[X86] Remove masking from the 512-bit packed floating point add/sub/mul/div builtins...

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@334359 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/include/clang/Basic/BuiltinsX86.def b/include/clang/Basic/BuiltinsX86.def
index 6be561fc0c5e5d8c0456355ef76da7abfb853d9b..fc580df0836049b0928fbbfd6f4747b1e831cae6 100644 (file)
--- a/include/clang/Basic/BuiltinsX86.def
+++ b/include/clang/Basic/BuiltinsX86.def
@@ -1087,14 +1087,14 @@ TARGET_BUILTIN(__builtin_ia32_pmulhrsw512, "V32sV32sV32s", "nc", "avx512bw")
 TARGET_BUILTIN(__builtin_ia32_pmulhuw512, "V32sV32sV32s", "nc", "avx512bw")
 TARGET_BUILTIN(__builtin_ia32_pmulhw512, "V32sV32sV32s", "nc", "avx512bw")
 
-TARGET_BUILTIN(__builtin_ia32_addpd512_mask, "V8dV8dV8dV8dUcIi", "nc", "avx512f")
-TARGET_BUILTIN(__builtin_ia32_addps512_mask, "V16fV16fV16fV16fUsIi", "nc", "avx512f")
-TARGET_BUILTIN(__builtin_ia32_divpd512_mask, "V8dV8dV8dV8dUcIi", "nc", "avx512f")
-TARGET_BUILTIN(__builtin_ia32_divps512_mask, "V16fV16fV16fV16fUsIi", "nc", "avx512f")
-TARGET_BUILTIN(__builtin_ia32_mulpd512_mask, "V8dV8dV8dV8dUcIi", "nc", "avx512f")
-TARGET_BUILTIN(__builtin_ia32_mulps512_mask, "V16fV16fV16fV16fUsIi", "nc", "avx512f")
-TARGET_BUILTIN(__builtin_ia32_subpd512_mask, "V8dV8dV8dV8dUcIi", "nc", "avx512f")
-TARGET_BUILTIN(__builtin_ia32_subps512_mask, "V16fV16fV16fV16fUsIi", "nc", "avx512f")
+TARGET_BUILTIN(__builtin_ia32_addpd512, "V8dV8dV8dIi", "nc", "avx512f")
+TARGET_BUILTIN(__builtin_ia32_addps512, "V16fV16fV16fIi", "nc", "avx512f")
+TARGET_BUILTIN(__builtin_ia32_divpd512, "V8dV8dV8dIi", "nc", "avx512f")
+TARGET_BUILTIN(__builtin_ia32_divps512, "V16fV16fV16fIi", "nc", "avx512f")
+TARGET_BUILTIN(__builtin_ia32_mulpd512, "V8dV8dV8dIi", "nc", "avx512f")
+TARGET_BUILTIN(__builtin_ia32_mulps512, "V16fV16fV16fIi", "nc", "avx512f")
+TARGET_BUILTIN(__builtin_ia32_subpd512, "V8dV8dV8dIi", "nc", "avx512f")
+TARGET_BUILTIN(__builtin_ia32_subps512, "V16fV16fV16fIi", "nc", "avx512f")
 
 TARGET_BUILTIN(__builtin_ia32_pmaddubsw512, "V32sV64cV64c", "nc", "avx512bw")
 TARGET_BUILTIN(__builtin_ia32_pmaddwd512, "V16iV32sV32s", "nc", "avx512bw")

diff --git a/lib/Headers/avx512fintrin.h b/lib/Headers/avx512fintrin.h
index 9daa559bb133d06daf75c39f6d1580d90a734172..a4c01019592ca38aa6fe27bb088e91b60b08a475 100644 (file)
--- a/lib/Headers/avx512fintrin.h
+++ b/lib/Headers/avx512fintrin.h
@@ -2060,40 +2060,32 @@ _mm512_maskz_add_ps(__mmask16 __U, __m512 __A, __m512 __B) {
 }
 
 #define _mm512_add_round_pd(A, B, R) \
-  (__m512d)__builtin_ia32_addpd512_mask((__v8df)(__m512d)(A), \
-                                        (__v8df)(__m512d)(B), \
-                                        (__v8df)_mm512_setzero_pd(), \
-                                        (__mmask8)-1, (int)(R))
+  (__m512d)__builtin_ia32_addpd512((__v8df)(__m512d)(A), \
+                                   (__v8df)(__m512d)(B), (int)(R))
 
 #define _mm512_mask_add_round_pd(W, U, A, B, R) \
-  (__m512d)__builtin_ia32_addpd512_mask((__v8df)(__m512d)(A), \
-                                        (__v8df)(__m512d)(B), \
-                                        (__v8df)(__m512d)(W), (__mmask8)(U), \
-                                        (int)(R))
+  (__m512d)__builtin_ia32_selectpd_512((__mmask8)(U), \
+                                   (__v8df)_mm512_add_round_pd((A), (B), (R)), \
+                                   (__v8df)(__m512d)(W));
 
 #define _mm512_maskz_add_round_pd(U, A, B, R) \
-  (__m512d)__builtin_ia32_addpd512_mask((__v8df)(__m512d)(A), \
-                                        (__v8df)(__m512d)(B), \
-                                        (__v8df)_mm512_setzero_pd(), \
-                                        (__mmask8)(U), (int)(R))
+  (__m512d)__builtin_ia32_selectpd_512((__mmask8)(U), \
+                                   (__v8df)_mm512_add_round_pd((A), (B), (R)), \
+                                   (__v8df)_mm512_setzero_pd());
 
 #define _mm512_add_round_ps(A, B, R) \
-  (__m512)__builtin_ia32_addps512_mask((__v16sf)(__m512)(A), \
-                                       (__v16sf)(__m512)(B), \
-                                       (__v16sf)_mm512_setzero_ps(), \
-                                       (__mmask16)-1, (int)(R))
+  (__m512)__builtin_ia32_addps512((__v16sf)(__m512)(A), \
+                                  (__v16sf)(__m512)(B), (int)(R))
 
 #define _mm512_mask_add_round_ps(W, U, A, B, R) \
-  (__m512)__builtin_ia32_addps512_mask((__v16sf)(__m512)(A), \
-                                       (__v16sf)(__m512)(B), \
-                                       (__v16sf)(__m512)(W), (__mmask16)(U), \
-                                       (int)(R))
+  (__m512)__builtin_ia32_selectps_512((__mmask16)(U), \
+                                  (__v16sf)_mm512_add_round_ps((A), (B), (R)), \
+                                  (__v16sf)(__m512)(W));
 
 #define _mm512_maskz_add_round_ps(U, A, B, R) \
-  (__m512)__builtin_ia32_addps512_mask((__v16sf)(__m512)(A), \
-                                       (__v16sf)(__m512)(B), \
-                                       (__v16sf)_mm512_setzero_ps(), \
-                                       (__mmask16)(U), (int)(R))
+  (__m512)__builtin_ia32_selectps_512((__mmask16)(U), \
+                                  (__v16sf)_mm512_add_round_ps((A), (B), (R)), \
+                                  (__v16sf)_mm512_setzero_ps());
 
 static __inline__ __m128 __DEFAULT_FN_ATTRS
 _mm_mask_sub_ss(__m128 __W, __mmask8 __U,__m128 __A, __m128 __B) {
@@ -2195,40 +2187,32 @@ _mm512_maskz_sub_ps(__mmask16 __U, __m512 __A, __m512 __B) {
 }
 
 #define _mm512_sub_round_pd(A, B, R) \
-  (__m512d)__builtin_ia32_subpd512_mask((__v8df)(__m512d)(A), \
-                                        (__v8df)(__m512d)(B), \
-                                        (__v8df)_mm512_setzero_pd(), \
-                                        (__mmask8)-1, (int)(R))
+  (__m512d)__builtin_ia32_subpd512((__v8df)(__m512d)(A), \
+                                   (__v8df)(__m512d)(B), (int)(R))
 
 #define _mm512_mask_sub_round_pd(W, U, A, B, R) \
-  (__m512d)__builtin_ia32_subpd512_mask((__v8df)(__m512d)(A), \
-                                        (__v8df)(__m512d)(B), \
-                                        (__v8df)(__m512d)(W), (__mmask8)(U), \
-                                        (int)(R))
+  (__m512d)__builtin_ia32_selectpd_512((__mmask8)(U), \
+                                   (__v8df)_mm512_sub_round_pd((A), (B), (R)), \
+                                   (__v8df)(__m512d)(W));
 
 #define _mm512_maskz_sub_round_pd(U, A, B, R) \
-  (__m512d)__builtin_ia32_subpd512_mask((__v8df)(__m512d)(A), \
-                                        (__v8df)(__m512d)(B), \
-                                        (__v8df)_mm512_setzero_pd(), \
-                                        (__mmask8)(U), (int)(R))
+  (__m512d)__builtin_ia32_selectpd_512((__mmask8)(U), \
+                                   (__v8df)_mm512_sub_round_pd((A), (B), (R)), \
+                                   (__v8df)_mm512_setzero_pd());
 
 #define _mm512_sub_round_ps(A, B, R) \
-  (__m512)__builtin_ia32_subps512_mask((__v16sf)(__m512)(A), \
-                                       (__v16sf)(__m512)(B), \
-                                       (__v16sf)_mm512_setzero_ps(), \
-                                       (__mmask16)-1, (int)(R))
+  (__m512)__builtin_ia32_subps512((__v16sf)(__m512)(A), \
+                                  (__v16sf)(__m512)(B), (int)(R))
 
-#define _mm512_mask_sub_round_ps(W, U, A, B, R)  \
-  (__m512)__builtin_ia32_subps512_mask((__v16sf)(__m512)(A), \
-                                       (__v16sf)(__m512)(B), \
-                                       (__v16sf)(__m512)(W), (__mmask16)(U), \
-                                       (int)(R))
+#define _mm512_mask_sub_round_ps(W, U, A, B, R) \
+  (__m512)__builtin_ia32_selectps_512((__mmask16)(U), \
+                                  (__v16sf)_mm512_sub_round_ps((A), (B), (R)), \
+                                  (__v16sf)(__m512)(W));
 
-#define _mm512_maskz_sub_round_ps(U, A, B, R)  \
-  (__m512)__builtin_ia32_subps512_mask((__v16sf)(__m512)(A), \
-                                       (__v16sf)(__m512)(B), \
-                                       (__v16sf)_mm512_setzero_ps(), \
-                                       (__mmask16)(U), (int)(R))
+#define _mm512_maskz_sub_round_ps(U, A, B, R) \
+  (__m512)__builtin_ia32_selectps_512((__mmask16)(U), \
+                                  (__v16sf)_mm512_sub_round_ps((A), (B), (R)), \
+                                  (__v16sf)_mm512_setzero_ps());
 
 static __inline__ __m128 __DEFAULT_FN_ATTRS
 _mm_mask_mul_ss(__m128 __W, __mmask8 __U,__m128 __A, __m128 __B) {
@@ -2330,40 +2314,32 @@ _mm512_maskz_mul_ps(__mmask16 __U, __m512 __A, __m512 __B) {
 }
 
 #define _mm512_mul_round_pd(A, B, R) \
-  (__m512d)__builtin_ia32_mulpd512_mask((__v8df)(__m512d)(A), \
-                                        (__v8df)(__m512d)(B), \
-                                        (__v8df)_mm512_setzero_pd(), \
-                                        (__mmask8)-1, (int)(R))
+  (__m512d)__builtin_ia32_mulpd512((__v8df)(__m512d)(A), \
+                                   (__v8df)(__m512d)(B), (int)(R))
 
 #define _mm512_mask_mul_round_pd(W, U, A, B, R) \
-  (__m512d)__builtin_ia32_mulpd512_mask((__v8df)(__m512d)(A), \
-                                        (__v8df)(__m512d)(B), \
-                                        (__v8df)(__m512d)(W), (__mmask8)(U), \
-                                        (int)(R))
+  (__m512d)__builtin_ia32_selectpd_512((__mmask8)(U), \
+                                   (__v8df)_mm512_mul_round_pd((A), (B), (R)), \
+                                   (__v8df)(__m512d)(W));
 
 #define _mm512_maskz_mul_round_pd(U, A, B, R) \
-  (__m512d)__builtin_ia32_mulpd512_mask((__v8df)(__m512d)(A), \
-                                        (__v8df)(__m512d)(B), \
-                                        (__v8df)_mm512_setzero_pd(), \
-                                        (__mmask8)(U), (int)(R))
+  (__m512d)__builtin_ia32_selectpd_512((__mmask8)(U), \
+                                   (__v8df)_mm512_mul_round_pd((A), (B), (R)), \
+                                   (__v8df)_mm512_setzero_pd());
 
 #define _mm512_mul_round_ps(A, B, R) \
-  (__m512)__builtin_ia32_mulps512_mask((__v16sf)(__m512)(A), \
-                                       (__v16sf)(__m512)(B), \
-                                       (__v16sf)_mm512_setzero_ps(), \
-                                       (__mmask16)-1, (int)(R))
+  (__m512)__builtin_ia32_mulps512((__v16sf)(__m512)(A), \
+                                  (__v16sf)(__m512)(B), (int)(R))
 
-#define _mm512_mask_mul_round_ps(W, U, A, B, R)  \
-  (__m512)__builtin_ia32_mulps512_mask((__v16sf)(__m512)(A), \
-                                       (__v16sf)(__m512)(B), \
-                                       (__v16sf)(__m512)(W), (__mmask16)(U), \
-                                       (int)(R))
+#define _mm512_mask_mul_round_ps(W, U, A, B, R) \
+  (__m512)__builtin_ia32_selectps_512((__mmask16)(U), \
+                                  (__v16sf)_mm512_mul_round_ps((A), (B), (R)), \
+                                  (__v16sf)(__m512)(W));
 
-#define _mm512_maskz_mul_round_ps(U, A, B, R)  \
-  (__m512)__builtin_ia32_mulps512_mask((__v16sf)(__m512)(A), \
-                                       (__v16sf)(__m512)(B), \
-                                       (__v16sf)_mm512_setzero_ps(), \
-                                       (__mmask16)(U), (int)(R))
+#define _mm512_maskz_mul_round_ps(U, A, B, R) \
+  (__m512)__builtin_ia32_selectps_512((__mmask16)(U), \
+                                  (__v16sf)_mm512_mul_round_ps((A), (B), (R)), \
+                                  (__v16sf)_mm512_setzero_ps());
 
 static __inline__ __m128 __DEFAULT_FN_ATTRS
 _mm_mask_div_ss(__m128 __W, __mmask8 __U,__m128 __A, __m128 __B) {
@@ -2478,40 +2454,32 @@ _mm512_maskz_div_ps(__mmask16 __U, __m512 __A, __m512 __B) {
 }
 
 #define _mm512_div_round_pd(A, B, R) \
-  (__m512d)__builtin_ia32_divpd512_mask((__v8df)(__m512d)(A), \
-                                        (__v8df)(__m512d)(B), \
-                                        (__v8df)_mm512_setzero_pd(), \
-                                        (__mmask8)-1, (int)(R))
+  (__m512d)__builtin_ia32_divpd512((__v8df)(__m512d)(A), \
+                                   (__v8df)(__m512d)(B), (int)(R))
 
 #define _mm512_mask_div_round_pd(W, U, A, B, R) \
-  (__m512d)__builtin_ia32_divpd512_mask((__v8df)(__m512d)(A), \
-                                        (__v8df)(__m512d)(B), \
-                                        (__v8df)(__m512d)(W), (__mmask8)(U), \
-                                        (int)(R))
+  (__m512d)__builtin_ia32_selectpd_512((__mmask8)(U), \
+                                   (__v8df)_mm512_div_round_pd((A), (B), (R)), \
+                                   (__v8df)(__m512d)(W));
 
 #define _mm512_maskz_div_round_pd(U, A, B, R) \
-  (__m512d)__builtin_ia32_divpd512_mask((__v8df)(__m512d)(A), \
-                                        (__v8df)(__m512d)(B), \
-                                        (__v8df)_mm512_setzero_pd(), \
-                                        (__mmask8)(U), (int)(R))
+  (__m512d)__builtin_ia32_selectpd_512((__mmask8)(U), \
+                                   (__v8df)_mm512_div_round_pd((A), (B), (R)), \
+                                   (__v8df)_mm512_setzero_pd());
 
 #define _mm512_div_round_ps(A, B, R) \
-  (__m512)__builtin_ia32_divps512_mask((__v16sf)(__m512)(A), \
-                                       (__v16sf)(__m512)(B), \
-                                       (__v16sf)_mm512_setzero_ps(), \
-                                       (__mmask16)-1, (int)(R))
+  (__m512)__builtin_ia32_divps512((__v16sf)(__m512)(A), \
+                                  (__v16sf)(__m512)(B), (int)(R))
 
-#define _mm512_mask_div_round_ps(W, U, A, B, R)  \
-  (__m512)__builtin_ia32_divps512_mask((__v16sf)(__m512)(A), \
-                                       (__v16sf)(__m512)(B), \
-                                       (__v16sf)(__m512)(W), (__mmask16)(U), \
-                                       (int)(R))
+#define _mm512_mask_div_round_ps(W, U, A, B, R) \
+  (__m512)__builtin_ia32_selectps_512((__mmask16)(U), \
+                                  (__v16sf)_mm512_div_round_ps((A), (B), (R)), \
+                                  (__v16sf)(__m512)(W));
 
-#define _mm512_maskz_div_round_ps(U, A, B, R)  \
-  (__m512)__builtin_ia32_divps512_mask((__v16sf)(__m512)(A), \
-                                       (__v16sf)(__m512)(B), \
-                                       (__v16sf)_mm512_setzero_ps(), \
-                                       (__mmask16)(U), (int)(R))
+#define _mm512_maskz_div_round_ps(U, A, B, R) \
+  (__m512)__builtin_ia32_selectps_512((__mmask16)(U), \
+                                  (__v16sf)_mm512_div_round_ps((A), (B), (R)), \
+                                  (__v16sf)_mm512_setzero_ps());
 
 #define _mm512_roundscale_ps(A, B) \
   (__m512)__builtin_ia32_rndscaleps_mask((__v16sf)(__m512)(A), (int)(B), \

diff --git a/lib/Sema/SemaChecking.cpp b/lib/Sema/SemaChecking.cpp
index ae9c2f021fb4103c3cd65ab76bbe2809300cb1c8..bf1e3a8b4d86d928204b4b4780c1697313053891 100644 (file)
--- a/lib/Sema/SemaChecking.cpp
+++ b/lib/Sema/SemaChecking.cpp
@@ -2359,6 +2359,14 @@ bool Sema::CheckX86BuiltinRoundingOrSAE(unsigned BuiltinID, CallExpr *TheCall) {
     ArgNum = 1;
     HasRC = true;
     break;
+  case X86::BI__builtin_ia32_addpd512:
+  case X86::BI__builtin_ia32_addps512:
+  case X86::BI__builtin_ia32_divpd512:
+  case X86::BI__builtin_ia32_divps512:
+  case X86::BI__builtin_ia32_mulpd512:
+  case X86::BI__builtin_ia32_mulps512:
+  case X86::BI__builtin_ia32_subpd512:
+  case X86::BI__builtin_ia32_subps512:
   case X86::BI__builtin_ia32_cvtsi2sd64:
   case X86::BI__builtin_ia32_cvtsi2ss32:
   case X86::BI__builtin_ia32_cvtsi2ss64:
@@ -2384,14 +2392,6 @@ bool Sema::CheckX86BuiltinRoundingOrSAE(unsigned BuiltinID, CallExpr *TheCall) {
     ArgNum = 3;
     HasRC = true;
     break;
-  case X86::BI__builtin_ia32_addpd512_mask:
-  case X86::BI__builtin_ia32_addps512_mask:
-  case X86::BI__builtin_ia32_divpd512_mask:
-  case X86::BI__builtin_ia32_divps512_mask:
-  case X86::BI__builtin_ia32_mulpd512_mask:
-  case X86::BI__builtin_ia32_mulps512_mask:
-  case X86::BI__builtin_ia32_subpd512_mask:
-  case X86::BI__builtin_ia32_subps512_mask:
   case X86::BI__builtin_ia32_addss_round_mask:
   case X86::BI__builtin_ia32_addsd_round_mask:
   case X86::BI__builtin_ia32_divss_round_mask:

diff --git a/test/CodeGen/avx512f-builtins.c b/test/CodeGen/avx512f-builtins.c
index dab1c419128ff3a4091c8348b003494fbc2567c7..89be9ae0d79ad77a1eba35189e456fce0a80046b 100644 (file)
--- a/test/CodeGen/avx512f-builtins.c
+++ b/test/CodeGen/avx512f-builtins.c
@@ -2229,17 +2229,19 @@ __m512i test_mm512_mask_mullox_epi64 (__m512i __W, __mmask8 __U, __m512i __A, __
 
 __m512d test_mm512_add_round_pd(__m512d __A, __m512d __B) {
   // CHECK-LABEL: @test_mm512_add_round_pd
-  // CHECK: @llvm.x86.avx512.mask.add.pd.512
+  // CHECK: @llvm.x86.avx512.add.pd.512
   return _mm512_add_round_pd(__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512d test_mm512_mask_add_round_pd(__m512d __W, __mmask8 __U, __m512d __A, __m512d __B) {
   // CHECK-LABEL: @test_mm512_mask_add_round_pd
-  // CHECK: @llvm.x86.avx512.mask.add.pd.512
+  // CHECK: @llvm.x86.avx512.add.pd.512
+  // CHECK: select <8 x i1> %{{.*}}, <8 x double> %{{.*}}, <8 x double> %{{.*}}
   return _mm512_mask_add_round_pd(__W,__U,__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512d test_mm512_maskz_add_round_pd(__mmask8 __U, __m512d __A, __m512d __B) {
   // CHECK-LABEL: @test_mm512_maskz_add_round_pd
-  // CHECK: @llvm.x86.avx512.mask.add.pd.512
+  // CHECK: @llvm.x86.avx512.add.pd.512
+  // CHECK: select <8 x i1> %{{.*}}, <8 x double> %{{.*}}, <8 x double> %{{.*}}
   return _mm512_maskz_add_round_pd(__U,__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512d test_mm512_mask_add_pd(__m512d __W, __mmask8 __U, __m512d __A, __m512d __B) {
@@ -2256,17 +2258,19 @@ __m512d test_mm512_maskz_add_pd(__mmask8 __U, __m512d __A, __m512d __B) {
 }
 __m512 test_mm512_add_round_ps(__m512 __A, __m512 __B) {
   // CHECK-LABEL: @test_mm512_add_round_ps
-  // CHECK: @llvm.x86.avx512.mask.add.ps.512
+  // CHECK: @llvm.x86.avx512.add.ps.512
   return _mm512_add_round_ps(__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512 test_mm512_mask_add_round_ps(__m512 __W, __mmask16 __U, __m512 __A, __m512 __B) {
   // CHECK-LABEL: @test_mm512_mask_add_round_ps
-  // CHECK: @llvm.x86.avx512.mask.add.ps.512
+  // CHECK: @llvm.x86.avx512.add.ps.512
+  // CHECK: select <16 x i1> %{{.*}}, <16 x float> %{{.*}}, <16 x float> %{{.*}}
   return _mm512_mask_add_round_ps(__W,__U,__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512 test_mm512_maskz_add_round_ps(__mmask16 __U, __m512 __A, __m512 __B) {
   // CHECK-LABEL: @test_mm512_maskz_add_round_ps
-  // CHECK: @llvm.x86.avx512.mask.add.ps.512
+  // CHECK: @llvm.x86.avx512.add.ps.512
+  // CHECK: select <16 x i1> %{{.*}}, <16 x float> %{{.*}}, <16 x float> %{{.*}}
   return _mm512_maskz_add_round_ps(__U,__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512 test_mm512_mask_add_ps(__m512 __W, __mmask16 __U, __m512 __A, __m512 __B) {
@@ -2333,17 +2337,19 @@ __m128d test_mm_maskz_add_sd(__mmask8 __U, __m128d __A, __m128d __B) {
 }
 __m512d test_mm512_sub_round_pd(__m512d __A, __m512d __B) {
   // CHECK-LABEL: @test_mm512_sub_round_pd
-  // CHECK: @llvm.x86.avx512.mask.sub.pd.512
+  // CHECK: @llvm.x86.avx512.sub.pd.512
   return _mm512_sub_round_pd(__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512d test_mm512_mask_sub_round_pd(__m512d __W, __mmask8 __U, __m512d __A, __m512d __B) {
   // CHECK-LABEL: @test_mm512_mask_sub_round_pd
-  // CHECK: @llvm.x86.avx512.mask.sub.pd.512
+  // CHECK: @llvm.x86.avx512.sub.pd.512
+  // CHECK: select <8 x i1> %{{.*}}, <8 x double> %{{.*}}, <8 x double> %{{.*}}
   return _mm512_mask_sub_round_pd(__W,__U,__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512d test_mm512_maskz_sub_round_pd(__mmask8 __U, __m512d __A, __m512d __B) {
   // CHECK-LABEL: @test_mm512_maskz_sub_round_pd
-  // CHECK: @llvm.x86.avx512.mask.sub.pd.512
+  // CHECK: @llvm.x86.avx512.sub.pd.512
+  // CHECK: select <8 x i1> %{{.*}}, <8 x double> %{{.*}}, <8 x double> %{{.*}}
   return _mm512_maskz_sub_round_pd(__U,__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512d test_mm512_mask_sub_pd(__m512d __W, __mmask8 __U, __m512d __A, __m512d __B) {
@@ -2360,17 +2366,19 @@ __m512d test_mm512_maskz_sub_pd(__mmask8 __U, __m512d __A, __m512d __B) {
 }
 __m512 test_mm512_sub_round_ps(__m512 __A, __m512 __B) {
   // CHECK-LABEL: @test_mm512_sub_round_ps
-  // CHECK: @llvm.x86.avx512.mask.sub.ps.512
+  // CHECK: @llvm.x86.avx512.sub.ps.512
   return _mm512_sub_round_ps(__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512 test_mm512_mask_sub_round_ps(__m512 __W, __mmask16 __U, __m512 __A, __m512 __B) {
   // CHECK-LABEL: @test_mm512_mask_sub_round_ps
-  // CHECK: @llvm.x86.avx512.mask.sub.ps.512
+  // CHECK: @llvm.x86.avx512.sub.ps.512
+  // CHECK: select <16 x i1> %{{.*}}, <16 x float> %{{.*}}, <16 x float> %{{.*}}
   return _mm512_mask_sub_round_ps(__W,__U,__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512 test_mm512_maskz_sub_round_ps(__mmask16 __U, __m512 __A, __m512 __B) {
   // CHECK-LABEL: @test_mm512_maskz_sub_round_ps
-  // CHECK: @llvm.x86.avx512.mask.sub.ps.512
+  // CHECK: @llvm.x86.avx512.sub.ps.512
+  // CHECK: select <16 x i1> %{{.*}}, <16 x float> %{{.*}}, <16 x float> %{{.*}}
   return _mm512_maskz_sub_round_ps(__U,__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512 test_mm512_mask_sub_ps(__m512 __W, __mmask16 __U, __m512 __A, __m512 __B) {
@@ -2437,17 +2445,19 @@ __m128d test_mm_maskz_sub_sd(__mmask8 __U, __m128d __A, __m128d __B) {
 }
 __m512d test_mm512_mul_round_pd(__m512d __A, __m512d __B) {
   // CHECK-LABEL: @test_mm512_mul_round_pd
-  // CHECK: @llvm.x86.avx512.mask.mul.pd.512
+  // CHECK: @llvm.x86.avx512.mul.pd.512
   return _mm512_mul_round_pd(__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512d test_mm512_mask_mul_round_pd(__m512d __W, __mmask8 __U, __m512d __A, __m512d __B) {
   // CHECK-LABEL: @test_mm512_mask_mul_round_pd
-  // CHECK: @llvm.x86.avx512.mask.mul.pd.512
+  // CHECK: @llvm.x86.avx512.mul.pd.512
+  // CHECK: select <8 x i1> %{{.*}}, <8 x double> %{{.*}}, <8 x double> %{{.*}}
   return _mm512_mask_mul_round_pd(__W,__U,__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512d test_mm512_maskz_mul_round_pd(__mmask8 __U, __m512d __A, __m512d __B) {
   // CHECK-LABEL: @test_mm512_maskz_mul_round_pd
-  // CHECK: @llvm.x86.avx512.mask.mul.pd.512
+  // CHECK: @llvm.x86.avx512.mul.pd.512
+  // CHECK: select <8 x i1> %{{.*}}, <8 x double> %{{.*}}, <8 x double> %{{.*}}
   return _mm512_maskz_mul_round_pd(__U,__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512d test_mm512_mask_mul_pd(__m512d __W, __mmask8 __U, __m512d __A, __m512d __B) {
@@ -2464,17 +2474,19 @@ __m512d test_mm512_maskz_mul_pd(__mmask8 __U, __m512d __A, __m512d __B) {
 }
 __m512 test_mm512_mul_round_ps(__m512 __A, __m512 __B) {
   // CHECK-LABEL: @test_mm512_mul_round_ps
-  // CHECK: @llvm.x86.avx512.mask.mul.ps.512
+  // CHECK: @llvm.x86.avx512.mul.ps.512
   return _mm512_mul_round_ps(__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512 test_mm512_mask_mul_round_ps(__m512 __W, __mmask16 __U, __m512 __A, __m512 __B) {
   // CHECK-LABEL: @test_mm512_mask_mul_round_ps
-  // CHECK: @llvm.x86.avx512.mask.mul.ps.512
+  // CHECK: @llvm.x86.avx512.mul.ps.512
+  // CHECK: select <16 x i1> %{{.*}}, <16 x float> %{{.*}}, <16 x float> %{{.*}}
   return _mm512_mask_mul_round_ps(__W,__U,__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512 test_mm512_maskz_mul_round_ps(__mmask16 __U, __m512 __A, __m512 __B) {
   // CHECK-LABEL: @test_mm512_maskz_mul_round_ps
-  // CHECK: @llvm.x86.avx512.mask.mul.ps.512
+  // CHECK: @llvm.x86.avx512.mul.ps.512
+  // CHECK: select <16 x i1> %{{.*}}, <16 x float> %{{.*}}, <16 x float> %{{.*}}
   return _mm512_maskz_mul_round_ps(__U,__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512 test_mm512_mask_mul_ps(__m512 __W, __mmask16 __U, __m512 __A, __m512 __B) {
@@ -2541,17 +2553,19 @@ __m128d test_mm_maskz_mul_sd(__mmask8 __U, __m128d __A, __m128d __B) {
 }
 __m512d test_mm512_div_round_pd(__m512d __A, __m512d __B) {
   // CHECK-LABEL: @test_mm512_div_round_pd
-  // CHECK: @llvm.x86.avx512.mask.div.pd.512
+  // CHECK: @llvm.x86.avx512.div.pd.512
   return _mm512_div_round_pd(__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512d test_mm512_mask_div_round_pd(__m512d __W, __mmask8 __U, __m512d __A, __m512d __B) {
   // CHECK-LABEL: @test_mm512_mask_div_round_pd
-  // CHECK: @llvm.x86.avx512.mask.div.pd.512
+  // CHECK: @llvm.x86.avx512.div.pd.512
+  // CHECK: select <8 x i1> %{{.*}}, <8 x double> %{{.*}}, <8 x double> %{{.*}}
   return _mm512_mask_div_round_pd(__W,__U,__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512d test_mm512_maskz_div_round_pd(__mmask8 __U, __m512d __A, __m512d __B) {
   // CHECK-LABEL: @test_mm512_maskz_div_round_pd
-  // CHECK: @llvm.x86.avx512.mask.div.pd.512
+  // CHECK: @llvm.x86.avx512.div.pd.512
+  // CHECK: select <8 x i1> %{{.*}}, <8 x double> %{{.*}}, <8 x double> %{{.*}}
   return _mm512_maskz_div_round_pd(__U,__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512d test_mm512_div_pd(__m512d __a, __m512d __b) {
@@ -2573,17 +2587,19 @@ __m512d test_mm512_maskz_div_pd(__mmask8 __U, __m512d __A, __m512d __B) {
 }
 __m512 test_mm512_div_round_ps(__m512 __A, __m512 __B) {
   // CHECK-LABEL: @test_mm512_div_round_ps
-  // CHECK: @llvm.x86.avx512.mask.div.ps.512
+  // CHECK: @llvm.x86.avx512.div.ps.512
   return _mm512_div_round_ps(__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512 test_mm512_mask_div_round_ps(__m512 __W, __mmask16 __U, __m512 __A, __m512 __B) {
   // CHECK-LABEL: @test_mm512_mask_div_round_ps
-  // CHECK: @llvm.x86.avx512.mask.div.ps.512
+  // CHECK: @llvm.x86.avx512.div.ps.512
+  // CHECK: select <16 x i1> %{{.*}}, <16 x float> %{{.*}}, <16 x float> %{{.*}}
   return _mm512_mask_div_round_ps(__W,__U,__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512 test_mm512_maskz_div_round_ps(__mmask16 __U, __m512 __A, __m512 __B) {
   // CHECK-LABEL: @test_mm512_maskz_div_round_ps
-  // CHECK: @llvm.x86.avx512.mask.div.ps.512
+  // CHECK: @llvm.x86.avx512.div.ps.512
+  // CHECK: select <16 x i1> %{{.*}}, <16 x float> %{{.*}}, <16 x float> %{{.*}}
   return _mm512_maskz_div_round_ps(__U,__A,__B,_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); 
 }
 __m512 test_mm512_div_ps(__m512 __A, __m512 __B) {