[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.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@333348 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/include/clang/Basic/BuiltinsX86.def b/include/clang/Basic/BuiltinsX86.def
index a3b78540e5836f17ac0e34002f0cb1edd090db40..f386078dae179338e84b34f905f31a0d7a99b11a 100644 (file)
--- a/include/clang/Basic/BuiltinsX86.def
+++ b/include/clang/Basic/BuiltinsX86.def
@@ -1461,18 +1461,12 @@ TARGET_BUILTIN(__builtin_ia32_movdqa64load128_mask, "V2LLiV2LLiC*V2LLiUc", "n",
 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")

diff --git a/lib/Headers/avx512ifmaintrin.h b/lib/Headers/avx512ifmaintrin.h
index 5defbaea8bcc1cdbaf579e0af7f0aee3c077482e..0018ef5de9afa83fad0395bfeadead28ae8659bb 100644 (file)
--- a/lib/Headers/avx512ifmaintrin.h
+++ b/lib/Headers/avx512ifmaintrin.h
@@ -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

diff --git a/lib/Headers/avx512ifmavlintrin.h b/lib/Headers/avx512ifmavlintrin.h
index 131ee5cb4f88e6094ea3db0414a0c6759931276c..934cfb433f97025db2218fabe4ff81bdde7a2687 100644 (file)
--- a/lib/Headers/avx512ifmavlintrin.h
+++ b/lib/Headers/avx512ifmavlintrin.h
@@ -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());
 }
 
 

diff --git a/test/CodeGen/avx512ifma-builtins.c b/test/CodeGen/avx512ifma-builtins.c
index 311d6989bf0df1d2658ccfc78feebb15747905a6..383cf5350428070fa7e4eb171d96a037f748692b 100644 (file)
--- a/test/CodeGen/avx512ifma-builtins.c
+++ b/test/CodeGen/avx512ifma-builtins.c
@@ -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); 
 }

diff --git a/test/CodeGen/avx512ifmavl-builtins.c b/test/CodeGen/avx512ifmavl-builtins.c
index 4aeec336ad943d9d4dfbf7288cd646e431e5f405..2fe9938eab7679982bfebf8d7ce5f055e523b9f6 100644 (file)
--- a/test/CodeGen/avx512ifmavl-builtins.c
+++ b/test/CodeGen/avx512ifmavl-builtins.c
@@ -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); 
 }