From: Craig Topper <craig.topper@intel.com>
Date: Sat, 16 Dec 2017 08:26:22 +0000 (+0000)
Subject: [X86] Implement kand/kandn/kor/kxor/kxnor/knot intrinsics using native IR.
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=d073eeed8abada8922c0566577e94c80a064c48b;p=clang

[X86] Implement kand/kandn/kor/kxor/kxnor/knot intrinsics using native IR.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@320919 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/CodeGen/CGBuiltin.cpp b/lib/CodeGen/CGBuiltin.cpp
index a1a2cef822..3ecd1c6697 100644
--- a/lib/CodeGen/CGBuiltin.cpp
+++ b/lib/CodeGen/CGBuiltin.cpp
@@ -7564,6 +7564,19 @@ static Value *EmitX86MaskedLoad(CodeGenFunction &CGF,
   return CGF.Builder.CreateMaskedLoad(Ops[0], Align, MaskVec, Ops[1]);
 }
 
+static Value *EmitX86MaskLogic(CodeGenFunction &CGF, Instruction::BinaryOps Opc,
+                              unsigned NumElts, SmallVectorImpl<Value *> &Ops,
+                              bool InvertLHS = false) {
+  Value *LHS = getMaskVecValue(CGF, Ops[0], NumElts);
+  Value *RHS = getMaskVecValue(CGF, Ops[1], NumElts);
+
+  if (InvertLHS)
+    LHS = CGF.Builder.CreateNot(LHS);
+
+  return CGF.Builder.CreateBitCast(CGF.Builder.CreateBinOp(Opc, LHS, RHS),
+                                  CGF.Builder.getIntNTy(std::max(NumElts, 8U)));
+}
+
 static Value *EmitX86SubVectorBroadcast(CodeGenFunction &CGF,
                                         SmallVectorImpl<Value *> &Ops,
                                         llvm::Type *DstTy,
@@ -8217,6 +8230,22 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
     return EmitX86MaskedCompare(*this, CC, false, Ops);
   }
 
+  case X86::BI__builtin_ia32_kandhi:
+    return EmitX86MaskLogic(*this, Instruction::And, 16, Ops);
+  case X86::BI__builtin_ia32_kandnhi:
+    return EmitX86MaskLogic(*this, Instruction::And, 16, Ops, true);
+  case X86::BI__builtin_ia32_korhi:
+    return EmitX86MaskLogic(*this, Instruction::Or, 16, Ops);
+  case X86::BI__builtin_ia32_kxnorhi:
+    return EmitX86MaskLogic(*this, Instruction::Xor, 16, Ops, true);
+  case X86::BI__builtin_ia32_kxorhi:
+    return EmitX86MaskLogic(*this, Instruction::Xor, 16, Ops);
+  case X86::BI__builtin_ia32_knothi: {
+    Ops[0] = getMaskVecValue(*this, Ops[0], 16);
+    return Builder.CreateBitCast(Builder.CreateNot(Ops[0]),
+                                 Builder.getInt16Ty());
+  }
+
   case X86::BI__builtin_ia32_vplzcntd_128_mask:
   case X86::BI__builtin_ia32_vplzcntd_256_mask:
   case X86::BI__builtin_ia32_vplzcntd_512_mask:
diff --git a/test/CodeGen/avx512f-builtins.c b/test/CodeGen/avx512f-builtins.c
index a4e9e3bc9d..9a3bf5a244 100644
--- a/test/CodeGen/avx512f-builtins.c
+++ b/test/CodeGen/avx512f-builtins.c
@@ -385,7 +385,9 @@ __m512d test_mm512_set1_pd(double d)
 __mmask16 test_mm512_knot(__mmask16 a)
 {
   // CHECK-LABEL: @test_mm512_knot
-  // CHECK: @llvm.x86.avx512.knot.w
+  // CHECK: [[IN:%.*]] = bitcast i16 %1 to <16 x i1>
+  // CHECK: [[NOT:%.*]] = xor <16 x i1> [[IN]], <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>
+  // CHECK: bitcast <16 x i1> [[NOT]] to i16
   return _mm512_knot(a);
 }
 
@@ -6211,22 +6213,38 @@ __m512i test_mm512_mask_permutexvar_epi32(__m512i __W, __mmask16 __M, __m512i __
   return _mm512_mask_permutexvar_epi32(__W, __M, __X, __Y); 
 }
 
-__mmask16 test_mm512_kand(__mmask16 __A, __mmask16 __B) {
+__mmask16 test_mm512_kand(__m512i __A, __m512i __B, __m512i __C, __m512i __D, __m512i __E, __m512i __F) {
   // CHECK-LABEL: @test_mm512_kand
-  // CHECK: @llvm.x86.avx512.kand.w
-  return _mm512_kand(__A, __B); 
+  // CHECK: [[LHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[RES:%.*]] = and <16 x i1> [[LHS]], [[RHS]]
+  // CHECK: bitcast <16 x i1> [[RES]] to i16
+  return _mm512_mask_cmpneq_epu32_mask(_mm512_kand(_mm512_cmpneq_epu32_mask(__A, __B),
+                                                   _mm512_cmpneq_epu32_mask(__C, __D)),
+                                                   __E, __F);
 }
 
-__mmask16 test_mm512_kandn(__mmask16 __A, __mmask16 __B) {
+__mmask16 test_mm512_kandn(__m512i __A, __m512i __B, __m512i __C, __m512i __D, __m512i __E, __m512i __F) {
   // CHECK-LABEL: @test_mm512_kandn
-  // CHECK: @llvm.x86.avx512.kandn.w
-  return _mm512_kandn(__A, __B); 
+  // CHECK: [[LHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[NOT:%.*]] = xor <16 x i1> [[LHS]], <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>
+  // CHECK: [[RES:%.*]] = and <16 x i1> [[NOT]], [[RHS]]
+  // CHECK: bitcast <16 x i1> [[RES]] to i16
+  return _mm512_mask_cmpneq_epu32_mask(_mm512_kandn(_mm512_cmpneq_epu32_mask(__A, __B),
+                                                    _mm512_cmpneq_epu32_mask(__C, __D)),
+                                                    __E, __F);
 }
 
-__mmask16 test_mm512_kor(__mmask16 __A, __mmask16 __B) {
+__mmask16 test_mm512_kor(__m512i __A, __m512i __B, __m512i __C, __m512i __D, __m512i __E, __m512i __F) {
   // CHECK-LABEL: @test_mm512_kor
-  // CHECK: @llvm.x86.avx512.kor.w
-  return _mm512_kor(__A, __B); 
+  // CHECK: [[LHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[RES:%.*]] = or <16 x i1> [[LHS]], [[RHS]]
+  // CHECK: bitcast <16 x i1> [[RES]] to i16
+  return _mm512_mask_cmpneq_epu32_mask(_mm512_kor(_mm512_cmpneq_epu32_mask(__A, __B),
+                                                  _mm512_cmpneq_epu32_mask(__C, __D)),
+                                                  __E, __F);
 }
 
 int test_mm512_kortestc(__mmask16 __A, __mmask16 __B) {
@@ -6254,16 +6272,27 @@ __mmask16 test_mm512_kunpackb(__m512i __A, __m512i __B, __m512i __C, __m512i __D
                                                        __E, __F);
 }
 
-__mmask16 test_mm512_kxnor(__mmask16 __A, __mmask16 __B) {
+__mmask16 test_mm512_kxnor(__m512i __A, __m512i __B, __m512i __C, __m512i __D, __m512i __E, __m512i __F) {
   // CHECK-LABEL: @test_mm512_kxnor
-  // CHECK: @llvm.x86.avx512.kxnor.w
-  return _mm512_kxnor(__A, __B); 
+  // CHECK: [[LHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[NOT:%.*]] = xor <16 x i1> [[LHS]], <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>
+  // CHECK: [[RES:%.*]] = xor <16 x i1> [[NOT]], [[RHS]]
+  // CHECK: bitcast <16 x i1> [[RES]] to i16
+  return _mm512_mask_cmpneq_epu32_mask(_mm512_kxnor(_mm512_cmpneq_epu32_mask(__A, __B),
+                                                    _mm512_cmpneq_epu32_mask(__C, __D)),
+                                                    __E, __F);
 }
 
-__mmask16 test_mm512_kxor(__mmask16 __A, __mmask16 __B) {
+__mmask16 test_mm512_kxor(__m512i __A, __m512i __B, __m512i __C, __m512i __D, __m512i __E, __m512i __F) {
   // CHECK-LABEL: @test_mm512_kxor
-  // CHECK: @llvm.x86.avx512.kxor.w
-  return _mm512_kxor(__A, __B); 
+  // CHECK: [[LHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[RHS:%.*]] = bitcast i16 %{{.*}} to <16 x i1>
+  // CHECK: [[RES:%.*]] = xor <16 x i1> [[LHS]], [[RHS]]
+  // CHECK: bitcast <16 x i1> [[RES]] to i16
+  return _mm512_mask_cmpneq_epu32_mask(_mm512_kxor(_mm512_cmpneq_epu32_mask(__A, __B),
+                                                   _mm512_cmpneq_epu32_mask(__C, __D)),
+                                                   __E, __F);
 }
 
 void test_mm512_stream_si512(__m512i * __P, __m512i __A) {