[X86] Add kshift intrinsics to match gcc and icc.

This adds the following intrinsics:
_kshiftli_mask8
_kshiftli_mask16
_kshiftli_mask32
_kshiftli_mask64
_kshiftri_mask8
_kshiftri_mask16
_kshiftri_mask32
_kshiftri_mask64

llvm-svn: 341234

5 files changed, 76 insertions, 0 deletions

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 7bb1a652d68..50e9b5ed099 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -9929,6 +9929,50 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
                                             "psrldq");
     return Builder.CreateBitCast(SV, ResultType, "cast");
   }
+  case X86::BI__builtin_ia32_kshiftliqi:
+  case X86::BI__builtin_ia32_kshiftlihi:
+  case X86::BI__builtin_ia32_kshiftlisi:
+  case X86::BI__builtin_ia32_kshiftlidi: {
+    unsigned ShiftVal = cast<llvm::ConstantInt>(Ops[1])->getZExtValue() & 0xff;
+    unsigned NumElts = Ops[0]->getType()->getIntegerBitWidth();
+
+    if (ShiftVal >= NumElts)
+      return llvm::Constant::getNullValue(Ops[0]->getType());
+
+    Value *In = getMaskVecValue(*this, Ops[0], NumElts);
+
+    uint32_t Indices[64];
+    for (unsigned i = 0; i != NumElts; ++i)
+      Indices[i] = NumElts + i - ShiftVal;
+
+    Value *Zero = llvm::Constant::getNullValue(In->getType());
+    Value *SV = Builder.CreateShuffleVector(Zero, In,
+                                            makeArrayRef(Indices, NumElts),
+                                            "kshiftl");
+    return Builder.CreateBitCast(SV, Ops[0]->getType());
+  }
+  case X86::BI__builtin_ia32_kshiftriqi:
+  case X86::BI__builtin_ia32_kshiftrihi:
+  case X86::BI__builtin_ia32_kshiftrisi:
+  case X86::BI__builtin_ia32_kshiftridi: {
+    unsigned ShiftVal = cast<llvm::ConstantInt>(Ops[1])->getZExtValue() & 0xff;
+    unsigned NumElts = Ops[0]->getType()->getIntegerBitWidth();
+
+    if (ShiftVal >= NumElts)
+      return llvm::Constant::getNullValue(Ops[0]->getType());
+
+    Value *In = getMaskVecValue(*this, Ops[0], NumElts);
+
+    uint32_t Indices[64];
+    for (unsigned i = 0; i != NumElts; ++i)
+      Indices[i] = i + ShiftVal;
+
+    Value *Zero = llvm::Constant::getNullValue(In->getType());
+    Value *SV = Builder.CreateShuffleVector(In, Zero,
+                                            makeArrayRef(Indices, NumElts),
+                                            "kshiftr");
+    return Builder.CreateBitCast(SV, Ops[0]->getType());
+  }
   case X86::BI__builtin_ia32_movnti:
   case X86::BI__builtin_ia32_movnti64:
   case X86::BI__builtin_ia32_movntsd:
diff --git a/clang/lib/Headers/avx512bwintrin.h b/clang/lib/Headers/avx512bwintrin.h
index fdf26765b6e..d7a03c961b3 100644
--- a/clang/lib/Headers/avx512bwintrin.h
+++ b/clang/lib/Headers/avx512bwintrin.h
@@ -155,6 +155,18 @@ _kadd_mask64(__mmask64 __A, __mmask64 __B)
   return (__mmask64)__builtin_ia32_kadddi((__mmask64)__A, (__mmask64)__B);
 }
 
+#define _kshiftli_mask32(A, I) \
+  (__mmask32)__builtin_ia32_kshiftlisi((__mmask32)(A), (unsigned int)(I))
+
+#define _kshiftri_mask32(A, I) \
+  (__mmask32)__builtin_ia32_kshiftrisi((__mmask32)(A), (unsigned int)(I))
+
+#define _kshiftli_mask64(A, I) \
+  (__mmask64)__builtin_ia32_kshiftlidi((__mmask64)(A), (unsigned int)(I))
+
+#define _kshiftri_mask64(A, I) \
+  (__mmask64)__builtin_ia32_kshiftridi((__mmask64)(A), (unsigned int)(I))
+
 /* Integer compare */
 
 #define _mm512_cmp_epi8_mask(a, b, p) \
diff --git a/clang/lib/Headers/avx512dqintrin.h b/clang/lib/Headers/avx512dqintrin.h
index a9a35794b43..b54e5474dfb 100644
--- a/clang/lib/Headers/avx512dqintrin.h
+++ b/clang/lib/Headers/avx512dqintrin.h
@@ -98,6 +98,12 @@ _kadd_mask16(__mmask16 __A, __mmask16 __B)
   return (__mmask16)__builtin_ia32_kaddhi((__mmask16)__A, (__mmask16)__B);
 }
 
+#define _kshiftli_mask8(A, I) \
+  (__mmask8)__builtin_ia32_kshiftliqi((__mmask8)(A), (unsigned int)(I))
+
+#define _kshiftri_mask8(A, I) \
+  (__mmask8)__builtin_ia32_kshiftriqi((__mmask8)(A), (unsigned int)(I))
+
 static __inline__ __m512i __DEFAULT_FN_ATTRS512
 _mm512_mullo_epi64 (__m512i __A, __m512i __B) {
   return (__m512i) ((__v8du) __A * (__v8du) __B);
diff --git a/clang/lib/Headers/avx512fintrin.h b/clang/lib/Headers/avx512fintrin.h
index b5ef07a425c..262f46cf319 100644
--- a/clang/lib/Headers/avx512fintrin.h
+++ b/clang/lib/Headers/avx512fintrin.h
@@ -8394,6 +8394,12 @@ _mm512_kxor (__mmask16 __A, __mmask16 __B)
 #define _kxnor_mask16 _mm512_kxnor
 #define _kxor_mask16 _mm512_kxor
 
+#define _kshiftli_mask16(A, I) \
+  (__mmask16)__builtin_ia32_kshiftlihi((__mmask16)(A), (unsigned int)(I))
+
+#define _kshiftri_mask16(A, I) \
+  (__mmask16)__builtin_ia32_kshiftrihi((__mmask16)(A), (unsigned int)(I))
+
 static __inline__ void __DEFAULT_FN_ATTRS512
 _mm512_stream_si512 (__m512i * __P, __m512i __A)
 {
diff --git a/clang/lib/Sema/SemaChecking.cpp b/clang/lib/Sema/SemaChecking.cpp
index 7826446b79a..ac7c046dffe 100644
--- a/clang/lib/Sema/SemaChecking.cpp
+++ b/clang/lib/Sema/SemaChecking.cpp
@@ -3626,6 +3626,14 @@ bool Sema::CheckX86BuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
   case X86::BI__builtin_ia32_psrldqi128_byteshift:
   case X86::BI__builtin_ia32_psrldqi256_byteshift:
   case X86::BI__builtin_ia32_psrldqi512_byteshift:
+  case X86::BI__builtin_ia32_kshiftliqi:
+  case X86::BI__builtin_ia32_kshiftlihi:
+  case X86::BI__builtin_ia32_kshiftlisi:
+  case X86::BI__builtin_ia32_kshiftlidi:
+  case X86::BI__builtin_ia32_kshiftriqi:
+  case X86::BI__builtin_ia32_kshiftrihi:
+  case X86::BI__builtin_ia32_kshiftrisi:
+  case X86::BI__builtin_ia32_kshiftridi:
     i = 1; l = 0; u = 255;
     break;
   case X86::BI__builtin_ia32_vperm2f128_pd256: