maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

Get rid of storelv4si builtin as it can be expressed directly. This is general 

goodness because it provides opportunites to cleanup things.  For example,

uint64_t t1(__m128i vA)
{
  uint64_t Alo;
  _mm_storel_epi64((__m128i*)&Alo, vA);
  return Alo;
}

was generating 

	movq	%xmm0, -8(%rbp)
	movq	-8(%rbp), %rax

and now generates

	movd	%xmm0, %rax

rdar://11282581

llvm-svn: 155924
This commit is contained in:
Chad Rosier 2012-05-01 18:11:51 +00:00
parent 8568fb91ea
commit 87622b8b84
4 changed files with 10 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
clang/include/clang/Basic/BuiltinsX86.def
View File

@ -303,8 +303,6 @@ BUILTIN(__builtin_ia32_lddqu, "V16ccC*", "")
BUILTIN(__builtin_ia32_palignr128, "V16cV16cV16cIc", "") BUILTIN(__builtin_ia32_palignr128, "V16cV16cV16cIc", "")
BUILTIN(__builtin_ia32_insertps128, "V4fV4fV4fi", "") BUILTIN(__builtin_ia32_insertps128, "V4fV4fV4fi", "")
BUILTIN(__builtin_ia32_storelv4si, "vV2i*V2LLi", "")
BUILTIN(__builtin_ia32_pblendvb128, "V16cV16cV16cV16c", "") BUILTIN(__builtin_ia32_pblendvb128, "V16cV16cV16cV16c", "")
BUILTIN(__builtin_ia32_pblendw128, "V8sV8sV8sIi", "") BUILTIN(__builtin_ia32_pblendw128, "V8sV8sV8sIi", "")
BUILTIN(__builtin_ia32_blendpd, "V2dV2dV2dIi", "") BUILTIN(__builtin_ia32_blendpd, "V2dV2dV2dIi", "")

5
clang/lib/Headers/emmintrin.h
View File

@ -1186,7 +1186,10 @@ _mm_maskmoveu_si128(__m128i d, __m128i n, char *p)
static __inline__ void __attribute__((__always_inline__, __nodebug__)) static __inline__ void __attribute__((__always_inline__, __nodebug__))
_mm_storel_epi64(__m128i *p, __m128i a) _mm_storel_epi64(__m128i *p, __m128i a)
{ {
__builtin_ia32_storelv4si((__v2si *)p, a); struct __mm_storel_epi64_struct {
long long u;
} __attribute__((__packed__, __may_alias__));
((struct __mm_storel_epi64_struct*)p)->u = a[0];
} }
static __inline__ void __attribute__((__always_inline__, __nodebug__)) static __inline__ void __attribute__((__always_inline__, __nodebug__))

1
clang/test/CodeGen/builtins-x86.c
View File

@ -344,7 +344,6 @@ void f0() {
tmp_V16c = __builtin_ia32_lddqu(tmp_cCp); tmp_V16c = __builtin_ia32_lddqu(tmp_cCp);
tmp_V2LLi = __builtin_ia32_palignr128(tmp_V2LLi, tmp_V2LLi, imm_i); tmp_V2LLi = __builtin_ia32_palignr128(tmp_V2LLi, tmp_V2LLi, imm_i);
tmp_V1LLi = __builtin_ia32_palignr(tmp_V1LLi, tmp_V1LLi, imm_i); tmp_V1LLi = __builtin_ia32_palignr(tmp_V1LLi, tmp_V1LLi, imm_i);
(void) __builtin_ia32_storelv4si(tmp_V2ip, tmp_V2LLi);
#ifdef USE_SSE4 #ifdef USE_SSE4
tmp_V16c = __builtin_ia32_pblendvb128(tmp_V16c, tmp_V16c, tmp_V16c); tmp_V16c = __builtin_ia32_pblendvb128(tmp_V16c, tmp_V16c, tmp_V16c);
tmp_V8s = __builtin_ia32_pblendw128(tmp_V8s, tmp_V8s, imm_i_0_256); tmp_V8s = __builtin_ia32_pblendw128(tmp_V8s, tmp_V8s, imm_i_0_256);

6
clang/test/CodeGen/sse-builtins.c
View File

@ -151,3 +151,9 @@ __m128d test_mm_round_sd(__m128d x, __m128d y) {
// CHECK: @llvm.x86.sse41.round.sd // CHECK: @llvm.x86.sse41.round.sd
return _mm_round_sd(x, y, 2); return _mm_round_sd(x, y, 2);
} }
void test_storel_epi64(__m128i x, void* y) {
// CHECK: define void @test_storel_epi64
// CHECK: store {{.*}} i64* {{.*}}, align 1{{$}}
_mm_storel_epi64(y, x);
}