/* 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.
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. */
2, /* vec_align_load_cost. */
3, /* vec_unalign_load_cost. */
3, /* vec_store_cost. */
- 6, /* cond_taken_branch_cost. */
- 1, /* cond_not_taken_branch_cost. */
+ 3, /* cond_taken_branch_cost. */
+ 2, /* cond_not_taken_branch_cost. */
};
struct processor_costs amdfam10_cost = {
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. */
2, /* vec_align_load_cost. */
2, /* vec_unalign_load_cost. */
2, /* vec_store_cost. */
- 6, /* cond_taken_branch_cost. */
+ 2, /* cond_taken_branch_cost. */
1, /* cond_not_taken_branch_cost. */
};
was set or cleared on the command line. */
static int ix86_isa_flags_explicit;
-/* Define a set of ISAs which aren't available for a given ISA. MMX
- and SSE ISAs are handled separately. */
+/* 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_3DNOW | OPTION_MASK_ISA_3DNOW_UNSET)
-#define OPTION_MASK_ISA_3DNOW_UNSET OPTION_MASK_ISA_3DNOW_A
+ (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_SSE2 | OPTION_MASK_ISA_SSE2_UNSET)
+ (OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_SSE2_UNSET)
#define OPTION_MASK_ISA_SSE2_UNSET \
- (OPTION_MASK_ISA_SSE3 | OPTION_MASK_ISA_SSE3_UNSET)
+ (OPTION_MASK_ISA_SSE2 | OPTION_MASK_ISA_SSE3_UNSET)
#define OPTION_MASK_ISA_SSE3_UNSET \
- (OPTION_MASK_ISA_SSSE3 | OPTION_MASK_ISA_SSSE3_UNSET)
+ (OPTION_MASK_ISA_SSE3 \
+ | OPTION_MASK_ISA_SSSE3_UNSET \
+ | OPTION_MASK_ISA_SSE4A_UNSET )
#define OPTION_MASK_ISA_SSSE3_UNSET \
- (OPTION_MASK_ISA_SSE4_1 | OPTION_MASK_ISA_SSE4_1_UNSET)
+ (OPTION_MASK_ISA_SSSE3 | OPTION_MASK_ISA_SSE4_1_UNSET)
#define OPTION_MASK_ISA_SSE4_1_UNSET \
- (OPTION_MASK_ISA_SSE4_2 | OPTION_MASK_ISA_SSE4_2_UNSET)
-#define OPTION_MASK_ISA_SSE4_2_UNSET OPTION_MASK_ISA_SSE4A
+ (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. -msse4 should be the same
- as -msse4.1 -msse4.2. -mno-sse4 should the same as -mno-sse4.1. */
-#define OPTION_MASK_ISA_SSE4 \
- (OPTION_MASK_ISA_SSE4_1 | 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_SSE4
+#define OPTION_MASK_ISA_SSE4A_UNSET \
+ (OPTION_MASK_ISA_SSE4A | OPTION_MASK_ISA_SSE5_UNSET)
-#define OPTION_MASK_ISA_SSE5_UNSET \
- (OPTION_MASK_ISA_3DNOW | OPTION_MASK_ISA_3DNOW_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;
switch (code)
{
case OPT_mmmx:
- ix86_isa_flags_explicit |= OPTION_MASK_ISA_MMX;
- if (!value)
+ if (value)
+ {
+ ix86_isa_flags |= OPTION_MASK_ISA_MMX_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_MMX_SET;
+ }
+ else
{
ix86_isa_flags &= ~OPTION_MASK_ISA_MMX_UNSET;
ix86_isa_flags_explicit |= OPTION_MASK_ISA_MMX_UNSET;
return true;
case OPT_m3dnow:
- ix86_isa_flags_explicit |= OPTION_MASK_ISA_3DNOW;
- if (!value)
+ if (value)
+ {
+ ix86_isa_flags |= OPTION_MASK_ISA_3DNOW_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_3DNOW_SET;
+ }
+ else
{
ix86_isa_flags &= ~OPTION_MASK_ISA_3DNOW_UNSET;
ix86_isa_flags_explicit |= OPTION_MASK_ISA_3DNOW_UNSET;
return false;
case OPT_msse:
- ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE;
- if (!value)
+ if (value)
+ {
+ 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:
- ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE2;
- if (!value)
+ if (value)
+ {
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE2_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE2_SET;
+ }
+ else
{
ix86_isa_flags &= ~OPTION_MASK_ISA_SSE2_UNSET;
ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE2_UNSET;
return true;
case OPT_msse3:
- ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE3;
- if (!value)
+ if (value)
+ {
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE3_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE3_SET;
+ }
+ else
{
ix86_isa_flags &= ~OPTION_MASK_ISA_SSE3_UNSET;
ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE3_UNSET;
return true;
case OPT_mssse3:
- ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSSE3;
- 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:
- ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE4_1;
- if (!value)
+ 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:
- ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE4_2;
- if (!value)
+ 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;
- ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE4;
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE4_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE4_SET;
return true;
case OPT_mno_sse4:
return true;
case OPT_msse4a:
- ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE4A;
- if (!value)
+ 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:
- ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE5;
- if (!value)
+ if (value)
+ {
+ ix86_isa_flags |= OPTION_MASK_ISA_SSE5_SET;
+ ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE5_SET;
+ }
+ else
{
ix86_isa_flags &= ~OPTION_MASK_ISA_SSE5_UNSET;
ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE5_UNSET;
if (!TARGET_80387)
target_flags |= MASK_NO_FANCY_MATH_387;
- /* Turn on SSE4A bultins for -msse5. */
- if (TARGET_SSE5)
- ix86_isa_flags |= OPTION_MASK_ISA_SSE4A;
-
- /* Turn on SSE4.1 builtins for -msse4.2. */
- if (TARGET_SSE4_2)
- ix86_isa_flags |= OPTION_MASK_ISA_SSE4_1;
-
- /* Turn on SSSE3 builtins for -msse4.1. */
- if (TARGET_SSE4_1)
- ix86_isa_flags |= OPTION_MASK_ISA_SSSE3;
-
- /* Turn on SSE3 builtins for -mssse3. */
- if (TARGET_SSSE3)
- ix86_isa_flags |= OPTION_MASK_ISA_SSE3;
-
- /* Turn on SSE3 builtins for -msse4a. */
- if (TARGET_SSE4A)
- ix86_isa_flags |= OPTION_MASK_ISA_SSE3;
-
- /* Turn on SSE2 builtins for -msse3. */
- if (TARGET_SSE3)
- ix86_isa_flags |= OPTION_MASK_ISA_SSE2;
-
- /* Turn on SSE builtins for -msse2. */
- if (TARGET_SSE2)
- ix86_isa_flags |= OPTION_MASK_ISA_SSE;
-
/* Turn on MMX builtins for -msse. */
if (TARGET_SSE)
{
x86_prefetch_sse = true;
}
- /* Turn on MMX builtins for 3Dnow. */
- if (TARGET_3DNOW)
- ix86_isa_flags |= OPTION_MASK_ISA_MMX;
-
/* Turn on popcnt instruction for -msse4.2 or -mabm. */
if (TARGET_SSE4_2 || TARGET_ABM)
x86_popcnt = true;
indirectly or considering a libcall. Otherwise return 0. */
static int
-ix86_function_sseregparm (const_tree type, const_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;
}
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;
+
+ /* 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);
/* 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);
}
}
SSE math is enabled or for functions with sseregparm attribute. */
if ((fn || fntype) && (mode == SFmode || mode == DFmode))
{
- int sse_level = ix86_function_sseregparm (fntype, fn);
+ int sse_level = ix86_function_sseregparm (fntype, fn, false);
if ((sse_level >= 1 && mode == SFmode)
|| (sse_level == 2 && mode == DFmode))
regno = FIRST_SSE_REG;
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);
}
}
- 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);
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;
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. */
/* 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;
+ }
+ 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;
+ }
default:
break;
}
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,
{ OPTION_MASK_ISA_SSE, CODE_FOR_addv4sf3, "__builtin_ia32_addps", IX86_BUILTIN_ADDPS, UNKNOWN, 0 },
{ OPTION_MASK_ISA_SSE, CODE_FOR_subv4sf3, "__builtin_ia32_subps", IX86_BUILTIN_SUBPS, UNKNOWN, 0 },
{ OPTION_MASK_ISA_SSE, CODE_FOR_mulv4sf3, "__builtin_ia32_mulps", IX86_BUILTIN_MULPS, UNKNOWN, 0 },
- { OPTION_MASK_ISA_SSE, CODE_FOR_divv4sf3, "__builtin_ia32_divps", IX86_BUILTIN_DIVPS, UNKNOWN, 0 },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_divv4sf3, "__builtin_ia32_divps", IX86_BUILTIN_DIVPS, UNKNOWN, 0 },
{ OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmaddv4sf3, "__builtin_ia32_addss", IX86_BUILTIN_ADDSS, UNKNOWN, 0 },
{ OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmsubv4sf3, "__builtin_ia32_subss", IX86_BUILTIN_SUBSS, UNKNOWN, 0 },
{ OPTION_MASK_ISA_SSE, CODE_FOR_sse_vmmulv4sf3, "__builtin_ia32_mulss", IX86_BUILTIN_MULSS, UNKNOWN, 0 },
{ OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_3DNOW_A, CODE_FOR_mmx_pmovmskb, 0, IX86_BUILTIN_PMOVMSKB, UNKNOWN, 0 },
{ OPTION_MASK_ISA_SSE, CODE_FOR_sse_movmskps, 0, IX86_BUILTIN_MOVMSKPS, UNKNOWN, 0 },
- { OPTION_MASK_ISA_SSE, CODE_FOR_sqrtv4sf2, 0, IX86_BUILTIN_SQRTPS, UNKNOWN, 0 },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sse_sqrtv4sf2, 0, IX86_BUILTIN_SQRTPS, UNKNOWN, 0 },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_sqrtv4sf2, 0, IX86_BUILTIN_SQRTPS_NR, UNKNOWN, 0 },
{ OPTION_MASK_ISA_SSE, CODE_FOR_sse_rsqrtv4sf2, 0, IX86_BUILTIN_RSQRTPS, UNKNOWN, 0 },
+ { OPTION_MASK_ISA_SSE, CODE_FOR_rsqrtv4sf2, 0, IX86_BUILTIN_RSQRTPS_NR, UNKNOWN, 0 },
{ OPTION_MASK_ISA_SSE, CODE_FOR_sse_rcpv4sf2, 0, IX86_BUILTIN_RCPPS, UNKNOWN, 0 },
{ OPTION_MASK_ISA_SSE, CODE_FOR_sse_cvtps2pi, 0, IX86_BUILTIN_CVTPS2PI, UNKNOWN, 0 },
def_builtin_const (OPTION_MASK_ISA_SSE, "__builtin_ia32_rcpps", v4sf_ftype_v4sf, IX86_BUILTIN_RCPPS);
def_builtin_const (OPTION_MASK_ISA_SSE, "__builtin_ia32_rcpss", v4sf_ftype_v4sf, IX86_BUILTIN_RCPSS);
def_builtin_const (OPTION_MASK_ISA_SSE, "__builtin_ia32_rsqrtps", v4sf_ftype_v4sf, IX86_BUILTIN_RSQRTPS);
+ def_builtin_const (OPTION_MASK_ISA_SSE, "__builtin_ia32_rsqrtps_nr", v4sf_ftype_v4sf, IX86_BUILTIN_RSQRTPS_NR);
def_builtin_const (OPTION_MASK_ISA_SSE, "__builtin_ia32_rsqrtss", v4sf_ftype_v4sf, IX86_BUILTIN_RSQRTSS);
ftype = build_function_type_list (float_type_node,
float_type_node,
NULL_TREE);
def_builtin_const (OPTION_MASK_ISA_SSE, "__builtin_ia32_rsqrtf", ftype, IX86_BUILTIN_RSQRTF);
def_builtin_const (OPTION_MASK_ISA_SSE, "__builtin_ia32_sqrtps", v4sf_ftype_v4sf, IX86_BUILTIN_SQRTPS);
+ def_builtin_const (OPTION_MASK_ISA_SSE, "__builtin_ia32_sqrtps_nr", v4sf_ftype_v4sf, IX86_BUILTIN_SQRTPS_NR);
def_builtin_const (OPTION_MASK_ISA_SSE, "__builtin_ia32_sqrtss", v4sf_ftype_v4sf, IX86_BUILTIN_SQRTSS);
def_builtin_const (OPTION_MASK_ISA_SSE, "__builtin_ia32_shufps", v4sf_ftype_v4sf_v4sf_int, IX86_BUILTIN_SHUFPS);
case BUILT_IN_SQRTF:
if (out_mode == SFmode && out_n == 4
&& in_mode == SFmode && in_n == 4)
- return ix86_builtins[IX86_BUILTIN_SQRTPS];
+ return ix86_builtins[IX86_BUILTIN_SQRTPS_NR];
break;
case BUILT_IN_LRINT:
switch (fn)
{
/* Vectorized version of sqrt to rsqrt conversion. */
- case IX86_BUILTIN_SQRTPS:
- return ix86_builtins[IX86_BUILTIN_RSQRTPS];
+ case IX86_BUILTIN_SQRTPS_NR:
+ return ix86_builtins[IX86_BUILTIN_RSQRTPS_NR];
default:
return NULL_TREE;
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 (ix86_cost->mmxsse_to_integer, 8);
+ return MAX (8, ix86_cost->mmxsse_to_integer);
if (MAYBE_FLOAT_CLASS_P (class1))
return ix86_cost->fp_move;
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))
{
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;
/* a / b = a * rcp(b) * (2.0 - b * rcp(b)) */
- /* x0 = 1./b estimate */
+ /* x0 = rcp(b) estimate */
emit_insn (gen_rtx_SET (VOIDmode, x0,
gen_rtx_UNSPEC (mode, gen_rtvec (1, b),
UNSPEC_RCP)));
void ix86_emit_swsqrtsf (rtx res, rtx a, enum machine_mode mode,
bool recip)
{
- rtx x0, e0, e1, e2, e3, three, half, zero, mask;
+ rtx x0, e0, e1, e2, e3, mthree, mhalf;
+ REAL_VALUE_TYPE r;
x0 = gen_reg_rtx (mode);
e0 = gen_reg_rtx (mode);
e2 = gen_reg_rtx (mode);
e3 = gen_reg_rtx (mode);
- three = CONST_DOUBLE_FROM_REAL_VALUE (dconst3, SFmode);
- half = CONST_DOUBLE_FROM_REAL_VALUE (dconsthalf, SFmode);
+ real_arithmetic (&r, NEGATE_EXPR, &dconst3, NULL);
+ mthree = CONST_DOUBLE_FROM_REAL_VALUE (r, SFmode);
- mask = gen_reg_rtx (mode);
+ real_arithmetic (&r, NEGATE_EXPR, &dconsthalf, NULL);
+ mhalf = CONST_DOUBLE_FROM_REAL_VALUE (r, SFmode);
if (VECTOR_MODE_P (mode))
{
- three = ix86_build_const_vector (SFmode, true, three);
- half = ix86_build_const_vector (SFmode, true, half);
+ mthree = ix86_build_const_vector (SFmode, true, mthree);
+ mhalf = ix86_build_const_vector (SFmode, true, mhalf);
}
- three = force_reg (mode, three);
- half = force_reg (mode, half);
-
- zero = force_reg (mode, CONST0_RTX(mode));
-
- /* sqrt(a) = 0.5 * a * rsqrtss(a) * (3.0 - a * rsqrtss(a) * rsqrtss(a))
- 1.0 / sqrt(a) = 0.5 * rsqrtss(a) * (3.0 - a * rsqrtss(a) * rsqrtss(a)) */
-
- /* Compare a to zero. */
- emit_insn (gen_rtx_SET (VOIDmode, mask,
- gen_rtx_NE (mode, a, zero)));
+ /* 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 = 1./sqrt(a) estimate */
+ /* x0 = rsqrt(a) estimate */
emit_insn (gen_rtx_SET (VOIDmode, x0,
gen_rtx_UNSPEC (mode, gen_rtvec (1, a),
UNSPEC_RSQRT)));
- /* Filter out infinity. */
- if (VECTOR_MODE_P (mode))
- emit_insn (gen_rtx_SET (VOIDmode, gen_lowpart (V4SFmode, x0),
- gen_rtx_AND (mode,
- gen_lowpart (V4SFmode, x0),
- gen_lowpart (V4SFmode, mask))));
- else
- emit_insn (gen_rtx_SET (VOIDmode, x0,
- gen_rtx_AND (mode, x0, mask)));
+
+ /* 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,
/* e1 = e0 * x0 */
emit_insn (gen_rtx_SET (VOIDmode, e1,
gen_rtx_MULT (mode, e0, x0)));
- /* e2 = 3. - e1 */
+
+ /* e2 = e1 - 3. */
+ mthree = force_reg (mode, mthree);
emit_insn (gen_rtx_SET (VOIDmode, e2,
- gen_rtx_MINUS (mode, three, e1)));
+ gen_rtx_PLUS (mode, e1, mthree)));
+
+ mhalf = force_reg (mode, mhalf);
if (recip)
- /* e3 = .5 * x0 */
+ /* e3 = -.5 * x0 */
emit_insn (gen_rtx_SET (VOIDmode, e3,
- gen_rtx_MULT (mode, half, x0)));
+ gen_rtx_MULT (mode, x0, mhalf)));
else
- /* e3 = .5 * e0 */
+ /* e3 = -.5 * e0 */
emit_insn (gen_rtx_SET (VOIDmode, e3,
- gen_rtx_MULT (mode, half, e0)));
+ gen_rtx_MULT (mode, e0, mhalf)));
/* ret = e2 * e3 */
emit_insn (gen_rtx_SET (VOIDmode, res,
gen_rtx_MULT (mode, e2, e3)));
OSDN Git Service
