#include "tm-constrs.h"
#include "params.h"
#include "cselib.h"
+#include "debug.h"
+#include "dwarf2out.h"
static rtx legitimize_dllimport_symbol (rtx, bool);
static enum calling_abi ix86_function_abi (const_tree);
\f
+#ifndef SUBTARGET32_DEFAULT_CPU
+#define SUBTARGET32_DEFAULT_CPU "i386"
+#endif
+
/* The svr4 ABI for the i386 says that records and unions are returned
in memory. */
#ifndef DEFAULT_PCC_STRUCT_RETURN
}
}
\f
-/* Return a string the documents the current -m options. The caller is
+/* Return a string that documents the current -m options. The caller is
responsible for freeing the string. */
static char *
{
{ "-m64", OPTION_MASK_ISA_64BIT },
{ "-mfma4", OPTION_MASK_ISA_FMA4 },
+ { "-mfma", OPTION_MASK_ISA_FMA },
{ "-mxop", OPTION_MASK_ISA_XOP },
{ "-mlwp", OPTION_MASK_ISA_LWP },
{ "-msse4a", OPTION_MASK_ISA_SSE4A },
{
int i;
unsigned int ix86_arch_mask, ix86_tune_mask;
+ const bool ix86_tune_specified = (ix86_tune_string != NULL);
const char *prefix;
const char *suffix;
const char *sw;
|| !strcmp (ix86_tune_string, "generic64")))
;
else if (!strncmp (ix86_tune_string, "generic", 7))
- error ("bad value (%s) for %stune=%s %s",
+ error ("bad value (%s) for %stune=%s %s",
ix86_tune_string, prefix, suffix, sw);
+ else if (!strcmp (ix86_tune_string, "x86-64"))
+ warning (OPT_Wdeprecated, "%stune=x86-64%s is deprecated. Use "
+ "%stune=k8%s or %stune=generic%s instead as appropriate.",
+ prefix, suffix, prefix, suffix, prefix, suffix);
}
else
{
ix86_tune_string = "generic32";
}
}
+
if (ix86_stringop_string)
{
if (!strcmp (ix86_stringop_string, "rep_byte"))
error ("bad value (%s) for %sstringop-strategy=%s %s",
ix86_stringop_string, prefix, suffix, sw);
}
- if (!strcmp (ix86_tune_string, "x86-64"))
- warning (OPT_Wdeprecated, "%stune=x86-64%s is deprecated. Use "
- "%stune=k8%s or %stune=generic%s instead as appropriate.",
- prefix, suffix, prefix, suffix, prefix, suffix);
if (!ix86_arch_string)
- ix86_arch_string = TARGET_64BIT ? "x86-64" : "i386";
+ ix86_arch_string = TARGET_64BIT ? "x86-64" : SUBTARGET32_DEFAULT_CPU;
else
ix86_arch_specified = 1;
- if (!strcmp (ix86_arch_string, "generic"))
- error ("generic CPU can be used only for %stune=%s %s",
- prefix, suffix, sw);
- if (!strncmp (ix86_arch_string, "generic", 7))
- error ("bad value (%s) for %sarch=%s %s",
- ix86_arch_string, prefix, suffix, sw);
-
/* Validate -mabi= value. */
if (ix86_abi_string)
{
break;
}
- if (i == pta_size)
+ if (!strcmp (ix86_arch_string, "generic"))
+ error ("generic CPU can be used only for %stune=%s %s",
+ prefix, suffix, sw);
+ else if (!strncmp (ix86_arch_string, "generic", 7) || i == pta_size)
error ("bad value (%s) for %sarch=%s %s",
ix86_arch_string, prefix, suffix, sw);
x86_prefetch_sse = true;
break;
}
- if (i == pta_size)
+
+ if (ix86_tune_specified && i == pta_size)
error ("bad value (%s) for %stune=%s %s",
ix86_tune_string, prefix, suffix, sw);
ix86_tls_dialect = TLS_DIALECT_GNU;
else if (strcmp (ix86_tls_dialect_string, "gnu2") == 0)
ix86_tls_dialect = TLS_DIALECT_GNU2;
- else if (strcmp (ix86_tls_dialect_string, "sun") == 0)
- ix86_tls_dialect = TLS_DIALECT_SUN;
else
error ("bad value (%s) for %stls-dialect=%s %s",
ix86_tls_dialect_string, prefix, suffix, sw);
if (TARGET_64BIT)
{
/* Do not warn when emulating the MS ABI. */
- if (TREE_CODE (*node) != FUNCTION_TYPE
+ if ((TREE_CODE (*node) != FUNCTION_TYPE
+ && TREE_CODE (*node) != METHOD_TYPE)
|| ix86_function_type_abi (*node) != MS_ABI)
warning (OPT_Wattributes, "%qE attribute ignored",
name);
/* This function generates code for -fpic that loads %ebx with
the return address of the caller and then returns. */
-void
-ix86_file_end (void)
+static void
+ix86_code_end (void)
{
rtx xops[2];
int regno;
for (regno = 0; regno < 8; ++regno)
{
char name[32];
+ tree decl;
if (! ((pic_labels_used >> regno) & 1))
continue;
get_pc_thunk_name (name, regno);
+ decl = build_decl (BUILTINS_LOCATION, FUNCTION_DECL,
+ get_identifier (name),
+ build_function_type (void_type_node, void_list_node));
+ DECL_RESULT (decl) = build_decl (BUILTINS_LOCATION, RESULT_DECL,
+ NULL_TREE, void_type_node);
+ TREE_PUBLIC (decl) = 1;
+ TREE_STATIC (decl) = 1;
+
#if TARGET_MACHO
if (TARGET_MACHO)
{
assemble_name (asm_out_file, name);
fputs ("\n", asm_out_file);
ASM_OUTPUT_LABEL (asm_out_file, name);
+ DECL_WEAK (decl) = 1;
}
else
#endif
if (USE_HIDDEN_LINKONCE)
{
- tree decl;
-
- decl = build_decl (BUILTINS_LOCATION,
- FUNCTION_DECL, get_identifier (name),
- error_mark_node);
- TREE_PUBLIC (decl) = 1;
- TREE_STATIC (decl) = 1;
DECL_COMDAT_GROUP (decl) = DECL_ASSEMBLER_NAME (decl);
(*targetm.asm_out.unique_section) (decl, 0);
ASM_OUTPUT_LABEL (asm_out_file, name);
}
+ DECL_INITIAL (decl) = make_node (BLOCK);
+ current_function_decl = decl;
+ init_function_start (decl);
+ first_function_block_is_cold = false;
+ /* Make sure unwind info is emitted for the thunk if needed. */
+ final_start_function (emit_barrier (), asm_out_file, 1);
+
xops[0] = gen_rtx_REG (Pmode, regno);
xops[1] = gen_rtx_MEM (Pmode, stack_pointer_rtx);
output_asm_insn ("mov%z0\t{%1, %0|%0, %1}", xops);
output_asm_insn ("ret", xops);
+ final_end_function ();
+ init_insn_lengths ();
+ free_after_compilation (cfun);
+ set_cfun (NULL);
+ current_function_decl = NULL;
}
-
- if (NEED_INDICATE_EXEC_STACK)
- file_end_indicate_exec_stack ();
}
/* Emit code for the SET_GOT patterns. */
if (!flag_pic)
output_asm_insn ("mov%z0\t{%2, %0|%0, %2}", xops);
else
- output_asm_insn ("call\t%a2", xops);
+ {
+ output_asm_insn ("call\t%a2", xops);
+#ifdef DWARF2_UNWIND_INFO
+ /* The call to next label acts as a push. */
+ if (dwarf2out_do_frame ())
+ {
+ rtx insn;
+ start_sequence ();
+ insn = emit_insn (gen_rtx_SET (VOIDmode, stack_pointer_rtx,
+ gen_rtx_PLUS (Pmode,
+ stack_pointer_rtx,
+ GEN_INT (-4))));
+ RTX_FRAME_RELATED_P (insn) = 1;
+ dwarf2out_frame_debug (insn, true);
+ end_sequence ();
+ }
+#endif
+ }
#if TARGET_MACHO
/* Output the Mach-O "canonical" label name ("Lxx$pb") here too. This
CODE_LABEL_NUMBER (XEXP (xops[2], 0)));
if (flag_pic)
- output_asm_insn ("pop%z0\t%0", xops);
+ {
+ output_asm_insn ("pop%z0\t%0", xops);
+#ifdef DWARF2_UNWIND_INFO
+ /* The pop is a pop and clobbers dest, but doesn't restore it
+ for unwind info purposes. */
+ if (dwarf2out_do_frame ())
+ {
+ rtx insn;
+ start_sequence ();
+ insn = emit_insn (gen_rtx_SET (VOIDmode, dest, const0_rtx));
+ dwarf2out_frame_debug (insn, true);
+ insn = emit_insn (gen_rtx_SET (VOIDmode, stack_pointer_rtx,
+ gen_rtx_PLUS (Pmode,
+ stack_pointer_rtx,
+ GEN_INT (4))));
+ RTX_FRAME_RELATED_P (insn) = 1;
+ dwarf2out_frame_debug (insn, true);
+ end_sequence ();
+ }
+#endif
+ }
}
else
{
get_pc_thunk_name (name, REGNO (dest));
pic_labels_used |= 1 << REGNO (dest);
+#ifdef DWARF2_UNWIND_INFO
+ /* Ensure all queued register saves are flushed before the
+ call. */
+ if (dwarf2out_do_frame ())
+ {
+ rtx insn;
+ start_sequence ();
+ insn = emit_barrier ();
+ end_sequence ();
+ dwarf2out_frame_debug (insn, false);
+ }
+#endif
xops[2] = gen_rtx_SYMBOL_REF (Pmode, ggc_strdup (name));
xops[2] = gen_rtx_MEM (QImode, xops[2]);
output_asm_insn ("call\t%X2", xops);
end_sequence ();
insn = emit_insn_before (seq, NEXT_INSN (entry_of_function ()));
- RTX_FRAME_RELATED_P (insn) = 1;
+ if (!optimize)
+ {
+ add_reg_note (insn, REG_CFA_SET_VDRAP, drap_vreg);
+ RTX_FRAME_RELATED_P (insn) = 1;
+ }
return drap_vreg;
}
else
ix86_cfa_state->reg == stack_pointer_rtx);
else
{
- /* Only valid for Win32. */
rtx eax = gen_rtx_REG (Pmode, AX_REG);
bool eax_live;
rtx t;
- gcc_assert (!TARGET_64BIT || cfun->machine->call_abi == MS_ABI);
-
if (cfun->machine->call_abi == MS_ABI)
eax_live = false;
else
break;
case UNSPEC_GOTTPOFF:
/* FIXME: This might be @TPOFF in Sun ld too. */
- fputs ("@GOTTPOFF", file);
+ fputs ("@gottpoff", file);
break;
case UNSPEC_TPOFF:
- fputs ("@TPOFF", file);
+ fputs ("@tpoff", file);
break;
case UNSPEC_NTPOFF:
if (TARGET_64BIT)
- fputs ("@TPOFF", file);
+ fputs ("@tpoff", file);
else
- fputs ("@NTPOFF", file);
+ fputs ("@ntpoff", file);
break;
case UNSPEC_DTPOFF:
- fputs ("@DTPOFF", file);
+ fputs ("@dtpoff", file);
break;
case UNSPEC_GOTNTPOFF:
if (TARGET_64BIT)
fputs (ASSEMBLER_DIALECT == ASM_ATT ?
- "@GOTTPOFF(%rip)": "@GOTTPOFF[rip]", file);
+ "@gottpoff(%rip)": "@gottpoff[rip]", file);
else
- fputs ("@GOTNTPOFF", file);
+ fputs ("@gotntpoff", file);
break;
case UNSPEC_INDNTPOFF:
- fputs ("@INDNTPOFF", file);
+ fputs ("@indntpoff", file);
break;
#if TARGET_MACHO
case UNSPEC_MACHOPIC_OFFSET:
{
fputs (ASM_LONG, file);
output_addr_const (file, x);
- fputs ("@DTPOFF", file);
+ fputs ("@dtpoff", file);
switch (size)
{
case 4:
ix86_delegitimize_address (rtx x)
{
rtx orig_x = delegitimize_mem_from_attrs (x);
+ /* addend is NULL or some rtx if x is something+GOTOFF where
+ something doesn't include the PIC register. */
+ rtx addend = NULL_RTX;
/* reg_addend is NULL or a multiple of some register. */
rtx reg_addend = NULL_RTX;
/* const_addend is NULL or a const_int. */
else if (ix86_pic_register_p (XEXP (reg_addend, 1)))
reg_addend = XEXP (reg_addend, 0);
else
- return orig_x;
- if (!REG_P (reg_addend)
- && GET_CODE (reg_addend) != MULT
- && GET_CODE (reg_addend) != ASHIFT)
- return orig_x;
+ {
+ reg_addend = NULL_RTX;
+ addend = XEXP (x, 0);
+ }
}
else
- return orig_x;
+ addend = XEXP (x, 0);
x = XEXP (XEXP (x, 1), 0);
if (GET_CODE (x) == PLUS
}
if (GET_CODE (x) == UNSPEC
- && ((XINT (x, 1) == UNSPEC_GOT && MEM_P (orig_x))
+ && ((XINT (x, 1) == UNSPEC_GOT && MEM_P (orig_x) && !addend)
|| (XINT (x, 1) == UNSPEC_GOTOFF && !MEM_P (orig_x))))
result = XVECEXP (x, 0, 0);
result = gen_rtx_CONST (Pmode, gen_rtx_PLUS (Pmode, result, const_addend));
if (reg_addend)
result = gen_rtx_PLUS (Pmode, reg_addend, result);
+ if (addend)
+ {
+ /* If the rest of original X doesn't involve the PIC register, add
+ addend and subtract pic_offset_table_rtx. This can happen e.g.
+ for code like:
+ leal (%ebx, %ecx, 4), %ecx
+ ...
+ movl foo@GOTOFF(%ecx), %edx
+ in which case we return (%ecx - %ebx) + foo. */
+ if (pic_offset_table_rtx)
+ result = gen_rtx_PLUS (Pmode, gen_rtx_MINUS (Pmode, copy_rtx (addend),
+ pic_offset_table_rtx),
+ result);
+ else
+ return orig_x;
+ }
return result;
}
L,W,B,Q,S,T -- print the opcode suffix for specified size of operand.
C -- print opcode suffix for set/cmov insn.
c -- like C, but print reversed condition
- E,e -- likewise, but for compare-and-branch fused insn.
F,f -- likewise, but for floating-point.
O -- if HAVE_AS_IX86_CMOV_SUN_SYNTAX, expand to "w.", "l." or "q.",
otherwise nothing
put_condition_code (GET_CODE (x), GET_MODE (XEXP (x, 0)), 1, 1, file);
return;
- case 'E':
- put_condition_code (GET_CODE (x), CCmode, 0, 0, file);
- return;
-
- case 'e':
- put_condition_code (GET_CODE (x), CCmode, 1, 0, file);
- return;
-
case 'H':
/* It doesn't actually matter what mode we use here, as we're
only going to use this for printing. */
case UNSPEC_GOTTPOFF:
output_addr_const (file, op);
/* FIXME: This might be @TPOFF in Sun ld. */
- fputs ("@GOTTPOFF", file);
+ fputs ("@gottpoff", file);
break;
case UNSPEC_TPOFF:
output_addr_const (file, op);
- fputs ("@TPOFF", file);
+ fputs ("@tpoff", file);
break;
case UNSPEC_NTPOFF:
output_addr_const (file, op);
if (TARGET_64BIT)
- fputs ("@TPOFF", file);
+ fputs ("@tpoff", file);
else
- fputs ("@NTPOFF", file);
+ fputs ("@ntpoff", file);
break;
case UNSPEC_DTPOFF:
output_addr_const (file, op);
- fputs ("@DTPOFF", file);
+ fputs ("@dtpoff", file);
break;
case UNSPEC_GOTNTPOFF:
output_addr_const (file, op);
if (TARGET_64BIT)
fputs (ASSEMBLER_DIALECT == ASM_ATT ?
- "@GOTTPOFF(%rip)" : "@GOTTPOFF[rip]", file);
+ "@gottpoff(%rip)" : "@gottpoff[rip]", file);
else
- fputs ("@GOTNTPOFF", file);
+ fputs ("@gotntpoff", file);
break;
case UNSPEC_INDNTPOFF:
output_addr_const (file, op);
- fputs ("@INDNTPOFF", file);
+ fputs ("@indntpoff", file);
break;
#if TARGET_MACHO
case UNSPEC_MACHOPIC_OFFSET:
{ OPTION_MASK_ISA_XOP, CODE_FOR_xop_pcom_tfv4si3, "__builtin_ia32_vpcomtrueud", IX86_BUILTIN_VPCOMTRUEUD, (enum rtx_code) PCOM_TRUE, (int)MULTI_ARG_2_SI_TF },
{ OPTION_MASK_ISA_XOP, CODE_FOR_xop_pcom_tfv2di3, "__builtin_ia32_vpcomtrueuq", IX86_BUILTIN_VPCOMTRUEUQ, (enum rtx_code) PCOM_TRUE, (int)MULTI_ARG_2_DI_TF },
- { OPTION_MASK_ISA_AVX, CODE_FOR_xop_vpermil2v2df3, "__builtin_ia32_vpermil2pd", IX86_BUILTIN_VPERMIL2PD, UNKNOWN, (int)MULTI_ARG_4_DF2_DI_I },
- { OPTION_MASK_ISA_AVX, CODE_FOR_xop_vpermil2v4sf3, "__builtin_ia32_vpermil2ps", IX86_BUILTIN_VPERMIL2PS, UNKNOWN, (int)MULTI_ARG_4_SF2_SI_I },
- { OPTION_MASK_ISA_AVX, CODE_FOR_xop_vpermil2v4df3, "__builtin_ia32_vpermil2pd256", IX86_BUILTIN_VPERMIL2PD256, UNKNOWN, (int)MULTI_ARG_4_DF2_DI_I1 },
- { OPTION_MASK_ISA_AVX, CODE_FOR_xop_vpermil2v8sf3, "__builtin_ia32_vpermil2ps256", IX86_BUILTIN_VPERMIL2PS256, UNKNOWN, (int)MULTI_ARG_4_SF2_SI_I1 },
+ { OPTION_MASK_ISA_XOP, CODE_FOR_xop_vpermil2v2df3, "__builtin_ia32_vpermil2pd", IX86_BUILTIN_VPERMIL2PD, UNKNOWN, (int)MULTI_ARG_4_DF2_DI_I },
+ { OPTION_MASK_ISA_XOP, CODE_FOR_xop_vpermil2v4sf3, "__builtin_ia32_vpermil2ps", IX86_BUILTIN_VPERMIL2PS, UNKNOWN, (int)MULTI_ARG_4_SF2_SI_I },
+ { OPTION_MASK_ISA_XOP, CODE_FOR_xop_vpermil2v4df3, "__builtin_ia32_vpermil2pd256", IX86_BUILTIN_VPERMIL2PD256, UNKNOWN, (int)MULTI_ARG_4_DF2_DI_I1 },
+ { OPTION_MASK_ISA_XOP, CODE_FOR_xop_vpermil2v8sf3, "__builtin_ia32_vpermil2ps256", IX86_BUILTIN_VPERMIL2PS256, UNKNOWN, (int)MULTI_ARG_4_SF2_SI_I1 },
};
nargs = 3;
nargs_constant = 2;
break;
- case MULTI_ARG_4_DF2_DI_I:
- case MULTI_ARG_4_DF2_DI_I1:
- case MULTI_ARG_4_SF2_SI_I:
- case MULTI_ARG_4_SF2_SI_I1:
+ case V2DF_FTYPE_V2DF_V2DF_V2DI_INT:
+ case V4DF_FTYPE_V4DF_V4DF_V4DI_INT:
+ case V4SF_FTYPE_V4SF_V4SF_V4SI_INT:
+ case V8SF_FTYPE_V8SF_V8SF_V8SI_INT:
nargs = 4;
nargs_constant = 1;
break;
if it is not available. */
static tree
-ix86_builtin_vectorized_function (unsigned int fn, tree type_out,
+ix86_builtin_vectorized_function (tree fndecl, tree type_out,
tree type_in)
{
enum machine_mode in_mode, out_mode;
int in_n, out_n;
+ enum built_in_function fn = DECL_FUNCTION_CODE (fndecl);
if (TREE_CODE (type_out) != VECTOR_TYPE
- || TREE_CODE (type_in) != VECTOR_TYPE)
+ || TREE_CODE (type_in) != VECTOR_TYPE
+ || DECL_BUILT_IN_CLASS (fndecl) != BUILT_IN_NORMAL)
return NULL_TREE;
out_mode = TYPE_MODE (TREE_TYPE (type_out));
if (!CONST_INT_P (er))
return 0;
ei = INTVAL (er);
- if (ei >= 2 * nelt)
+ if (ei >= nelt)
return 0;
ipar[i] = ei;
}
fprintf (file, "\t.indirect_symbol %s\n", symbol_name);
fprintf (file, ASM_LONG "%s\n", binder_name);
}
-
-void
-darwin_x86_file_end (void)
-{
- darwin_file_end ();
- ix86_file_end ();
-}
#endif /* TARGET_MACHO */
/* Order the registers for register allocator. */
*(*this + vcall_offset) should be added to THIS. */
static void
-x86_output_mi_thunk (FILE *file ATTRIBUTE_UNUSED,
+x86_output_mi_thunk (FILE *file,
tree thunk ATTRIBUTE_UNUSED, HOST_WIDE_INT delta,
HOST_WIDE_INT vcall_offset, tree function)
{
rtx this_param = x86_this_parameter (function);
rtx this_reg, tmp;
+ /* Make sure unwind info is emitted for the thunk if needed. */
+ final_start_function (emit_barrier (), file, 1);
+
/* 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_param))
/* Adjust the this parameter by a fixed constant. */
if (delta)
{
- /* Make things pretty and `subl $4,%eax' rather than `addl $-4,%eax'.
- Exceptions: -128 encodes smaller than 128, so swap sign and op. */
- bool sub = delta < 0 || delta == 128;
- xops[0] = GEN_INT (sub ? -delta : delta);
+ xops[0] = GEN_INT (delta);
xops[1] = this_reg ? this_reg : this_param;
if (TARGET_64BIT)
{
xops[0] = tmp;
xops[1] = this_param;
}
- if (sub)
+ if (x86_maybe_negate_const_int (&xops[0], DImode))
output_asm_insn ("sub{q}\t{%0, %1|%1, %0}", xops);
else
output_asm_insn ("add{q}\t{%0, %1|%1, %0}", xops);
}
- else if (sub)
+ else if (x86_maybe_negate_const_int (&xops[0], SImode))
output_asm_insn ("sub{l}\t{%0, %1|%1, %0}", xops);
else
output_asm_insn ("add{l}\t{%0, %1|%1, %0}", xops);
output_asm_insn ("jmp\t{*}%1", xops);
}
}
+ final_end_function ();
}
static void
if (TARGET_64BIT)
{
#ifndef NO_PROFILE_COUNTERS
- fprintf (file, "\tleaq\t" LPREFIX "P%d@(%%rip),%%r11\n", labelno);
+ fprintf (file, "\tleaq\t" LPREFIX "P%d(%%rip),%%r11\n", labelno);
#endif
if (DEFAULT_ABI == SYSV_ABI && flag_pic)
extended_reg_mentioned_1, NULL);
}
+/* If profitable, negate (without causing overflow) integer constant
+ of mode MODE at location LOC. Return true in this case. */
+bool
+x86_maybe_negate_const_int (rtx *loc, enum machine_mode mode)
+{
+ HOST_WIDE_INT val;
+
+ if (!CONST_INT_P (*loc))
+ return false;
+
+ switch (mode)
+ {
+ case DImode:
+ /* DImode x86_64 constants must fit in 32 bits. */
+ gcc_assert (x86_64_immediate_operand (*loc, mode));
+
+ mode = SImode;
+ break;
+
+ case SImode:
+ case HImode:
+ case QImode:
+ break;
+
+ default:
+ gcc_unreachable ();
+ }
+
+ /* Avoid overflows. */
+ if (mode_signbit_p (mode, *loc))
+ return false;
+
+ val = INTVAL (*loc);
+
+ /* Make things pretty and `subl $4,%eax' rather than `addl $-4,%eax'.
+ Exceptions: -128 encodes smaller than 128, so swap sign and op. */
+ if ((val < 0 && val != -128)
+ || val == 128)
+ {
+ *loc = GEN_INT (-val);
+ return true;
+ }
+
+ return false;
+}
+
/* Generate an unsigned DImode/SImode to FP conversion. This is the same code
optabs would emit if we didn't have TFmode patterns. */
do_subreg:
vmode = V8HImode;
target = gen_lowpart (vmode, target);
- op0 = gen_lowpart (vmode, target);
- op1 = gen_lowpart (vmode, target);
+ op0 = gen_lowpart (vmode, op0);
+ op1 = gen_lowpart (vmode, op1);
break;
default:
}
/* This matches five different patterns with the different modes. */
- x = gen_rtx_VEC_MERGE (vmode, op0, op1, GEN_INT (mask));
+ x = gen_rtx_VEC_MERGE (vmode, op1, op0, GEN_INT (mask));
x = gen_rtx_SET (VOIDmode, target, x);
emit_insn (x);
input where SEL+CONCAT may not. */
if (d->op0 == d->op1)
{
- if (expand_vselect (d->target, d->op0, d->perm, nelt))
+ int mask = nelt - 1;
+
+ for (i = 0; i < nelt; i++)
+ perm2[i] = d->perm[i] & mask;
+
+ if (expand_vselect (d->target, d->op0, perm2, nelt))
return true;
/* There are plenty of patterns in sse.md that are written for
every other permutation operand. */
for (i = 0; i < nelt; i += 2)
{
- perm2[i] = d->perm[i];
- perm2[i+1] = d->perm[i+1] + nelt;
+ perm2[i] = d->perm[i] & mask;
+ perm2[i + 1] = (d->perm[i + 1] & mask) + nelt;
}
if (expand_vselect_vconcat (d->target, d->op0, d->op0, perm2, nelt))
return true;
/* Recognize shufps, which means adding {0, 0, nelt, nelt}. */
if (nelt >= 4)
{
- memcpy (perm2, d->perm, nelt);
- for (i = 2; i < nelt; i += 4)
+ for (i = 0; i < nelt; i += 4)
{
- perm2[i+0] += nelt;
- perm2[i+1] += nelt;
+ perm2[i + 0] = d->perm[i + 0] & mask;
+ perm2[i + 1] = d->perm[i + 1] & mask;
+ perm2[i + 2] = (d->perm[i + 2] & mask) + nelt;
+ perm2[i + 3] = (d->perm[i + 3] & mask) + nelt;
}
if (expand_vselect_vconcat (d->target, d->op0, d->op0, perm2, nelt))
#undef TARGET_CAN_ELIMINATE
#define TARGET_CAN_ELIMINATE ix86_can_eliminate
+#undef TARGET_ASM_CODE_END
+#define TARGET_ASM_CODE_END ix86_code_end
+
struct gcc_target targetm = TARGET_INITIALIZER;
\f
#include "gt-i386.h"
OSDN Git Service
