[AVX-512] Remove masked 128/256-bit palignr builtins. We can just use a select in...

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@284920 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/include/clang/Basic/BuiltinsX86.def b/include/clang/Basic/BuiltinsX86.def
index 8e3fa3d74bb1bfca519d6eed707b0a9b6275161e..af520105d3d267f2267cd41b6563dfa6c672f171 100644 (file)
--- a/include/clang/Basic/BuiltinsX86.def
+++ b/include/clang/Basic/BuiltinsX86.def
@@ -1961,8 +1961,6 @@ TARGET_BUILTIN(__builtin_ia32_kxnorhi, "UsUsUs","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_kxorhi, "UsUsUs","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_movntdqa512, "V8LLiV8LLi*","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_palignr512_mask, "V64cV64cV64cIiV64cULLi","","avx512bw")
-TARGET_BUILTIN(__builtin_ia32_palignr128_mask, "V16cV16cV16cIiV16cUs","","avx512bw,avx512vl")
-TARGET_BUILTIN(__builtin_ia32_palignr256_mask, "V32cV32cV32cIiV32cUi","","avx512bw,avx512vl")
 TARGET_BUILTIN(__builtin_ia32_dbpsadbw128_mask, "V8sV16cV16cIiV8sUc","","avx512bw,avx512vl")
 TARGET_BUILTIN(__builtin_ia32_dbpsadbw256_mask, "V16sV32cV32cIiV16sUs","","avx512bw,avx512vl")
 TARGET_BUILTIN(__builtin_ia32_dbpsadbw512_mask, "V32sV64cV64cIiV32sUi","","avx512bw")

diff --git a/lib/CodeGen/CGBuiltin.cpp b/lib/CodeGen/CGBuiltin.cpp
index 8bb9a465864306de1356e1629be232c70f56916b..d78744246cc1a9d666c15b07d12263b1c07a4886 100644 (file)
--- a/lib/CodeGen/CGBuiltin.cpp
+++ b/lib/CodeGen/CGBuiltin.cpp
@@ -7346,8 +7346,6 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
   }
   case X86::BI__builtin_ia32_palignr128:
   case X86::BI__builtin_ia32_palignr256:
-  case X86::BI__builtin_ia32_palignr128_mask:
-  case X86::BI__builtin_ia32_palignr256_mask:
   case X86::BI__builtin_ia32_palignr512_mask: {
     unsigned ShiftVal = cast<llvm::ConstantInt>(Ops[2])->getZExtValue();
 

diff --git a/lib/Headers/avx512vlbwintrin.h b/lib/Headers/avx512vlbwintrin.h
index ff4528500db73d53c4ed1e1eece7c730ddcdf6fe..b6c2c9a9968a7d0041037feae386b5bbd117cbb8 100644 (file)
--- a/lib/Headers/avx512vlbwintrin.h
+++ b/lib/Headers/avx512vlbwintrin.h
@@ -3313,28 +3313,24 @@ _mm256_mask_permutexvar_epi16 (__m256i __W, __mmask16 __M, __m256i __A,
 }
 
 #define _mm_mask_alignr_epi8(W, U, A, B, N) __extension__ ({ \
-  (__m128i)__builtin_ia32_palignr128_mask((__v16qi)(__m128i)(A), \
-                                          (__v16qi)(__m128i)(B), (int)(N), \
-                                          (__v16qi)(__m128i)(W), \
-                                          (__mmask16)(U)); })
+  (__m128i)__builtin_ia32_selectb_128((__mmask16)(U), \
+                                      _mm_alignr_epi8((A), (B), (int)(N)), \
+                                      (__v16qi)(__m128i)(W)); })
 
 #define _mm_maskz_alignr_epi8(U, A, B, N) __extension__ ({ \
-  (__m128i)__builtin_ia32_palignr128_mask((__v16qi)(__m128i)(A), \
-                                          (__v16qi)(__m128i)(B), (int)(N), \
-                                          (__v16qi)_mm_setzero_si128(), \
-                                          (__mmask16)(U)); })
+  (__m128i)__builtin_ia32_selectb_128((__mmask16)(U), \
+                                      _mm_alignr_epi8((A), (B), (int)(N)), \
+                                      (__v16qi)_mm_setzero_si128()); })
 
 #define _mm256_mask_alignr_epi8(W, U, A, B, N) __extension__ ({ \
-  (__m256i)__builtin_ia32_palignr256_mask((__v32qi)(__m256i)(A), \
-                                          (__v32qi)(__m256i)(B), (int)(N), \
-                                          (__v32qi)(__m256i)(W), \
-                                          (__mmask32)(U)); })
+  (__m256i)__builtin_ia32_selectb_256((__mmask32)(U), \
+                                      _mm256_alignr_epi8((A), (B), (int)(N)), \
+                                      (__v32qi)(__m256i)(W)); })
 
 #define _mm256_maskz_alignr_epi8(U, A, B, N) __extension__ ({ \
-  (__m256i)__builtin_ia32_palignr256_mask((__v32qi)(__m256i)(A), \
-                                          (__v32qi)(__m256i)(B), (int)(N), \
-                                          (__v32qi)_mm256_setzero_si256(), \
-                                          (__mmask32)(U)); })
+  (__m256i)__builtin_ia32_selectb_256((__mmask32)(U), \
+                                      _mm256_alignr_epi8((A), (B), (int)(N)), \
+                                      (__v32qi)_mm256_setzero_si256()); })
 
 #define _mm_dbsad_epu8(A, B, imm) __extension__ ({ \
   (__m128i)__builtin_ia32_dbpsadbw128_mask((__v16qi)(__m128i)(A), \

diff --git a/lib/Sema/SemaChecking.cpp b/lib/Sema/SemaChecking.cpp
index 1b0d7f38859105f2624007d2924a3efbd0beeb0b..501f93ed6292fa41abb67c68a240ca8b73747905 100644 (file)
--- a/lib/Sema/SemaChecking.cpp
+++ b/lib/Sema/SemaChecking.cpp
@@ -2167,8 +2167,6 @@ bool Sema::CheckX86BuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
     break;
   case X86::BI__builtin_ia32_palignr128:
   case X86::BI__builtin_ia32_palignr256:
-  case X86::BI__builtin_ia32_palignr128_mask:
-  case X86::BI__builtin_ia32_palignr256_mask:
   case X86::BI__builtin_ia32_palignr512_mask:
   case X86::BI__builtin_ia32_alignq512_mask:
   case X86::BI__builtin_ia32_alignd512_mask: