r333348 - [X86] Remove mask from avx512ifma builtins. Use a select instruction instead.
Craig Topper via cfe-commits
cfe-commits at lists.llvm.org
Sat May 26 11:55:26 PDT 2018
Author: ctopper
Date: Sat May 26 11:55:26 2018
New Revision: 333348
URL: http://llvm.org/viewvc/llvm-project?rev=333348&view=rev
Log:
[X86] Remove mask from avx512ifma builtins. Use a select instruction instead.
This reduces from 12 builtins to 6 since we no longer need a mask and maskz version.
Modified:
cfe/trunk/include/clang/Basic/BuiltinsX86.def
cfe/trunk/lib/Headers/avx512ifmaintrin.h
cfe/trunk/lib/Headers/avx512ifmavlintrin.h
cfe/trunk/test/CodeGen/avx512ifma-builtins.c
cfe/trunk/test/CodeGen/avx512ifmavl-builtins.c
Modified: cfe/trunk/include/clang/Basic/BuiltinsX86.def
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/BuiltinsX86.def?rev=333348&r1=333347&r2=333348&view=diff
==============================================================================
--- cfe/trunk/include/clang/Basic/BuiltinsX86.def (original)
+++ cfe/trunk/include/clang/Basic/BuiltinsX86.def Sat May 26 11:55:26 2018
@@ -1461,18 +1461,12 @@ TARGET_BUILTIN(__builtin_ia32_movdqa64lo
TARGET_BUILTIN(__builtin_ia32_movdqa64load256_mask, "V4LLiV4LLiC*V4LLiUc", "n", "avx512vl")
TARGET_BUILTIN(__builtin_ia32_movdqa64store128_mask, "vV2LLi*V2LLiUc", "n", "avx512f")
TARGET_BUILTIN(__builtin_ia32_movdqa64store256_mask, "vV4LLi*V4LLiUc", "n", "avx512f")
-TARGET_BUILTIN(__builtin_ia32_vpmadd52huq512_mask, "V8LLiV8LLiV8LLiV8LLiUc", "nc", "avx512ifma")
-TARGET_BUILTIN(__builtin_ia32_vpmadd52huq512_maskz, "V8LLiV8LLiV8LLiV8LLiUc", "nc", "avx512ifma")
-TARGET_BUILTIN(__builtin_ia32_vpmadd52luq512_mask, "V8LLiV8LLiV8LLiV8LLiUc", "nc", "avx512ifma")
-TARGET_BUILTIN(__builtin_ia32_vpmadd52luq512_maskz, "V8LLiV8LLiV8LLiV8LLiUc", "nc", "avx512ifma")
-TARGET_BUILTIN(__builtin_ia32_vpmadd52huq128_mask, "V2LLiV2LLiV2LLiV2LLiUc", "nc", "avx512ifma,avx512vl")
-TARGET_BUILTIN(__builtin_ia32_vpmadd52huq128_maskz, "V2LLiV2LLiV2LLiV2LLiUc", "nc", "avx512ifma,avx512vl")
-TARGET_BUILTIN(__builtin_ia32_vpmadd52huq256_mask, "V4LLiV4LLiV4LLiV4LLiUc", "nc", "avx512ifma,avx512vl")
-TARGET_BUILTIN(__builtin_ia32_vpmadd52huq256_maskz, "V4LLiV4LLiV4LLiV4LLiUc", "nc", "avx512ifma,avx512vl")
-TARGET_BUILTIN(__builtin_ia32_vpmadd52luq128_mask, "V2LLiV2LLiV2LLiV2LLiUc", "nc", "avx512ifma,avx512vl")
-TARGET_BUILTIN(__builtin_ia32_vpmadd52luq128_maskz, "V2LLiV2LLiV2LLiV2LLiUc", "nc", "avx512ifma,avx512vl")
-TARGET_BUILTIN(__builtin_ia32_vpmadd52luq256_mask, "V4LLiV4LLiV4LLiV4LLiUc", "nc", "avx512ifma,avx512vl")
-TARGET_BUILTIN(__builtin_ia32_vpmadd52luq256_maskz, "V4LLiV4LLiV4LLiV4LLiUc", "nc", "avx512ifma,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_vpmadd52huq512, "V8LLiV8LLiV8LLiV8LLi", "nc", "avx512ifma")
+TARGET_BUILTIN(__builtin_ia32_vpmadd52luq512, "V8LLiV8LLiV8LLiV8LLi", "nc", "avx512ifma")
+TARGET_BUILTIN(__builtin_ia32_vpmadd52huq128, "V2LLiV2LLiV2LLiV2LLi", "nc", "avx512ifma,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_vpmadd52huq256, "V4LLiV4LLiV4LLiV4LLi", "nc", "avx512ifma,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_vpmadd52luq128, "V2LLiV2LLiV2LLiV2LLi", "nc", "avx512ifma,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_vpmadd52luq256, "V4LLiV4LLiV4LLiV4LLi", "nc", "avx512ifma,avx512vl")
TARGET_BUILTIN(__builtin_ia32_vpermi2varqi512_mask, "V64cV64cV64cV64cULLi", "nc", "avx512vbmi")
TARGET_BUILTIN(__builtin_ia32_vpermt2varqi512_mask, "V64cV64cV64cV64cULLi", "nc", "avx512vbmi")
TARGET_BUILTIN(__builtin_ia32_vpermt2varqi512_maskz, "V64cV64cV64cV64cULLi", "nc", "avx512vbmi")
Modified: cfe/trunk/lib/Headers/avx512ifmaintrin.h
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Headers/avx512ifmaintrin.h?rev=333348&r1=333347&r2=333348&view=diff
==============================================================================
--- cfe/trunk/lib/Headers/avx512ifmaintrin.h (original)
+++ cfe/trunk/lib/Headers/avx512ifmaintrin.h Sat May 26 11:55:26 2018
@@ -34,57 +34,47 @@
static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_madd52hi_epu64 (__m512i __X, __m512i __Y, __m512i __Z)
{
- return (__m512i) __builtin_ia32_vpmadd52huq512_mask ((__v8di) __X,
- (__v8di) __Y,
- (__v8di) __Z,
- (__mmask8) -1);
+ return (__m512i)__builtin_ia32_vpmadd52huq512((__v8di) __X, (__v8di) __Y,
+ (__v8di) __Z);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_mask_madd52hi_epu64 (__m512i __W, __mmask8 __M, __m512i __X,
- __m512i __Y)
+_mm512_mask_madd52hi_epu64 (__m512i __W, __mmask8 __M, __m512i __X, __m512i __Y)
{
- return (__m512i) __builtin_ia32_vpmadd52huq512_mask ((__v8di) __W,
- (__v8di) __X,
- (__v8di) __Y,
- (__mmask8) __M);
+ return (__m512i)__builtin_ia32_selectq_512(__M,
+ (__v8di)_mm512_madd52hi_epu64(__W, __X, __Y),
+ (__v8di)__W);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_maskz_madd52hi_epu64 (__mmask8 __M, __m512i __X, __m512i __Y, __m512i __Z)
{
- return (__m512i) __builtin_ia32_vpmadd52huq512_maskz ((__v8di) __X,
- (__v8di) __Y,
- (__v8di) __Z,
- (__mmask8) __M);
+ return (__m512i)__builtin_ia32_selectq_512(__M,
+ (__v8di)_mm512_madd52hi_epu64(__X, __Y, __Z),
+ (__v8di)_mm512_setzero_si512());
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_madd52lo_epu64 (__m512i __X, __m512i __Y, __m512i __Z)
{
- return (__m512i) __builtin_ia32_vpmadd52luq512_mask ((__v8di) __X,
- (__v8di) __Y,
- (__v8di) __Z,
- (__mmask8) -1);
+ return (__m512i)__builtin_ia32_vpmadd52luq512((__v8di) __X, (__v8di) __Y,
+ (__v8di) __Z);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_mask_madd52lo_epu64 (__m512i __W, __mmask8 __M, __m512i __X,
- __m512i __Y)
+_mm512_mask_madd52lo_epu64 (__m512i __W, __mmask8 __M, __m512i __X, __m512i __Y)
{
- return (__m512i) __builtin_ia32_vpmadd52luq512_mask ((__v8di) __W,
- (__v8di) __X,
- (__v8di) __Y,
- (__mmask8) __M);
+ return (__m512i)__builtin_ia32_selectq_512(__M,
+ (__v8di)_mm512_madd52lo_epu64(__W, __X, __Y),
+ (__v8di)__W);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_maskz_madd52lo_epu64 (__mmask8 __M, __m512i __X, __m512i __Y, __m512i __Z)
{
- return (__m512i) __builtin_ia32_vpmadd52luq512_maskz ((__v8di) __X,
- (__v8di) __Y,
- (__v8di) __Z,
- (__mmask8) __M);
+ return (__m512i)__builtin_ia32_selectq_512(__M,
+ (__v8di)_mm512_madd52lo_epu64(__X, __Y, __Z),
+ (__v8di)_mm512_setzero_si512());
}
#undef __DEFAULT_FN_ATTRS
Modified: cfe/trunk/lib/Headers/avx512ifmavlintrin.h
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Headers/avx512ifmavlintrin.h?rev=333348&r1=333347&r2=333348&view=diff
==============================================================================
--- cfe/trunk/lib/Headers/avx512ifmavlintrin.h (original)
+++ cfe/trunk/lib/Headers/avx512ifmavlintrin.h Sat May 26 11:55:26 2018
@@ -36,111 +36,93 @@
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_madd52hi_epu64 (__m128i __X, __m128i __Y, __m128i __Z)
{
- return (__m128i) __builtin_ia32_vpmadd52huq128_mask ((__v2di) __X,
- (__v2di) __Y,
- (__v2di) __Z,
- (__mmask8) -1);
+ return (__m128i)__builtin_ia32_vpmadd52huq128((__v2di) __X, (__v2di) __Y,
+ (__v2di) __Z);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_mask_madd52hi_epu64 (__m128i __W, __mmask8 __M, __m128i __X, __m128i __Y)
{
- return (__m128i) __builtin_ia32_vpmadd52huq128_mask ((__v2di) __W,
- (__v2di) __X,
- (__v2di) __Y,
- (__mmask8) __M);
+ return (__m128i)__builtin_ia32_selectq_128(__M,
+ (__v2di)_mm_madd52hi_epu64(__W, __X, __Y),
+ (__v2di)__W);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_maskz_madd52hi_epu64 (__mmask8 __M, __m128i __X, __m128i __Y, __m128i __Z)
{
- return (__m128i) __builtin_ia32_vpmadd52huq128_maskz ((__v2di) __X,
- (__v2di) __Y,
- (__v2di) __Z,
- (__mmask8) __M);
+ return (__m128i)__builtin_ia32_selectq_128(__M,
+ (__v2di)_mm_madd52hi_epu64(__X, __Y, __Z),
+ (__v2di)_mm_setzero_si128());
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
_mm256_madd52hi_epu64 (__m256i __X, __m256i __Y, __m256i __Z)
{
- return (__m256i) __builtin_ia32_vpmadd52huq256_mask ((__v4di) __X,
- (__v4di) __Y,
- (__v4di) __Z,
- (__mmask8) -1);
+ return (__m256i)__builtin_ia32_vpmadd52huq256((__v4di)__X, (__v4di)__Y,
+ (__v4di)__Z);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
-_mm256_mask_madd52hi_epu64 (__m256i __W, __mmask8 __M, __m256i __X,
- __m256i __Y)
+_mm256_mask_madd52hi_epu64 (__m256i __W, __mmask8 __M, __m256i __X, __m256i __Y)
{
- return (__m256i) __builtin_ia32_vpmadd52huq256_mask ((__v4di) __W,
- (__v4di) __X,
- (__v4di) __Y,
- (__mmask8) __M);
+ return (__m256i)__builtin_ia32_selectq_256(__M,
+ (__v4di)_mm256_madd52hi_epu64(__W, __X, __Y),
+ (__v4di)__W);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
_mm256_maskz_madd52hi_epu64 (__mmask8 __M, __m256i __X, __m256i __Y, __m256i __Z)
{
- return (__m256i) __builtin_ia32_vpmadd52huq256_maskz ((__v4di) __X,
- (__v4di) __Y,
- (__v4di) __Z,
- (__mmask8) __M);
+ return (__m256i)__builtin_ia32_selectq_256(__M,
+ (__v4di)_mm256_madd52hi_epu64(__X, __Y, __Z),
+ (__v4di)_mm256_setzero_si256());
}
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_madd52lo_epu64 (__m128i __X, __m128i __Y, __m128i __Z)
{
- return (__m128i) __builtin_ia32_vpmadd52luq128_mask ((__v2di) __X,
- (__v2di) __Y,
- (__v2di) __Z,
- (__mmask8) -1);
+ return (__m128i)__builtin_ia32_vpmadd52luq128((__v2di)__X, (__v2di)__Y,
+ (__v2di)__Z);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_mask_madd52lo_epu64 (__m128i __W, __mmask8 __M, __m128i __X, __m128i __Y)
{
- return (__m128i) __builtin_ia32_vpmadd52luq128_mask ((__v2di) __W,
- (__v2di) __X,
- (__v2di) __Y,
- (__mmask8) __M);
+ return (__m128i)__builtin_ia32_selectq_128(__M,
+ (__v2di)_mm_madd52lo_epu64(__W, __X, __Y),
+ (__v2di)__W);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_maskz_madd52lo_epu64 (__mmask8 __M, __m128i __X, __m128i __Y, __m128i __Z)
{
- return (__m128i) __builtin_ia32_vpmadd52luq128_maskz ((__v2di) __X,
- (__v2di) __Y,
- (__v2di) __Z,
- (__mmask8) __M);
+ return (__m128i)__builtin_ia32_selectq_128(__M,
+ (__v2di)_mm_madd52lo_epu64(__X, __Y, __Z),
+ (__v2di)_mm_setzero_si128());
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
_mm256_madd52lo_epu64 (__m256i __X, __m256i __Y, __m256i __Z)
{
- return (__m256i) __builtin_ia32_vpmadd52luq256_mask ((__v4di) __X,
- (__v4di) __Y,
- (__v4di) __Z,
- (__mmask8) -1);
+ return (__m256i)__builtin_ia32_vpmadd52luq256((__v4di)__X, (__v4di)__Y,
+ (__v4di)__Z);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
-_mm256_mask_madd52lo_epu64 (__m256i __W, __mmask8 __M, __m256i __X,
- __m256i __Y)
+_mm256_mask_madd52lo_epu64 (__m256i __W, __mmask8 __M, __m256i __X, __m256i __Y)
{
- return (__m256i) __builtin_ia32_vpmadd52luq256_mask ((__v4di) __W,
- (__v4di) __X,
- (__v4di) __Y,
- (__mmask8) __M);
+ return (__m256i)__builtin_ia32_selectq_256(__M,
+ (__v4di)_mm256_madd52lo_epu64(__W, __X, __Y),
+ (__v4di)__W);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
_mm256_maskz_madd52lo_epu64 (__mmask8 __M, __m256i __X, __m256i __Y, __m256i __Z)
{
- return (__m256i) __builtin_ia32_vpmadd52luq256_maskz ((__v4di) __X,
- (__v4di) __Y,
- (__v4di) __Z,
- (__mmask8) __M);
+ return (__m256i)__builtin_ia32_selectq_256(__M,
+ (__v4di)_mm256_madd52lo_epu64(__X, __Y, __Z),
+ (__v4di)_mm256_setzero_si256());
}
Modified: cfe/trunk/test/CodeGen/avx512ifma-builtins.c
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGen/avx512ifma-builtins.c?rev=333348&r1=333347&r2=333348&view=diff
==============================================================================
--- cfe/trunk/test/CodeGen/avx512ifma-builtins.c (original)
+++ cfe/trunk/test/CodeGen/avx512ifma-builtins.c Sat May 26 11:55:26 2018
@@ -5,36 +5,40 @@
__m512i test_mm512_madd52hi_epu64(__m512i __X, __m512i __Y, __m512i __Z) {
// CHECK-LABEL: @test_mm512_madd52hi_epu64
- // CHECK: @llvm.x86.avx512.mask.vpmadd52h.uq.512
+ // CHECK: @llvm.x86.avx512.vpmadd52h.uq.512
return _mm512_madd52hi_epu64(__X, __Y, __Z);
}
__m512i test_mm512_mask_madd52hi_epu64(__m512i __W, __mmask8 __M, __m512i __X, __m512i __Y) {
// CHECK-LABEL: @test_mm512_mask_madd52hi_epu64
- // CHECK: @llvm.x86.avx512.mask.vpmadd52h.uq.512
+ // CHECK: @llvm.x86.avx512.vpmadd52h.uq.512
+ // CHECK: select <8 x i1> %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}}
return _mm512_mask_madd52hi_epu64(__W, __M, __X, __Y);
}
__m512i test_mm512_maskz_madd52hi_epu64(__mmask8 __M, __m512i __X, __m512i __Y, __m512i __Z) {
// CHECK-LABEL: @test_mm512_maskz_madd52hi_epu64
- // CHECK: @llvm.x86.avx512.maskz.vpmadd52h.uq.512
+ // CHECK: @llvm.x86.avx512.vpmadd52h.uq.512
+ // CHECK: select <8 x i1> %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}}
return _mm512_maskz_madd52hi_epu64(__M, __X, __Y, __Z);
}
__m512i test_mm512_madd52lo_epu64(__m512i __X, __m512i __Y, __m512i __Z) {
// CHECK-LABEL: @test_mm512_madd52lo_epu64
- // CHECK: @llvm.x86.avx512.mask.vpmadd52l.uq.512
+ // CHECK: @llvm.x86.avx512.vpmadd52l.uq.512
return _mm512_madd52lo_epu64(__X, __Y, __Z);
}
__m512i test_mm512_mask_madd52lo_epu64(__m512i __W, __mmask8 __M, __m512i __X, __m512i __Y) {
// CHECK-LABEL: @test_mm512_mask_madd52lo_epu64
- // CHECK: @llvm.x86.avx512.mask.vpmadd52l.uq.512
+ // CHECK: @llvm.x86.avx512.vpmadd52l.uq.512
+ // CHECK: select <8 x i1> %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}}
return _mm512_mask_madd52lo_epu64(__W, __M, __X, __Y);
}
__m512i test_mm512_maskz_madd52lo_epu64(__mmask8 __M, __m512i __X, __m512i __Y, __m512i __Z) {
// CHECK-LABEL: @test_mm512_maskz_madd52lo_epu64
- // CHECK: @llvm.x86.avx512.mask.vpmadd52l.uq.512
+ // CHECK: @llvm.x86.avx512.vpmadd52l.uq.512
+ // CHECK: select <8 x i1> %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}}
return _mm512_maskz_madd52lo_epu64(__M, __X, __Y, __Z);
}
Modified: cfe/trunk/test/CodeGen/avx512ifmavl-builtins.c
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGen/avx512ifmavl-builtins.c?rev=333348&r1=333347&r2=333348&view=diff
==============================================================================
--- cfe/trunk/test/CodeGen/avx512ifmavl-builtins.c (original)
+++ cfe/trunk/test/CodeGen/avx512ifmavl-builtins.c Sat May 26 11:55:26 2018
@@ -4,72 +4,80 @@
__m128i test_mm_madd52hi_epu64(__m128i __X, __m128i __Y, __m128i __Z) {
// CHECK-LABEL: @test_mm_madd52hi_epu64
- // CHECK: @llvm.x86.avx512.mask.vpmadd52h.uq.128
+ // CHECK: @llvm.x86.avx512.vpmadd52h.uq.128
return _mm_madd52hi_epu64(__X, __Y, __Z);
}
__m128i test_mm_mask_madd52hi_epu64(__m128i __W, __mmask8 __M, __m128i __X, __m128i __Y) {
// CHECK-LABEL: @test_mm_mask_madd52hi_epu64
- // CHECK: @llvm.x86.avx512.mask.vpmadd52h.uq.128
+ // CHECK: @llvm.x86.avx512.vpmadd52h.uq.128
+ // CHECK: select <2 x i1> %{{.*}}, <2 x i64> %{{.*}}, <2 x i64> %{{.*}}
return _mm_mask_madd52hi_epu64(__W, __M, __X, __Y);
}
__m128i test_mm_maskz_madd52hi_epu64(__mmask8 __M, __m128i __X, __m128i __Y, __m128i __Z) {
// CHECK-LABEL: @test_mm_maskz_madd52hi_epu64
- // CHECK: @llvm.x86.avx512.maskz.vpmadd52h.uq.128
+ // CHECK: @llvm.x86.avx512.vpmadd52h.uq.128
+ // CHECK: select <2 x i1> %{{.*}}, <2 x i64> %{{.*}}, <2 x i64> %{{.*}}
return _mm_maskz_madd52hi_epu64(__M, __X, __Y, __Z);
}
__m256i test_mm256_madd52hi_epu64(__m256i __X, __m256i __Y, __m256i __Z) {
// CHECK-LABEL: @test_mm256_madd52hi_epu64
- // CHECK: @llvm.x86.avx512.mask.vpmadd52h.uq.256
+ // CHECK: @llvm.x86.avx512.vpmadd52h.uq.256
return _mm256_madd52hi_epu64(__X, __Y, __Z);
}
__m256i test_mm256_mask_madd52hi_epu64(__m256i __W, __mmask8 __M, __m256i __X, __m256i __Y) {
// CHECK-LABEL: @test_mm256_mask_madd52hi_epu64
- // CHECK: @llvm.x86.avx512.mask.vpmadd52h.uq.256
+ // CHECK: @llvm.x86.avx512.vpmadd52h.uq.256
+ // CHECK: select <4 x i1> %{{.*}}, <4 x i64> %{{.*}}, <4 x i64> %{{.*}}
return _mm256_mask_madd52hi_epu64(__W, __M, __X, __Y);
}
__m256i test_mm256_maskz_madd52hi_epu64(__mmask8 __M, __m256i __X, __m256i __Y, __m256i __Z) {
// CHECK-LABEL: @test_mm256_maskz_madd52hi_epu64
- // CHECK: @llvm.x86.avx512.maskz.vpmadd52h.uq.256
+ // CHECK: @llvm.x86.avx512.vpmadd52h.uq.256
+ // CHECK: select <4 x i1> %{{.*}}, <4 x i64> %{{.*}}, <4 x i64> %{{.*}}
return _mm256_maskz_madd52hi_epu64(__M, __X, __Y, __Z);
}
__m128i test_mm_madd52lo_epu64(__m128i __X, __m128i __Y, __m128i __Z) {
// CHECK-LABEL: @test_mm_madd52lo_epu64
- // CHECK: @llvm.x86.avx512.mask.vpmadd52l.uq.128
+ // CHECK: @llvm.x86.avx512.vpmadd52l.uq.128
return _mm_madd52lo_epu64(__X, __Y, __Z);
}
__m128i test_mm_mask_madd52lo_epu64(__m128i __W, __mmask8 __M, __m128i __X, __m128i __Y) {
// CHECK-LABEL: @test_mm_mask_madd52lo_epu64
- // CHECK: @llvm.x86.avx512.mask.vpmadd52l.uq.128
+ // CHECK: @llvm.x86.avx512.vpmadd52l.uq.128
+ // CHECK: select <2 x i1> %{{.*}}, <2 x i64> %{{.*}}, <2 x i64> %{{.*}}
return _mm_mask_madd52lo_epu64(__W, __M, __X, __Y);
}
__m128i test_mm_maskz_madd52lo_epu64(__mmask8 __M, __m128i __X, __m128i __Y, __m128i __Z) {
// CHECK-LABEL: @test_mm_maskz_madd52lo_epu64
- // CHECK: @llvm.x86.avx512.maskz.vpmadd52l.uq.128
+ // CHECK: @llvm.x86.avx512.vpmadd52l.uq.128
+ // CHECK: select <2 x i1> %{{.*}}, <2 x i64> %{{.*}}, <2 x i64> %{{.*}}
return _mm_maskz_madd52lo_epu64(__M, __X, __Y, __Z);
}
__m256i test_mm256_madd52lo_epu64(__m256i __X, __m256i __Y, __m256i __Z) {
// CHECK-LABEL: @test_mm256_madd52lo_epu64
- // CHECK: @llvm.x86.avx512.mask.vpmadd52l.uq.256
+ // CHECK: @llvm.x86.avx512.vpmadd52l.uq.256
return _mm256_madd52lo_epu64(__X, __Y, __Z);
}
__m256i test_mm256_mask_madd52lo_epu64(__m256i __W, __mmask8 __M, __m256i __X, __m256i __Y) {
// CHECK-LABEL: @test_mm256_mask_madd52lo_epu64
- // CHECK: @llvm.x86.avx512.mask.vpmadd52l.uq.256
+ // CHECK: @llvm.x86.avx512.vpmadd52l.uq.256
+ // CHECK: select <4 x i1> %{{.*}}, <4 x i64> %{{.*}}, <4 x i64> %{{.*}}
return _mm256_mask_madd52lo_epu64(__W, __M, __X, __Y);
}
__m256i test_mm256_maskz_madd52lo_epu64(__mmask8 __M, __m256i __X, __m256i __Y, __m256i __Z) {
// CHECK-LABEL: @test_mm256_maskz_madd52lo_epu64
- // CHECK: @llvm.x86.avx512.mask.vpmadd52l.uq.256
+ // CHECK: @llvm.x86.avx512.vpmadd52l.uq.256
+ // CHECK: select <4 x i1> %{{.*}}, <4 x i64> %{{.*}}, <4 x i64> %{{.*}}
return _mm256_maskz_madd52lo_epu64(__M, __X, __Y, __Z);
}
More information about the cfe-commits
mailing list