r284925 - [AVX-512] Remove masked 128/256-bit pshufb builtins. Replace with a select and the older unmaksed builtins.

Craig Topper via cfe-commits cfe-commits at lists.llvm.org
Sat Oct 22 14:24:42 PDT 2016


Author: ctopper
Date: Sat Oct 22 16:24:42 2016
New Revision: 284925

URL: http://llvm.org/viewvc/llvm-project?rev=284925&view=rev
Log:
[AVX-512] Remove masked 128/256-bit pshufb builtins. Replace with a select and the older unmaksed builtins.

Modified:
    cfe/trunk/include/clang/Basic/BuiltinsX86.def
    cfe/trunk/lib/Headers/avx512vlbwintrin.h
    cfe/trunk/test/CodeGen/avx512vlbw-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=284925&r1=284924&r2=284925&view=diff
==============================================================================
--- cfe/trunk/include/clang/Basic/BuiltinsX86.def (original)
+++ cfe/trunk/include/clang/Basic/BuiltinsX86.def Sat Oct 22 16:24:42 2016
@@ -1156,8 +1156,6 @@ TARGET_BUILTIN(__builtin_ia32_pminub128_
 TARGET_BUILTIN(__builtin_ia32_pminub256_mask, "V32cV32cV32cV32cUi", "", "avx512vl,avx512bw")
 TARGET_BUILTIN(__builtin_ia32_pminuw128_mask, "V8sV8sV8sV8sUc", "", "avx512vl,avx512bw")
 TARGET_BUILTIN(__builtin_ia32_pminuw256_mask, "V16sV16sV16sV16sUs", "", "avx512vl,avx512bw")
-TARGET_BUILTIN(__builtin_ia32_pshufb128_mask, "V16cV16cV16cV16cUs", "", "avx512vl,avx512bw")
-TARGET_BUILTIN(__builtin_ia32_pshufb256_mask, "V32cV32cV32cV32cUi", "", "avx512vl,avx512bw")
 TARGET_BUILTIN(__builtin_ia32_psubsb128_mask, "V16cV16cV16cV16cUs", "", "avx512vl,avx512bw")
 TARGET_BUILTIN(__builtin_ia32_psubsb256_mask, "V32cV32cV32cV32cUi", "", "avx512vl,avx512bw")
 TARGET_BUILTIN(__builtin_ia32_psubsw128_mask, "V8sV8sV8sV8sUc", "", "avx512vl,avx512bw")

Modified: cfe/trunk/lib/Headers/avx512vlbwintrin.h
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Headers/avx512vlbwintrin.h?rev=284925&r1=284924&r2=284925&view=diff
==============================================================================
--- cfe/trunk/lib/Headers/avx512vlbwintrin.h (original)
+++ cfe/trunk/lib/Headers/avx512vlbwintrin.h Sat Oct 22 16:24:42 2016
@@ -1531,41 +1531,35 @@ _mm256_mask_min_epu16 (__m256i __W, __mm
 }
 
 static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_mask_shuffle_epi8 (__m128i __W, __mmask16 __U, __m128i __A,
-           __m128i __B)
+_mm_mask_shuffle_epi8(__m128i __W, __mmask16 __U, __m128i __A, __m128i __B)
 {
-  return (__m128i) __builtin_ia32_pshufb128_mask ((__v16qi) __A,
-               (__v16qi) __B,
-               (__v16qi) __W,
-               (__mmask16) __U);
+  return (__m128i)__builtin_ia32_selectb_128((__mmask16)__U,
+                                            (__v16qi)_mm_shuffle_epi8(__A, __B),
+                                            (__v16qi)__W);
 }
 
 static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_maskz_shuffle_epi8 (__mmask16 __U, __m128i __A, __m128i __B)
+_mm_maskz_shuffle_epi8(__mmask16 __U, __m128i __A, __m128i __B)
 {
-  return (__m128i) __builtin_ia32_pshufb128_mask ((__v16qi) __A,
-               (__v16qi) __B,
-               (__v16qi) _mm_setzero_si128 (),
-               (__mmask16) __U);
+  return (__m128i)__builtin_ia32_selectb_128((__mmask16)__U,
+                                            (__v16qi)_mm_shuffle_epi8(__A, __B),
+                                            (__v16qi)_mm_setzero_si128());
 }
 
 static __inline__ __m256i __DEFAULT_FN_ATTRS
-_mm256_mask_shuffle_epi8 (__m256i __W, __mmask32 __U, __m256i __A,
-        __m256i __B)
+_mm256_mask_shuffle_epi8(__m256i __W, __mmask32 __U, __m256i __A, __m256i __B)
 {
-  return (__m256i) __builtin_ia32_pshufb256_mask ((__v32qi) __A,
-               (__v32qi) __B,
-               (__v32qi) __W,
-               (__mmask32) __U);
+  return (__m256i)__builtin_ia32_selectb_256((__mmask32)__U,
+                                         (__v32qi)_mm256_shuffle_epi8(__A, __B),
+                                         (__v32qi)__W);
 }
 
 static __inline__ __m256i __DEFAULT_FN_ATTRS
-_mm256_maskz_shuffle_epi8 (__mmask32 __U, __m256i __A, __m256i __B)
+_mm256_maskz_shuffle_epi8(__mmask32 __U, __m256i __A, __m256i __B)
 {
-  return (__m256i) __builtin_ia32_pshufb256_mask ((__v32qi) __A,
-               (__v32qi) __B,
-               (__v32qi) _mm256_setzero_si256 (),
-               (__mmask32) __U);
+  return (__m256i)__builtin_ia32_selectb_256((__mmask32)__U,
+                                         (__v32qi)_mm256_shuffle_epi8(__A, __B),
+                                         (__v32qi)_mm256_setzero_si256());
 }
 
 static __inline__ __m128i __DEFAULT_FN_ATTRS

Modified: cfe/trunk/test/CodeGen/avx512vlbw-builtins.c
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGen/avx512vlbw-builtins.c?rev=284925&r1=284924&r2=284925&view=diff
==============================================================================
--- cfe/trunk/test/CodeGen/avx512vlbw-builtins.c (original)
+++ cfe/trunk/test/CodeGen/avx512vlbw-builtins.c Sat Oct 22 16:24:42 2016
@@ -1323,22 +1323,26 @@ __m256i test_mm256_mask_min_epu16(__m256
 }
 __m128i test_mm_mask_shuffle_epi8(__m128i __W, __mmask16 __U, __m128i __A,           __m128i __B) {
   // CHECK-LABEL: @test_mm_mask_shuffle_epi8
-  // CHECK: @llvm.x86.avx512.mask.pshuf.b.128
+  // CHECK: @llvm.x86.ssse3.pshuf.b
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i8> %{{.*}}, <16 x i8> %{{.*}}
   return _mm_mask_shuffle_epi8(__W,__U,__A,__B); 
 }
 __m128i test_mm_maskz_shuffle_epi8(__mmask16 __U, __m128i __A, __m128i __B) {
   // CHECK-LABEL: @test_mm_maskz_shuffle_epi8
-  // CHECK: @llvm.x86.avx512.mask.pshuf.b.128
+  // CHECK: @llvm.x86.ssse3.pshuf.b
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i8> %{{.*}}, <16 x i8> %{{.*}}
   return _mm_maskz_shuffle_epi8(__U,__A,__B); 
 }
 __m256i test_mm256_mask_shuffle_epi8(__m256i __W, __mmask32 __U, __m256i __A,        __m256i __B) {
   // CHECK-LABEL: @test_mm256_mask_shuffle_epi8
-  // CHECK: @llvm.x86.avx512.mask.pshuf.b.256
+  // CHECK: @llvm.x86.avx2.pshuf.b
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i8> %{{.*}}, <32 x i8> %{{.*}}
   return _mm256_mask_shuffle_epi8(__W,__U,__A,__B); 
 }
 __m256i test_mm256_maskz_shuffle_epi8(__mmask32 __U, __m256i __A, __m256i __B) {
   // CHECK-LABEL: @test_mm256_maskz_shuffle_epi8
-  // CHECK: @llvm.x86.avx512.mask.pshuf.b.256
+  // CHECK: @llvm.x86.avx2.pshuf.b
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i8> %{{.*}}, <32 x i8> %{{.*}}
   return _mm256_maskz_shuffle_epi8(__U,__A,__B); 
 }
 __m128i test_mm_mask_subs_epi8(__m128i __W, __mmask16 __U, __m128i __A,        __m128i __B) {




More information about the cfe-commits mailing list