From: Craig Topper <craig.topper@gmail.com>
Date: Tue, 27 Dec 2016 03:46:16 +0000 (+0000)
Subject: [AVX-512] Replace masked 512-bit pmuldq and pmuludq builtins with the newly added... 
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=ece2a94816301a0cc9570a6aae9748d8971ca2bd;p=clang

[AVX-512] Replace masked 512-bit pmuldq and pmuludq builtins with the newly added unmasked versions and selects.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@290575 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Headers/avx512fintrin.h b/lib/Headers/avx512fintrin.h
index 30ebfc0430..e6a7217c89 100644
--- a/lib/Headers/avx512fintrin.h
+++ b/lib/Headers/avx512fintrin.h
@@ -1422,7 +1422,7 @@ _mm512_mul_epi32(__m512i __X, __m512i __Y)
 static __inline __m512i __DEFAULT_FN_ATTRS
 _mm512_mask_mul_epi32(__m512i __W, __mmask8 __M, __m512i __X, __m512i __Y)
 {
-  return (__m512i)__builtin_ia32_selectq_512((__mmask8)__U,
+  return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
                                              (__v8di)_mm512_mul_epi32(__X, __Y),
                                              (__v8di)__W);
 }
@@ -1430,7 +1430,7 @@ _mm512_mask_mul_epi32(__m512i __W, __mmask8 __M, __m512i __X, __m512i __Y)
 static __inline __m512i __DEFAULT_FN_ATTRS
 _mm512_maskz_mul_epi32(__mmask8 __M, __m512i __X, __m512i __Y)
 {
-  return (__m512i)__builtin_ia32_selectq_512((__mmask8)__U,
+  return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
                                              (__v8di)_mm512_mul_epi32(__X, __Y),
                                              (__v8di)_mm512_setzero_si512 ());
 }
@@ -1444,7 +1444,7 @@ _mm512_mul_epu32(__m512i __X, __m512i __Y)
 static __inline __m512i __DEFAULT_FN_ATTRS
 _mm512_mask_mul_epu32(__m512i __W, __mmask8 __M, __m512i __X, __m512i __Y)
 {
-  return (__m512i)__builtin_ia32_selectq_512((__mmask8)__U,
+  return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
                                              (__v8di)_mm512_mul_epu32(__X, __Y),
                                              (__v8di)__W);
 }
@@ -1452,7 +1452,7 @@ _mm512_mask_mul_epu32(__m512i __W, __mmask8 __M, __m512i __X, __m512i __Y)
 static __inline __m512i __DEFAULT_FN_ATTRS
 _mm512_maskz_mul_epu32(__mmask8 __M, __m512i __X, __m512i __Y)
 {
-  return (__m512i)__builtin_ia32_selectq_512((__mmask8)__U,
+  return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
                                              (__v8di)_mm512_mul_epu32(__X, __Y),
                                              (__v8di)_mm512_setzero_si512 ());
 }
diff --git a/test/CodeGen/avx512f-builtins.c b/test/CodeGen/avx512f-builtins.c
index ebd162835f..760783af1c 100644
--- a/test/CodeGen/avx512f-builtins.c
+++ b/test/CodeGen/avx512f-builtins.c
@@ -1867,29 +1867,45 @@ __m512i test_mm512_add_epi64(__m512i __A, __m512i __B) {
   return _mm512_add_epi64(__A,__B);
 }
 
+__m512i test_mm512_mul_epi32(__m512i __A, __m512i __B) {
+  //CHECK-LABEL: @test_mm512_mul_epi32
+  //CHECK: @llvm.x86.avx512.pmul.dq.512
+  return _mm512_mul_epi32(__A,__B);
+}
+
 __m512i test_mm512_maskz_mul_epi32 (__mmask16 __k,__m512i __A, __m512i __B) {
   //CHECK-LABEL: @test_mm512_maskz_mul_epi32
-  //CHECK: @llvm.x86.avx512.mask.pmul.dq.512
+  //CHECK: @llvm.x86.avx512.pmul.dq.512
+  //CHECK: select <8 x i1> %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}}
   return _mm512_maskz_mul_epi32(__k,__A,__B);
 }
 
 __m512i test_mm512_mask_mul_epi32 (__mmask16 __k,__m512i __A, __m512i __B,
                                    __m512i __src) {
   //CHECK-LABEL: @test_mm512_mask_mul_epi32
-  //CHECK: @llvm.x86.avx512.mask.pmul.dq.512
+  //CHECK: @llvm.x86.avx512.pmul.dq.512
+  //CHECK: select <8 x i1> %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}}
   return _mm512_mask_mul_epi32(__src,__k,__A,__B);
 }
 
+__m512i test_mm512_mul_epu32 (__m512i __A, __m512i __B) {
+  //CHECK-LABEL: @test_mm512_mul_epu32
+  //CHECK: @llvm.x86.avx512.pmulu.dq.512
+  return _mm512_mul_epu32(__A,__B);
+}
+
 __m512i test_mm512_maskz_mul_epu32 (__mmask16 __k,__m512i __A, __m512i __B) {
   //CHECK-LABEL: @test_mm512_maskz_mul_epu32
-  //CHECK: @llvm.x86.avx512.mask.pmulu.dq.512
+  //CHECK: @llvm.x86.avx512.pmulu.dq.512
+  //CHECK: select <8 x i1> %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}}
   return _mm512_maskz_mul_epu32(__k,__A,__B);
 }
 
 __m512i test_mm512_mask_mul_epu32 (__mmask16 __k,__m512i __A, __m512i __B, 
                                    __m512i __src) {
   //CHECK-LABEL: @test_mm512_mask_mul_epu32
-  //CHECK: @llvm.x86.avx512.mask.pmulu.dq.512
+  //CHECK: @llvm.x86.avx512.pmulu.dq.512
+  //CHECK: select <8 x i1> %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}}
   return _mm512_mask_mul_epu32(__src,__k,__A,__B);
 }