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

These aren't documented in the Intel Intrinsics Guide, but are supported by gcc and icc.

Includes these intrinsics:
_ktestc_mask8_u8, _ktestz_mask8_u8, _ktest_mask8_u8
_ktestc_mask16_u8, _ktestz_mask16_u8, _ktest_mask16_u8
_ktestc_mask32_u8, _ktestz_mask32_u8, _ktest_mask32_u8
_ktestc_mask64_u8, _ktestz_mask64_u8, _ktest_mask64_u8

llvm-svn: 341265

diff --git a/clang/include/clang/Basic/BuiltinsX86.def b/clang/include/clang/Basic/BuiltinsX86.def
index 11c3305..db81f58 100644 (file)
--- a/clang/include/clang/Basic/BuiltinsX86.def
+++ b/clang/include/clang/Basic/BuiltinsX86.def
@@ -1761,6 +1761,14 @@ TARGET_BUILTIN(__builtin_ia32_kortestcsi, "iUiUi", "nc", "avx512bw")
 TARGET_BUILTIN(__builtin_ia32_kortestzsi, "iUiUi", "nc", "avx512bw")
 TARGET_BUILTIN(__builtin_ia32_kortestcdi, "iULLiULLi", "nc", "avx512bw")
 TARGET_BUILTIN(__builtin_ia32_kortestzdi, "iULLiULLi", "nc", "avx512bw")
+TARGET_BUILTIN(__builtin_ia32_ktestcqi, "iUcUc", "nc", "avx512dq")
+TARGET_BUILTIN(__builtin_ia32_ktestzqi, "iUcUc", "nc", "avx512dq")
+TARGET_BUILTIN(__builtin_ia32_ktestchi, "iUsUs", "nc", "avx512dq")
+TARGET_BUILTIN(__builtin_ia32_ktestzhi, "iUsUs", "nc", "avx512dq")
+TARGET_BUILTIN(__builtin_ia32_ktestcsi, "iUiUi", "nc", "avx512bw")
+TARGET_BUILTIN(__builtin_ia32_ktestzsi, "iUiUi", "nc", "avx512bw")
+TARGET_BUILTIN(__builtin_ia32_ktestcdi, "iULLiULLi", "nc", "avx512bw")
+TARGET_BUILTIN(__builtin_ia32_ktestzdi, "iULLiULLi", "nc", "avx512bw")
 TARGET_BUILTIN(__builtin_ia32_kunpckhi, "UsUsUs", "nc", "avx512f")
 TARGET_BUILTIN(__builtin_ia32_kxnorqi, "UcUcUc", "nc", "avx512dq")
 TARGET_BUILTIN(__builtin_ia32_kxnorhi, "UsUsUs", "nc", "avx512f")

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 58f8b7d..789f01c 100644 (file)
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -10075,6 +10075,50 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
     return Builder.CreateZExt(Cmp, ConvertType(E->getType()));
   }
 
+  case X86::BI__builtin_ia32_ktestcqi:
+  case X86::BI__builtin_ia32_ktestzqi:
+  case X86::BI__builtin_ia32_ktestchi:
+  case X86::BI__builtin_ia32_ktestzhi:
+  case X86::BI__builtin_ia32_ktestcsi:
+  case X86::BI__builtin_ia32_ktestzsi:
+  case X86::BI__builtin_ia32_ktestcdi:
+  case X86::BI__builtin_ia32_ktestzdi: {
+    Intrinsic::ID IID;
+    switch (BuiltinID) {
+    default: llvm_unreachable("Unsupported intrinsic!");
+    case X86::BI__builtin_ia32_ktestcqi:
+      IID = Intrinsic::x86_avx512_ktestc_b;
+      break;
+    case X86::BI__builtin_ia32_ktestzqi:
+      IID = Intrinsic::x86_avx512_ktestz_b;
+      break;
+    case X86::BI__builtin_ia32_ktestchi:
+      IID = Intrinsic::x86_avx512_ktestc_w;
+      break;
+    case X86::BI__builtin_ia32_ktestzhi:
+      IID = Intrinsic::x86_avx512_ktestz_w;
+      break;
+    case X86::BI__builtin_ia32_ktestcsi:
+      IID = Intrinsic::x86_avx512_ktestc_d;
+      break;
+    case X86::BI__builtin_ia32_ktestzsi:
+      IID = Intrinsic::x86_avx512_ktestz_d;
+      break;
+    case X86::BI__builtin_ia32_ktestcdi:
+      IID = Intrinsic::x86_avx512_ktestc_q;
+      break;
+    case X86::BI__builtin_ia32_ktestzdi:
+      IID = Intrinsic::x86_avx512_ktestz_q;
+      break;
+    }
+
+    unsigned NumElts = Ops[0]->getType()->getIntegerBitWidth();
+    Value *LHS = getMaskVecValue(*this, Ops[0], NumElts);
+    Value *RHS = getMaskVecValue(*this, Ops[1], NumElts);
+    Function *Intr = CGM.getIntrinsic(IID);
+    return Builder.CreateCall(Intr, {LHS, RHS});
+  }
+
   case X86::BI__builtin_ia32_kaddqi:
   case X86::BI__builtin_ia32_kaddhi:
   case X86::BI__builtin_ia32_kaddsi:

diff --git a/clang/lib/Headers/avx512bwintrin.h b/clang/lib/Headers/avx512bwintrin.h
index 740f5cd..bff5b97 100644 (file)
--- a/clang/lib/Headers/avx512bwintrin.h
+++ b/clang/lib/Headers/avx512bwintrin.h
@@ -143,6 +143,42 @@ _kortest_mask64_u8(__mmask64 __A, __mmask64 __B, unsigned char *__C) {
   return (unsigned char)__builtin_ia32_kortestzdi(__A, __B);
 }
 
+static __inline__ unsigned char __DEFAULT_FN_ATTRS
+_ktestc_mask32_u8(__mmask32 __A, __mmask32 __B)
+{
+  return (unsigned char)__builtin_ia32_ktestcsi(__A, __B);
+}
+
+static __inline__ unsigned char __DEFAULT_FN_ATTRS
+_ktestz_mask32_u8(__mmask32 __A, __mmask32 __B)
+{
+  return (unsigned char)__builtin_ia32_ktestzsi(__A, __B);
+}
+
+static __inline__ unsigned char __DEFAULT_FN_ATTRS
+_ktest_mask32_u8(__mmask32 __A, __mmask32 __B, unsigned char *__C) {
+  *__C = (unsigned char)__builtin_ia32_ktestcsi(__A, __B);
+  return (unsigned char)__builtin_ia32_ktestzsi(__A, __B);
+}
+
+static __inline__ unsigned char __DEFAULT_FN_ATTRS
+_ktestc_mask64_u8(__mmask64 __A, __mmask64 __B)
+{
+  return (unsigned char)__builtin_ia32_ktestcdi(__A, __B);
+}
+
+static __inline__ unsigned char __DEFAULT_FN_ATTRS
+_ktestz_mask64_u8(__mmask64 __A, __mmask64 __B)
+{
+  return (unsigned char)__builtin_ia32_ktestzdi(__A, __B);
+}
+
+static __inline__ unsigned char __DEFAULT_FN_ATTRS
+_ktest_mask64_u8(__mmask64 __A, __mmask64 __B, unsigned char *__C) {
+  *__C = (unsigned char)__builtin_ia32_ktestcdi(__A, __B);
+  return (unsigned char)__builtin_ia32_ktestzdi(__A, __B);
+}
+
 static __inline__ __mmask32 __DEFAULT_FN_ATTRS
 _kadd_mask32(__mmask32 __A, __mmask32 __B)
 {

diff --git a/clang/lib/Headers/avx512dqintrin.h b/clang/lib/Headers/avx512dqintrin.h
index 7e025e9..6e6c293 100644 (file)
--- a/clang/lib/Headers/avx512dqintrin.h
+++ b/clang/lib/Headers/avx512dqintrin.h
@@ -86,6 +86,42 @@ _kortest_mask8_u8(__mmask8 __A, __mmask8 __B, unsigned char *__C) {
   return (unsigned char)__builtin_ia32_kortestzqi(__A, __B);
 }
 
+static __inline__ unsigned char __DEFAULT_FN_ATTRS
+_ktestc_mask8_u8(__mmask8 __A, __mmask8 __B)
+{
+  return (unsigned char)__builtin_ia32_ktestcqi(__A, __B);
+}
+
+static __inline__ unsigned char __DEFAULT_FN_ATTRS
+_ktestz_mask8_u8(__mmask8 __A, __mmask8 __B)
+{
+  return (unsigned char)__builtin_ia32_ktestzqi(__A, __B);
+}
+
+static __inline__ unsigned char __DEFAULT_FN_ATTRS
+_ktest_mask8_u8(__mmask8 __A, __mmask8 __B, unsigned char *__C) {
+  *__C = (unsigned char)__builtin_ia32_ktestcqi(__A, __B);
+  return (unsigned char)__builtin_ia32_ktestzqi(__A, __B);
+}
+
+static __inline__ unsigned char __DEFAULT_FN_ATTRS
+_ktestc_mask16_u8(__mmask16 __A, __mmask16 __B)
+{
+  return (unsigned char)__builtin_ia32_ktestchi(__A, __B);
+}
+
+static __inline__ unsigned char __DEFAULT_FN_ATTRS
+_ktestz_mask16_u8(__mmask16 __A, __mmask16 __B)
+{
+  return (unsigned char)__builtin_ia32_ktestzhi(__A, __B);
+}
+
+static __inline__ unsigned char __DEFAULT_FN_ATTRS
+_ktest_mask16_u8(__mmask16 __A, __mmask16 __B, unsigned char *__C) {
+  *__C = (unsigned char)__builtin_ia32_ktestchi(__A, __B);
+  return (unsigned char)__builtin_ia32_ktestzhi(__A, __B);
+}
+
 static __inline__ __mmask8 __DEFAULT_FN_ATTRS
 _kadd_mask8(__mmask8 __A, __mmask8 __B)
 {

diff --git a/clang/test/CodeGen/avx512bw-builtins.c b/clang/test/CodeGen/avx512bw-builtins.c
index 22bc974..01476c9 100644 (file)
--- a/clang/test/CodeGen/avx512bw-builtins.c
+++ b/clang/test/CodeGen/avx512bw-builtins.c
@@ -226,6 +226,74 @@ unsigned char test_kortest_mask64_u8(__m512i __A, __m512i __B, __m512i __C, __m5
                             _mm512_cmpneq_epu8_mask(__C, __D), CF);
 }
 
+unsigned char test_ktestz_mask32_u8(__m512i __A, __m512i __B, __m512i __C, __m512i __D) {
+  // CHECK-LABEL: @test_ktestz_mask32_u8
+  // CHECK: [[LHS:%.*]] = bitcast i32 %{{.*}} to <32 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i32 %{{.*}} to <32 x i1>
+  // CHECK: [[RES:%.*]] = call i32 @llvm.x86.avx512.ktestz.d(<32 x i1> [[LHS]], <32 x i1> [[RHS]])
+  // CHECK: trunc i32 [[RES]] to i8
+  return _ktestz_mask32_u8(_mm512_cmpneq_epu16_mask(__A, __B),
+                           _mm512_cmpneq_epu16_mask(__C, __D));
+}
+
+unsigned char test_ktestc_mask32_u8(__m512i __A, __m512i __B, __m512i __C, __m512i __D) {
+  // CHECK-LABEL: @test_ktestc_mask32_u8
+  // CHECK: [[LHS:%.*]] = bitcast i32 %{{.*}} to <32 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i32 %{{.*}} to <32 x i1>
+  // CHECK: [[RES:%.*]] = call i32 @llvm.x86.avx512.ktestc.d(<32 x i1> [[LHS]], <32 x i1> [[RHS]])
+  // CHECK: trunc i32 [[RES]] to i8
+  return _ktestc_mask32_u8(_mm512_cmpneq_epu16_mask(__A, __B),
+                           _mm512_cmpneq_epu16_mask(__C, __D));
+}
+
+unsigned char test_ktest_mask32_u8(__m512i __A, __m512i __B, __m512i __C, __m512i __D, unsigned char *CF) {
+  // CHECK-LABEL: @test_ktest_mask32_u8
+  // CHECK: [[LHS:%.*]] = bitcast i32 %{{.*}} to <32 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i32 %{{.*}} to <32 x i1>
+  // CHECK: [[RES:%.*]] = call i32 @llvm.x86.avx512.ktestc.d(<32 x i1> [[LHS]], <32 x i1> [[RHS]])
+  // CHECK: trunc i32 [[RES]] to i8
+  // CHECK: [[LHS:%.*]] = bitcast i32 %{{.*}} to <32 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i32 %{{.*}} to <32 x i1>
+  // CHECK: [[RES:%.*]] = call i32 @llvm.x86.avx512.ktestz.d(<32 x i1> [[LHS]], <32 x i1> [[RHS]])
+  // CHECK: trunc i32 [[RES]] to i8
+  return _ktest_mask32_u8(_mm512_cmpneq_epu16_mask(__A, __B),
+                          _mm512_cmpneq_epu16_mask(__C, __D), CF);
+}
+
+unsigned char test_ktestz_mask64_u8(__m512i __A, __m512i __B, __m512i __C, __m512i __D) {
+  // CHECK-LABEL: @test_ktestz_mask64_u8
+  // CHECK: [[LHS:%.*]] = bitcast i64 %{{.*}} to <64 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i64 %{{.*}} to <64 x i1>
+  // CHECK: [[RES:%.*]] = call i32 @llvm.x86.avx512.ktestz.q(<64 x i1> [[LHS]], <64 x i1> [[RHS]])
+  // CHECK: trunc i32 [[RES]] to i8
+  return _ktestz_mask64_u8(_mm512_cmpneq_epu8_mask(__A, __B),
+                           _mm512_cmpneq_epu8_mask(__C, __D));
+}
+
+unsigned char test_ktestc_mask64_u8(__m512i __A, __m512i __B, __m512i __C, __m512i __D) {
+  // CHECK-LABEL: @test_ktestc_mask64_u8
+  // CHECK: [[LHS:%.*]] = bitcast i64 %{{.*}} to <64 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i64 %{{.*}} to <64 x i1>
+  // CHECK: [[RES:%.*]] = call i32 @llvm.x86.avx512.ktestc.q(<64 x i1> [[LHS]], <64 x i1> [[RHS]])
+  // CHECK: trunc i32 [[RES]] to i8
+  return _ktestc_mask64_u8(_mm512_cmpneq_epu8_mask(__A, __B),
+                           _mm512_cmpneq_epu8_mask(__C, __D));
+}
+
+unsigned char test_ktest_mask64_u8(__m512i __A, __m512i __B, __m512i __C, __m512i __D, unsigned char *CF) {
+  // CHECK-LABEL: @test_ktest_mask64_u8
+  // CHECK: [[LHS:%.*]] = bitcast i64 %{{.*}} to <64 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i64 %{{.*}} to <64 x i1>
+  // CHECK: [[RES:%.*]] = call i32 @llvm.x86.avx512.ktestc.q(<64 x i1> [[LHS]], <64 x i1> [[RHS]])
+  // CHECK: trunc i32 [[RES]] to i8
+  // CHECK: [[LHS:%.*]] = bitcast i64 %{{.*}} to <64 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i64 %{{.*}} to <64 x i1>
+  // CHECK: [[RES:%.*]] = call i32 @llvm.x86.avx512.ktestz.q(<64 x i1> [[LHS]], <64 x i1> [[RHS]])
+  // CHECK: trunc i32 [[RES]] to i8
+  return _ktest_mask64_u8(_mm512_cmpneq_epu8_mask(__A, __B),
+                          _mm512_cmpneq_epu8_mask(__C, __D), CF);
+}
+
 __mmask32 test_kadd_mask32(__m512i __A, __m512i __B, __m512i __C, __m512i __D, __m512i __E, __m512i __F) {
   // CHECK-LABEL: @test_kadd_mask32
   // CHECK: [[LHS:%.*]] = bitcast i32 %{{.*}} to <32 x i1>

diff --git a/clang/test/CodeGen/avx512dq-builtins.c b/clang/test/CodeGen/avx512dq-builtins.c
index 77f5833..6227a83 100644 (file)
--- a/clang/test/CodeGen/avx512dq-builtins.c
+++ b/clang/test/CodeGen/avx512dq-builtins.c
@@ -114,6 +114,74 @@ unsigned char test_kortest_mask8_u8(__m512i __A, __m512i __B, __m512i __C, __m51
                            _mm512_cmpneq_epu64_mask(__C, __D), CF);
 }
 
+unsigned char test_ktestz_mask8_u8(__m512i __A, __m512i __B, __m512i __C, __m512i __D) {
+  // CHECK-LABEL: @test_ktestz_mask8_u8
+  // CHECK: [[LHS:%.*]] = bitcast i8 %{{.*}} to <8 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i8 %{{.*}} to <8 x i1>
+  // CHECK: [[RES:%.*]] = call i32 @llvm.x86.avx512.ktestz.b(<8 x i1> [[LHS]], <8 x i1> [[RHS]])
+  // CHECK: trunc i32 [[RES]] to i8
+  return _ktestz_mask8_u8(_mm512_cmpneq_epu64_mask(__A, __B),
+                          _mm512_cmpneq_epu64_mask(__C, __D));
+}
+
+unsigned char test_ktestc_mask8_u8(__m512i __A, __m512i __B, __m512i __C, __m512i __D) {
+  // CHECK-LABEL: @test_ktestc_mask8_u8
+  // CHECK: [[LHS:%.*]] = bitcast i8 %{{.*}} to <8 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i8 %{{.*}} to <8 x i1>
+  // CHECK: [[RES:%.*]] = call i32 @llvm.x86.avx512.ktestc.b(<8 x i1> [[LHS]], <8 x i1> [[RHS]])
+  // CHECK: trunc i32 [[RES]] to i8
+  return _ktestc_mask8_u8(_mm512_cmpneq_epu64_mask(__A, __B),
+                          _mm512_cmpneq_epu64_mask(__C, __D));
+}
+
+unsigned char test_ktest_mask8_u8(__m512i __A, __m512i __B, __m512i __C, __m512i __D, unsigned char *CF) {
+  // CHECK-LABEL: @test_ktest_mask8_u8
+  // CHECK: [[LHS:%.*]] = bitcast i8 %{{.*}} to <8 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i8 %{{.*}} to <8 x i1>
+  // CHECK: [[RES:%.*]] = call i32 @llvm.x86.avx512.ktestc.b(<8 x i1> [[LHS]], <8 x i1> [[RHS]])
+  // CHECK: trunc i32 [[RES]] to i8
+  // CHECK: [[LHS:%.*]] = bitcast i8 %{{.*}} to <8 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i8 %{{.*}} to <8 x i1>
+  // CHECK: [[RES:%.*]] = call i32 @llvm.x86.avx512.ktestz.b(<8 x i1> [[LHS]], <8 x i1> [[RHS]])
+  // CHECK: trunc i32 [[RES]] to i8
+  return _ktest_mask8_u8(_mm512_cmpneq_epu64_mask(__A, __B),
+                         _mm512_cmpneq_epu64_mask(__C, __D), CF);
+}
+
+unsigned char test_ktestz_mask16_u8(__m512i __A, __m512i __B, __m512i __C, __m512i __D) {
+  // CHECK-LABEL: @test_ktestz_mask16_u8
+  // CHECK: [[LHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[RES:%.*]] = call i32 @llvm.x86.avx512.ktestz.w(<16 x i1> [[LHS]], <16 x i1> [[RHS]])
+  // CHECK: trunc i32 [[RES]] to i8
+  return _ktestz_mask16_u8(_mm512_cmpneq_epu32_mask(__A, __B),
+                           _mm512_cmpneq_epu32_mask(__C, __D));
+}
+
+unsigned char test_ktestc_mask16_u8(__m512i __A, __m512i __B, __m512i __C, __m512i __D) {
+  // CHECK-LABEL: @test_ktestc_mask16_u8
+  // CHECK: [[LHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[RES:%.*]] = call i32 @llvm.x86.avx512.ktestc.w(<16 x i1> [[LHS]], <16 x i1> [[RHS]])
+  // CHECK: trunc i32 [[RES]] to i8
+  return _ktestc_mask16_u8(_mm512_cmpneq_epu32_mask(__A, __B),
+                           _mm512_cmpneq_epu32_mask(__C, __D));
+}
+
+unsigned char test_ktest_mask16_u8(__m512i __A, __m512i __B, __m512i __C, __m512i __D, unsigned char *CF) {
+  // CHECK-LABEL: @test_ktest_mask16_u8
+  // CHECK: [[LHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[RES:%.*]] = call i32 @llvm.x86.avx512.ktestc.w(<16 x i1> [[LHS]], <16 x i1> [[RHS]])
+  // CHECK: trunc i32 [[RES]] to i8
+  // CHECK: [[LHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[RES:%.*]] = call i32 @llvm.x86.avx512.ktestz.w(<16 x i1> [[LHS]], <16 x i1> [[RHS]])
+  // CHECK: trunc i32 [[RES]] to i8
+  return _ktest_mask16_u8(_mm512_cmpneq_epu32_mask(__A, __B),
+                          _mm512_cmpneq_epu32_mask(__C, __D), CF);
+}
+
 __mmask8 test_kadd_mask8(__m512i __A, __m512i __B, __m512i __C, __m512i __D, __m512i __E, __m512i __F) {
   // CHECK-LABEL: @test_kadd_mask8
   // CHECK: [[LHS:%.*]] = bitcast i8 %{{.*}} to <8 x i1>