "__builtin_ia32_loaddqu", v16qi_ftype_pcchar, IX86_BUILTIN_LOADDQU);
- def_builtin (MASK_SSE2, "__builtin_ia32_storedqu", void_ftype_pchar_v16qi, IX86_BUILTIN_STOREDQU);
-
- def_builtin (MASK_SSE2, "__builtin_ia32_pmuludq", di_ftype_v2si_v2si, IX86_BUILTIN_PMULUDQ);
- def_builtin (MASK_SSE2, "__builtin_ia32_pmuludq128", v2di_ftype_v4si_v4si, IX86_BUILTIN_PMULUDQ128);
-
- def_builtin (MASK_SSE2, "__builtin_ia32_pslldqi128", v2di_ftype_v2di_int, IX86_BUILTIN_PSLLDQI128);
- def_builtin (MASK_SSE2, "__builtin_ia32_psllwi128", v8hi_ftype_v8hi_int, IX86_BUILTIN_PSLLWI128);
- def_builtin (MASK_SSE2, "__builtin_ia32_pslldi128", v4si_ftype_v4si_int, IX86_BUILTIN_PSLLDI128);
- def_builtin (MASK_SSE2, "__builtin_ia32_psllqi128", v2di_ftype_v2di_int, IX86_BUILTIN_PSLLQI128);
- def_builtin (MASK_SSE2, "__builtin_ia32_psllw128", v8hi_ftype_v8hi_v8hi, IX86_BUILTIN_PSLLW128);
- def_builtin (MASK_SSE2, "__builtin_ia32_pslld128", v4si_ftype_v4si_v4si, IX86_BUILTIN_PSLLD128);
- def_builtin (MASK_SSE2, "__builtin_ia32_psllq128", v2di_ftype_v2di_v2di, IX86_BUILTIN_PSLLQ128);
-
- def_builtin (MASK_SSE2, "__builtin_ia32_psrldqi128", v2di_ftype_v2di_int, IX86_BUILTIN_PSRLDQI128);
- def_builtin (MASK_SSE2, "__builtin_ia32_psrlwi128", v8hi_ftype_v8hi_int, IX86_BUILTIN_PSRLWI128);
- def_builtin (MASK_SSE2, "__builtin_ia32_psrldi128", v4si_ftype_v4si_int, IX86_BUILTIN_PSRLDI128);
- def_builtin (MASK_SSE2, "__builtin_ia32_psrlqi128", v2di_ftype_v2di_int, IX86_BUILTIN_PSRLQI128);
- def_builtin (MASK_SSE2, "__builtin_ia32_psrlw128", v8hi_ftype_v8hi_v8hi, IX86_BUILTIN_PSRLW128);
- def_builtin (MASK_SSE2, "__builtin_ia32_psrld128", v4si_ftype_v4si_v4si, IX86_BUILTIN_PSRLD128);
- def_builtin (MASK_SSE2, "__builtin_ia32_psrlq128", v2di_ftype_v2di_v2di, IX86_BUILTIN_PSRLQ128);
-
- def_builtin (MASK_SSE2, "__builtin_ia32_psrawi128", v8hi_ftype_v8hi_int, IX86_BUILTIN_PSRAWI128);
- def_builtin (MASK_SSE2, "__builtin_ia32_psradi128", v4si_ftype_v4si_int, IX86_BUILTIN_PSRADI128);
- def_builtin (MASK_SSE2, "__builtin_ia32_psraw128", v8hi_ftype_v8hi_v8hi, IX86_BUILTIN_PSRAW128);
- def_builtin (MASK_SSE2, "__builtin_ia32_psrad128", v4si_ftype_v4si_v4si, IX86_BUILTIN_PSRAD128);
-
- def_builtin (MASK_SSE2, "__builtin_ia32_pmaddwd128", v4si_ftype_v8hi_v8hi, IX86_BUILTIN_PMADDWD128);
-
- /* Prescott New Instructions. */
- def_builtin (MASK_SSE3, "__builtin_ia32_monitor",
- void_ftype_pcvoid_unsigned_unsigned,
- IX86_BUILTIN_MONITOR);
- def_builtin (MASK_SSE3, "__builtin_ia32_mwait",
- void_ftype_unsigned_unsigned,
- IX86_BUILTIN_MWAIT);
- def_builtin (MASK_SSE3, "__builtin_ia32_lddqu",
- v16qi_ftype_pcchar, IX86_BUILTIN_LDDQU);
-
- /* SSSE3. */
- def_builtin (MASK_SSSE3, "__builtin_ia32_palignr128",
- v2di_ftype_v2di_v2di_int, IX86_BUILTIN_PALIGNR128);
- def_builtin (MASK_SSSE3, "__builtin_ia32_palignr", di_ftype_di_di_int,
- IX86_BUILTIN_PALIGNR);
-
- /* AMDFAM10 SSE4A New built-ins */
- def_builtin (MASK_SSE4A, "__builtin_ia32_movntsd",
- void_ftype_pdouble_v2df, IX86_BUILTIN_MOVNTSD);
- def_builtin (MASK_SSE4A, "__builtin_ia32_movntss",
- void_ftype_pfloat_v4sf, IX86_BUILTIN_MOVNTSS);
- def_builtin (MASK_SSE4A, "__builtin_ia32_extrqi",
- v2di_ftype_v2di_unsigned_unsigned, IX86_BUILTIN_EXTRQI);
- def_builtin (MASK_SSE4A, "__builtin_ia32_extrq",
- v2di_ftype_v2di_v16qi, IX86_BUILTIN_EXTRQ);
- def_builtin (MASK_SSE4A, "__builtin_ia32_insertqi",
- v2di_ftype_v2di_v2di_unsigned_unsigned, IX86_BUILTIN_INSERTQI);
- def_builtin (MASK_SSE4A, "__builtin_ia32_insertq",
- v2di_ftype_v2di_v2di, IX86_BUILTIN_INSERTQ);
-
- /* Access to the vec_init patterns. */
- ftype = build_function_type_list (V2SI_type_node, integer_type_node,
- integer_type_node, NULL_TREE);
- def_builtin (MASK_MMX, "__builtin_ia32_vec_init_v2si",
- ftype, IX86_BUILTIN_VEC_INIT_V2SI);
-
+ case V2SI_FTYPE_V4SF:
+ type = v2si_ftype_v4sf;
+ break;
+ case V2SI_FTYPE_V2DF:
+ type = v2si_ftype_v2df;
+ break;
+ case V2SI_FTYPE_V2SF:
+ type = v2si_ftype_v2sf;
+ break;
+ case V2DF_FTYPE_V4SF:
+ type = v2df_ftype_v4sf;
+ break;
+ case V2DF_FTYPE_V2DF:
+ case V2DF_FTYPE_V2DF_VEC_MERGE:
+ type = v2df_ftype_v2df;
+ break;
+ case V2DF_FTYPE_V2SI:
+ type = v2df_ftype_v2si;
+ break;
+ case V2DF_FTYPE_V4SI:
+ type = v2df_ftype_v4si;
+ break;
+ case V2SF_FTYPE_V2SF:
+ type = v2sf_ftype_v2sf;
+ break;
+ case V2SF_FTYPE_V2SI:
+ type = v2sf_ftype_v2si;
+ break;
+ case V16QI_FTYPE_V16QI_V16QI:
+ type = v16qi_ftype_v16qi_v16qi;
+ break;
+ case V16QI_FTYPE_V8HI_V8HI:
+ type = v16qi_ftype_v8hi_v8hi;
+ break;
+ case V8QI_FTYPE_V8QI_V8QI:
+ type = v8qi_ftype_v8qi_v8qi;
+ break;
+ case V8QI_FTYPE_V4HI_V4HI:
+ type = v8qi_ftype_v4hi_v4hi;
+ break;
+ case V8HI_FTYPE_V8HI_V8HI:
+ case V8HI_FTYPE_V8HI_V8HI_COUNT:
+ type = v8hi_ftype_v8hi_v8hi;
+ break;
+ case V8HI_FTYPE_V16QI_V16QI:
+ type = v8hi_ftype_v16qi_v16qi;
+ break;
+ case V8HI_FTYPE_V4SI_V4SI:
+ type = v8hi_ftype_v4si_v4si;
+ break;
+ case V8HI_FTYPE_V8HI_SI_COUNT:
+ type = v8hi_ftype_v8hi_int;
+ break;
+ case V4SI_FTYPE_V4SI_V4SI:
+ case V4SI_FTYPE_V4SI_V4SI_COUNT:
+ type = v4si_ftype_v4si_v4si;
+ break;
+ case V4SI_FTYPE_V8HI_V8HI:
+ type = v4si_ftype_v8hi_v8hi;
+ break;
+ case V4SI_FTYPE_V4SF_V4SF:
+ type = v4si_ftype_v4sf_v4sf;
+ break;
+ case V4SI_FTYPE_V2DF_V2DF:
+ type = v4si_ftype_v2df_v2df;
+ break;
+ case V4SI_FTYPE_V4SI_SI_COUNT:
+ type = v4si_ftype_v4si_int;
+ break;
+ case V4HI_FTYPE_V4HI_V4HI:
+ case V4HI_FTYPE_V4HI_V4HI_COUNT:
+ type = v4hi_ftype_v4hi_v4hi;
+ break;
+ case V4HI_FTYPE_V8QI_V8QI:
+ type = v4hi_ftype_v8qi_v8qi;
+ break;
+ case V4HI_FTYPE_V2SI_V2SI:
+ type = v4hi_ftype_v2si_v2si;
+ break;
+ case V4HI_FTYPE_V4HI_SI_COUNT:
+ type = v4hi_ftype_v4hi_int;
+ break;
+ case V4SF_FTYPE_V4SF_V4SF:
+ case V4SF_FTYPE_V4SF_V4SF_SWAP:
+ type = v4sf_ftype_v4sf_v4sf;
+ break;
+ case V4SF_FTYPE_V4SF_V2SI:
+ type = v4sf_ftype_v4sf_v2si;
+ break;
+ case V4SF_FTYPE_V4SF_V2DF:
+ type = v4sf_ftype_v4sf_v2df;
+ break;
+ case V4SF_FTYPE_V4SF_DI:
+ type = v4sf_ftype_v4sf_int64;
+ break;
+ case V4SF_FTYPE_V4SF_SI:
+ type = v4sf_ftype_v4sf_int;
+ break;
+ case V2DI_FTYPE_V2DI_V2DI:
+ case V2DI_FTYPE_V2DI_V2DI_COUNT:
+ type = v2di_ftype_v2di_v2di;
+ break;
+ case V2DI_FTYPE_V16QI_V16QI:
+ type = v2di_ftype_v16qi_v16qi;
+ break;
+ case V2DI_FTYPE_V4SI_V4SI:
+ type = v2di_ftype_v4si_v4si;
+ break;
+ case V2DI_FTYPE_V2DI_V16QI:
+ type = v2di_ftype_v2di_v16qi;
+ break;
+ case V2DI_FTYPE_V2DF_V2DF:
+ type = v2di_ftype_v2df_v2df;
+ break;
+ case V2DI_FTYPE_V2DI_SI_COUNT:
+ type = v2di_ftype_v2di_int;
+ break;
+ case V2SI_FTYPE_V2SI_V2SI:
+ case V2SI_FTYPE_V2SI_V2SI_COUNT:
+ type = v2si_ftype_v2si_v2si;
+ break;
+ case V2SI_FTYPE_V4HI_V4HI:
+ type = v2si_ftype_v4hi_v4hi;
+ break;
+ case V2SI_FTYPE_V2SF_V2SF:
+ type = v2si_ftype_v2sf_v2sf;
+ break;
+ case V2SI_FTYPE_V2SI_SI_COUNT:
+ type = v2si_ftype_v2si_int;
+ break;
+ case V2DF_FTYPE_V2DF_V2DF:
+ case V2DF_FTYPE_V2DF_V2DF_SWAP:
+ type = v2df_ftype_v2df_v2df;
+ break;
+ case V2DF_FTYPE_V2DF_V4SF:
+ type = v2df_ftype_v2df_v4sf;
+ break;
+ case V2DF_FTYPE_V2DF_DI:
+ type = v2df_ftype_v2df_int64;
+ break;
+ case V2DF_FTYPE_V2DF_SI:
+ type = v2df_ftype_v2df_int;
+ break;
+ case V2SF_FTYPE_V2SF_V2SF:
+ type = v2sf_ftype_v2sf_v2sf;
+ break;
+ case V1DI_FTYPE_V1DI_V1DI:
+ case V1DI_FTYPE_V1DI_V1DI_COUNT:
+ type = v1di_ftype_v1di_v1di;
+ break;
+ case V1DI_FTYPE_V8QI_V8QI:
+ type = v1di_ftype_v8qi_v8qi;
+ break;
+ case V1DI_FTYPE_V2SI_V2SI:
+ type = v1di_ftype_v2si_v2si;
+ break;
+ case V1DI_FTYPE_V1DI_SI_COUNT:
+ type = v1di_ftype_v1di_int;
+ break;
+ case UINT64_FTYPE_UINT64_UINT64:
+ type = uint64_ftype_uint64_uint64;
+ break;
+ case UINT_FTYPE_UINT_UINT:
+ type = unsigned_ftype_unsigned_unsigned;
+ break;
+ case UINT_FTYPE_UINT_USHORT:
+ type = unsigned_ftype_unsigned_ushort;
+ break;
+ case UINT_FTYPE_UINT_UCHAR:
+ type = unsigned_ftype_unsigned_uchar;
+ break;
+ case V8HI_FTYPE_V8HI_INT:
+ type = v8hi_ftype_v8hi_int;
+ break;
+ case V4SI_FTYPE_V4SI_INT:
+ type = v4si_ftype_v4si_int;
+ break;
+ case V4HI_FTYPE_V4HI_INT:
+ type = v4hi_ftype_v4hi_int;
+ break;
+ case V4SF_FTYPE_V4SF_INT:
+ type = v4sf_ftype_v4sf_int;
+ break;
+ case V2DI_FTYPE_V2DI_INT:
+ case V2DI2TI_FTYPE_V2DI_INT:
+ type = v2di_ftype_v2di_int;
+ break;
+ case V2DF_FTYPE_V2DF_INT:
+ type = v2df_ftype_v2df_int;
+ break;
+ case V16QI_FTYPE_V16QI_V16QI_V16QI:
+ type = v16qi_ftype_v16qi_v16qi_v16qi;
+ break;
+ case V4SF_FTYPE_V4SF_V4SF_V4SF:
+ type = v4sf_ftype_v4sf_v4sf_v4sf;
+ break;
+ case V2DF_FTYPE_V2DF_V2DF_V2DF:
+ type = v2df_ftype_v2df_v2df_v2df;
+ break;
+ case V16QI_FTYPE_V16QI_V16QI_INT:
+ type = v16qi_ftype_v16qi_v16qi_int;
+ break;
+ case V8HI_FTYPE_V8HI_V8HI_INT:
+ type = v8hi_ftype_v8hi_v8hi_int;
+ break;
+ case V4SI_FTYPE_V4SI_V4SI_INT:
+ type = v4si_ftype_v4si_v4si_int;
+ break;
+ case V4SF_FTYPE_V4SF_V4SF_INT:
+ type = v4sf_ftype_v4sf_v4sf_int;
+ break;
+ case V2DI_FTYPE_V2DI_V2DI_INT:
+ case V2DI2TI_FTYPE_V2DI_V2DI_INT:
+ type = v2di_ftype_v2di_v2di_int;
+ break;
+ case V2DF_FTYPE_V2DF_V2DF_INT:
+ type = v2df_ftype_v2df_v2df_int;
+ break;
+ case V2DI_FTYPE_V2DI_UINT_UINT:
+ type = v2di_ftype_v2di_unsigned_unsigned;
+ break;
+ case V2DI_FTYPE_V2DI_V2DI_UINT_UINT:
+ type = v2di_ftype_v2di_v2di_unsigned_unsigned;
+ break;
+ case V1DI2DI_FTYPE_V1DI_V1DI_INT:
+ type = v1di_ftype_v1di_v1di_int;
+ break;
+ default:
+ gcc_unreachable ();
+ }
+
+ def_builtin_const (d->mask, d->name, type, d->code);
+ }
+
+ /* pcmpestr[im] insns. */
+ for (i = 0, d = bdesc_pcmpestr;
+ i < ARRAY_SIZE (bdesc_pcmpestr);
+ i++, d++)
+ {
+ if (d->code == IX86_BUILTIN_PCMPESTRM128)
+ ftype = v16qi_ftype_v16qi_int_v16qi_int_int;
+ else
+ ftype = int_ftype_v16qi_int_v16qi_int_int;
+ def_builtin_const (d->mask, d->name, ftype, d->code);
+ }
+
+ /* pcmpistr[im] insns. */
+ for (i = 0, d = bdesc_pcmpistr;
+ i < ARRAY_SIZE (bdesc_pcmpistr);
+ i++, d++)
+ {
+ if (d->code == IX86_BUILTIN_PCMPISTRM128)
+ ftype = v16qi_ftype_v16qi_v16qi_int;
+ else
+ ftype = int_ftype_v16qi_v16qi_int;
+ def_builtin_const (d->mask, d->name, ftype, d->code);
+ }
+
+ /* comi/ucomi insns. */
+ for (i = 0, d = bdesc_comi; i < ARRAY_SIZE (bdesc_comi); i++, d++)
+ if (d->mask == OPTION_MASK_ISA_SSE2)
+ def_builtin_const (d->mask, d->name, int_ftype_v2df_v2df, d->code);
+ else
+ def_builtin_const (d->mask, d->name, int_ftype_v4sf_v4sf, d->code);
+
+ /* SSE */
+ def_builtin (OPTION_MASK_ISA_SSE, "__builtin_ia32_ldmxcsr", void_ftype_unsigned, IX86_BUILTIN_LDMXCSR);
+ def_builtin (OPTION_MASK_ISA_SSE, "__builtin_ia32_stmxcsr", unsigned_ftype_void, IX86_BUILTIN_STMXCSR);
+
+ /* SSE or 3DNow!A */
+ def_builtin (OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_3DNOW_A, "__builtin_ia32_maskmovq", void_ftype_v8qi_v8qi_pchar, IX86_BUILTIN_MASKMOVQ);
+
+ /* SSE2 */
+ def_builtin (OPTION_MASK_ISA_SSE2, "__builtin_ia32_maskmovdqu", void_ftype_v16qi_v16qi_pchar, IX86_BUILTIN_MASKMOVDQU);
+
+ def_builtin (OPTION_MASK_ISA_SSE2, "__builtin_ia32_clflush", void_ftype_pcvoid, IX86_BUILTIN_CLFLUSH);
+ x86_mfence = def_builtin (OPTION_MASK_ISA_SSE2, "__builtin_ia32_mfence", void_ftype_void, IX86_BUILTIN_MFENCE);
+
+ /* SSE3. */
+ def_builtin (OPTION_MASK_ISA_SSE3, "__builtin_ia32_monitor", void_ftype_pcvoid_unsigned_unsigned, IX86_BUILTIN_MONITOR);
+ def_builtin (OPTION_MASK_ISA_SSE3, "__builtin_ia32_mwait", void_ftype_unsigned_unsigned, IX86_BUILTIN_MWAIT);
+
+ /* AES */
+ if (TARGET_AES)
+ {
+ /* Define AES built-in functions only if AES is enabled. */
+ def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_aesenc128", v2di_ftype_v2di_v2di, IX86_BUILTIN_AESENC128);
+ def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_aesenclast128", v2di_ftype_v2di_v2di, IX86_BUILTIN_AESENCLAST128);
+ def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_aesdec128", v2di_ftype_v2di_v2di, IX86_BUILTIN_AESDEC128);
+ def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_aesdeclast128", v2di_ftype_v2di_v2di, IX86_BUILTIN_AESDECLAST128);
+ def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_aesimc128", v2di_ftype_v2di, IX86_BUILTIN_AESIMC128);
+ def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_aeskeygenassist128", v2di_ftype_v2di_int, IX86_BUILTIN_AESKEYGENASSIST128);
+ }
+
+ /* PCLMUL */
+ if (TARGET_PCLMUL)
+ {
+ /* Define PCLMUL built-in function only if PCLMUL is enabled. */
+ def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_pclmulqdq128", v2di_ftype_v2di_v2di_int, IX86_BUILTIN_PCLMULQDQ128);
+ }
+
+ /* Access to the vec_init patterns. */
+ ftype = build_function_type_list (V2SI_type_node, integer_type_node,
+ integer_type_node, NULL_TREE);
+ def_builtin_const (OPTION_MASK_ISA_MMX, "__builtin_ia32_vec_init_v2si", ftype, IX86_BUILTIN_VEC_INIT_V2SI);
+
ftype = build_function_type_list (V4HI_type_node, short_integer_type_node,
short_integer_type_node,
short_integer_type_node,
short_integer_type_node, NULL_TREE);
- def_builtin (MASK_MMX, "__builtin_ia32_vec_init_v4hi",
- ftype, IX86_BUILTIN_VEC_INIT_V4HI);
+ def_builtin_const (OPTION_MASK_ISA_MMX, "__builtin_ia32_vec_init_v4hi", ftype, IX86_BUILTIN_VEC_INIT_V4HI);
ftype = build_function_type_list (V8QI_type_node, char_type_node,
char_type_node, char_type_node,
char_type_node, char_type_node,
char_type_node, char_type_node,
char_type_node, NULL_TREE);
- def_builtin (MASK_MMX, "__builtin_ia32_vec_init_v8qi",
- ftype, IX86_BUILTIN_VEC_INIT_V8QI);
+ def_builtin_const (OPTION_MASK_ISA_MMX, "__builtin_ia32_vec_init_v8qi", ftype, IX86_BUILTIN_VEC_INIT_V8QI);
/* Access to the vec_extract patterns. */
ftype = build_function_type_list (double_type_node, V2DF_type_node,
integer_type_node, NULL_TREE);
- def_builtin (MASK_SSE, "__builtin_ia32_vec_ext_v2df",
- ftype, IX86_BUILTIN_VEC_EXT_V2DF);
+ def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_vec_ext_v2df", ftype, IX86_BUILTIN_VEC_EXT_V2DF);
ftype = build_function_type_list (long_long_integer_type_node,
V2DI_type_node, integer_type_node,
NULL_TREE);
- def_builtin (MASK_SSE, "__builtin_ia32_vec_ext_v2di",
- ftype, IX86_BUILTIN_VEC_EXT_V2DI);
+ def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_vec_ext_v2di", ftype, IX86_BUILTIN_VEC_EXT_V2DI);
ftype = build_function_type_list (float_type_node, V4SF_type_node,
integer_type_node, NULL_TREE);
- def_builtin (MASK_SSE, "__builtin_ia32_vec_ext_v4sf",
- ftype, IX86_BUILTIN_VEC_EXT_V4SF);
+ def_builtin_const (OPTION_MASK_ISA_SSE, "__builtin_ia32_vec_ext_v4sf", ftype, IX86_BUILTIN_VEC_EXT_V4SF);
ftype = build_function_type_list (intSI_type_node, V4SI_type_node,
integer_type_node, NULL_TREE);
- def_builtin (MASK_SSE, "__builtin_ia32_vec_ext_v4si",
- ftype, IX86_BUILTIN_VEC_EXT_V4SI);
+ def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_vec_ext_v4si", ftype, IX86_BUILTIN_VEC_EXT_V4SI);
ftype = build_function_type_list (intHI_type_node, V8HI_type_node,
integer_type_node, NULL_TREE);
- def_builtin (MASK_SSE, "__builtin_ia32_vec_ext_v8hi",
- ftype, IX86_BUILTIN_VEC_EXT_V8HI);
+ def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_vec_ext_v8hi", ftype, IX86_BUILTIN_VEC_EXT_V8HI);
ftype = build_function_type_list (intHI_type_node, V4HI_type_node,
integer_type_node, NULL_TREE);
- def_builtin (MASK_SSE | MASK_3DNOW_A, "__builtin_ia32_vec_ext_v4hi",
- ftype, IX86_BUILTIN_VEC_EXT_V4HI);
+ def_builtin_const (OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_3DNOW_A, "__builtin_ia32_vec_ext_v4hi", ftype, IX86_BUILTIN_VEC_EXT_V4HI);
ftype = build_function_type_list (intSI_type_node, V2SI_type_node,
integer_type_node, NULL_TREE);
- def_builtin (MASK_MMX, "__builtin_ia32_vec_ext_v2si",
- ftype, IX86_BUILTIN_VEC_EXT_V2SI);
+ def_builtin_const (OPTION_MASK_ISA_MMX, "__builtin_ia32_vec_ext_v2si", ftype, IX86_BUILTIN_VEC_EXT_V2SI);
+
+ ftype = build_function_type_list (intQI_type_node, V16QI_type_node,
+ integer_type_node, NULL_TREE);
+ def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_vec_ext_v16qi", ftype, IX86_BUILTIN_VEC_EXT_V16QI);
/* Access to the vec_set patterns. */
+ ftype = build_function_type_list (V2DI_type_node, V2DI_type_node,
+ intDI_type_node,
+ integer_type_node, NULL_TREE);
+ def_builtin_const (OPTION_MASK_ISA_SSE4_1 | OPTION_MASK_ISA_64BIT, "__builtin_ia32_vec_set_v2di", ftype, IX86_BUILTIN_VEC_SET_V2DI);
+
+ ftype = build_function_type_list (V4SF_type_node, V4SF_type_node,
+ float_type_node,
+ integer_type_node, NULL_TREE);
+ def_builtin_const (OPTION_MASK_ISA_SSE4_1, "__builtin_ia32_vec_set_v4sf", ftype, IX86_BUILTIN_VEC_SET_V4SF);
+
+ ftype = build_function_type_list (V4SI_type_node, V4SI_type_node,
+ intSI_type_node,
+ integer_type_node, NULL_TREE);
+ def_builtin_const (OPTION_MASK_ISA_SSE4_1, "__builtin_ia32_vec_set_v4si", ftype, IX86_BUILTIN_VEC_SET_V4SI);
+
ftype = build_function_type_list (V8HI_type_node, V8HI_type_node,
intHI_type_node,
integer_type_node, NULL_TREE);
- def_builtin (MASK_SSE, "__builtin_ia32_vec_set_v8hi",
- ftype, IX86_BUILTIN_VEC_SET_V8HI);
+ def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_vec_set_v8hi", ftype, IX86_BUILTIN_VEC_SET_V8HI);
ftype = build_function_type_list (V4HI_type_node, V4HI_type_node,
intHI_type_node,
integer_type_node, NULL_TREE);
- def_builtin (MASK_SSE | MASK_3DNOW_A, "__builtin_ia32_vec_set_v4hi",
- ftype, IX86_BUILTIN_VEC_SET_V4HI);
+ def_builtin_const (OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_3DNOW_A, "__builtin_ia32_vec_set_v4hi", ftype, IX86_BUILTIN_VEC_SET_V4HI);
+
+ ftype = build_function_type_list (V16QI_type_node, V16QI_type_node,
+ intQI_type_node,
+ integer_type_node, NULL_TREE);
+ def_builtin_const (OPTION_MASK_ISA_SSE4_1, "__builtin_ia32_vec_set_v16qi", ftype, IX86_BUILTIN_VEC_SET_V16QI);
+
+ /* Add SSE5 multi-arg argument instructions */
+ for (i = 0, d = bdesc_multi_arg; i < ARRAY_SIZE (bdesc_multi_arg); i++, d++)
+ {
+ tree mtype = NULL_TREE;
+
+ if (d->name == 0)
+ continue;
+
+ switch ((enum multi_arg_type)d->flag)
+ {
+ case MULTI_ARG_3_SF: mtype = v4sf_ftype_v4sf_v4sf_v4sf; break;
+ case MULTI_ARG_3_DF: mtype = v2df_ftype_v2df_v2df_v2df; break;
+ case MULTI_ARG_3_DI: mtype = v2di_ftype_v2di_v2di_v2di; break;
+ case MULTI_ARG_3_SI: mtype = v4si_ftype_v4si_v4si_v4si; break;
+ case MULTI_ARG_3_SI_DI: mtype = v4si_ftype_v4si_v4si_v2di; break;
+ case MULTI_ARG_3_HI: mtype = v8hi_ftype_v8hi_v8hi_v8hi; break;
+ case MULTI_ARG_3_HI_SI: mtype = v8hi_ftype_v8hi_v8hi_v4si; break;
+ case MULTI_ARG_3_QI: mtype = v16qi_ftype_v16qi_v16qi_v16qi; break;
+ case MULTI_ARG_3_PERMPS: mtype = v4sf_ftype_v4sf_v4sf_v16qi; break;
+ case MULTI_ARG_3_PERMPD: mtype = v2df_ftype_v2df_v2df_v16qi; break;
+ case MULTI_ARG_2_SF: mtype = v4sf_ftype_v4sf_v4sf; break;
+ case MULTI_ARG_2_DF: mtype = v2df_ftype_v2df_v2df; break;
+ case MULTI_ARG_2_DI: mtype = v2di_ftype_v2di_v2di; break;
+ case MULTI_ARG_2_SI: mtype = v4si_ftype_v4si_v4si; break;
+ case MULTI_ARG_2_HI: mtype = v8hi_ftype_v8hi_v8hi; break;
+ case MULTI_ARG_2_QI: mtype = v16qi_ftype_v16qi_v16qi; break;
+ case MULTI_ARG_2_DI_IMM: mtype = v2di_ftype_v2di_si; break;
+ case MULTI_ARG_2_SI_IMM: mtype = v4si_ftype_v4si_si; break;
+ case MULTI_ARG_2_HI_IMM: mtype = v8hi_ftype_v8hi_si; break;
+ case MULTI_ARG_2_QI_IMM: mtype = v16qi_ftype_v16qi_si; break;
+ case MULTI_ARG_2_SF_CMP: mtype = v4sf_ftype_v4sf_v4sf; break;
+ case MULTI_ARG_2_DF_CMP: mtype = v2df_ftype_v2df_v2df; break;
+ case MULTI_ARG_2_DI_CMP: mtype = v2di_ftype_v2di_v2di; break;
+ case MULTI_ARG_2_SI_CMP: mtype = v4si_ftype_v4si_v4si; break;
+ case MULTI_ARG_2_HI_CMP: mtype = v8hi_ftype_v8hi_v8hi; break;
+ case MULTI_ARG_2_QI_CMP: mtype = v16qi_ftype_v16qi_v16qi; break;
+ case MULTI_ARG_2_SF_TF: mtype = v4sf_ftype_v4sf_v4sf; break;
+ case MULTI_ARG_2_DF_TF: mtype = v2df_ftype_v2df_v2df; break;
+ case MULTI_ARG_2_DI_TF: mtype = v2di_ftype_v2di_v2di; break;
+ case MULTI_ARG_2_SI_TF: mtype = v4si_ftype_v4si_v4si; break;
+ case MULTI_ARG_2_HI_TF: mtype = v8hi_ftype_v8hi_v8hi; break;
+ case MULTI_ARG_2_QI_TF: mtype = v16qi_ftype_v16qi_v16qi; break;
+ case MULTI_ARG_1_SF: mtype = v4sf_ftype_v4sf; break;
+ case MULTI_ARG_1_DF: mtype = v2df_ftype_v2df; break;
+ case MULTI_ARG_1_DI: mtype = v2di_ftype_v2di; break;
+ case MULTI_ARG_1_SI: mtype = v4si_ftype_v4si; break;
+ case MULTI_ARG_1_HI: mtype = v8hi_ftype_v8hi; break;
+ case MULTI_ARG_1_QI: mtype = v16qi_ftype_v16qi; break;
+ case MULTI_ARG_1_SI_DI: mtype = v2di_ftype_v4si; break;
+ case MULTI_ARG_1_HI_DI: mtype = v2di_ftype_v8hi; break;
+ case MULTI_ARG_1_HI_SI: mtype = v4si_ftype_v8hi; break;
+ case MULTI_ARG_1_QI_DI: mtype = v2di_ftype_v16qi; break;
+ case MULTI_ARG_1_QI_SI: mtype = v4si_ftype_v16qi; break;
+ case MULTI_ARG_1_QI_HI: mtype = v8hi_ftype_v16qi; break;
+ case MULTI_ARG_1_PH2PS: mtype = v4sf_ftype_v4hi; break;
+ case MULTI_ARG_1_PS2PH: mtype = v4hi_ftype_v4sf; break;
+ case MULTI_ARG_UNKNOWN:
+ default:
+ gcc_unreachable ();
+ }
+
+ if (mtype)
+ def_builtin_const (d->mask, d->name, mtype, d->code);
+ }
}
static void
static rtx
ix86_expand_binop_builtin (enum insn_code icode, tree exp, rtx target)
{
- rtx pat, xops[3];
+ rtx pat;
tree arg0 = CALL_EXPR_ARG (exp, 0);
tree arg1 = CALL_EXPR_ARG (exp, 1);
rtx op0 = expand_normal (arg0);
op1 = gen_lowpart (TImode, x);
}
- /* The insn must want input operands in the same modes as the
- result. */
- gcc_assert ((GET_MODE (op0) == mode0 || GET_MODE (op0) == VOIDmode)
- && (GET_MODE (op1) == mode1 || GET_MODE (op1) == VOIDmode));
-
if (!(*insn_data[icode].operand[1].predicate) (op0, mode0))
op0 = copy_to_mode_reg (mode0, op0);
if (!(*insn_data[icode].operand[2].predicate) (op1, mode1))
op1 = copy_to_mode_reg (mode1, op1);
- /* ??? Using ix86_fixup_binary_operands is problematic when
- we've got mismatched modes. Fake it. */
-
- xops[0] = target;
- xops[1] = op0;
- xops[2] = op1;
-
- if (tmode == mode0 && tmode == mode1)
- {
- target = ix86_fixup_binary_operands (UNKNOWN, tmode, xops);
- op0 = xops[1];
- op1 = xops[2];
- }
- else if (optimize || !ix86_binary_operator_ok (UNKNOWN, tmode, xops))
- {
- op0 = force_reg (mode0, op0);
- op1 = force_reg (mode1, op1);
- target = gen_reg_rtx (tmode);
- }
-
pat = GEN_FCN (icode) (target, op0, op1);
if (! pat)
return 0;
+
emit_insn (pat);
+
return target;
}
-/* Subroutine of ix86_expand_builtin to take care of stores. */
+/* Subroutine of ix86_expand_builtin to take care of 2-4 argument insns. */
static rtx
-ix86_expand_store_builtin (enum insn_code icode, tree exp)
+ix86_expand_multi_arg_builtin (enum insn_code icode, tree exp, rtx target,
+ enum multi_arg_type m_type,
+ enum insn_code sub_code)
{
rtx pat;
- tree arg0 = CALL_EXPR_ARG (exp, 0);
- tree arg1 = CALL_EXPR_ARG (exp, 1);
- rtx op0 = expand_normal (arg0);
- rtx op1 = expand_normal (arg1);
- enum machine_mode mode0 = insn_data[icode].operand[0].mode;
- enum machine_mode mode1 = insn_data[icode].operand[1].mode;
+ int i;
+ int nargs;
+ bool comparison_p = false;
+ bool tf_p = false;
+ bool last_arg_constant = false;
+ int num_memory = 0;
+ struct {
+ rtx op;
+ enum machine_mode mode;
+ } args[4];
- if (VECTOR_MODE_P (mode1))
- op1 = safe_vector_operand (op1, mode1);
+ enum machine_mode tmode = insn_data[icode].operand[0].mode;
- op0 = gen_rtx_MEM (mode0, copy_to_mode_reg (Pmode, op0));
- op1 = copy_to_mode_reg (mode1, op1);
+ switch (m_type)
+ {
+ case MULTI_ARG_3_SF:
+ case MULTI_ARG_3_DF:
+ case MULTI_ARG_3_DI:
+ case MULTI_ARG_3_SI:
+ case MULTI_ARG_3_SI_DI:
+ case MULTI_ARG_3_HI:
+ case MULTI_ARG_3_HI_SI:
+ case MULTI_ARG_3_QI:
+ case MULTI_ARG_3_PERMPS:
+ case MULTI_ARG_3_PERMPD:
+ nargs = 3;
+ break;
- pat = GEN_FCN (icode) (op0, op1);
- if (pat)
- emit_insn (pat);
- return 0;
-}
+ case MULTI_ARG_2_SF:
+ case MULTI_ARG_2_DF:
+ case MULTI_ARG_2_DI:
+ case MULTI_ARG_2_SI:
+ case MULTI_ARG_2_HI:
+ case MULTI_ARG_2_QI:
+ nargs = 2;
+ break;
-/* Subroutine of ix86_expand_builtin to take care of unop insns. */
+ case MULTI_ARG_2_DI_IMM:
+ case MULTI_ARG_2_SI_IMM:
+ case MULTI_ARG_2_HI_IMM:
+ case MULTI_ARG_2_QI_IMM:
+ nargs = 2;
+ last_arg_constant = true;
+ break;
-static rtx
-ix86_expand_unop_builtin (enum insn_code icode, tree exp,
- rtx target, int do_load)
-{
- rtx pat;
- tree arg0 = CALL_EXPR_ARG (exp, 0);
- rtx op0 = expand_normal (arg0);
- enum machine_mode tmode = insn_data[icode].operand[0].mode;
- enum machine_mode mode0 = insn_data[icode].operand[1].mode;
+ case MULTI_ARG_1_SF:
+ case MULTI_ARG_1_DF:
+ case MULTI_ARG_1_DI:
+ case MULTI_ARG_1_SI:
+ case MULTI_ARG_1_HI:
+ case MULTI_ARG_1_QI:
+ case MULTI_ARG_1_SI_DI:
+ case MULTI_ARG_1_HI_DI:
+ case MULTI_ARG_1_HI_SI:
+ case MULTI_ARG_1_QI_DI:
+ case MULTI_ARG_1_QI_SI:
+ case MULTI_ARG_1_QI_HI:
+ case MULTI_ARG_1_PH2PS:
+ case MULTI_ARG_1_PS2PH:
+ nargs = 1;
+ break;
+
+ case MULTI_ARG_2_SF_CMP:
+ case MULTI_ARG_2_DF_CMP:
+ case MULTI_ARG_2_DI_CMP:
+ case MULTI_ARG_2_SI_CMP:
+ case MULTI_ARG_2_HI_CMP:
+ case MULTI_ARG_2_QI_CMP:
+ nargs = 2;
+ comparison_p = true;
+ break;
+
+ case MULTI_ARG_2_SF_TF:
+ case MULTI_ARG_2_DF_TF:
+ case MULTI_ARG_2_DI_TF:
+ case MULTI_ARG_2_SI_TF:
+ case MULTI_ARG_2_HI_TF:
+ case MULTI_ARG_2_QI_TF:
+ nargs = 2;
+ tf_p = true;
+ break;
+
+ case MULTI_ARG_UNKNOWN:
+ default:
+ gcc_unreachable ();
+ }
if (optimize || !target
|| GET_MODE (target) != tmode
|| ! (*insn_data[icode].operand[0].predicate) (target, tmode))
target = gen_reg_rtx (tmode);
- if (do_load)
- op0 = gen_rtx_MEM (mode0, copy_to_mode_reg (Pmode, op0));
- else
+
+ gcc_assert (nargs <= 4);
+
+ for (i = 0; i < nargs; i++)
{
- if (VECTOR_MODE_P (mode0))
- op0 = safe_vector_operand (op0, mode0);
+ tree arg = CALL_EXPR_ARG (exp, i);
+ rtx op = expand_normal (arg);
+ int adjust = (comparison_p) ? 1 : 0;
+ enum machine_mode mode = insn_data[icode].operand[i+adjust+1].mode;
- if ((optimize && !register_operand (op0, mode0))
- || ! (*insn_data[icode].operand[1].predicate) (op0, mode0))
- op0 = copy_to_mode_reg (mode0, op0);
+ if (last_arg_constant && i == nargs-1)
+ {
+ if (GET_CODE (op) != CONST_INT)
+ {
+ error ("last argument must be an immediate");
+ return gen_reg_rtx (tmode);
+ }
+ }
+ else
+ {
+ if (VECTOR_MODE_P (mode))
+ op = safe_vector_operand (op, mode);
+
+ /* If we aren't optimizing, only allow one memory operand to be
+ generated. */
+ if (memory_operand (op, mode))
+ num_memory++;
+
+ gcc_assert (GET_MODE (op) == mode || GET_MODE (op) == VOIDmode);
+
+ if (optimize
+ || ! (*insn_data[icode].operand[i+adjust+1].predicate) (op, mode)
+ || num_memory > 1)
+ op = force_reg (mode, op);
+ }
+
+ args[i].op = op;
+ args[i].mode = mode;
+ }
+
+ switch (nargs)
+ {
+ case 1:
+ pat = GEN_FCN (icode) (target, args[0].op);
+ break;
+
+ case 2:
+ if (tf_p)
+ pat = GEN_FCN (icode) (target, args[0].op, args[1].op,
+ GEN_INT ((int)sub_code));
+ else if (! comparison_p)
+ pat = GEN_FCN (icode) (target, args[0].op, args[1].op);
+ else
+ {
+ rtx cmp_op = gen_rtx_fmt_ee (sub_code, GET_MODE (target),
+ args[0].op,
+ args[1].op);
+
+ pat = GEN_FCN (icode) (target, cmp_op, args[0].op, args[1].op);
+ }
+ break;
+
+ case 3:
+ pat = GEN_FCN (icode) (target, args[0].op, args[1].op, args[2].op);
+ break;
+
+ default:
+ gcc_unreachable ();
}
- pat = GEN_FCN (icode) (target, op0);
if (! pat)
return 0;
+
emit_insn (pat);
return target;
}
-/* Subroutine of ix86_expand_builtin to take care of three special unop insns:
- sqrtss, rsqrtss, rcpss. */
+/* Subroutine of ix86_expand_args_builtin to take care of scalar unop
+ insns with vec_merge. */
static rtx
-ix86_expand_unop1_builtin (enum insn_code icode, tree exp, rtx target)
+ix86_expand_unop_vec_merge_builtin (enum insn_code icode, tree exp,
+ rtx target)
{
rtx pat;
tree arg0 = CALL_EXPR_ARG (exp, 0);
/* Subroutine of ix86_expand_builtin to take care of comparison insns. */
static rtx
-ix86_expand_sse_compare (const struct builtin_description *d, tree exp,
- rtx target)
+ix86_expand_sse_compare (const struct builtin_description *d,
+ tree exp, rtx target, bool swap)
{
rtx pat;
tree arg0 = CALL_EXPR_ARG (exp, 0);
/* Swap operands if we have a comparison that isn't available in
hardware. */
- if (d->flag & BUILTIN_DESC_SWAP_OPERANDS)
+ if (swap)
{
rtx tmp = gen_reg_rtx (mode1);
emit_move_insn (tmp, op1);
tree arg1 = CALL_EXPR_ARG (exp, 1);
rtx op0 = expand_normal (arg0);
rtx op1 = expand_normal (arg1);
- rtx op2;
enum machine_mode mode0 = insn_data[d->icode].operand[0].mode;
enum machine_mode mode1 = insn_data[d->icode].operand[1].mode;
enum rtx_code comparison = d->comparison;
|| !(*insn_data[d->icode].operand[1].predicate) (op1, mode1))
op1 = copy_to_mode_reg (mode1, op1);
- op2 = gen_rtx_fmt_ee (comparison, mode0, op0, op1);
pat = GEN_FCN (d->icode) (op0, op1);
if (! pat)
return 0;
return SUBREG_REG (target);
}
-/* Return the integer constant in ARG. Constrain it to be in the range
- of the subparts of VEC_TYPE; issue an error if not. */
+/* Subroutine of ix86_expand_builtin to take care of ptest insns. */
-static int
-get_element_number (tree vec_type, tree arg)
+static rtx
+ix86_expand_sse_ptest (const struct builtin_description *d, tree exp,
+ rtx target)
{
- unsigned HOST_WIDE_INT elt, max = TYPE_VECTOR_SUBPARTS (vec_type) - 1;
-
- if (!host_integerp (arg, 1)
- || (elt = tree_low_cst (arg, 1), elt > max))
- {
- error ("selector must be an integer constant in the range 0..%wi", max);
- return 0;
- }
+ rtx pat;
+ tree arg0 = CALL_EXPR_ARG (exp, 0);
+ tree arg1 = CALL_EXPR_ARG (exp, 1);
+ rtx op0 = expand_normal (arg0);
+ rtx op1 = expand_normal (arg1);
+ enum machine_mode mode0 = insn_data[d->icode].operand[0].mode;
+ enum machine_mode mode1 = insn_data[d->icode].operand[1].mode;
+ enum rtx_code comparison = d->comparison;
- return elt;
-}
+ if (VECTOR_MODE_P (mode0))
+ op0 = safe_vector_operand (op0, mode0);
+ if (VECTOR_MODE_P (mode1))
+ op1 = safe_vector_operand (op1, mode1);
-/* A subroutine of ix86_expand_builtin. These builtins are a wrapper around
- ix86_expand_vector_init. We DO have language-level syntax for this, in
- the form of (type){ init-list }. Except that since we can't place emms
- instructions from inside the compiler, we can't allow the use of MMX
- registers unless the user explicitly asks for it. So we do *not* define
- vec_set/vec_extract/vec_init patterns for MMX modes in mmx.md. Instead
- we have builtins invoked by mmintrin.h that gives us license to emit
- these sorts of instructions. */
+ target = gen_reg_rtx (SImode);
+ emit_move_insn (target, const0_rtx);
+ target = gen_rtx_SUBREG (QImode, target, 0);
-static rtx
-ix86_expand_vec_init_builtin (tree type, tree exp, rtx target)
-{
- enum machine_mode tmode = TYPE_MODE (type);
- enum machine_mode inner_mode = GET_MODE_INNER (tmode);
- int i, n_elt = GET_MODE_NUNITS (tmode);
- rtvec v = rtvec_alloc (n_elt);
-
- gcc_assert (VECTOR_MODE_P (tmode));
- gcc_assert (call_expr_nargs (exp) == n_elt);
-
- for (i = 0; i < n_elt; ++i)
- {
- rtx x = expand_normal (CALL_EXPR_ARG (exp, i));
- RTVEC_ELT (v, i) = gen_lowpart (inner_mode, x);
- }
+ if ((optimize && !register_operand (op0, mode0))
+ || !(*insn_data[d->icode].operand[0].predicate) (op0, mode0))
+ op0 = copy_to_mode_reg (mode0, op0);
+ if ((optimize && !register_operand (op1, mode1))
+ || !(*insn_data[d->icode].operand[1].predicate) (op1, mode1))
+ op1 = copy_to_mode_reg (mode1, op1);
- if (!target || !register_operand (target, tmode))
- target = gen_reg_rtx (tmode);
+ pat = GEN_FCN (d->icode) (op0, op1);
+ if (! pat)
+ return 0;
+ emit_insn (pat);
+ emit_insn (gen_rtx_SET (VOIDmode,
+ gen_rtx_STRICT_LOW_PART (VOIDmode, target),
+ gen_rtx_fmt_ee (comparison, QImode,
+ SET_DEST (pat),
+ const0_rtx)));
- ix86_expand_vector_init (true, target, gen_rtx_PARALLEL (tmode, v));
- return target;
+ return SUBREG_REG (target);
}
-/* A subroutine of ix86_expand_builtin. These builtins are a wrapper around
- ix86_expand_vector_extract. They would be redundant (for non-MMX) if we
- had a language-level syntax for referencing vector elements. */
+/* Subroutine of ix86_expand_builtin to take care of pcmpestr[im] insns. */
static rtx
-ix86_expand_vec_ext_builtin (tree exp, rtx target)
+ix86_expand_sse_pcmpestr (const struct builtin_description *d,
+ tree exp, rtx target)
{
- enum machine_mode tmode, mode0;
- tree arg0, arg1;
- int elt;
- rtx op0;
+ rtx pat;
+ tree arg0 = CALL_EXPR_ARG (exp, 0);
+ tree arg1 = CALL_EXPR_ARG (exp, 1);
+ tree arg2 = CALL_EXPR_ARG (exp, 2);
+ tree arg3 = CALL_EXPR_ARG (exp, 3);
+ tree arg4 = CALL_EXPR_ARG (exp, 4);
+ rtx scratch0, scratch1;
+ rtx op0 = expand_normal (arg0);
+ rtx op1 = expand_normal (arg1);
+ rtx op2 = expand_normal (arg2);
+ rtx op3 = expand_normal (arg3);
+ rtx op4 = expand_normal (arg4);
+ enum machine_mode tmode0, tmode1, modev2, modei3, modev4, modei5, modeimm;
+
+ tmode0 = insn_data[d->icode].operand[0].mode;
+ tmode1 = insn_data[d->icode].operand[1].mode;
+ modev2 = insn_data[d->icode].operand[2].mode;
+ modei3 = insn_data[d->icode].operand[3].mode;
+ modev4 = insn_data[d->icode].operand[4].mode;
+ modei5 = insn_data[d->icode].operand[5].mode;
+ modeimm = insn_data[d->icode].operand[6].mode;
+
+ if (VECTOR_MODE_P (modev2))
+ op0 = safe_vector_operand (op0, modev2);
+ if (VECTOR_MODE_P (modev4))
+ op2 = safe_vector_operand (op2, modev4);
+
+ if (! (*insn_data[d->icode].operand[2].predicate) (op0, modev2))
+ op0 = copy_to_mode_reg (modev2, op0);
+ if (! (*insn_data[d->icode].operand[3].predicate) (op1, modei3))
+ op1 = copy_to_mode_reg (modei3, op1);
+ if ((optimize && !register_operand (op2, modev4))
+ || !(*insn_data[d->icode].operand[4].predicate) (op2, modev4))
+ op2 = copy_to_mode_reg (modev4, op2);
+ if (! (*insn_data[d->icode].operand[5].predicate) (op3, modei5))
+ op3 = copy_to_mode_reg (modei5, op3);
+
+ if (! (*insn_data[d->icode].operand[6].predicate) (op4, modeimm))
+ {
+ error ("the fifth argument must be a 8-bit immediate");
+ return const0_rtx;
+ }
- arg0 = CALL_EXPR_ARG (exp, 0);
- arg1 = CALL_EXPR_ARG (exp, 1);
+ if (d->code == IX86_BUILTIN_PCMPESTRI128)
+ {
+ if (optimize || !target
+ || GET_MODE (target) != tmode0
+ || ! (*insn_data[d->icode].operand[0].predicate) (target, tmode0))
+ target = gen_reg_rtx (tmode0);
- op0 = expand_normal (arg0);
- elt = get_element_number (TREE_TYPE (arg0), arg1);
+ scratch1 = gen_reg_rtx (tmode1);
- tmode = TYPE_MODE (TREE_TYPE (TREE_TYPE (arg0)));
- mode0 = TYPE_MODE (TREE_TYPE (arg0));
- gcc_assert (VECTOR_MODE_P (mode0));
+ pat = GEN_FCN (d->icode) (target, scratch1, op0, op1, op2, op3, op4);
+ }
+ else if (d->code == IX86_BUILTIN_PCMPESTRM128)
+ {
+ if (optimize || !target
+ || GET_MODE (target) != tmode1
+ || ! (*insn_data[d->icode].operand[1].predicate) (target, tmode1))
+ target = gen_reg_rtx (tmode1);
- op0 = force_reg (mode0, op0);
+ scratch0 = gen_reg_rtx (tmode0);
- if (optimize || !target || !register_operand (target, tmode))
- target = gen_reg_rtx (tmode);
+ pat = GEN_FCN (d->icode) (scratch0, target, op0, op1, op2, op3, op4);
+ }
+ else
+ {
+ gcc_assert (d->flag);
- ix86_expand_vector_extract (true, target, op0, elt);
+ scratch0 = gen_reg_rtx (tmode0);
+ scratch1 = gen_reg_rtx (tmode1);
- return target;
+ pat = GEN_FCN (d->icode) (scratch0, scratch1, op0, op1, op2, op3, op4);
+ }
+
+ if (! pat)
+ return 0;
+
+ emit_insn (pat);
+
+ if (d->flag)
+ {
+ target = gen_reg_rtx (SImode);
+ emit_move_insn (target, const0_rtx);
+ target = gen_rtx_SUBREG (QImode, target, 0);
+
+ emit_insn
+ (gen_rtx_SET (VOIDmode, gen_rtx_STRICT_LOW_PART (VOIDmode, target),
+ gen_rtx_fmt_ee (EQ, QImode,
+ gen_rtx_REG ((enum machine_mode) d->flag,
+ FLAGS_REG),
+ const0_rtx)));
+ return SUBREG_REG (target);
+ }
+ else
+ return target;
}
-/* A subroutine of ix86_expand_builtin. These builtins are a wrapper around
- ix86_expand_vector_set. They would be redundant (for non-MMX) if we had
- a language-level syntax for referencing vector elements. */
+
+/* Subroutine of ix86_expand_builtin to take care of pcmpistr[im] insns. */
static rtx
-ix86_expand_vec_set_builtin (tree exp)
+ix86_expand_sse_pcmpistr (const struct builtin_description *d,
+ tree exp, rtx target)
{
- enum machine_mode tmode, mode1;
- tree arg0, arg1, arg2;
- int elt;
- rtx op0, op1;
+ rtx pat;
+ tree arg0 = CALL_EXPR_ARG (exp, 0);
+ tree arg1 = CALL_EXPR_ARG (exp, 1);
+ tree arg2 = CALL_EXPR_ARG (exp, 2);
+ rtx scratch0, scratch1;
+ rtx op0 = expand_normal (arg0);
+ rtx op1 = expand_normal (arg1);
+ rtx op2 = expand_normal (arg2);
+ enum machine_mode tmode0, tmode1, modev2, modev3, modeimm;
- arg0 = CALL_EXPR_ARG (exp, 0);
- arg1 = CALL_EXPR_ARG (exp, 1);
- arg2 = CALL_EXPR_ARG (exp, 2);
+ tmode0 = insn_data[d->icode].operand[0].mode;
+ tmode1 = insn_data[d->icode].operand[1].mode;
+ modev2 = insn_data[d->icode].operand[2].mode;
+ modev3 = insn_data[d->icode].operand[3].mode;
+ modeimm = insn_data[d->icode].operand[4].mode;
- tmode = TYPE_MODE (TREE_TYPE (arg0));
- mode1 = TYPE_MODE (TREE_TYPE (TREE_TYPE (arg0)));
- gcc_assert (VECTOR_MODE_P (tmode));
+ if (VECTOR_MODE_P (modev2))
+ op0 = safe_vector_operand (op0, modev2);
+ if (VECTOR_MODE_P (modev3))
+ op1 = safe_vector_operand (op1, modev3);
- op0 = expand_expr (arg0, NULL_RTX, tmode, 0);
- op1 = expand_expr (arg1, NULL_RTX, mode1, 0);
- elt = get_element_number (TREE_TYPE (arg0), arg2);
+ if (! (*insn_data[d->icode].operand[2].predicate) (op0, modev2))
+ op0 = copy_to_mode_reg (modev2, op0);
+ if ((optimize && !register_operand (op1, modev3))
+ || !(*insn_data[d->icode].operand[3].predicate) (op1, modev3))
+ op1 = copy_to_mode_reg (modev3, op1);
- if (GET_MODE (op1) != mode1 && GET_MODE (op1) != VOIDmode)
- op1 = convert_modes (mode1, GET_MODE (op1), op1, true);
+ if (! (*insn_data[d->icode].operand[4].predicate) (op2, modeimm))
+ {
+ error ("the third argument must be a 8-bit immediate");
+ return const0_rtx;
+ }
- op0 = force_reg (tmode, op0);
- op1 = force_reg (mode1, op1);
+ if (d->code == IX86_BUILTIN_PCMPISTRI128)
+ {
+ if (optimize || !target
+ || GET_MODE (target) != tmode0
+ || ! (*insn_data[d->icode].operand[0].predicate) (target, tmode0))
+ target = gen_reg_rtx (tmode0);
+
+ scratch1 = gen_reg_rtx (tmode1);
+
+ pat = GEN_FCN (d->icode) (target, scratch1, op0, op1, op2);
+ }
+ else if (d->code == IX86_BUILTIN_PCMPISTRM128)
+ {
+ if (optimize || !target
+ || GET_MODE (target) != tmode1
+ || ! (*insn_data[d->icode].operand[1].predicate) (target, tmode1))
+ target = gen_reg_rtx (tmode1);
+
+ scratch0 = gen_reg_rtx (tmode0);
+
+ pat = GEN_FCN (d->icode) (scratch0, target, op0, op1, op2);
+ }
+ else
+ {
+ gcc_assert (d->flag);
+
+ scratch0 = gen_reg_rtx (tmode0);
+ scratch1 = gen_reg_rtx (tmode1);
+
+ pat = GEN_FCN (d->icode) (scratch0, scratch1, op0, op1, op2);
+ }
- ix86_expand_vector_set (true, op0, op1, elt);
+ if (! pat)
+ return 0;
+
+ emit_insn (pat);
- return op0;
+ if (d->flag)
+ {
+ target = gen_reg_rtx (SImode);
+ emit_move_insn (target, const0_rtx);
+ target = gen_rtx_SUBREG (QImode, target, 0);
+
+ emit_insn
+ (gen_rtx_SET (VOIDmode, gen_rtx_STRICT_LOW_PART (VOIDmode, target),
+ gen_rtx_fmt_ee (EQ, QImode,
+ gen_rtx_REG ((enum machine_mode) d->flag,
+ FLAGS_REG),
+ const0_rtx)));
+ return SUBREG_REG (target);
+ }
+ else
+ return target;
}
-/* Expand an expression EXP that calls a built-in function,
- with result going to TARGET if that's convenient
- (and in mode MODE if that's convenient).
- SUBTARGET may be used as the target for computing one of EXP's operands.
- IGNORE is nonzero if the value is to be ignored. */
+/* Subroutine of ix86_expand_builtin to take care of insns with
+ variable number of operands. */
static rtx
-ix86_expand_builtin (tree exp, rtx target, rtx subtarget ATTRIBUTE_UNUSED,
- enum machine_mode mode ATTRIBUTE_UNUSED,
- int ignore ATTRIBUTE_UNUSED)
+ix86_expand_args_builtin (const struct builtin_description *d,
+ tree exp, rtx target)
{
- const struct builtin_description *d;
- size_t i;
- enum insn_code icode;
- tree fndecl = TREE_OPERAND (CALL_EXPR_FN (exp), 0);
- tree arg0, arg1, arg2, arg3;
- rtx op0, op1, op2, op3, pat;
- enum machine_mode tmode, mode0, mode1, mode2, mode3, mode4;
- unsigned int fcode = DECL_FUNCTION_CODE (fndecl);
+ rtx pat, real_target;
+ unsigned int i, nargs;
+ unsigned int nargs_constant = 0;
+ int num_memory = 0;
+ struct
+ {
+ rtx op;
+ enum machine_mode mode;
+ } args[4];
+ bool last_arg_count = false;
+ enum insn_code icode = d->icode;
+ const struct insn_data *insn_p = &insn_data[icode];
+ enum machine_mode tmode = insn_p->operand[0].mode;
+ enum machine_mode rmode = VOIDmode;
+ bool swap = false;
+ enum rtx_code comparison = d->comparison;
- switch (fcode)
+ switch ((enum ix86_builtin_type) d->flag)
+ {
+ case INT_FTYPE_V2DI_V2DI_PTEST:
+ return ix86_expand_sse_ptest (d, exp, target);
+ case FLOAT128_FTYPE_FLOAT128:
+ case FLOAT_FTYPE_FLOAT:
+ case INT64_FTYPE_V4SF:
+ case INT64_FTYPE_V2DF:
+ case INT_FTYPE_V16QI:
+ case INT_FTYPE_V8QI:
+ case INT_FTYPE_V4SF:
+ case INT_FTYPE_V2DF:
+ case V16QI_FTYPE_V16QI:
+ case V8HI_FTYPE_V8HI:
+ case V8HI_FTYPE_V16QI:
+ case V8QI_FTYPE_V8QI:
+ case V4SI_FTYPE_V4SI:
+ case V4SI_FTYPE_V16QI:
+ case V4SI_FTYPE_V4SF:
+ case V4SI_FTYPE_V8HI:
+ case V4SI_FTYPE_V2DF:
+ case V4HI_FTYPE_V4HI:
+ case V4SF_FTYPE_V4SF:
+ case V4SF_FTYPE_V4SI:
+ case V4SF_FTYPE_V2DF:
+ case V2DI_FTYPE_V2DI:
+ case V2DI_FTYPE_V16QI:
+ case V2DI_FTYPE_V8HI:
+ case V2DI_FTYPE_V4SI:
+ case V2DF_FTYPE_V2DF:
+ case V2DF_FTYPE_V4SI:
+ case V2DF_FTYPE_V4SF:
+ case V2DF_FTYPE_V2SI:
+ case V2SI_FTYPE_V2SI:
+ case V2SI_FTYPE_V4SF:
+ case V2SI_FTYPE_V2SF:
+ case V2SI_FTYPE_V2DF:
+ case V2SF_FTYPE_V2SF:
+ case V2SF_FTYPE_V2SI:
+ nargs = 1;
+ break;
+ case V4SF_FTYPE_V4SF_VEC_MERGE:
+ case V2DF_FTYPE_V2DF_VEC_MERGE:
+ return ix86_expand_unop_vec_merge_builtin (icode, exp, target);
+ case FLOAT128_FTYPE_FLOAT128_FLOAT128:
+ case V16QI_FTYPE_V16QI_V16QI:
+ case V16QI_FTYPE_V8HI_V8HI:
+ case V8QI_FTYPE_V8QI_V8QI:
+ case V8QI_FTYPE_V4HI_V4HI:
+ case V8HI_FTYPE_V8HI_V8HI:
+ case V8HI_FTYPE_V16QI_V16QI:
+ case V8HI_FTYPE_V4SI_V4SI:
+ case V4SI_FTYPE_V4SI_V4SI:
+ case V4SI_FTYPE_V8HI_V8HI:
+ case V4SI_FTYPE_V4SF_V4SF:
+ case V4SI_FTYPE_V2DF_V2DF:
+ case V4HI_FTYPE_V4HI_V4HI:
+ case V4HI_FTYPE_V8QI_V8QI:
+ case V4HI_FTYPE_V2SI_V2SI:
+ case V4SF_FTYPE_V4SF_V4SF:
+ case V4SF_FTYPE_V4SF_V2SI:
+ case V4SF_FTYPE_V4SF_V2DF:
+ case V4SF_FTYPE_V4SF_DI:
+ case V4SF_FTYPE_V4SF_SI:
+ case V2DI_FTYPE_V2DI_V2DI:
+ case V2DI_FTYPE_V16QI_V16QI:
+ case V2DI_FTYPE_V4SI_V4SI:
+ case V2DI_FTYPE_V2DI_V16QI:
+ case V2DI_FTYPE_V2DF_V2DF:
+ case V2SI_FTYPE_V2SI_V2SI:
+ case V2SI_FTYPE_V4HI_V4HI:
+ case V2SI_FTYPE_V2SF_V2SF:
+ case V2DF_FTYPE_V2DF_V2DF:
+ case V2DF_FTYPE_V2DF_V4SF:
+ case V2DF_FTYPE_V2DF_DI:
+ case V2DF_FTYPE_V2DF_SI:
+ case V2SF_FTYPE_V2SF_V2SF:
+ case V1DI_FTYPE_V1DI_V1DI:
+ case V1DI_FTYPE_V8QI_V8QI:
+ case V1DI_FTYPE_V2SI_V2SI:
+ if (comparison == UNKNOWN)
+ return ix86_expand_binop_builtin (icode, exp, target);
+ nargs = 2;
+ break;
+ case V4SF_FTYPE_V4SF_V4SF_SWAP:
+ case V2DF_FTYPE_V2DF_V2DF_SWAP:
+ gcc_assert (comparison != UNKNOWN);
+ nargs = 2;
+ swap = true;
+ break;
+ case V8HI_FTYPE_V8HI_V8HI_COUNT:
+ case V8HI_FTYPE_V8HI_SI_COUNT:
+ case V4SI_FTYPE_V4SI_V4SI_COUNT:
+ case V4SI_FTYPE_V4SI_SI_COUNT:
+ case V4HI_FTYPE_V4HI_V4HI_COUNT:
+ case V4HI_FTYPE_V4HI_SI_COUNT:
+ case V2DI_FTYPE_V2DI_V2DI_COUNT:
+ case V2DI_FTYPE_V2DI_SI_COUNT:
+ case V2SI_FTYPE_V2SI_V2SI_COUNT:
+ case V2SI_FTYPE_V2SI_SI_COUNT:
+ case V1DI_FTYPE_V1DI_V1DI_COUNT:
+ case V1DI_FTYPE_V1DI_SI_COUNT:
+ nargs = 2;
+ last_arg_count = true;
+ break;
+ case UINT64_FTYPE_UINT64_UINT64:
+ case UINT_FTYPE_UINT_UINT:
+ case UINT_FTYPE_UINT_USHORT:
+ case UINT_FTYPE_UINT_UCHAR:
+ nargs = 2;
+ break;
+ case V2DI2TI_FTYPE_V2DI_INT:
+ nargs = 2;
+ rmode = V2DImode;
+ nargs_constant = 1;
+ break;
+ case V8HI_FTYPE_V8HI_INT:
+ case V4SI_FTYPE_V4SI_INT:
+ case V4HI_FTYPE_V4HI_INT:
+ case V4SF_FTYPE_V4SF_INT:
+ case V2DI_FTYPE_V2DI_INT:
+ case V2DF_FTYPE_V2DF_INT:
+ nargs = 2;
+ nargs_constant = 1;
+ break;
+ case V16QI_FTYPE_V16QI_V16QI_V16QI:
+ case V4SF_FTYPE_V4SF_V4SF_V4SF:
+ case V2DF_FTYPE_V2DF_V2DF_V2DF:
+ nargs = 3;
+ break;
+ case V16QI_FTYPE_V16QI_V16QI_INT:
+ case V8HI_FTYPE_V8HI_V8HI_INT:
+ case V4SI_FTYPE_V4SI_V4SI_INT:
+ case V4SF_FTYPE_V4SF_V4SF_INT:
+ case V2DI_FTYPE_V2DI_V2DI_INT:
+ case V2DF_FTYPE_V2DF_V2DF_INT:
+ nargs = 3;
+ nargs_constant = 1;
+ break;
+ case V2DI2TI_FTYPE_V2DI_V2DI_INT:
+ nargs = 3;
+ rmode = V2DImode;
+ nargs_constant = 1;
+ break;
+ case V1DI2DI_FTYPE_V1DI_V1DI_INT:
+ nargs = 3;
+ rmode = DImode;
+ nargs_constant = 1;
+ break;
+ case V2DI_FTYPE_V2DI_UINT_UINT:
+ nargs = 3;
+ nargs_constant = 2;
+ break;
+ case V2DI_FTYPE_V2DI_V2DI_UINT_UINT:
+ nargs = 4;
+ nargs_constant = 2;
+ break;
+ default:
+ gcc_unreachable ();
+ }
+
+ gcc_assert (nargs <= ARRAY_SIZE (args));
+
+ if (comparison != UNKNOWN)
{
- case IX86_BUILTIN_EMMS:
- emit_insn (gen_mmx_emms ());
- return 0;
+ gcc_assert (nargs == 2);
+ return ix86_expand_sse_compare (d, exp, target, swap);
+ }
- case IX86_BUILTIN_SFENCE:
- emit_insn (gen_sse_sfence ());
- return 0;
+ if (rmode == VOIDmode || rmode == tmode)
+ {
+ if (optimize
+ || target == 0
+ || GET_MODE (target) != tmode
+ || ! (*insn_p->operand[0].predicate) (target, tmode))
+ target = gen_reg_rtx (tmode);
+ real_target = target;
+ }
+ else
+ {
+ target = gen_reg_rtx (rmode);
+ real_target = simplify_gen_subreg (tmode, target, rmode, 0);
+ }
- case IX86_BUILTIN_MASKMOVQ:
- case IX86_BUILTIN_MASKMOVDQU:
- icode = (fcode == IX86_BUILTIN_MASKMOVQ
- ? CODE_FOR_mmx_maskmovq
- : CODE_FOR_sse2_maskmovdqu);
- /* Note the arg order is different from the operand order. */
- arg1 = CALL_EXPR_ARG (exp, 0);
- arg2 = CALL_EXPR_ARG (exp, 1);
- arg0 = CALL_EXPR_ARG (exp, 2);
- op0 = expand_normal (arg0);
- op1 = expand_normal (arg1);
- op2 = expand_normal (arg2);
- mode0 = insn_data[icode].operand[0].mode;
- mode1 = insn_data[icode].operand[1].mode;
- mode2 = insn_data[icode].operand[2].mode;
+ for (i = 0; i < nargs; i++)
+ {
+ tree arg = CALL_EXPR_ARG (exp, i);
+ rtx op = expand_normal (arg);
+ enum machine_mode mode = insn_p->operand[i + 1].mode;
+ bool match = (*insn_p->operand[i + 1].predicate) (op, mode);
- op0 = force_reg (Pmode, op0);
- op0 = gen_rtx_MEM (mode1, op0);
+ if (last_arg_count && (i + 1) == nargs)
+ {
+ /* SIMD shift insns take either an 8-bit immediate or
+ register as count. But builtin functions take int as
+ count. If count doesn't match, we put it in register. */
+ if (!match)
+ {
+ op = simplify_gen_subreg (SImode, op, GET_MODE (op), 0);
+ if (!(*insn_p->operand[i + 1].predicate) (op, mode))
+ op = copy_to_reg (op);
+ }
+ }
+ else if ((nargs - i) <= nargs_constant)
+ {
+ if (!match)
+ switch (icode)
+ {
+ case CODE_FOR_sse4_1_roundpd:
+ case CODE_FOR_sse4_1_roundps:
+ case CODE_FOR_sse4_1_roundsd:
+ case CODE_FOR_sse4_1_roundss:
+ case CODE_FOR_sse4_1_blendps:
+ error ("the last argument must be a 4-bit immediate");
+ return const0_rtx;
+
+ case CODE_FOR_sse4_1_blendpd:
+ error ("the last argument must be a 2-bit immediate");
+ return const0_rtx;
+
+ default:
+ switch (nargs_constant)
+ {
+ case 2:
+ if ((nargs - i) == nargs_constant)
+ {
+ error ("the next to last argument must be an 8-bit immediate");
+ break;
+ }
+ case 1:
+ error ("the last argument must be an 8-bit immediate");
+ break;
+ default:
+ gcc_unreachable ();
+ }
+ return const0_rtx;
+ }
+ }
+ else
+ {
+ if (VECTOR_MODE_P (mode))
+ op = safe_vector_operand (op, mode);
- if (! (*insn_data[icode].operand[0].predicate) (op0, mode0))
- op0 = copy_to_mode_reg (mode0, op0);
- if (! (*insn_data[icode].operand[1].predicate) (op1, mode1))
- op1 = copy_to_mode_reg (mode1, op1);
- if (! (*insn_data[icode].operand[2].predicate) (op2, mode2))
- op2 = copy_to_mode_reg (mode2, op2);
- pat = GEN_FCN (icode) (op0, op1, op2);
- if (! pat)
- return 0;
- emit_insn (pat);
- return 0;
+ /* If we aren't optimizing, only allow one memory operand to
+ be generated. */
+ if (memory_operand (op, mode))
+ num_memory++;
- case IX86_BUILTIN_SQRTSS:
- return ix86_expand_unop1_builtin (CODE_FOR_sse_vmsqrtv4sf2, exp, target);
- case IX86_BUILTIN_RSQRTSS:
- return ix86_expand_unop1_builtin (CODE_FOR_sse_vmrsqrtv4sf2, exp, target);
- case IX86_BUILTIN_RCPSS:
- return ix86_expand_unop1_builtin (CODE_FOR_sse_vmrcpv4sf2, exp, target);
-
- case IX86_BUILTIN_LOADUPS:
- return ix86_expand_unop_builtin (CODE_FOR_sse_movups, exp, target, 1);
-
- case IX86_BUILTIN_STOREUPS:
- return ix86_expand_store_builtin (CODE_FOR_sse_movups, exp);
-
- case IX86_BUILTIN_LOADHPS:
- case IX86_BUILTIN_LOADLPS:
- case IX86_BUILTIN_LOADHPD:
- case IX86_BUILTIN_LOADLPD:
- icode = (fcode == IX86_BUILTIN_LOADHPS ? CODE_FOR_sse_loadhps
- : fcode == IX86_BUILTIN_LOADLPS ? CODE_FOR_sse_loadlps
- : fcode == IX86_BUILTIN_LOADHPD ? CODE_FOR_sse2_loadhpd
- : CODE_FOR_sse2_loadlpd);
- arg0 = CALL_EXPR_ARG (exp, 0);
- arg1 = CALL_EXPR_ARG (exp, 1);
- op0 = expand_normal (arg0);
- op1 = expand_normal (arg1);
- tmode = insn_data[icode].operand[0].mode;
- mode0 = insn_data[icode].operand[1].mode;
- mode1 = insn_data[icode].operand[2].mode;
+ if (GET_MODE (op) == mode || GET_MODE (op) == VOIDmode)
+ {
+ if (optimize || !match || num_memory > 1)
+ op = copy_to_mode_reg (mode, op);
+ }
+ else
+ {
+ op = copy_to_reg (op);
+ op = simplify_gen_subreg (mode, op, GET_MODE (op), 0);
+ }
+ }
- op0 = force_reg (mode0, op0);
- op1 = gen_rtx_MEM (mode1, copy_to_mode_reg (Pmode, op1));
- if (optimize || target == 0
- || GET_MODE (target) != tmode
- || !register_operand (target, tmode))
- target = gen_reg_rtx (tmode);
- pat = GEN_FCN (icode) (target, op0, op1);
- if (! pat)
- return 0;
- emit_insn (pat);
- return target;
+ args[i].op = op;
+ args[i].mode = mode;
+ }
- case IX86_BUILTIN_STOREHPS:
- case IX86_BUILTIN_STORELPS:
- icode = (fcode == IX86_BUILTIN_STOREHPS ? CODE_FOR_sse_storehps
- : CODE_FOR_sse_storelps);
- arg0 = CALL_EXPR_ARG (exp, 0);
- arg1 = CALL_EXPR_ARG (exp, 1);
- op0 = expand_normal (arg0);
- op1 = expand_normal (arg1);
- mode0 = insn_data[icode].operand[0].mode;
- mode1 = insn_data[icode].operand[1].mode;
+ switch (nargs)
+ {
+ case 1:
+ pat = GEN_FCN (icode) (real_target, args[0].op);
+ break;
+ case 2:
+ pat = GEN_FCN (icode) (real_target, args[0].op, args[1].op);
+ break;
+ case 3:
+ pat = GEN_FCN (icode) (real_target, args[0].op, args[1].op,
+ args[2].op);
+ break;
+ case 4:
+ pat = GEN_FCN (icode) (real_target, args[0].op, args[1].op,
+ args[2].op, args[3].op);
+ break;
+ default:
+ gcc_unreachable ();
+ }
- op0 = gen_rtx_MEM (mode0, copy_to_mode_reg (Pmode, op0));
- op1 = force_reg (mode1, op1);
+ if (! pat)
+ return 0;
- pat = GEN_FCN (icode) (op0, op1);
- if (! pat)
- return 0;
- emit_insn (pat);
- return const0_rtx;
+ emit_insn (pat);
+ return target;
+}
- case IX86_BUILTIN_MOVNTPS:
- return ix86_expand_store_builtin (CODE_FOR_sse_movntv4sf, exp);
- case IX86_BUILTIN_MOVNTQ:
- return ix86_expand_store_builtin (CODE_FOR_sse_movntdi, exp);
+/* Subroutine of ix86_expand_builtin to take care of special insns
+ with variable number of operands. */
- case IX86_BUILTIN_LDMXCSR:
- op0 = expand_normal (CALL_EXPR_ARG (exp, 0));
- target = assign_386_stack_local (SImode, SLOT_TEMP);
- emit_move_insn (target, op0);
- emit_insn (gen_sse_ldmxcsr (target));
- return 0;
+static rtx
+ix86_expand_special_args_builtin (const struct builtin_description *d,
+ tree exp, rtx target)
+{
+ tree arg;
+ rtx pat, op;
+ unsigned int i, nargs, arg_adjust, memory;
+ struct
+ {
+ rtx op;
+ enum machine_mode mode;
+ } args[2];
+ enum insn_code icode = d->icode;
+ bool last_arg_constant = false;
+ const struct insn_data *insn_p = &insn_data[icode];
+ enum machine_mode tmode = insn_p->operand[0].mode;
+ enum { load, store } class;
- case IX86_BUILTIN_STMXCSR:
- target = assign_386_stack_local (SImode, SLOT_TEMP);
- emit_insn (gen_sse_stmxcsr (target));
- return copy_to_mode_reg (SImode, target);
+ switch ((enum ix86_special_builtin_type) d->flag)
+ {
+ case VOID_FTYPE_VOID:
+ emit_insn (GEN_FCN (icode) (target));
+ return 0;
+ case V2DI_FTYPE_PV2DI:
+ case V16QI_FTYPE_PCCHAR:
+ case V4SF_FTYPE_PCFLOAT:
+ case V2DF_FTYPE_PCDOUBLE:
+ nargs = 1;
+ class = load;
+ memory = 0;
+ break;
+ case VOID_FTYPE_PV2SF_V4SF:
+ case VOID_FTYPE_PV2DI_V2DI:
+ case VOID_FTYPE_PCHAR_V16QI:
+ case VOID_FTYPE_PFLOAT_V4SF:
+ case VOID_FTYPE_PDOUBLE_V2DF:
+ case VOID_FTYPE_PDI_DI:
+ case VOID_FTYPE_PINT_INT:
+ nargs = 1;
+ class = store;
+ /* Reserve memory operand for target. */
+ memory = ARRAY_SIZE (args);
+ break;
+ case V4SF_FTYPE_V4SF_PCV2SF:
+ case V2DF_FTYPE_V2DF_PCDOUBLE:
+ nargs = 2;
+ class = load;
+ memory = 1;
+ break;
+ default:
+ gcc_unreachable ();
+ }
- case IX86_BUILTIN_SHUFPS:
- case IX86_BUILTIN_SHUFPD:
- icode = (fcode == IX86_BUILTIN_SHUFPS
- ? CODE_FOR_sse_shufps
- : CODE_FOR_sse2_shufpd);
- arg0 = CALL_EXPR_ARG (exp, 0);
- arg1 = CALL_EXPR_ARG (exp, 1);
- arg2 = CALL_EXPR_ARG (exp, 2);
- op0 = expand_normal (arg0);
- op1 = expand_normal (arg1);
- op2 = expand_normal (arg2);
- tmode = insn_data[icode].operand[0].mode;
- mode0 = insn_data[icode].operand[1].mode;
- mode1 = insn_data[icode].operand[2].mode;
- mode2 = insn_data[icode].operand[3].mode;
+ gcc_assert (nargs <= ARRAY_SIZE (args));
- if (! (*insn_data[icode].operand[1].predicate) (op0, mode0))
- op0 = copy_to_mode_reg (mode0, op0);
- if ((optimize && !register_operand (op1, mode1))
- || !(*insn_data[icode].operand[2].predicate) (op1, mode1))
- op1 = copy_to_mode_reg (mode1, op1);
- if (! (*insn_data[icode].operand[3].predicate) (op2, mode2))
- {
- /* @@@ better error message */
- error ("mask must be an immediate");
- return gen_reg_rtx (tmode);
- }
- if (optimize || target == 0
+ if (class == store)
+ {
+ arg = CALL_EXPR_ARG (exp, 0);
+ op = expand_normal (arg);
+ gcc_assert (target == 0);
+ target = gen_rtx_MEM (tmode, copy_to_mode_reg (Pmode, op));
+ arg_adjust = 1;
+ }
+ else
+ {
+ arg_adjust = 0;
+ if (optimize
+ || target == 0
|| GET_MODE (target) != tmode
- || ! (*insn_data[icode].operand[0].predicate) (target, tmode))
+ || ! (*insn_p->operand[0].predicate) (target, tmode))
target = gen_reg_rtx (tmode);
- pat = GEN_FCN (icode) (target, op0, op1, op2);
- if (! pat)
- return 0;
- emit_insn (pat);
- return target;
-
- case IX86_BUILTIN_PSHUFW:
- case IX86_BUILTIN_PSHUFD:
- case IX86_BUILTIN_PSHUFHW:
- case IX86_BUILTIN_PSHUFLW:
- icode = ( fcode == IX86_BUILTIN_PSHUFHW ? CODE_FOR_sse2_pshufhw
- : fcode == IX86_BUILTIN_PSHUFLW ? CODE_FOR_sse2_pshuflw
- : fcode == IX86_BUILTIN_PSHUFD ? CODE_FOR_sse2_pshufd
- : CODE_FOR_mmx_pshufw);
- arg0 = CALL_EXPR_ARG (exp, 0);
- arg1 = CALL_EXPR_ARG (exp, 1);
- op0 = expand_normal (arg0);
- op1 = expand_normal (arg1);
- tmode = insn_data[icode].operand[0].mode;
- mode1 = insn_data[icode].operand[1].mode;
- mode2 = insn_data[icode].operand[2].mode;
+ }
+
+ for (i = 0; i < nargs; i++)
+ {
+ enum machine_mode mode = insn_p->operand[i + 1].mode;
+ bool match;
- if (! (*insn_data[icode].operand[1].predicate) (op0, mode1))
- op0 = copy_to_mode_reg (mode1, op0);
- if (! (*insn_data[icode].operand[2].predicate) (op1, mode2))
+ arg = CALL_EXPR_ARG (exp, i + arg_adjust);
+ op = expand_normal (arg);
+ match = (*insn_p->operand[i + 1].predicate) (op, mode);
+
+ if (last_arg_constant && (i + 1) == nargs)
{
- /* @@@ better error message */
- error ("mask must be an immediate");
- return const0_rtx;
+ if (!match)
+ switch (icode)
+ {
+ default:
+ error ("the last argument must be an 8-bit immediate");
+ return const0_rtx;
+ }
}
- if (target == 0
- || GET_MODE (target) != tmode
- || ! (*insn_data[icode].operand[0].predicate) (target, tmode))
- target = gen_reg_rtx (tmode);
- pat = GEN_FCN (icode) (target, op0, op1);
- if (! pat)
- return 0;
- emit_insn (pat);
- return target;
-
- case IX86_BUILTIN_PSLLWI128:
- icode = CODE_FOR_ashlv8hi3;
- goto do_pshifti;
- case IX86_BUILTIN_PSLLDI128:
- icode = CODE_FOR_ashlv4si3;
- goto do_pshifti;
- case IX86_BUILTIN_PSLLQI128:
- icode = CODE_FOR_ashlv2di3;
- goto do_pshifti;
- case IX86_BUILTIN_PSRAWI128:
- icode = CODE_FOR_ashrv8hi3;
- goto do_pshifti;
- case IX86_BUILTIN_PSRADI128:
- icode = CODE_FOR_ashrv4si3;
- goto do_pshifti;
- case IX86_BUILTIN_PSRLWI128:
- icode = CODE_FOR_lshrv8hi3;
- goto do_pshifti;
- case IX86_BUILTIN_PSRLDI128:
- icode = CODE_FOR_lshrv4si3;
- goto do_pshifti;
- case IX86_BUILTIN_PSRLQI128:
- icode = CODE_FOR_lshrv2di3;
- goto do_pshifti;
- do_pshifti:
- arg0 = CALL_EXPR_ARG (exp, 0);
- arg1 = CALL_EXPR_ARG (exp, 1);
- op0 = expand_normal (arg0);
- op1 = expand_normal (arg1);
-
- if (!CONST_INT_P (op1))
+ else
{
- error ("shift must be an immediate");
- return const0_rtx;
+ if (i == memory)
+ {
+ /* This must be the memory operand. */
+ op = gen_rtx_MEM (mode, copy_to_mode_reg (Pmode, op));
+ gcc_assert (GET_MODE (op) == mode
+ || GET_MODE (op) == VOIDmode);
+ }
+ else
+ {
+ /* This must be register. */
+ if (VECTOR_MODE_P (mode))
+ op = safe_vector_operand (op, mode);
+
+ gcc_assert (GET_MODE (op) == mode
+ || GET_MODE (op) == VOIDmode);
+ op = copy_to_mode_reg (mode, op);
+ }
}
- if (INTVAL (op1) < 0 || INTVAL (op1) > 255)
- op1 = GEN_INT (255);
- tmode = insn_data[icode].operand[0].mode;
- mode1 = insn_data[icode].operand[1].mode;
- if (! (*insn_data[icode].operand[1].predicate) (op0, mode1))
- op0 = copy_to_reg (op0);
+ args[i].op = op;
+ args[i].mode = mode;
+ }
- target = gen_reg_rtx (tmode);
- pat = GEN_FCN (icode) (target, op0, op1);
- if (!pat)
- return 0;
- emit_insn (pat);
- return target;
-
- case IX86_BUILTIN_PSLLW128:
- icode = CODE_FOR_ashlv8hi3;
- goto do_pshift;
- case IX86_BUILTIN_PSLLD128:
- icode = CODE_FOR_ashlv4si3;
- goto do_pshift;
- case IX86_BUILTIN_PSLLQ128:
- icode = CODE_FOR_ashlv2di3;
- goto do_pshift;
- case IX86_BUILTIN_PSRAW128:
- icode = CODE_FOR_ashrv8hi3;
- goto do_pshift;
- case IX86_BUILTIN_PSRAD128:
- icode = CODE_FOR_ashrv4si3;
- goto do_pshift;
- case IX86_BUILTIN_PSRLW128:
- icode = CODE_FOR_lshrv8hi3;
- goto do_pshift;
- case IX86_BUILTIN_PSRLD128:
- icode = CODE_FOR_lshrv4si3;
- goto do_pshift;
- case IX86_BUILTIN_PSRLQ128:
- icode = CODE_FOR_lshrv2di3;
- goto do_pshift;
- do_pshift:
- arg0 = CALL_EXPR_ARG (exp, 0);
- arg1 = CALL_EXPR_ARG (exp, 1);
- op0 = expand_normal (arg0);
- op1 = expand_normal (arg1);
+ switch (nargs)
+ {
+ case 1:
+ pat = GEN_FCN (icode) (target, args[0].op);
+ break;
+ case 2:
+ pat = GEN_FCN (icode) (target, args[0].op, args[1].op);
+ break;
+ default:
+ gcc_unreachable ();
+ }
- tmode = insn_data[icode].operand[0].mode;
- mode1 = insn_data[icode].operand[1].mode;
+ if (! pat)
+ return 0;
+ emit_insn (pat);
+ return class == store ? 0 : target;
+}
- if (! (*insn_data[icode].operand[1].predicate) (op0, mode1))
- op0 = copy_to_reg (op0);
+/* Return the integer constant in ARG. Constrain it to be in the range
+ of the subparts of VEC_TYPE; issue an error if not. */
- op1 = simplify_gen_subreg (TImode, op1, GET_MODE (op1), 0);
- if (! (*insn_data[icode].operand[2].predicate) (op1, TImode))
- op1 = copy_to_reg (op1);
+static int
+get_element_number (tree vec_type, tree arg)
+{
+ unsigned HOST_WIDE_INT elt, max = TYPE_VECTOR_SUBPARTS (vec_type) - 1;
- target = gen_reg_rtx (tmode);
- pat = GEN_FCN (icode) (target, op0, op1);
- if (!pat)
- return 0;
- emit_insn (pat);
- return target;
+ if (!host_integerp (arg, 1)
+ || (elt = tree_low_cst (arg, 1), elt > max))
+ {
+ error ("selector must be an integer constant in the range 0..%wi", max);
+ return 0;
+ }
- case IX86_BUILTIN_PSLLDQI128:
- case IX86_BUILTIN_PSRLDQI128:
- icode = (fcode == IX86_BUILTIN_PSLLDQI128 ? CODE_FOR_sse2_ashlti3
- : CODE_FOR_sse2_lshrti3);
- arg0 = CALL_EXPR_ARG (exp, 0);
- arg1 = CALL_EXPR_ARG (exp, 1);
- op0 = expand_normal (arg0);
- op1 = expand_normal (arg1);
- tmode = insn_data[icode].operand[0].mode;
- mode1 = insn_data[icode].operand[1].mode;
- mode2 = insn_data[icode].operand[2].mode;
+ return elt;
+}
- if (! (*insn_data[icode].operand[1].predicate) (op0, mode1))
- {
- op0 = copy_to_reg (op0);
- op0 = simplify_gen_subreg (mode1, op0, GET_MODE (op0), 0);
- }
- if (! (*insn_data[icode].operand[2].predicate) (op1, mode2))
- {
- error ("shift must be an immediate");
- return const0_rtx;
- }
- target = gen_reg_rtx (V2DImode);
- pat = GEN_FCN (icode) (simplify_gen_subreg (tmode, target, V2DImode, 0),
- op0, op1);
- if (! pat)
- return 0;
- emit_insn (pat);
- return target;
+/* A subroutine of ix86_expand_builtin. These builtins are a wrapper around
+ ix86_expand_vector_init. We DO have language-level syntax for this, in
+ the form of (type){ init-list }. Except that since we can't place emms
+ instructions from inside the compiler, we can't allow the use of MMX
+ registers unless the user explicitly asks for it. So we do *not* define
+ vec_set/vec_extract/vec_init patterns for MMX modes in mmx.md. Instead
+ we have builtins invoked by mmintrin.h that gives us license to emit
+ these sorts of instructions. */
+
+static rtx
+ix86_expand_vec_init_builtin (tree type, tree exp, rtx target)
+{
+ enum machine_mode tmode = TYPE_MODE (type);
+ enum machine_mode inner_mode = GET_MODE_INNER (tmode);
+ int i, n_elt = GET_MODE_NUNITS (tmode);
+ rtvec v = rtvec_alloc (n_elt);
+
+ gcc_assert (VECTOR_MODE_P (tmode));
+ gcc_assert (call_expr_nargs (exp) == n_elt);
- case IX86_BUILTIN_FEMMS:
- emit_insn (gen_mmx_femms ());
- return NULL_RTX;
+ for (i = 0; i < n_elt; ++i)
+ {
+ rtx x = expand_normal (CALL_EXPR_ARG (exp, i));
+ RTVEC_ELT (v, i) = gen_lowpart (inner_mode, x);
+ }
- case IX86_BUILTIN_PAVGUSB:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_uavgv8qi3, exp, target);
+ if (!target || !register_operand (target, tmode))
+ target = gen_reg_rtx (tmode);
- case IX86_BUILTIN_PF2ID:
- return ix86_expand_unop_builtin (CODE_FOR_mmx_pf2id, exp, target, 0);
+ ix86_expand_vector_init (true, target, gen_rtx_PARALLEL (tmode, v));
+ return target;
+}
- case IX86_BUILTIN_PFACC:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_haddv2sf3, exp, target);
+/* A subroutine of ix86_expand_builtin. These builtins are a wrapper around
+ ix86_expand_vector_extract. They would be redundant (for non-MMX) if we
+ had a language-level syntax for referencing vector elements. */
- case IX86_BUILTIN_PFADD:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_addv2sf3, exp, target);
+static rtx
+ix86_expand_vec_ext_builtin (tree exp, rtx target)
+{
+ enum machine_mode tmode, mode0;
+ tree arg0, arg1;
+ int elt;
+ rtx op0;
- case IX86_BUILTIN_PFCMPEQ:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_eqv2sf3, exp, target);
+ arg0 = CALL_EXPR_ARG (exp, 0);
+ arg1 = CALL_EXPR_ARG (exp, 1);
- case IX86_BUILTIN_PFCMPGE:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_gev2sf3, exp, target);
+ op0 = expand_normal (arg0);
+ elt = get_element_number (TREE_TYPE (arg0), arg1);
- case IX86_BUILTIN_PFCMPGT:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_gtv2sf3, exp, target);
+ tmode = TYPE_MODE (TREE_TYPE (TREE_TYPE (arg0)));
+ mode0 = TYPE_MODE (TREE_TYPE (arg0));
+ gcc_assert (VECTOR_MODE_P (mode0));
- case IX86_BUILTIN_PFMAX:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_smaxv2sf3, exp, target);
+ op0 = force_reg (mode0, op0);
- case IX86_BUILTIN_PFMIN:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_sminv2sf3, exp, target);
+ if (optimize || !target || !register_operand (target, tmode))
+ target = gen_reg_rtx (tmode);
- case IX86_BUILTIN_PFMUL:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_mulv2sf3, exp, target);
+ ix86_expand_vector_extract (true, target, op0, elt);
- case IX86_BUILTIN_PFRCP:
- return ix86_expand_unop_builtin (CODE_FOR_mmx_rcpv2sf2, exp, target, 0);
+ return target;
+}
- case IX86_BUILTIN_PFRCPIT1:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_rcpit1v2sf3, exp, target);
+/* A subroutine of ix86_expand_builtin. These builtins are a wrapper around
+ ix86_expand_vector_set. They would be redundant (for non-MMX) if we had
+ a language-level syntax for referencing vector elements. */
- case IX86_BUILTIN_PFRCPIT2:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_rcpit2v2sf3, exp, target);
+static rtx
+ix86_expand_vec_set_builtin (tree exp)
+{
+ enum machine_mode tmode, mode1;
+ tree arg0, arg1, arg2;
+ int elt;
+ rtx op0, op1, target;
- case IX86_BUILTIN_PFRSQIT1:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_rsqit1v2sf3, exp, target);
+ arg0 = CALL_EXPR_ARG (exp, 0);
+ arg1 = CALL_EXPR_ARG (exp, 1);
+ arg2 = CALL_EXPR_ARG (exp, 2);
- case IX86_BUILTIN_PFRSQRT:
- return ix86_expand_unop_builtin (CODE_FOR_mmx_rsqrtv2sf2, exp, target, 0);
+ tmode = TYPE_MODE (TREE_TYPE (arg0));
+ mode1 = TYPE_MODE (TREE_TYPE (TREE_TYPE (arg0)));
+ gcc_assert (VECTOR_MODE_P (tmode));
- case IX86_BUILTIN_PFSUB:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_subv2sf3, exp, target);
+ op0 = expand_expr (arg0, NULL_RTX, tmode, EXPAND_NORMAL);
+ op1 = expand_expr (arg1, NULL_RTX, mode1, EXPAND_NORMAL);
+ elt = get_element_number (TREE_TYPE (arg0), arg2);
- case IX86_BUILTIN_PFSUBR:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_subrv2sf3, exp, target);
+ if (GET_MODE (op1) != mode1 && GET_MODE (op1) != VOIDmode)
+ op1 = convert_modes (mode1, GET_MODE (op1), op1, true);
- case IX86_BUILTIN_PI2FD:
- return ix86_expand_unop_builtin (CODE_FOR_mmx_floatv2si2, exp, target, 0);
+ op0 = force_reg (tmode, op0);
+ op1 = force_reg (mode1, op1);
- case IX86_BUILTIN_PMULHRW:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_pmulhrwv4hi3, exp, target);
+ /* OP0 is the source of these builtin functions and shouldn't be
+ modified. Create a copy, use it and return it as target. */
+ target = gen_reg_rtx (tmode);
+ emit_move_insn (target, op0);
+ ix86_expand_vector_set (true, target, op1, elt);
- case IX86_BUILTIN_PF2IW:
- return ix86_expand_unop_builtin (CODE_FOR_mmx_pf2iw, exp, target, 0);
+ return target;
+}
- case IX86_BUILTIN_PFNACC:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_hsubv2sf3, exp, target);
+/* Expand an expression EXP that calls a built-in function,
+ with result going to TARGET if that's convenient
+ (and in mode MODE if that's convenient).
+ SUBTARGET may be used as the target for computing one of EXP's operands.
+ IGNORE is nonzero if the value is to be ignored. */
- case IX86_BUILTIN_PFPNACC:
- return ix86_expand_binop_builtin (CODE_FOR_mmx_addsubv2sf3, exp, target);
+static rtx
+ix86_expand_builtin (tree exp, rtx target, rtx subtarget ATTRIBUTE_UNUSED,
+ enum machine_mode mode ATTRIBUTE_UNUSED,
+ int ignore ATTRIBUTE_UNUSED)
+{
+ const struct builtin_description *d;
+ size_t i;
+ enum insn_code icode;
+ tree fndecl = TREE_OPERAND (CALL_EXPR_FN (exp), 0);
+ tree arg0, arg1, arg2;
+ rtx op0, op1, op2, pat;
+ enum machine_mode mode0, mode1, mode2;
+ unsigned int fcode = DECL_FUNCTION_CODE (fndecl);
- case IX86_BUILTIN_PI2FW:
- return ix86_expand_unop_builtin (CODE_FOR_mmx_pi2fw, exp, target, 0);
+ switch (fcode)
+ {
+ case IX86_BUILTIN_MASKMOVQ:
+ case IX86_BUILTIN_MASKMOVDQU:
+ icode = (fcode == IX86_BUILTIN_MASKMOVQ
+ ? CODE_FOR_mmx_maskmovq
+ : CODE_FOR_sse2_maskmovdqu);
+ /* Note the arg order is different from the operand order. */
+ arg1 = CALL_EXPR_ARG (exp, 0);
+ arg2 = CALL_EXPR_ARG (exp, 1);
+ arg0 = CALL_EXPR_ARG (exp, 2);
+ op0 = expand_normal (arg0);
+ op1 = expand_normal (arg1);
+ op2 = expand_normal (arg2);
+ mode0 = insn_data[icode].operand[0].mode;
+ mode1 = insn_data[icode].operand[1].mode;
+ mode2 = insn_data[icode].operand[2].mode;
- case IX86_BUILTIN_PSWAPDSI:
- return ix86_expand_unop_builtin (CODE_FOR_mmx_pswapdv2si2, exp, target, 0);
+ op0 = force_reg (Pmode, op0);
+ op0 = gen_rtx_MEM (mode1, op0);
- case IX86_BUILTIN_PSWAPDSF:
- return ix86_expand_unop_builtin (CODE_FOR_mmx_pswapdv2sf2, exp, target, 0);
+ if (! (*insn_data[icode].operand[0].predicate) (op0, mode0))
+ op0 = copy_to_mode_reg (mode0, op0);
+ if (! (*insn_data[icode].operand[1].predicate) (op1, mode1))
+ op1 = copy_to_mode_reg (mode1, op1);
+ if (! (*insn_data[icode].operand[2].predicate) (op2, mode2))
+ op2 = copy_to_mode_reg (mode2, op2);
+ pat = GEN_FCN (icode) (op0, op1, op2);
+ if (! pat)
+ return 0;
+ emit_insn (pat);
+ return 0;
- case IX86_BUILTIN_SQRTSD:
- return ix86_expand_unop1_builtin (CODE_FOR_sse2_vmsqrtv2df2, exp, target);
- case IX86_BUILTIN_LOADUPD:
- return ix86_expand_unop_builtin (CODE_FOR_sse2_movupd, exp, target, 1);
- case IX86_BUILTIN_STOREUPD:
- return ix86_expand_store_builtin (CODE_FOR_sse2_movupd, exp);
+ case IX86_BUILTIN_LDMXCSR:
+ op0 = expand_normal (CALL_EXPR_ARG (exp, 0));
+ target = assign_386_stack_local (SImode, SLOT_VIRTUAL);
+ emit_move_insn (target, op0);
+ emit_insn (gen_sse_ldmxcsr (target));
+ return 0;
- case IX86_BUILTIN_MFENCE:
- emit_insn (gen_sse2_mfence ());
- return 0;
- case IX86_BUILTIN_LFENCE:
- emit_insn (gen_sse2_lfence ());
- return 0;
+ case IX86_BUILTIN_STMXCSR:
+ target = assign_386_stack_local (SImode, SLOT_VIRTUAL);
+ emit_insn (gen_sse_stmxcsr (target));
+ return copy_to_mode_reg (SImode, target);
case IX86_BUILTIN_CLFLUSH:
arg0 = CALL_EXPR_ARG (exp, 0);
emit_insn (gen_sse2_clflush (op0));
return 0;
- case IX86_BUILTIN_MOVNTPD:
- return ix86_expand_store_builtin (CODE_FOR_sse2_movntv2df, exp);
- case IX86_BUILTIN_MOVNTDQ:
- return ix86_expand_store_builtin (CODE_FOR_sse2_movntv2di, exp);
- case IX86_BUILTIN_MOVNTI:
- return ix86_expand_store_builtin (CODE_FOR_sse2_movntsi, exp);
-
- case IX86_BUILTIN_LOADDQU:
- return ix86_expand_unop_builtin (CODE_FOR_sse2_movdqu, exp, target, 1);
- case IX86_BUILTIN_STOREDQU:
- return ix86_expand_store_builtin (CODE_FOR_sse2_movdqu, exp);
-
case IX86_BUILTIN_MONITOR:
arg0 = CALL_EXPR_ARG (exp, 0);
arg1 = CALL_EXPR_ARG (exp, 1);
emit_insn (gen_sse3_mwait (op0, op1));
return 0;
- case IX86_BUILTIN_LDDQU:
- return ix86_expand_unop_builtin (CODE_FOR_sse3_lddqu, exp,
- target, 1);
-
- case IX86_BUILTIN_PALIGNR:
- case IX86_BUILTIN_PALIGNR128:
- if (fcode == IX86_BUILTIN_PALIGNR)
- {
- icode = CODE_FOR_ssse3_palignrdi;
- mode = DImode;
- }
- else
- {
- icode = CODE_FOR_ssse3_palignrti;
- mode = V2DImode;
- }
- arg0 = CALL_EXPR_ARG (exp, 0);
- arg1 = CALL_EXPR_ARG (exp, 1);
- arg2 = CALL_EXPR_ARG (exp, 2);
- op0 = expand_expr (arg0, NULL_RTX, VOIDmode, 0);
- op1 = expand_expr (arg1, NULL_RTX, VOIDmode, 0);
- op2 = expand_expr (arg2, NULL_RTX, VOIDmode, 0);
- tmode = insn_data[icode].operand[0].mode;
- mode1 = insn_data[icode].operand[1].mode;
- mode2 = insn_data[icode].operand[2].mode;
- mode3 = insn_data[icode].operand[3].mode;
-
- if (! (*insn_data[icode].operand[1].predicate) (op0, mode1))
- {
- op0 = copy_to_reg (op0);
- op0 = simplify_gen_subreg (mode1, op0, GET_MODE (op0), 0);
- }
- if (! (*insn_data[icode].operand[2].predicate) (op1, mode2))
- {
- op1 = copy_to_reg (op1);
- op1 = simplify_gen_subreg (mode2, op1, GET_MODE (op1), 0);
- }
- if (! (*insn_data[icode].operand[3].predicate) (op2, mode3))
- {
- error ("shift must be an immediate");
- return const0_rtx;
- }
- target = gen_reg_rtx (mode);
- pat = GEN_FCN (icode) (simplify_gen_subreg (tmode, target, mode, 0),
- op0, op1, op2);
- if (! pat)
- return 0;
- emit_insn (pat);
- return target;
-
- case IX86_BUILTIN_MOVNTSD:
- return ix86_expand_store_builtin (CODE_FOR_sse4a_vmmovntv2df, exp);
-
- case IX86_BUILTIN_MOVNTSS:
- return ix86_expand_store_builtin (CODE_FOR_sse4a_vmmovntv4sf, exp);
-
- case IX86_BUILTIN_INSERTQ:
- case IX86_BUILTIN_EXTRQ:
- icode = (fcode == IX86_BUILTIN_EXTRQ
- ? CODE_FOR_sse4a_extrq
- : CODE_FOR_sse4a_insertq);
- arg0 = CALL_EXPR_ARG (exp, 0);
- arg1 = CALL_EXPR_ARG (exp, 1);
- op0 = expand_normal (arg0);
- op1 = expand_normal (arg1);
- tmode = insn_data[icode].operand[0].mode;
- mode1 = insn_data[icode].operand[1].mode;
- mode2 = insn_data[icode].operand[2].mode;
- if (! (*insn_data[icode].operand[1].predicate) (op0, mode1))
- op0 = copy_to_mode_reg (mode1, op0);
- if (! (*insn_data[icode].operand[2].predicate) (op1, mode2))
- op1 = copy_to_mode_reg (mode2, op1);
- if (optimize || target == 0
- || GET_MODE (target) != tmode
- || ! (*insn_data[icode].operand[0].predicate) (target, tmode))
- target = gen_reg_rtx (tmode);
- pat = GEN_FCN (icode) (target, op0, op1);
- if (! pat)
- return NULL_RTX;
- emit_insn (pat);
- return target;
-
- case IX86_BUILTIN_EXTRQI:
- icode = CODE_FOR_sse4a_extrqi;
- arg0 = CALL_EXPR_ARG (exp, 0);
- arg1 = CALL_EXPR_ARG (exp, 1);
- arg2 = CALL_EXPR_ARG (exp, 2);
- op0 = expand_normal (arg0);
- op1 = expand_normal (arg1);
- op2 = expand_normal (arg2);
- tmode = insn_data[icode].operand[0].mode;
- mode1 = insn_data[icode].operand[1].mode;
- mode2 = insn_data[icode].operand[2].mode;
- mode3 = insn_data[icode].operand[3].mode;
- if (! (*insn_data[icode].operand[1].predicate) (op0, mode1))
- op0 = copy_to_mode_reg (mode1, op0);
- if (! (*insn_data[icode].operand[2].predicate) (op1, mode2))
- {
- error ("index mask must be an immediate");
- return gen_reg_rtx (tmode);
- }
- if (! (*insn_data[icode].operand[3].predicate) (op2, mode3))
- {
- error ("length mask must be an immediate");
- return gen_reg_rtx (tmode);
- }
- if (optimize || target == 0
- || GET_MODE (target) != tmode
- || ! (*insn_data[icode].operand[0].predicate) (target, tmode))
- target = gen_reg_rtx (tmode);
- pat = GEN_FCN (icode) (target, op0, op1, op2);
- if (! pat)
- return NULL_RTX;
- emit_insn (pat);
- return target;
-
- case IX86_BUILTIN_INSERTQI:
- icode = CODE_FOR_sse4a_insertqi;
- arg0 = CALL_EXPR_ARG (exp, 0);
- arg1 = CALL_EXPR_ARG (exp, 1);
- arg2 = CALL_EXPR_ARG (exp, 2);
- arg3 = CALL_EXPR_ARG (exp, 3);
- op0 = expand_normal (arg0);
- op1 = expand_normal (arg1);
- op2 = expand_normal (arg2);
- op3 = expand_normal (arg3);
- tmode = insn_data[icode].operand[0].mode;
- mode1 = insn_data[icode].operand[1].mode;
- mode2 = insn_data[icode].operand[2].mode;
- mode3 = insn_data[icode].operand[3].mode;
- mode4 = insn_data[icode].operand[4].mode;
-
- if (! (*insn_data[icode].operand[1].predicate) (op0, mode1))
- op0 = copy_to_mode_reg (mode1, op0);
-
- if (! (*insn_data[icode].operand[2].predicate) (op1, mode2))
- op1 = copy_to_mode_reg (mode2, op1);
-
- if (! (*insn_data[icode].operand[3].predicate) (op2, mode3))
- {
- error ("index mask must be an immediate");
- return gen_reg_rtx (tmode);
- }
- if (! (*insn_data[icode].operand[4].predicate) (op3, mode4))
- {
- error ("length mask must be an immediate");
- return gen_reg_rtx (tmode);
- }
- if (optimize || target == 0
- || GET_MODE (target) != tmode
- || ! (*insn_data[icode].operand[0].predicate) (target, tmode))
- target = gen_reg_rtx (tmode);
- pat = GEN_FCN (icode) (target, op0, op1, op2, op3);
- if (! pat)
- return NULL_RTX;
- emit_insn (pat);
- return target;
-
case IX86_BUILTIN_VEC_INIT_V2SI:
case IX86_BUILTIN_VEC_INIT_V4HI:
case IX86_BUILTIN_VEC_INIT_V8QI:
case IX86_BUILTIN_VEC_EXT_V8HI:
case IX86_BUILTIN_VEC_EXT_V2SI:
case IX86_BUILTIN_VEC_EXT_V4HI:
+ case IX86_BUILTIN_VEC_EXT_V16QI:
return ix86_expand_vec_ext_builtin (exp, target);
+ case IX86_BUILTIN_VEC_SET_V2DI:
+ case IX86_BUILTIN_VEC_SET_V4SF:
+ case IX86_BUILTIN_VEC_SET_V4SI:
case IX86_BUILTIN_VEC_SET_V8HI:
case IX86_BUILTIN_VEC_SET_V4HI:
+ case IX86_BUILTIN_VEC_SET_V16QI:
return ix86_expand_vec_set_builtin (exp);
+ case IX86_BUILTIN_INFQ:
+ {
+ REAL_VALUE_TYPE inf;
+ rtx tmp;
+
+ real_inf (&inf);
+ tmp = CONST_DOUBLE_FROM_REAL_VALUE (inf, mode);
+
+ tmp = validize_mem (force_const_mem (mode, tmp));
+
+ if (target == 0)
+ target = gen_reg_rtx (mode);
+
+ emit_move_insn (target, tmp);
+ return target;
+ }
+
default:
break;
}
- for (i = 0, d = bdesc_2arg; i < ARRAY_SIZE (bdesc_2arg); i++, d++)
+ for (i = 0, d = bdesc_special_args;
+ i < ARRAY_SIZE (bdesc_special_args);
+ i++, d++)
if (d->code == fcode)
- {
- /* Compares are treated specially. */
- if (d->icode == CODE_FOR_sse_maskcmpv4sf3
- || d->icode == CODE_FOR_sse_vmmaskcmpv4sf3
- || d->icode == CODE_FOR_sse2_maskcmpv2df3
- || d->icode == CODE_FOR_sse2_vmmaskcmpv2df3)
- return ix86_expand_sse_compare (d, exp, target);
-
- return ix86_expand_binop_builtin (d->icode, exp, target);
- }
+ return ix86_expand_special_args_builtin (d, exp, target);
+
+ for (i = 0, d = bdesc_args;
+ i < ARRAY_SIZE (bdesc_args);
+ i++, d++)
+ if (d->code == fcode)
+ return ix86_expand_args_builtin (d, exp, target);
+
+ for (i = 0, d = bdesc_comi; i < ARRAY_SIZE (bdesc_comi); i++, d++)
+ if (d->code == fcode)
+ return ix86_expand_sse_comi (d, exp, target);
+
+ for (i = 0, d = bdesc_pcmpestr;
+ i < ARRAY_SIZE (bdesc_pcmpestr);
+ i++, d++)
+ if (d->code == fcode)
+ return ix86_expand_sse_pcmpestr (d, exp, target);
- for (i = 0, d = bdesc_1arg; i < ARRAY_SIZE (bdesc_1arg); i++, d++)
+ for (i = 0, d = bdesc_pcmpistr;
+ i < ARRAY_SIZE (bdesc_pcmpistr);
+ i++, d++)
if (d->code == fcode)
- return ix86_expand_unop_builtin (d->icode, exp, target, 0);
+ return ix86_expand_sse_pcmpistr (d, exp, target);
+
+ for (i = 0, d = bdesc_multi_arg; i < ARRAY_SIZE (bdesc_multi_arg); i++, d++)
+ if (d->code == fcode)
+ return ix86_expand_multi_arg_builtin (d->icode, exp, target,
+ (enum multi_arg_type)d->flag,
+ d->comparison);
+
+ gcc_unreachable ();
+}
+
+/* Returns a function decl for a vectorized version of the builtin function
+ with builtin function code FN and the result vector type TYPE, or NULL_TREE
+ if it is not available. */
+
+static tree
+ix86_builtin_vectorized_function (unsigned int fn, tree type_out,
+ tree type_in)
+{
+ enum machine_mode in_mode, out_mode;
+ int in_n, out_n;
+
+ if (TREE_CODE (type_out) != VECTOR_TYPE
+ || TREE_CODE (type_in) != VECTOR_TYPE)
+ return NULL_TREE;
+
+ out_mode = TYPE_MODE (TREE_TYPE (type_out));
+ out_n = TYPE_VECTOR_SUBPARTS (type_out);
+ in_mode = TYPE_MODE (TREE_TYPE (type_in));
+ in_n = TYPE_VECTOR_SUBPARTS (type_in);
+
+ switch (fn)
+ {
+ case BUILT_IN_SQRT:
+ if (out_mode == DFmode && out_n == 2
+ && in_mode == DFmode && in_n == 2)
+ return ix86_builtins[IX86_BUILTIN_SQRTPD];
+ break;
+
+ case BUILT_IN_SQRTF:
+ if (out_mode == SFmode && out_n == 4
+ && in_mode == SFmode && in_n == 4)
+ return ix86_builtins[IX86_BUILTIN_SQRTPS_NR];
+ break;
+
+ case BUILT_IN_LRINT:
+ if (out_mode == SImode && out_n == 4
+ && in_mode == DFmode && in_n == 2)
+ return ix86_builtins[IX86_BUILTIN_VEC_PACK_SFIX];
+ break;
+
+ case BUILT_IN_LRINTF:
+ if (out_mode == SImode && out_n == 4
+ && in_mode == SFmode && in_n == 4)
+ return ix86_builtins[IX86_BUILTIN_CVTPS2DQ];
+ break;
+
+ default:
+ ;
+ }
+
+ /* Dispatch to a handler for a vectorization library. */
+ if (ix86_veclib_handler)
+ return (*ix86_veclib_handler)(fn, type_out, type_in);
+
+ return NULL_TREE;
+}
+
+/* Handler for an SVML-style interface to
+ a library with vectorized intrinsics. */
+
+static tree
+ix86_veclibabi_svml (enum built_in_function fn, tree type_out, tree type_in)
+{
+ char name[20];
+ tree fntype, new_fndecl, args;
+ unsigned arity;
+ const char *bname;
+ enum machine_mode el_mode, in_mode;
+ int n, in_n;
+
+ /* The SVML is suitable for unsafe math only. */
+ if (!flag_unsafe_math_optimizations)
+ return NULL_TREE;
+
+ el_mode = TYPE_MODE (TREE_TYPE (type_out));
+ n = TYPE_VECTOR_SUBPARTS (type_out);
+ in_mode = TYPE_MODE (TREE_TYPE (type_in));
+ in_n = TYPE_VECTOR_SUBPARTS (type_in);
+ if (el_mode != in_mode
+ || n != in_n)
+ return NULL_TREE;
+
+ switch (fn)
+ {
+ case BUILT_IN_EXP:
+ case BUILT_IN_LOG:
+ case BUILT_IN_LOG10:
+ case BUILT_IN_POW:
+ case BUILT_IN_TANH:
+ case BUILT_IN_TAN:
+ case BUILT_IN_ATAN:
+ case BUILT_IN_ATAN2:
+ case BUILT_IN_ATANH:
+ case BUILT_IN_CBRT:
+ case BUILT_IN_SINH:
+ case BUILT_IN_SIN:
+ case BUILT_IN_ASINH:
+ case BUILT_IN_ASIN:
+ case BUILT_IN_COSH:
+ case BUILT_IN_COS:
+ case BUILT_IN_ACOSH:
+ case BUILT_IN_ACOS:
+ if (el_mode != DFmode || n != 2)
+ return NULL_TREE;
+ break;
+
+ case BUILT_IN_EXPF:
+ case BUILT_IN_LOGF:
+ case BUILT_IN_LOG10F:
+ case BUILT_IN_POWF:
+ case BUILT_IN_TANHF:
+ case BUILT_IN_TANF:
+ case BUILT_IN_ATANF:
+ case BUILT_IN_ATAN2F:
+ case BUILT_IN_ATANHF:
+ case BUILT_IN_CBRTF:
+ case BUILT_IN_SINHF:
+ case BUILT_IN_SINF:
+ case BUILT_IN_ASINHF:
+ case BUILT_IN_ASINF:
+ case BUILT_IN_COSHF:
+ case BUILT_IN_COSF:
+ case BUILT_IN_ACOSHF:
+ case BUILT_IN_ACOSF:
+ if (el_mode != SFmode || n != 4)
+ return NULL_TREE;
+ break;
+
+ default:
+ return NULL_TREE;
+ }
+
+ bname = IDENTIFIER_POINTER (DECL_NAME (implicit_built_in_decls[fn]));
+
+ if (fn == BUILT_IN_LOGF)
+ strcpy (name, "vmlsLn4");
+ else if (fn == BUILT_IN_LOG)
+ strcpy (name, "vmldLn2");
+ else if (n == 4)
+ {
+ sprintf (name, "vmls%s", bname+10);
+ name[strlen (name)-1] = '4';
+ }
+ else
+ sprintf (name, "vmld%s2", bname+10);
+
+ /* Convert to uppercase. */
+ name[4] &= ~0x20;
+
+ arity = 0;
+ for (args = DECL_ARGUMENTS (implicit_built_in_decls[fn]); args;
+ args = TREE_CHAIN (args))
+ arity++;
+
+ if (arity == 1)
+ fntype = build_function_type_list (type_out, type_in, NULL);
+ else
+ fntype = build_function_type_list (type_out, type_in, type_in, NULL);
- for (i = 0, d = bdesc_comi; i < ARRAY_SIZE (bdesc_comi); i++, d++)
- if (d->code == fcode)
- return ix86_expand_sse_comi (d, exp, target);
+ /* Build a function declaration for the vectorized function. */
+ new_fndecl = build_decl (FUNCTION_DECL, get_identifier (name), fntype);
+ TREE_PUBLIC (new_fndecl) = 1;
+ DECL_EXTERNAL (new_fndecl) = 1;
+ DECL_IS_NOVOPS (new_fndecl) = 1;
+ TREE_READONLY (new_fndecl) = 1;
- gcc_unreachable ();
+ return new_fndecl;
}
-/* Returns a function decl for a vectorized version of the builtin function
- with builtin function code FN and the result vector type TYPE, or NULL_TREE
- if it is not available. */
+/* Handler for an ACML-style interface to
+ a library with vectorized intrinsics. */
static tree
-ix86_builtin_vectorized_function (enum built_in_function fn, tree type_out,
- tree type_in)
-{
- enum machine_mode in_mode, out_mode;
- int in_n, out_n;
-
- if (TREE_CODE (type_out) != VECTOR_TYPE
- || TREE_CODE (type_in) != VECTOR_TYPE)
+ix86_veclibabi_acml (enum built_in_function fn, tree type_out, tree type_in)
+{
+ char name[20] = "__vr.._";
+ tree fntype, new_fndecl, args;
+ unsigned arity;
+ const char *bname;
+ enum machine_mode el_mode, in_mode;
+ int n, in_n;
+
+ /* The ACML is 64bits only and suitable for unsafe math only as
+ it does not correctly support parts of IEEE with the required
+ precision such as denormals. */
+ if (!TARGET_64BIT
+ || !flag_unsafe_math_optimizations)
return NULL_TREE;
- out_mode = TYPE_MODE (TREE_TYPE (type_out));
- out_n = TYPE_VECTOR_SUBPARTS (type_out);
+ el_mode = TYPE_MODE (TREE_TYPE (type_out));
+ n = TYPE_VECTOR_SUBPARTS (type_out);
in_mode = TYPE_MODE (TREE_TYPE (type_in));
in_n = TYPE_VECTOR_SUBPARTS (type_in);
+ if (el_mode != in_mode
+ || n != in_n)
+ return NULL_TREE;
switch (fn)
{
- case BUILT_IN_SQRT:
- if (out_mode == DFmode && out_n == 2
- && in_mode == DFmode && in_n == 2)
- return ix86_builtins[IX86_BUILTIN_SQRTPD];
- return NULL_TREE;
-
- case BUILT_IN_SQRTF:
- if (out_mode == SFmode && out_n == 4
- && in_mode == SFmode && in_n == 4)
- return ix86_builtins[IX86_BUILTIN_SQRTPS];
- return NULL_TREE;
+ case BUILT_IN_SIN:
+ case BUILT_IN_COS:
+ case BUILT_IN_EXP:
+ case BUILT_IN_LOG:
+ case BUILT_IN_LOG2:
+ case BUILT_IN_LOG10:
+ name[4] = 'd';
+ name[5] = '2';
+ if (el_mode != DFmode
+ || n != 2)
+ return NULL_TREE;
+ break;
- case BUILT_IN_LRINTF:
- if (out_mode == SImode && out_n == 4
- && in_mode == SFmode && in_n == 4)
- return ix86_builtins[IX86_BUILTIN_CVTPS2DQ];
- return NULL_TREE;
+ case BUILT_IN_SINF:
+ case BUILT_IN_COSF:
+ case BUILT_IN_EXPF:
+ case BUILT_IN_POWF:
+ case BUILT_IN_LOGF:
+ case BUILT_IN_LOG2F:
+ case BUILT_IN_LOG10F:
+ name[4] = 's';
+ name[5] = '4';
+ if (el_mode != SFmode
+ || n != 4)
+ return NULL_TREE;
+ break;
default:
- ;
+ return NULL_TREE;
}
- return NULL_TREE;
+ bname = IDENTIFIER_POINTER (DECL_NAME (implicit_built_in_decls[fn]));
+ sprintf (name + 7, "%s", bname+10);
+
+ arity = 0;
+ for (args = DECL_ARGUMENTS (implicit_built_in_decls[fn]); args;
+ args = TREE_CHAIN (args))
+ arity++;
+
+ if (arity == 1)
+ fntype = build_function_type_list (type_out, type_in, NULL);
+ else
+ fntype = build_function_type_list (type_out, type_in, type_in, NULL);
+
+ /* Build a function declaration for the vectorized function. */
+ new_fndecl = build_decl (FUNCTION_DECL, get_identifier (name), fntype);
+ TREE_PUBLIC (new_fndecl) = 1;
+ DECL_EXTERNAL (new_fndecl) = 1;
+ DECL_IS_NOVOPS (new_fndecl) = 1;
+ TREE_READONLY (new_fndecl) = 1;
+
+ return new_fndecl;
}
+
/* Returns a decl of a function that implements conversion of the
input vector of type TYPE, or NULL_TREE if it is not available. */
static tree
-ix86_builtin_conversion (enum tree_code code, tree type)
+ix86_vectorize_builtin_conversion (unsigned int code, tree type)
{
if (TREE_CODE (type) != VECTOR_TYPE)
return NULL_TREE;
-
+
switch (code)
{
case FLOAT_EXPR:
}
}
+/* Returns a code for a target-specific builtin that implements
+ reciprocal of the function, or NULL_TREE if not available. */
+
+static tree
+ix86_builtin_reciprocal (unsigned int fn, bool md_fn,
+ bool sqrt ATTRIBUTE_UNUSED)
+{
+ if (! (TARGET_SSE_MATH && TARGET_RECIP && !optimize_size
+ && flag_finite_math_only && !flag_trapping_math
+ && flag_unsafe_math_optimizations))
+ return NULL_TREE;
+
+ if (md_fn)
+ /* Machine dependent builtins. */
+ switch (fn)
+ {
+ /* Vectorized version of sqrt to rsqrt conversion. */
+ case IX86_BUILTIN_SQRTPS_NR:
+ return ix86_builtins[IX86_BUILTIN_RSQRTPS_NR];
+
+ default:
+ return NULL_TREE;
+ }
+ else
+ /* Normal builtins. */
+ switch (fn)
+ {
+ /* Sqrt to rsqrt conversion. */
+ case BUILT_IN_SQRTF:
+ return ix86_builtins[IX86_BUILTIN_RSQRTF];
+
+ default:
+ return NULL_TREE;
+ }
+}
+
/* Store OPERAND to the memory after reload is completed. This means
that we can't easily use assign_stack_local. */
rtx
Narrow ALL_REGS to GENERAL_REGS. This supports allowing movsf and
movdf to do mem-to-mem moves through integer regs. */
enum reg_class
-ix86_preferred_reload_class (rtx x, enum reg_class class)
+ix86_preferred_reload_class (rtx x, enum reg_class regclass)
{
enum machine_mode mode = GET_MODE (x);
/* We're only allowed to return a subclass of CLASS. Many of the
following checks fail for NO_REGS, so eliminate that early. */
- if (class == NO_REGS)
+ if (regclass == NO_REGS)
return NO_REGS;
/* All classes can load zeros. */
if (x == CONST0_RTX (mode))
- return class;
+ return regclass;
/* Force constants into memory if we are loading a (nonzero) constant into
an MMX or SSE register. This is because there are no MMX/SSE instructions
to load from a constant. */
if (CONSTANT_P (x)
- && (MAYBE_MMX_CLASS_P (class) || MAYBE_SSE_CLASS_P (class)))
+ && (MAYBE_MMX_CLASS_P (regclass) || MAYBE_SSE_CLASS_P (regclass)))
return NO_REGS;
/* Prefer SSE regs only, if we can use them for math. */
if (TARGET_SSE_MATH && !TARGET_MIX_SSE_I387 && SSE_FLOAT_MODE_P (mode))
- return SSE_CLASS_P (class) ? class : NO_REGS;
+ return SSE_CLASS_P (regclass) ? regclass : NO_REGS;
/* Floating-point constants need more complex checks. */
if (GET_CODE (x) == CONST_DOUBLE && GET_MODE (x) != VOIDmode)
{
/* General regs can load everything. */
- if (reg_class_subset_p (class, GENERAL_REGS))
- return class;
+ if (reg_class_subset_p (regclass, GENERAL_REGS))
+ return regclass;
/* Floats can load 0 and 1 plus some others. Note that we eliminated
zero above. We only want to wind up preferring 80387 registers if
&& standard_80387_constant_p (x))
{
/* Limit class to non-sse. */
- if (class == FLOAT_SSE_REGS)
+ if (regclass == FLOAT_SSE_REGS)
return FLOAT_REGS;
- if (class == FP_TOP_SSE_REGS)
+ if (regclass == FP_TOP_SSE_REGS)
return FP_TOP_REG;
- if (class == FP_SECOND_SSE_REGS)
+ if (regclass == FP_SECOND_SSE_REGS)
return FP_SECOND_REG;
- if (class == FLOAT_INT_REGS || class == FLOAT_REGS)
- return class;
+ if (regclass == FLOAT_INT_REGS || regclass == FLOAT_REGS)
+ return regclass;
}
return NO_REGS;
(plus soft-fp const_int). Which can only be computed into general
regs. */
if (GET_CODE (x) == PLUS)
- return reg_class_subset_p (class, GENERAL_REGS) ? class : NO_REGS;
+ return reg_class_subset_p (regclass, GENERAL_REGS) ? regclass : NO_REGS;
/* QImode constants are easy to load, but non-constant QImode data
must go into Q_REGS. */
if (GET_MODE (x) == QImode && !CONSTANT_P (x))
{
- if (reg_class_subset_p (class, Q_REGS))
- return class;
- if (reg_class_subset_p (Q_REGS, class))
+ if (reg_class_subset_p (regclass, Q_REGS))
+ return regclass;
+ if (reg_class_subset_p (Q_REGS, regclass))
return Q_REGS;
return NO_REGS;
}
- return class;
+ return regclass;
}
/* Discourage putting floating-point values in SSE registers unless
SSE math is being used, and likewise for the 387 registers. */
enum reg_class
-ix86_preferred_output_reload_class (rtx x, enum reg_class class)
+ix86_preferred_output_reload_class (rtx x, enum reg_class regclass)
{
enum machine_mode mode = GET_MODE (x);
alternative: if reload cannot do this, it will still use its choice. */
mode = GET_MODE (x);
if (TARGET_SSE_MATH && SSE_FLOAT_MODE_P (mode))
- return MAYBE_SSE_CLASS_P (class) ? SSE_REGS : NO_REGS;
+ return MAYBE_SSE_CLASS_P (regclass) ? SSE_REGS : NO_REGS;
- if (TARGET_80387 && SCALAR_FLOAT_MODE_P (mode))
+ if (X87_FLOAT_MODE_P (mode))
{
- if (class == FP_TOP_SSE_REGS)
+ if (regclass == FP_TOP_SSE_REGS)
return FP_TOP_REG;
- else if (class == FP_SECOND_SSE_REGS)
+ else if (regclass == FP_SECOND_SSE_REGS)
return FP_SECOND_REG;
else
- return FLOAT_CLASS_P (class) ? class : NO_REGS;
+ return FLOAT_CLASS_P (regclass) ? regclass : NO_REGS;
+ }
+
+ return regclass;
+}
+
+static enum reg_class
+ix86_secondary_reload (bool in_p, rtx x, enum reg_class class,
+ enum machine_mode mode,
+ secondary_reload_info *sri ATTRIBUTE_UNUSED)
+{
+ /* QImode spills from non-QI registers require
+ intermediate register on 32bit targets. */
+ if (!in_p && mode == QImode && class == NON_Q_REGS
+ && !TARGET_64BIT)
+ {
+ int regno;
+
+ if (REG_P (x))
+ regno = REGNO (x);
+ else
+ regno = -1;
+
+ if (regno >= FIRST_PSEUDO_REGISTER || GET_CODE (x) == SUBREG)
+ regno = true_regnum (x);
+
+ /* Return Q_REGS if the operand is in memory. */
+ if (regno == -1)
+ return Q_REGS;
}
- return class;
+ return NO_REGS;
}
/* If we are copying between general and FP registers, we need a memory
location. The same is true for SSE and MMX registers.
+ To optimize register_move_cost performance, allow inline variant.
+
The macro can't work reliably when one of the CLASSES is class containing
registers from multiple units (SSE, MMX, integer). We avoid this by never
combining those units in single alternative in the machine description.
When STRICT is false, we are being called from REGISTER_MOVE_COST, so do not
enforce these sanity checks. */
-int
-ix86_secondary_memory_needed (enum reg_class class1, enum reg_class class2,
+static inline int
+inline_secondary_memory_needed (enum reg_class class1, enum reg_class class2,
enum machine_mode mode, int strict)
{
if (MAYBE_FLOAT_CLASS_P (class1) != FLOAT_CLASS_P (class1)
return false;
}
+int
+ix86_secondary_memory_needed (enum reg_class class1, enum reg_class class2,
+ enum machine_mode mode, int strict)
+{
+ return inline_secondary_memory_needed (class1, class2, mode, strict);
+}
+
/* Return true if the registers in CLASS cannot represent the change from
modes FROM to TO. */
bool
ix86_cannot_change_mode_class (enum machine_mode from, enum machine_mode to,
- enum reg_class class)
+ enum reg_class regclass)
{
if (from == to)
return false;
/* x87 registers can't do subreg at all, as all values are reformatted
to extended precision. */
- if (MAYBE_FLOAT_CLASS_P (class))
+ if (MAYBE_FLOAT_CLASS_P (regclass))
return true;
- if (MAYBE_SSE_CLASS_P (class) || MAYBE_MMX_CLASS_P (class))
+ if (MAYBE_SSE_CLASS_P (regclass) || MAYBE_MMX_CLASS_P (regclass))
{
/* Vector registers do not support QI or HImode loads. If we don't
disallow a change to these modes, reload will assume it's ok to
return false;
}
+/* Return the cost of moving data of mode M between a
+ register and memory. A value of 2 is the default; this cost is
+ relative to those in `REGISTER_MOVE_COST'.
+
+ This function is used extensively by register_move_cost that is used to
+ build tables at startup. Make it inline in this case.
+ When IN is 2, return maximum of in and out move cost.
+
+ If moving between registers and memory is more expensive than
+ between two registers, you should define this macro to express the
+ relative cost.
+
+ Model also increased moving costs of QImode registers in non
+ Q_REGS classes.
+ */
+static inline int
+inline_memory_move_cost (enum machine_mode mode, enum reg_class regclass,
+ int in)
+{
+ int cost;
+ if (FLOAT_CLASS_P (regclass))
+ {
+ int index;
+ switch (mode)
+ {
+ case SFmode:
+ index = 0;
+ break;
+ case DFmode:
+ index = 1;
+ break;
+ case XFmode:
+ index = 2;
+ break;
+ default:
+ return 100;
+ }
+ if (in == 2)
+ return MAX (ix86_cost->fp_load [index], ix86_cost->fp_store [index]);
+ return in ? ix86_cost->fp_load [index] : ix86_cost->fp_store [index];
+ }
+ if (SSE_CLASS_P (regclass))
+ {
+ int index;
+ switch (GET_MODE_SIZE (mode))
+ {
+ case 4:
+ index = 0;
+ break;
+ case 8:
+ index = 1;
+ break;
+ case 16:
+ index = 2;
+ break;
+ default:
+ return 100;
+ }
+ if (in == 2)
+ return MAX (ix86_cost->sse_load [index], ix86_cost->sse_store [index]);
+ return in ? ix86_cost->sse_load [index] : ix86_cost->sse_store [index];
+ }
+ if (MMX_CLASS_P (regclass))
+ {
+ int index;
+ switch (GET_MODE_SIZE (mode))
+ {
+ case 4:
+ index = 0;
+ break;
+ case 8:
+ index = 1;
+ break;
+ default:
+ return 100;
+ }
+ if (in)
+ return MAX (ix86_cost->mmx_load [index], ix86_cost->mmx_store [index]);
+ return in ? ix86_cost->mmx_load [index] : ix86_cost->mmx_store [index];
+ }
+ switch (GET_MODE_SIZE (mode))
+ {
+ case 1:
+ if (Q_CLASS_P (regclass) || TARGET_64BIT)
+ {
+ if (!in)
+ return ix86_cost->int_store[0];
+ if (TARGET_PARTIAL_REG_DEPENDENCY && !optimize_size)
+ cost = ix86_cost->movzbl_load;
+ else
+ cost = ix86_cost->int_load[0];
+ if (in == 2)
+ return MAX (cost, ix86_cost->int_store[0]);
+ return cost;
+ }
+ else
+ {
+ if (in == 2)
+ return MAX (ix86_cost->movzbl_load, ix86_cost->int_store[0] + 4);
+ if (in)
+ return ix86_cost->movzbl_load;
+ else
+ return ix86_cost->int_store[0] + 4;
+ }
+ break;
+ case 2:
+ if (in == 2)
+ return MAX (ix86_cost->int_load[1], ix86_cost->int_store[1]);
+ return in ? ix86_cost->int_load[1] : ix86_cost->int_store[1];
+ default:
+ /* Compute number of 32bit moves needed. TFmode is moved as XFmode. */
+ if (mode == TFmode)
+ mode = XFmode;
+ if (in == 2)
+ cost = MAX (ix86_cost->int_load[2] , ix86_cost->int_store[2]);
+ else if (in)
+ cost = ix86_cost->int_load[2];
+ else
+ cost = ix86_cost->int_store[2];
+ return (cost * (((int) GET_MODE_SIZE (mode)
+ + UNITS_PER_WORD - 1) / UNITS_PER_WORD));
+ }
+}
+
+int
+ix86_memory_move_cost (enum machine_mode mode, enum reg_class regclass, int in)
+{
+ return inline_memory_move_cost (mode, regclass, in);
+}
+
+
/* Return the cost of moving data from a register in class CLASS1 to
one in class CLASS2.
by load. In order to avoid bad register allocation choices, we need
for this to be *at least* as high as the symmetric MEMORY_MOVE_COST. */
- if (ix86_secondary_memory_needed (class1, class2, mode, 0))
+ if (inline_secondary_memory_needed (class1, class2, mode, 0))
{
int cost = 1;
- cost += MAX (MEMORY_MOVE_COST (mode, class1, 0),
- MEMORY_MOVE_COST (mode, class1, 1));
- cost += MAX (MEMORY_MOVE_COST (mode, class2, 0),
- MEMORY_MOVE_COST (mode, class2, 1));
+ cost += inline_memory_move_cost (mode, class1, 2);
+ cost += inline_memory_move_cost (mode, class2, 2);
/* In case of copying from general_purpose_register we may emit multiple
stores followed by single load causing memory size mismatch stall.
/* Moves between SSE/MMX and integer unit are expensive. */
if (MMX_CLASS_P (class1) != MMX_CLASS_P (class2)
|| SSE_CLASS_P (class1) != SSE_CLASS_P (class2))
- return ix86_cost->mmxsse_to_integer;
+
+ /* ??? By keeping returned value relatively high, we limit the number
+ of moves between integer and MMX/SSE registers for all targets.
+ Additionally, high value prevents problem with x86_modes_tieable_p(),
+ where integer modes in MMX/SSE registers are not tieable
+ because of missing QImode and HImode moves to, from or between
+ MMX/SSE registers. */
+ return MAX (8, ix86_cost->mmxsse_to_integer);
+
if (MAYBE_FLOAT_CLASS_P (class1))
return ix86_cost->fp_move;
if (MAYBE_SSE_CLASS_P (class1))
return 1;
else if (VALID_FP_MODE_P (mode))
return 1;
+ else if (VALID_DFP_MODE_P (mode))
+ return 1;
/* Lots of MMX code casts 8 byte vector modes to DImode. If we then go
on to use that value in smaller contexts, this can easily force a
pseudo to be allocated to GENERAL_REGS. Since this is no worse than
/* If MODE2 is appropriate for an MMX register, then tie
with any other mode acceptable to MMX registers. */
if (GET_MODE_SIZE (mode2) == 8
- && ix86_hard_regno_mode_ok (FIRST_MMX_REG, mode2))
- return (GET_MODE_SIZE (mode1) == 8
- && ix86_hard_regno_mode_ok (FIRST_MMX_REG, mode1));
-
- return false;
-}
-
-/* Return the cost of moving data of mode M between a
- register and memory. A value of 2 is the default; this cost is
- relative to those in `REGISTER_MOVE_COST'.
-
- If moving between registers and memory is more expensive than
- between two registers, you should define this macro to express the
- relative cost.
-
- Model also increased moving costs of QImode registers in non
- Q_REGS classes.
- */
-int
-ix86_memory_move_cost (enum machine_mode mode, enum reg_class class, int in)
-{
- if (FLOAT_CLASS_P (class))
- {
- int index;
- switch (mode)
- {
- case SFmode:
- index = 0;
- break;
- case DFmode:
- index = 1;
- break;
- case XFmode:
- index = 2;
- break;
- default:
- return 100;
- }
- return in ? ix86_cost->fp_load [index] : ix86_cost->fp_store [index];
- }
- if (SSE_CLASS_P (class))
- {
- int index;
- switch (GET_MODE_SIZE (mode))
- {
- case 4:
- index = 0;
- break;
- case 8:
- index = 1;
- break;
- case 16:
- index = 2;
- break;
- default:
- return 100;
- }
- return in ? ix86_cost->sse_load [index] : ix86_cost->sse_store [index];
- }
- if (MMX_CLASS_P (class))
- {
- int index;
- switch (GET_MODE_SIZE (mode))
- {
- case 4:
- index = 0;
- break;
- case 8:
- index = 1;
- break;
- default:
- return 100;
- }
- return in ? ix86_cost->mmx_load [index] : ix86_cost->mmx_store [index];
- }
- switch (GET_MODE_SIZE (mode))
- {
- case 1:
- if (in)
- return (Q_CLASS_P (class) ? ix86_cost->int_load[0]
- : ix86_cost->movzbl_load);
- else
- return (Q_CLASS_P (class) ? ix86_cost->int_store[0]
- : ix86_cost->int_store[0] + 4);
- break;
- case 2:
- return in ? ix86_cost->int_load[1] : ix86_cost->int_store[1];
- default:
- /* Compute number of 32bit moves needed. TFmode is moved as XFmode. */
- if (mode == TFmode)
- mode = XFmode;
- return ((in ? ix86_cost->int_load[2] : ix86_cost->int_store[2])
- * (((int) GET_MODE_SIZE (mode)
- + UNITS_PER_WORD - 1) / UNITS_PER_WORD));
- }
+ && ix86_hard_regno_mode_ok (FIRST_MMX_REG, mode2))
+ return (GET_MODE_SIZE (mode1) == 8
+ && ix86_hard_regno_mode_ok (FIRST_MMX_REG, mode1));
+
+ return false;
}
/* Compute a (partial) cost for rtx X. Return true if the complete
scanned. In either case, *TOTAL contains the cost result. */
static bool
-ix86_rtx_costs (rtx x, int code, int outer_code, int *total)
+ix86_rtx_costs (rtx x, int code, int outer_code_i, int *total)
{
+ enum rtx_code outer_code = (enum rtx_code) outer_code_i;
enum machine_mode mode = GET_MODE (x);
switch (code)
return false;
case MULT:
- if (FLOAT_MODE_P (mode))
+ if (SSE_FLOAT_MODE_P (mode) && TARGET_SSE_MATH)
+ {
+ /* ??? SSE scalar cost should be used here. */
+ *total = ix86_cost->fmul;
+ return false;
+ }
+ else if (X87_FLOAT_MODE_P (mode))
+ {
+ *total = ix86_cost->fmul;
+ return false;
+ }
+ else if (FLOAT_MODE_P (mode))
{
+ /* ??? SSE vector cost should be used here. */
*total = ix86_cost->fmul;
return false;
}
nbits = 7;
/* Compute costs correctly for widening multiplication. */
- if ((GET_CODE (op0) == SIGN_EXTEND || GET_CODE (op1) == ZERO_EXTEND)
+ if ((GET_CODE (op0) == SIGN_EXTEND || GET_CODE (op0) == ZERO_EXTEND)
&& GET_MODE_SIZE (GET_MODE (XEXP (op0, 0))) * 2
== GET_MODE_SIZE (mode))
{
case UDIV:
case MOD:
case UMOD:
- if (FLOAT_MODE_P (mode))
+ if (SSE_FLOAT_MODE_P (mode) && TARGET_SSE_MATH)
+ /* ??? SSE cost should be used here. */
+ *total = ix86_cost->fdiv;
+ else if (X87_FLOAT_MODE_P (mode))
+ *total = ix86_cost->fdiv;
+ else if (FLOAT_MODE_P (mode))
+ /* ??? SSE vector cost should be used here. */
*total = ix86_cost->fdiv;
else
*total = ix86_cost->divide[MODE_INDEX (mode)];
return false;
case PLUS:
- if (FLOAT_MODE_P (mode))
- *total = ix86_cost->fadd;
- else if (GET_MODE_CLASS (mode) == MODE_INT
+ if (GET_MODE_CLASS (mode) == MODE_INT
&& GET_MODE_BITSIZE (mode) <= GET_MODE_BITSIZE (Pmode))
{
if (GET_CODE (XEXP (x, 0)) == PLUS
/* FALLTHRU */
case MINUS:
- if (FLOAT_MODE_P (mode))
+ if (SSE_FLOAT_MODE_P (mode) && TARGET_SSE_MATH)
+ {
+ /* ??? SSE cost should be used here. */
+ *total = ix86_cost->fadd;
+ return false;
+ }
+ else if (X87_FLOAT_MODE_P (mode))
+ {
+ *total = ix86_cost->fadd;
+ return false;
+ }
+ else if (FLOAT_MODE_P (mode))
{
+ /* ??? SSE vector cost should be used here. */
*total = ix86_cost->fadd;
return false;
}
/* FALLTHRU */
case NEG:
- if (FLOAT_MODE_P (mode))
+ if (SSE_FLOAT_MODE_P (mode) && TARGET_SSE_MATH)
+ {
+ /* ??? SSE cost should be used here. */
+ *total = ix86_cost->fchs;
+ return false;
+ }
+ else if (X87_FLOAT_MODE_P (mode))
+ {
+ *total = ix86_cost->fchs;
+ return false;
+ }
+ else if (FLOAT_MODE_P (mode))
{
+ /* ??? SSE vector cost should be used here. */
*total = ix86_cost->fchs;
return false;
}
return false;
case FLOAT_EXTEND:
- if (!TARGET_SSE_MATH
- || mode == XFmode
- || (mode == DFmode && !TARGET_SSE2))
+ if (!(SSE_FLOAT_MODE_P (mode) && TARGET_SSE_MATH))
*total = 0;
return false;
case ABS:
- if (FLOAT_MODE_P (mode))
+ if (SSE_FLOAT_MODE_P (mode) && TARGET_SSE_MATH)
+ /* ??? SSE cost should be used here. */
+ *total = ix86_cost->fabs;
+ else if (X87_FLOAT_MODE_P (mode))
+ *total = ix86_cost->fabs;
+ else if (FLOAT_MODE_P (mode))
+ /* ??? SSE vector cost should be used here. */
*total = ix86_cost->fabs;
return false;
case SQRT:
- if (FLOAT_MODE_P (mode))
+ if (SSE_FLOAT_MODE_P (mode) && TARGET_SSE_MATH)
+ /* ??? SSE cost should be used here. */
+ *total = ix86_cost->fsqrt;
+ else if (X87_FLOAT_MODE_P (mode))
+ *total = ix86_cost->fsqrt;
+ else if (FLOAT_MODE_P (mode))
+ /* ??? SSE vector cost should be used here. */
*total = ix86_cost->fsqrt;
return false;
}
static bool
-ix86_ms_bitfield_layout_p (tree record_type)
+ix86_ms_bitfield_layout_p (const_tree record_type)
{
return (TARGET_MS_BITFIELD_LAYOUT &&
!lookup_attribute ("gcc_struct", TYPE_ATTRIBUTES (record_type)))
{
tree type = TREE_TYPE (function);
bool aggr = aggregate_value_p (TREE_TYPE (type), type) != 0;
+ int nregs;
if (TARGET_64BIT)
{
return gen_rtx_REG (DImode, parm_regs[aggr]);
}
- if (ix86_function_regparm (type, function) > 0
- && !type_has_variadic_args_p (type))
+ nregs = ix86_function_regparm (type, function);
+
+ if (nregs > 0 && !stdarg_p (type))
{
- int regno = 0;
+ int regno;
+
if (lookup_attribute ("fastcall", TYPE_ATTRIBUTES (type)))
- regno = 2;
+ regno = aggr ? DX_REG : CX_REG;
+ else
+ {
+ regno = AX_REG;
+ if (aggr)
+ {
+ regno = DX_REG;
+ if (nregs == 1)
+ return gen_rtx_MEM (SImode,
+ plus_constant (stack_pointer_rtx, 4));
+ }
+ }
return gen_rtx_REG (SImode, regno);
}
/* Determine whether x86_output_mi_thunk can succeed. */
static bool
-x86_can_output_mi_thunk (tree thunk ATTRIBUTE_UNUSED,
+x86_can_output_mi_thunk (const_tree thunk ATTRIBUTE_UNUSED,
HOST_WIDE_INT delta ATTRIBUTE_UNUSED,
- HOST_WIDE_INT vcall_offset, tree function)
+ HOST_WIDE_INT vcall_offset, const_tree function)
{
/* 64-bit can handle anything. */
if (TARGET_64BIT)
HOST_WIDE_INT vcall_offset, tree function)
{
rtx xops[3];
- rtx this = x86_this_parameter (function);
+ rtx this_param = x86_this_parameter (function);
rtx this_reg, tmp;
/* If VCALL_OFFSET, we'll need THIS in a register. Might as well
pull it in now and let DELTA benefit. */
- if (REG_P (this))
- this_reg = this;
+ if (REG_P (this_param))
+ this_reg = this_param;
else if (vcall_offset)
{
/* Put the this parameter into %eax. */
- xops[0] = this;
- xops[1] = this_reg = gen_rtx_REG (Pmode, 0);
- output_asm_insn ("mov{l}\t{%0, %1|%1, %0}", xops);
+ xops[0] = this_param;
+ xops[1] = this_reg = gen_rtx_REG (Pmode, AX_REG);
+ if (TARGET_64BIT)
+ output_asm_insn ("mov{q}\t{%0, %1|%1, %0}", xops);
+ else
+ output_asm_insn ("mov{l}\t{%0, %1|%1, %0}", xops);
}
else
this_reg = NULL_RTX;
if (delta)
{
xops[0] = GEN_INT (delta);
- xops[1] = this_reg ? this_reg : this;
+ xops[1] = this_reg ? this_reg : this_param;
if (TARGET_64BIT)
{
if (!x86_64_general_operand (xops[0], DImode))
xops[1] = tmp;
output_asm_insn ("mov{q}\t{%1, %0|%0, %1}", xops);
xops[0] = tmp;
- xops[1] = this;
+ xops[1] = this_param;
}
output_asm_insn ("add{q}\t{%0, %1|%1, %0}", xops);
}
tmp = gen_rtx_REG (DImode, R10_REG);
else
{
- int tmp_regno = 2 /* ECX */;
+ int tmp_regno = CX_REG;
if (lookup_attribute ("fastcall",
TYPE_ATTRIBUTES (TREE_TYPE (function))))
- tmp_regno = 0 /* EAX */;
+ tmp_regno = AX_REG;
tmp = gen_rtx_REG (SImode, tmp_regno);
}
}
/* If necessary, drop THIS back to its stack slot. */
- if (this_reg && this_reg != this)
+ if (this_reg && this_reg != this_param)
{
xops[0] = this_reg;
- xops[1] = this;
- output_asm_insn ("mov{l}\t{%0, %1|%1, %0}", xops);
+ xops[1] = this_param;
+ if (TARGET_64BIT)
+ output_asm_insn ("mov{q}\t{%0, %1|%1, %0}", xops);
+ else
+ output_asm_insn ("mov{l}\t{%0, %1|%1, %0}", xops);
}
xops[0] = XEXP (DECL_RTL (function), 0);
else
#endif /* TARGET_MACHO */
{
- tmp = gen_rtx_REG (SImode, 2 /* ECX */);
+ tmp = gen_rtx_REG (SImode, CX_REG);
output_set_got (tmp, NULL_RTX);
xops[1] = tmp;
if (X86_FILE_START_FLTUSED)
fputs ("\t.global\t__fltused\n", asm_out_file);
if (ix86_asm_dialect == ASM_INTEL)
- fputs ("\t.intel_syntax\n", asm_out_file);
+ fputs ("\t.intel_syntax noprefix\n", asm_out_file);
}
int
enum machine_mode vsimode;
rtx new_target;
rtx x, tmp;
+ bool use_vector_set = false;
+
+ switch (mode)
+ {
+ case V2DImode:
+ use_vector_set = TARGET_64BIT && TARGET_SSE4_1;
+ break;
+ case V16QImode:
+ case V4SImode:
+ case V4SFmode:
+ use_vector_set = TARGET_SSE4_1;
+ break;
+ case V8HImode:
+ use_vector_set = TARGET_SSE2;
+ break;
+ case V4HImode:
+ use_vector_set = TARGET_SSE || TARGET_3DNOW_A;
+ default:
+ break;
+ }
+
+ if (use_vector_set)
+ {
+ emit_insn (gen_rtx_SET (VOIDmode, target, CONST0_RTX (mode)));
+ var = force_reg (GET_MODE_INNER (mode), var);
+ ix86_expand_vector_set (mmx_ok, target, var, one_var);
+ return true;
+ }
switch (mode)
{
/* For V4SF and V4SI, we implement a concat of two V2 vectors.
Recurse to load the two halves. */
- op0 = gen_reg_rtx (half_mode);
- v = gen_rtvec (2, XVECEXP (vals, 0, 0), XVECEXP (vals, 0, 1));
- ix86_expand_vector_init (false, op0, gen_rtx_PARALLEL (half_mode, v));
-
op1 = gen_reg_rtx (half_mode);
v = gen_rtvec (2, XVECEXP (vals, 0, 2), XVECEXP (vals, 0, 3));
ix86_expand_vector_init (false, op1, gen_rtx_PARALLEL (half_mode, v));
+ op0 = gen_reg_rtx (half_mode);
+ v = gen_rtvec (2, XVECEXP (vals, 0, 0), XVECEXP (vals, 0, 1));
+ ix86_expand_vector_init (false, op0, gen_rtx_PARALLEL (half_mode, v));
+
use_vec_concat = true;
}
break;
case V8HImode:
+ if (TARGET_SSE2)
+ {
+ rtx ops[4];
+ unsigned int i, j;
+
+ for (i = 0; i < ARRAY_SIZE (ops); i++)
+ {
+ /* Extend the odd elment from HImode to SImode using
+ a paradoxical SUBREG. */
+ op0 = gen_reg_rtx (SImode);
+ emit_move_insn (op0, gen_lowpart (SImode,
+ XVECEXP (vals, 0,
+ i + i)));
+
+ /* Insert the SImode value as low element of V4SImode
+ vector. */
+ op1 = gen_reg_rtx (V4SImode);
+ op0 = gen_rtx_VEC_MERGE (V4SImode,
+ gen_rtx_VEC_DUPLICATE (V4SImode,
+ op0),
+ CONST0_RTX (V4SImode),
+ const1_rtx);
+ emit_insn (gen_rtx_SET (VOIDmode, op1, op0));
+
+ /* Cast the V4SImode vector back to a V8HImode vector. */
+ op0 = gen_reg_rtx (mode);
+ emit_move_insn (op0, gen_lowpart (mode, op1));
+
+ /* Load even HI elements into the second positon. */
+ emit_insn (gen_vec_setv8hi (op0, XVECEXP (vals, 0,
+ i + i + 1),
+ const1_rtx));
+
+ /* Cast V8HImode vector to V4SImode vector. */
+ ops[i] = gen_reg_rtx (V4SImode);
+ emit_move_insn (ops[i], gen_lowpart (V4SImode, op0));
+ }
+
+ /* Interleave low V4SIs. */
+ for (i = j = 0; i < ARRAY_SIZE (ops); i += 2, j++)
+ {
+ op0 = gen_reg_rtx (V4SImode);
+ emit_insn (gen_vec_interleave_lowv4si (op0, ops[i],
+ ops[i + 1]));
+
+ /* Cast V4SImode vectors to V2DImode vectors. */
+ op1 = gen_reg_rtx (V2DImode);
+ emit_move_insn (op1, gen_lowpart (V2DImode, op0));
+ ops[j] = op1;
+ }
+
+ /* Interleave low V2DIs. */
+ op0 = gen_reg_rtx (V2DImode);
+ emit_insn (gen_vec_interleave_lowv2di (op0, ops[0], ops[1]));
+
+ /* Cast the V2DImode vector back to a V8HImode vector. */
+ emit_insn (gen_rtx_SET (VOIDmode, target,
+ gen_lowpart (mode, op0)));
+ return;
+ }
+
case V16QImode:
+ if (TARGET_SSE4_1)
+ {
+ rtx ops[8];
+ unsigned int i, j;
+
+ for (i = 0; i < ARRAY_SIZE (ops); i++)
+ {
+ /* Extend the odd elment from QImode to SImode using
+ a paradoxical SUBREG. */
+ op0 = gen_reg_rtx (SImode);
+ emit_move_insn (op0, gen_lowpart (SImode,
+ XVECEXP (vals, 0,
+ i + i)));
+
+ /* Insert the SImode value as low element of V4SImode
+ vector. */
+ op1 = gen_reg_rtx (V4SImode);
+ op0 = gen_rtx_VEC_MERGE (V4SImode,
+ gen_rtx_VEC_DUPLICATE (V4SImode,
+ op0),
+ CONST0_RTX (V4SImode),
+ const1_rtx);
+ emit_insn (gen_rtx_SET (VOIDmode, op1, op0));
+
+ /* Cast the V4SImode vector back to a V16QImode vector. */
+ op0 = gen_reg_rtx (mode);
+ emit_move_insn (op0, gen_lowpart (mode, op1));
+
+ /* Load even QI elements into the second positon. */
+ emit_insn (gen_vec_setv16qi (op0, XVECEXP (vals, 0,
+ i + i + 1),
+ const1_rtx));
+
+ /* Cast V16QImode vector to V8HImode vector. */
+ ops[i] = gen_reg_rtx (V8HImode);
+ emit_move_insn (ops[i], gen_lowpart (V8HImode, op0));
+ }
+
+ /* Interleave low V8HIs. */
+ for (i = j = 0; i < ARRAY_SIZE (ops); i += 2, j++)
+ {
+ op0 = gen_reg_rtx (V8HImode);
+ emit_insn (gen_vec_interleave_lowv8hi (op0, ops[i],
+ ops[i + 1]));
+
+ /* Cast V8HImode vector to V4SImode vector. */
+ op1 = gen_reg_rtx (V4SImode);
+ emit_move_insn (op1, gen_lowpart (V4SImode, op0));
+ ops[j] = op1;
+ }
+
+ /* Interleave low V4SIs. */
+ for (i = j = 0; i < ARRAY_SIZE (ops) / 2; i += 2, j++)
+ {
+ op0 = gen_reg_rtx (V4SImode);
+ emit_insn (gen_vec_interleave_lowv4si (op0, ops[i],
+ ops[i + 1]));
+
+ /* Cast V4SImode vectors to V2DImode vectors. */
+ op1 = gen_reg_rtx (V2DImode);
+ emit_move_insn (op1, gen_lowpart (V2DImode, op0));
+ ops[j] = op1;
+ }
+
+ /* Interleave low V2DIs. */
+ op0 = gen_reg_rtx (V2DImode);
+ emit_insn (gen_vec_interleave_lowv2di (op0, ops[0], ops[1]));
+
+ /* Cast the V2DImode vector back to a V8HImode vector. */
+ emit_insn (gen_rtx_SET (VOIDmode, target,
+ gen_lowpart (mode, op0)));
+ return;
+ }
+
case V4HImode:
case V8QImode:
break;
if (use_vec_concat)
{
- if (!register_operand (op0, half_mode))
- op0 = force_reg (half_mode, op0);
if (!register_operand (op1, half_mode))
op1 = force_reg (half_mode, op1);
+ if (!register_operand (op0, half_mode))
+ op0 = force_reg (half_mode, op0);
emit_insn (gen_rtx_SET (VOIDmode, target,
gen_rtx_VEC_CONCAT (mode, op0, op1)));
for (i = 0; i < n_elts; ++i)
{
x = XVECEXP (vals, 0, i);
- if (!CONSTANT_P (x))
+ if (!(CONST_INT_P (x)
+ || GET_CODE (x) == CONST_DOUBLE
+ || GET_CODE (x) == CONST_FIXED))
n_var++, one_var = i;
else if (x != CONST0_RTX (inner_mode))
all_const_zero = false;
}
break;
- case V2DFmode:
case V2DImode:
+ use_vec_merge = TARGET_SSE4_1;
+ if (use_vec_merge)
+ break;
+
+ case V2DFmode:
{
rtx op0, op1;
return;
case V4SFmode:
+ use_vec_merge = TARGET_SSE4_1;
+ if (use_vec_merge)
+ break;
+
switch (elt)
{
case 0:
break;
case V4SImode:
+ use_vec_merge = TARGET_SSE4_1;
+ if (use_vec_merge)
+ break;
+
/* Element 0 handled by vec_merge below. */
if (elt == 0)
{
break;
case V16QImode:
+ use_vec_merge = TARGET_SSE4_1;
+ break;
+
case V8QImode:
default:
break;
break;
case V4SFmode:
+ use_vec_extr = TARGET_SSE4_1;
+ if (use_vec_extr)
+ break;
+
switch (elt)
{
case 0:
break;
case V4SImode:
+ use_vec_extr = TARGET_SSE4_1;
+ if (use_vec_extr)
+ break;
+
if (TARGET_SSE2)
{
switch (elt)
break;
case V16QImode:
+ use_vec_extr = TARGET_SSE4_1;
+ break;
+
case V8QImode:
/* ??? Could extract the appropriate HImode element and shift. */
default:
tmp = gen_rtx_VEC_SELECT (inner_mode, vec, tmp);
/* Let the rtl optimizers know about the zero extension performed. */
- if (inner_mode == HImode)
+ if (inner_mode == QImode || inner_mode == HImode)
{
tmp = gen_rtx_ZERO_EXTEND (SImode, tmp);
target = gen_lowpart (SImode, target);
{
if (DECIMAL_FLOAT_MODE_P (mode))
return true;
+ else if (mode == TFmode)
+ return TARGET_64BIT;
else
return default_scalar_mode_supported_p (mode);
}
return false;
}
+/* Target hook for c_mode_for_suffix. */
+static enum machine_mode
+ix86_c_mode_for_suffix (char suffix)
+{
+ if (TARGET_64BIT && suffix == 'q')
+ return TFmode;
+ if (TARGET_MMX && suffix == 'w')
+ return XFmode;
+
+ return VOIDmode;
+}
+
/* Worker function for TARGET_MD_ASM_CLOBBERS.
We do this in the new i386 backend to maintain source compatibility
return clobbers;
}
-/* Implementes target vector targetm.asm.encode_section_info. This
+/* Implements target vector targetm.asm.encode_section_info. This
is not used by netware. */
static void ATTRIBUTE_UNUSED
emit_label (label2);
}
+/* Output code to perform a Newton-Rhapson approximation of a single precision
+ floating point divide [http://en.wikipedia.org/wiki/N-th_root_algorithm]. */
+
+void ix86_emit_swdivsf (rtx res, rtx a, rtx b, enum machine_mode mode)
+{
+ rtx x0, x1, e0, e1, two;
+
+ x0 = gen_reg_rtx (mode);
+ e0 = gen_reg_rtx (mode);
+ e1 = gen_reg_rtx (mode);
+ x1 = gen_reg_rtx (mode);
+
+ two = CONST_DOUBLE_FROM_REAL_VALUE (dconst2, SFmode);
+
+ if (VECTOR_MODE_P (mode))
+ two = ix86_build_const_vector (SFmode, true, two);
+
+ two = force_reg (mode, two);
+
+ /* a / b = a * rcp(b) * (2.0 - b * rcp(b)) */
+
+ /* x0 = rcp(b) estimate */
+ emit_insn (gen_rtx_SET (VOIDmode, x0,
+ gen_rtx_UNSPEC (mode, gen_rtvec (1, b),
+ UNSPEC_RCP)));
+ /* e0 = x0 * b */
+ emit_insn (gen_rtx_SET (VOIDmode, e0,
+ gen_rtx_MULT (mode, x0, b)));
+ /* e1 = 2. - e0 */
+ emit_insn (gen_rtx_SET (VOIDmode, e1,
+ gen_rtx_MINUS (mode, two, e0)));
+ /* x1 = x0 * e1 */
+ emit_insn (gen_rtx_SET (VOIDmode, x1,
+ gen_rtx_MULT (mode, x0, e1)));
+ /* res = a * x1 */
+ emit_insn (gen_rtx_SET (VOIDmode, res,
+ gen_rtx_MULT (mode, a, x1)));
+}
+
+/* Output code to perform a Newton-Rhapson approximation of a
+ single precision floating point [reciprocal] square root. */
+
+void ix86_emit_swsqrtsf (rtx res, rtx a, enum machine_mode mode,
+ bool recip)
+{
+ rtx x0, e0, e1, e2, e3, mthree, mhalf;
+ REAL_VALUE_TYPE r;
+
+ x0 = gen_reg_rtx (mode);
+ e0 = gen_reg_rtx (mode);
+ e1 = gen_reg_rtx (mode);
+ e2 = gen_reg_rtx (mode);
+ e3 = gen_reg_rtx (mode);
+
+ real_from_integer (&r, VOIDmode, -3, -1, 0);
+ mthree = CONST_DOUBLE_FROM_REAL_VALUE (r, SFmode);
+
+ real_arithmetic (&r, NEGATE_EXPR, &dconsthalf, NULL);
+ mhalf = CONST_DOUBLE_FROM_REAL_VALUE (r, SFmode);
+
+ if (VECTOR_MODE_P (mode))
+ {
+ mthree = ix86_build_const_vector (SFmode, true, mthree);
+ mhalf = ix86_build_const_vector (SFmode, true, mhalf);
+ }
+
+ /* sqrt(a) = -0.5 * a * rsqrtss(a) * (a * rsqrtss(a) * rsqrtss(a) - 3.0)
+ rsqrt(a) = -0.5 * rsqrtss(a) * (a * rsqrtss(a) * rsqrtss(a) - 3.0) */
+
+ /* x0 = rsqrt(a) estimate */
+ emit_insn (gen_rtx_SET (VOIDmode, x0,
+ gen_rtx_UNSPEC (mode, gen_rtvec (1, a),
+ UNSPEC_RSQRT)));
+
+ /* If (a == 0.0) Filter out infinity to prevent NaN for sqrt(0.0). */
+ if (!recip)
+ {
+ rtx zero, mask;
+
+ zero = gen_reg_rtx (mode);
+ mask = gen_reg_rtx (mode);
+
+ zero = force_reg (mode, CONST0_RTX(mode));
+ emit_insn (gen_rtx_SET (VOIDmode, mask,
+ gen_rtx_NE (mode, zero, a)));
+
+ emit_insn (gen_rtx_SET (VOIDmode, x0,
+ gen_rtx_AND (mode, x0, mask)));
+ }
+
+ /* e0 = x0 * a */
+ emit_insn (gen_rtx_SET (VOIDmode, e0,
+ gen_rtx_MULT (mode, x0, a)));
+ /* e1 = e0 * x0 */
+ emit_insn (gen_rtx_SET (VOIDmode, e1,
+ gen_rtx_MULT (mode, e0, x0)));
+
+ /* e2 = e1 - 3. */
+ mthree = force_reg (mode, mthree);
+ emit_insn (gen_rtx_SET (VOIDmode, e2,
+ gen_rtx_PLUS (mode, e1, mthree)));
+
+ mhalf = force_reg (mode, mhalf);
+ if (recip)
+ /* e3 = -.5 * x0 */
+ emit_insn (gen_rtx_SET (VOIDmode, e3,
+ gen_rtx_MULT (mode, x0, mhalf)));
+ else
+ /* e3 = -.5 * e0 */
+ emit_insn (gen_rtx_SET (VOIDmode, e3,
+ gen_rtx_MULT (mode, e0, mhalf)));
+ /* ret = e2 * e3 */
+ emit_insn (gen_rtx_SET (VOIDmode, res,
+ gen_rtx_MULT (mode, e2, e3)));
+}
+
/* Solaris implementation of TARGET_ASM_NAMED_SECTION. */
static void ATTRIBUTE_UNUSED
/* Return the mangling of TYPE if it is an extended fundamental type. */
static const char *
-ix86_mangle_fundamental_type (tree type)
+ix86_mangle_type (const_tree type)
{
+ type = TYPE_MAIN_VARIANT (type);
+
+ if (TREE_CODE (type) != VOID_TYPE && TREE_CODE (type) != BOOLEAN_TYPE
+ && TREE_CODE (type) != INTEGER_TYPE && TREE_CODE (type) != REAL_TYPE)
+ return NULL;
+
switch (TYPE_MODE (type))
{
case TFmode:
/* load nextafter (0.5, 0.0) */
fmt = REAL_MODE_FORMAT (mode);
- real_2expN (&half_minus_pred_half, -(fmt->p) - 1);
+ real_2expN (&half_minus_pred_half, -(fmt->p) - 1, mode);
REAL_ARITHMETIC (pred_half, MINUS_EXPR, dconsthalf, half_minus_pred_half);
/* adj = copysign (0.5, op1) */
/* load nextafter (0.5, 0.0) */
fmt = REAL_MODE_FORMAT (mode);
- real_2expN (&half_minus_pred_half, -(fmt->p) - 1);
+ real_2expN (&half_minus_pred_half, -(fmt->p) - 1, mode);
REAL_ARITHMETIC (pred_half, MINUS_EXPR, dconsthalf, half_minus_pred_half);
/* xa = xa + 0.5 */
}
\f
+/* Validate whether a SSE5 instruction is valid or not.
+ OPERANDS is the array of operands.
+ NUM is the number of operands.
+ USES_OC0 is true if the instruction uses OC0 and provides 4 variants.
+ NUM_MEMORY is the maximum number of memory operands to accept. */
+
+bool
+ix86_sse5_valid_op_p (rtx operands[], rtx insn ATTRIBUTE_UNUSED, int num,
+ bool uses_oc0, int num_memory)
+{
+ int mem_mask;
+ int mem_count;
+ int i;
+
+ /* Count the number of memory arguments */
+ mem_mask = 0;
+ mem_count = 0;
+ for (i = 0; i < num; i++)
+ {
+ enum machine_mode mode = GET_MODE (operands[i]);
+ if (register_operand (operands[i], mode))
+ ;
+
+ else if (memory_operand (operands[i], mode))
+ {
+ mem_mask |= (1 << i);
+ mem_count++;
+ }
+
+ else
+ {
+ rtx pattern = PATTERN (insn);
+
+ /* allow 0 for pcmov */
+ if (GET_CODE (pattern) != SET
+ || GET_CODE (SET_SRC (pattern)) != IF_THEN_ELSE
+ || i < 2
+ || operands[i] != CONST0_RTX (mode))
+ return false;
+ }
+ }
+
+ /* Special case pmacsdq{l,h} where we allow the 3rd argument to be
+ a memory operation. */
+ if (num_memory < 0)
+ {
+ num_memory = -num_memory;
+ if ((mem_mask & (1 << (num-1))) != 0)
+ {
+ mem_mask &= ~(1 << (num-1));
+ mem_count--;
+ }
+ }
+
+ /* If there were no memory operations, allow the insn */
+ if (mem_mask == 0)
+ return true;
+
+ /* Do not allow the destination register to be a memory operand. */
+ else if (mem_mask & (1 << 0))
+ return false;
+
+ /* If there are too many memory operations, disallow the instruction. While
+ the hardware only allows 1 memory reference, before register allocation
+ for some insns, we allow two memory operations sometimes in order to allow
+ code like the following to be optimized:
+
+ float fmadd (float *a, float *b, float *c) { return (*a * *b) + *c; }
+
+ or similar cases that are vectorized into using the fmaddss
+ instruction. */
+ else if (mem_count > num_memory)
+ return false;
+
+ /* Don't allow more than one memory operation if not optimizing. */
+ else if (mem_count > 1 && !optimize)
+ return false;
+
+ else if (num == 4 && mem_count == 1)
+ {
+ /* formats (destination is the first argument), example fmaddss:
+ xmm1, xmm1, xmm2, xmm3/mem
+ xmm1, xmm1, xmm2/mem, xmm3
+ xmm1, xmm2, xmm3/mem, xmm1
+ xmm1, xmm2/mem, xmm3, xmm1 */
+ if (uses_oc0)
+ return ((mem_mask == (1 << 1))
+ || (mem_mask == (1 << 2))
+ || (mem_mask == (1 << 3)));
+
+ /* format, example pmacsdd:
+ xmm1, xmm2, xmm3/mem, xmm1 */
+ else
+ return (mem_mask == (1 << 2));
+ }
+
+ else if (num == 4 && num_memory == 2)
+ {
+ /* If there are two memory operations, we can load one of the memory ops
+ into the destination register. This is for optimizing the
+ multiply/add ops, which the combiner has optimized both the multiply
+ and the add insns to have a memory operation. We have to be careful
+ that the destination doesn't overlap with the inputs. */
+ rtx op0 = operands[0];
+
+ if (reg_mentioned_p (op0, operands[1])
+ || reg_mentioned_p (op0, operands[2])
+ || reg_mentioned_p (op0, operands[3]))
+ return false;
+
+ /* formats (destination is the first argument), example fmaddss:
+ xmm1, xmm1, xmm2, xmm3/mem
+ xmm1, xmm1, xmm2/mem, xmm3
+ xmm1, xmm2, xmm3/mem, xmm1
+ xmm1, xmm2/mem, xmm3, xmm1
+
+ For the oc0 case, we will load either operands[1] or operands[3] into
+ operands[0], so any combination of 2 memory operands is ok. */
+ if (uses_oc0)
+ return true;
+
+ /* format, example pmacsdd:
+ xmm1, xmm2, xmm3/mem, xmm1
+
+ For the integer multiply/add instructions be more restrictive and
+ require operands[2] and operands[3] to be the memory operands. */
+ else
+ return (mem_mask == ((1 << 2) | (1 << 3)));
+ }
+
+ else if (num == 3 && num_memory == 1)
+ {
+ /* formats, example protb:
+ xmm1, xmm2, xmm3/mem
+ xmm1, xmm2/mem, xmm3 */
+ if (uses_oc0)
+ return ((mem_mask == (1 << 1)) || (mem_mask == (1 << 2)));
+
+ /* format, example comeq:
+ xmm1, xmm2, xmm3/mem */
+ else
+ return (mem_mask == (1 << 2));
+ }
+
+ else
+ gcc_unreachable ();
+
+ return false;
+}
+
+\f
+/* Fixup an SSE5 instruction that has 2 memory input references into a form the
+ hardware will allow by using the destination register to load one of the
+ memory operations. Presently this is used by the multiply/add routines to
+ allow 2 memory references. */
+
+void
+ix86_expand_sse5_multiple_memory (rtx operands[],
+ int num,
+ enum machine_mode mode)
+{
+ rtx op0 = operands[0];
+ if (num != 4
+ || memory_operand (op0, mode)
+ || reg_mentioned_p (op0, operands[1])
+ || reg_mentioned_p (op0, operands[2])
+ || reg_mentioned_p (op0, operands[3]))
+ gcc_unreachable ();
+
+ /* For 2 memory operands, pick either operands[1] or operands[3] to move into
+ the destination register. */
+ if (memory_operand (operands[1], mode))
+ {
+ emit_move_insn (op0, operands[1]);
+ operands[1] = op0;
+ }
+ else if (memory_operand (operands[3], mode))
+ {
+ emit_move_insn (op0, operands[3]);
+ operands[3] = op0;
+ }
+ else
+ gcc_unreachable ();
+
+ return;
+}
+
+\f
/* Table of valid machine attributes. */
static const struct attribute_spec ix86_attribute_table[] =
{
{ NULL, 0, 0, false, false, false, NULL }
};
+/* Implement targetm.vectorize.builtin_vectorization_cost. */
+static int
+x86_builtin_vectorization_cost (bool runtime_test)
+{
+ /* If the branch of the runtime test is taken - i.e. - the vectorized
+ version is skipped - this incurs a misprediction cost (because the
+ vectorized version is expected to be the fall-through). So we subtract
+ the latency of a mispredicted branch from the costs that are incured
+ when the vectorized version is executed.
+
+ TODO: The values in individual target tables have to be tuned or new
+ fields may be needed. For eg. on K8, the default branch path is the
+ not-taken path. If the taken path is predicted correctly, the minimum
+ penalty of going down the taken-path is 1 cycle. If the taken-path is
+ not predicted correctly, then the minimum penalty is 10 cycles. */
+
+ if (runtime_test)
+ {
+ return (-(ix86_cost->cond_taken_branch_cost));
+ }
+ else
+ return 0;
+}
+
/* Initialize the GCC target structure. */
#undef TARGET_ATTRIBUTE_TABLE
#define TARGET_ATTRIBUTE_TABLE ix86_attribute_table
#define TARGET_EXPAND_BUILTIN ix86_expand_builtin
#undef TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION
-#define TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION ix86_builtin_vectorized_function
+#define TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION \
+ ix86_builtin_vectorized_function
+
#undef TARGET_VECTORIZE_BUILTIN_CONVERSION
-#define TARGET_VECTORIZE_BUILTIN_CONVERSION ix86_builtin_conversion
+#define TARGET_VECTORIZE_BUILTIN_CONVERSION ix86_vectorize_builtin_conversion
+
+#undef TARGET_BUILTIN_RECIPROCAL
+#define TARGET_BUILTIN_RECIPROCAL ix86_builtin_reciprocal
#undef TARGET_ASM_FUNCTION_EPILOGUE
#define TARGET_ASM_FUNCTION_EPILOGUE ix86_output_function_epilogue
#undef TARGET_CANNOT_FORCE_CONST_MEM
#define TARGET_CANNOT_FORCE_CONST_MEM ix86_cannot_force_const_mem
#undef TARGET_USE_BLOCKS_FOR_CONSTANT_P
-#define TARGET_USE_BLOCKS_FOR_CONSTANT_P hook_bool_mode_rtx_true
+#define TARGET_USE_BLOCKS_FOR_CONSTANT_P hook_bool_mode_const_rtx_true
#undef TARGET_DELEGITIMIZE_ADDRESS
#define TARGET_DELEGITIMIZE_ADDRESS ix86_delegitimize_address
#undef TARGET_DEFAULT_TARGET_FLAGS
#define TARGET_DEFAULT_TARGET_FLAGS \
(TARGET_DEFAULT \
- | TARGET_64BIT_DEFAULT \
| TARGET_SUBTARGET_DEFAULT \
| TARGET_TLS_DIRECT_SEG_REFS_DEFAULT)
#undef TARGET_BUILD_BUILTIN_VA_LIST
#define TARGET_BUILD_BUILTIN_VA_LIST ix86_build_builtin_va_list
+#undef TARGET_EXPAND_BUILTIN_VA_START
+#define TARGET_EXPAND_BUILTIN_VA_START ix86_va_start
+
#undef TARGET_MD_ASM_CLOBBERS
#define TARGET_MD_ASM_CLOBBERS ix86_md_asm_clobbers
#undef TARGET_PROMOTE_PROTOTYPES
-#define TARGET_PROMOTE_PROTOTYPES hook_bool_tree_true
+#define TARGET_PROMOTE_PROTOTYPES hook_bool_const_tree_true
#undef TARGET_STRUCT_VALUE_RTX
#define TARGET_STRUCT_VALUE_RTX ix86_struct_value_rtx
#undef TARGET_SETUP_INCOMING_VARARGS
#undef TARGET_VECTOR_MODE_SUPPORTED_P
#define TARGET_VECTOR_MODE_SUPPORTED_P ix86_vector_mode_supported_p
+#undef TARGET_C_MODE_FOR_SUFFIX
+#define TARGET_C_MODE_FOR_SUFFIX ix86_c_mode_for_suffix
+
#ifdef HAVE_AS_TLS
#undef TARGET_ASM_OUTPUT_DWARF_DTPREL
#define TARGET_ASM_OUTPUT_DWARF_DTPREL i386_output_dwarf_dtprel
#define TARGET_INSERT_ATTRIBUTES SUBTARGET_INSERT_ATTRIBUTES
#endif
-#undef TARGET_MANGLE_FUNDAMENTAL_TYPE
-#define TARGET_MANGLE_FUNDAMENTAL_TYPE ix86_mangle_fundamental_type
+#undef TARGET_MANGLE_TYPE
+#define TARGET_MANGLE_TYPE ix86_mangle_type
#undef TARGET_STACK_PROTECT_FAIL
#define TARGET_STACK_PROTECT_FAIL ix86_stack_protect_fail
#undef TARGET_FUNCTION_VALUE
#define TARGET_FUNCTION_VALUE ix86_function_value
+#undef TARGET_SECONDARY_RELOAD
+#define TARGET_SECONDARY_RELOAD ix86_secondary_reload
+
+#undef TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST
+#define TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST x86_builtin_vectorization_cost
+
struct gcc_target targetm = TARGET_INITIALIZER;
\f
#include "gt-i386.h"
OSDN Git Service
