r289345 - [AVX-512] Remove masking from 512-bit pshufb builtin. The backend now has a version without masking so wrap it with select.

[Craig Topper via cfe-commits]

Author: ctopper
Date: Sat Dec 10 17:09:52 2016
New Revision: 289345

URL: http://llvm.org/viewvc/llvm-project?rev=289345&view=rev
Log:
[AVX-512] Remove masking from 512-bit pshufb builtin. The backend now has a version without masking so wrap it with select.

This will allow the backend to constant fold these to generic shuffle vectors like 128-bit and 256-bit without having to working about handling masking.

Modified:
    cfe/trunk/include/clang/Basic/BuiltinsX86.def
    cfe/trunk/lib/Headers/avx512bwintrin.h
    cfe/trunk/test/CodeGen/avx512bw-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=289345&r1=289344&r2=289345&view=diff
==============================================================================
--- cfe/trunk/include/clang/Basic/BuiltinsX86.def (original)
+++ cfe/trunk/include/clang/Basic/BuiltinsX86.def Sat Dec 10 17:09:52 2016
@@ -1086,7 +1086,7 @@ TARGET_BUILTIN(__builtin_ia32_pminsb512_
 TARGET_BUILTIN(__builtin_ia32_pminsw512_mask, "V32sV32sV32sV32sUi", "", "avx512bw")
 TARGET_BUILTIN(__builtin_ia32_pminub512_mask, "V64cV64cV64cV64cULLi", "", "avx512bw")
 TARGET_BUILTIN(__builtin_ia32_pminuw512_mask, "V32sV32sV32sV32sUi", "", "avx512bw")
-TARGET_BUILTIN(__builtin_ia32_pshufb512_mask, "V64cV64cV64cV64cULLi", "", "avx512bw")
+TARGET_BUILTIN(__builtin_ia32_pshufb512, "V64cV64cV64c", "", "avx512bw")
 TARGET_BUILTIN(__builtin_ia32_psubsb512_mask, "V64cV64cV64cV64cULLi", "", "avx512bw")
 TARGET_BUILTIN(__builtin_ia32_psubsw512_mask, "V32sV32sV32sV32sUi", "", "avx512bw")
 TARGET_BUILTIN(__builtin_ia32_psubusb512_mask, "V64cV64cV64cV64cULLi", "", "avx512bw")

Modified: cfe/trunk/lib/Headers/avx512bwintrin.h
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Headers/avx512bwintrin.h?rev=289345&r1=289344&r2=289345&view=diff
==============================================================================
--- cfe/trunk/lib/Headers/avx512bwintrin.h (original)
+++ cfe/trunk/lib/Headers/avx512bwintrin.h Sat Dec 10 17:09:52 2016
@@ -1008,31 +1008,25 @@ _mm512_mask_min_epu16 (__m512i __W, __mm
 }
 
 static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_shuffle_epi8 (__m512i __A, __m512i __B)
+_mm512_shuffle_epi8(__m512i __A, __m512i __B)
 {
-  return (__m512i) __builtin_ia32_pshufb512_mask ((__v64qi) __A,
-              (__v64qi) __B,
-              (__v64qi) _mm512_setzero_qi(),
-              (__mmask64) -1);
+  return (__m512i)__builtin_ia32_pshufb512((__v64qi)__A,(__v64qi)__B);
 }
 
 static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_mask_shuffle_epi8 (__m512i __W, __mmask64 __U, __m512i __A,
-        __m512i __B)
+_mm512_mask_shuffle_epi8(__m512i __W, __mmask64 __U, __m512i __A, __m512i __B)
 {
-  return (__m512i) __builtin_ia32_pshufb512_mask ((__v64qi) __A,
-              (__v64qi) __B,
-              (__v64qi) __W,
-              (__mmask64) __U);
+  return (__m512i)__builtin_ia32_selectb_512((__mmask64)__U,
+                                         (__v64qi)_mm512_shuffle_epi8(__A, __B),
+                                         (__v64qi)__W);
 }
 
 static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_maskz_shuffle_epi8 (__mmask64 __U, __m512i __A, __m512i __B)
+_mm512_maskz_shuffle_epi8(__mmask64 __U, __m512i __A, __m512i __B)
 {
-  return (__m512i) __builtin_ia32_pshufb512_mask ((__v64qi) __A,
-              (__v64qi) __B,
-              (__v64qi) _mm512_setzero_qi(),
-              (__mmask64) __U);
+  return (__m512i)__builtin_ia32_selectb_512((__mmask64)__U,
+                                         (__v64qi)_mm512_shuffle_epi8(__A, __B),
+                                         (__v64qi)_mm512_setzero_qi());
 }
 
 static __inline__ __m512i __DEFAULT_FN_ATTRS

Modified: cfe/trunk/test/CodeGen/avx512bw-builtins.c
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGen/avx512bw-builtins.c?rev=289345&r1=289344&r2=289345&view=diff
==============================================================================
--- cfe/trunk/test/CodeGen/avx512bw-builtins.c (original)
+++ cfe/trunk/test/CodeGen/avx512bw-builtins.c Sat Dec 10 17:09:52 2016
@@ -820,17 +820,19 @@ __m512i test_mm512_mask_min_epu16(__m512
 }
 __m512i test_mm512_shuffle_epi8(__m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_shuffle_epi8
-  // CHECK: @llvm.x86.avx512.mask.pshuf.b.512
+  // CHECK: @llvm.x86.avx512.pshuf.b.512
   return _mm512_shuffle_epi8(__A,__B); 
 }
 __m512i test_mm512_mask_shuffle_epi8(__m512i __W, __mmask64 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_mask_shuffle_epi8
-  // CHECK: @llvm.x86.avx512.mask.pshuf.b.512
+  // CHECK: @llvm.x86.avx512.pshuf.b.512
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i8> %{{.*}}, <64 x i8> %{{.*}}
   return _mm512_mask_shuffle_epi8(__W,__U,__A,__B); 
 }
 __m512i test_mm512_maskz_shuffle_epi8(__mmask64 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_maskz_shuffle_epi8
-  // CHECK: @llvm.x86.avx512.mask.pshuf.b.512
+  // CHECK: @llvm.x86.avx512.pshuf.b.512
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i8> %{{.*}}, <64 x i8> %{{.*}}
   return _mm512_maskz_shuffle_epi8(__U,__A,__B); 
 }
 __m512i test_mm512_subs_epi8(__m512i __A, __m512i __B) {