[AVX-512] Fix the valid immediates for the scatter/gather prefetch intrinsics.

The immediate should be 1 or 2, not 0 or 1. This was found while adding bounds checking to clang. In fact the existing clang builtin test failed if we ran it all the way to assembly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@297591 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index b74bf9a947cea07cb0f713904fc839832b23eaaf..cf8304e565abdbc2f128fcc0d4a09fc16e8b0820 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -20203,8 +20203,9 @@ static SDValue LowerINTRINSIC_W_CHAIN(SDValue Op, const X86Subtarget &Subtarget,
   case PREFETCH: {
     SDValue Hint = Op.getOperand(6);
     unsigned HintVal = cast<ConstantSDNode>(Hint)->getZExtValue();
-    assert(HintVal < 2 && "Wrong prefetch hint in intrinsic: should be 0 or 1");
-    unsigned Opcode = (HintVal ? IntrData->Opc1 : IntrData->Opc0);
+    assert((HintVal == 1 || HintVal == 2) &&
+           "Wrong prefetch hint in intrinsic: should be 1 or 2");
+    unsigned Opcode = (HintVal != 1 ? IntrData->Opc1 : IntrData->Opc0);
     SDValue Chain = Op.getOperand(0);
     SDValue Mask  = Op.getOperand(2);
     SDValue Index = Op.getOperand(3);

diff --git a/test/CodeGen/X86/avx512-gather-scatter-intrin.ll b/test/CodeGen/X86/avx512-gather-scatter-intrin.ll
index a9b649b12f0b0e6e32eddde1f5abbeef0f11b242..20f6de29246ed9700514d8d5825bee59b7f647d7 100644 (file)
--- a/test/CodeGen/X86/avx512-gather-scatter-intrin.ll
+++ b/test/CodeGen/X86/avx512-gather-scatter-intrin.ll
@@ -284,10 +284,10 @@ define void @prefetch(<8 x i64> %ind, i8* %base) {
 ; CHECK-NEXT:    vscatterpf1qps (%rdi,%zmm0,2) {%k1}
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
-  call void @llvm.x86.avx512.gatherpf.qps.512(i8 -1, <8 x i64> %ind, i8* %base, i32 4, i32 0)
-  call void @llvm.x86.avx512.gatherpf.qps.512(i8 0, <8 x i64> %ind, i8* %base, i32 4, i32 1)
-  call void @llvm.x86.avx512.scatterpf.qps.512(i8 1, <8 x i64> %ind, i8* %base, i32 2, i32 0)
-  call void @llvm.x86.avx512.scatterpf.qps.512(i8 120, <8 x i64> %ind, i8* %base, i32 2, i32 1)
+  call void @llvm.x86.avx512.gatherpf.qps.512(i8 -1, <8 x i64> %ind, i8* %base, i32 4, i32 1)
+  call void @llvm.x86.avx512.gatherpf.qps.512(i8 0, <8 x i64> %ind, i8* %base, i32 4, i32 2)
+  call void @llvm.x86.avx512.scatterpf.qps.512(i8 1, <8 x i64> %ind, i8* %base, i32 2, i32 1)
+  call void @llvm.x86.avx512.scatterpf.qps.512(i8 120, <8 x i64> %ind, i8* %base, i32 2, i32 2)
   ret void
 }