[X86] Use select instrution and fpextend in the implementation of _mm512_mask_cvtps_p...

llvm-svn: 332213

diff --git a/clang/lib/Headers/avx512fintrin.h b/clang/lib/Headers/avx512fintrin.h
index 01f764d..e891b6c 100644 (file)
--- a/clang/lib/Headers/avx512fintrin.h
+++ b/clang/lib/Headers/avx512fintrin.h
@@ -9317,20 +9317,17 @@ _mm512_cvtps_pd (__m256 __A)
 static __inline__ __m512d __DEFAULT_FN_ATTRS
 _mm512_mask_cvtps_pd (__m512d __W, __mmask8 __U, __m256 __A)
 {
-  return (__m512d) __builtin_ia32_cvtps2pd512_mask ((__v8sf) __A,
-                (__v8df) __W,
-                (__mmask8) __U,
-                _MM_FROUND_CUR_DIRECTION);
+  return (__m512d)__builtin_ia32_selectpd_512((__mmask8)__U,
+                                              (__v8df)_mm512_cvtps_pd(__A),
+                                              (__v8df)__W);
 }
 
 static __inline__ __m512d __DEFAULT_FN_ATTRS
 _mm512_maskz_cvtps_pd (__mmask8 __U, __m256 __A)
 {
-  return (__m512d) __builtin_ia32_cvtps2pd512_mask ((__v8sf) __A,
-                (__v8df)
-                _mm512_setzero_pd (),
-                (__mmask8) __U,
-                _MM_FROUND_CUR_DIRECTION);
+  return (__m512d)__builtin_ia32_selectpd_512((__mmask8)__U,
+                                              (__v8df)_mm512_cvtps_pd(__A),
+                                              (__v8df)_mm512_setzero_pd());
 }
 
 static __inline__ __m512 __DEFAULT_FN_ATTRS

diff --git a/clang/test/CodeGen/avx512f-builtins.c b/clang/test/CodeGen/avx512f-builtins.c
index 3bef192..b84740d 100644 (file)
--- a/clang/test/CodeGen/avx512f-builtins.c
+++ b/clang/test/CodeGen/avx512f-builtins.c
@@ -6643,20 +6643,23 @@ __m512d test_mm512_cvtpslo_pd(__m512 __A) {
 
 __m512d test_mm512_mask_cvtps_pd(__m512d __W, __mmask8 __U, __m256 __A) {
   // CHECK-LABEL: @test_mm512_mask_cvtps_pd
-  // CHECK: @llvm.x86.avx512.mask.cvtps2pd.512
+  // CHECK: fpext <8 x float> %{{.*}} to <8 x double>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x double> %{{.*}}, <8 x double> %{{.*}}
   return _mm512_mask_cvtps_pd(__W, __U, __A); 
 }
 
 __m512d test_mm512_mask_cvtpslo_pd(__m512d __W, __mmask8 __U, __m512 __A) {
   // CHECK-LABEL: @test_mm512_mask_cvtpslo_pd
   // CHECK: shufflevector <16 x float> %{{.*}}, <16 x float> %{{.*}}, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-  // CHECK: @llvm.x86.avx512.mask.cvtps2pd.512
+  // CHECK: fpext <8 x float> %{{.*}} to <8 x double>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x double> %{{.*}}, <8 x double> %{{.*}}
   return _mm512_mask_cvtpslo_pd(__W, __U, __A);
 }
 
 __m512d test_mm512_maskz_cvtps_pd(__mmask8 __U, __m256 __A) {
   // CHECK-LABEL: @test_mm512_maskz_cvtps_pd
-  // CHECK: @llvm.x86.avx512.mask.cvtps2pd.512
+  // CHECK: fpext <8 x float> %{{.*}} to <8 x double>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x double> %{{.*}}, <8 x double> %{{.*}}
   return _mm512_maskz_cvtps_pd(__U, __A); 
 }
 __m512d test_mm512_mask_mov_pd(__m512d __W, __mmask8 __U, __m512d __A) {