[NVPTX] Fix case where a sext load of an i1 type may produce an

ld.u1 instead of an ld.u8.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182924 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/NVPTX/NVPTXISelDAGToDAG.cpp b/lib/Target/NVPTX/NVPTXISelDAGToDAG.cpp
index 120d7161067ec2fca0653ba6d1f04b6c0930f560..40871d6d2d45afd63961fef1f52f181f450732ea 100644 (file)
--- a/lib/Target/NVPTX/NVPTXISelDAGToDAG.cpp
+++ b/lib/Target/NVPTX/NVPTXISelDAGToDAG.cpp
@@ -205,7 +205,8 @@ SDNode *NVPTXDAGToDAGISel::SelectLoad(SDNode *N) {
   //          type is integer
   // Float  : ISD::NON_EXTLOAD or ISD::EXTLOAD and the type is float
   MVT ScalarVT = SimpleVT.getScalarType();
-  unsigned fromTypeWidth = ScalarVT.getSizeInBits();
+  // Read at least 8 bits (predicates are stored as 8-bit values)
+  unsigned fromTypeWidth = std::max(8U, ScalarVT.getSizeInBits());
   unsigned int fromType;
   if ((LD->getExtensionType() == ISD::SEXTLOAD))
     fromType = NVPTX::PTXLdStInstCode::Signed;
@@ -430,7 +431,8 @@ SDNode *NVPTXDAGToDAGISel::SelectLoadVector(SDNode *N) {
   //          type is integer
   // Float  : ISD::NON_EXTLOAD or ISD::EXTLOAD and the type is float
   MVT ScalarVT = SimpleVT.getScalarType();
-  unsigned FromTypeWidth = ScalarVT.getSizeInBits();
+  // Read at least 8 bits (predicates are stored as 8-bit values)
+  unsigned FromTypeWidth = std::max(8U, ScalarVT.getSizeInBits());
   unsigned int FromType;
   // The last operand holds the original LoadSDNode::getExtensionType() value
   unsigned ExtensionType = cast<ConstantSDNode>(

diff --git a/test/CodeGen/NVPTX/load-sext-i1.ll b/test/CodeGen/NVPTX/load-sext-i1.ll
new file mode 100644 (file)
index 0000000..c9b2e97
--- /dev/null
+++ b/test/CodeGen/NVPTX/load-sext-i1.ll
@@ -0,0 +1,14 @@
+; RUN: llc < %s -march=nvptx -mcpu=sm_20 -drvcuda | FileCheck %s
+
+target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v16:16:16-v32:32:32-v64:64:64-v128:128:128-n16:32:64"
+
+
+define void @main(i1* %a1, i32 %a2, i32* %arg3) {
+; CHECK: ld.u8
+; CHECK-NOT: ld.u1
+  %t1 = getelementptr i1* %a1, i32 %a2
+  %t2 = load i1* %t1
+  %t3 = sext i1 %t2 to i32
+  store i32 %t3, i32* %arg3
+  ret void
+}