[NVPTX] Fix bugs related to isSingleValueType

[jingyue]

Summary

With isSingleValueType starting to treat vector types as single-value types,
code that uses this interface needs to be updated.

[jingyue]

Added a minor inline comment.

lib/Target/NVPTX/NVPTXAsmPrinter.cpp

1352

TD->getPrefTypeAlignment(Ty) already handles vector types in a seemingly more reasonable way. For instance, if the element type of Ty is <5 x float>, the deleted code here returns 20, but getPrefTypeAlignment returns 32.

[jholewinski]

LGTM.

[jingyue]

Rebase against master

The format of metadata is updated.

lib/Target/NVPTX/NVPTXAsmPrinter.cpp

if (isManaged(*GVar)) {
     O << " .attribute(.managed)";
   }
 
   if (GVar->getAlignment() == 0)
     O << " .align " << (int) TD->getPrefTypeAlignment(ETy);
   else
     O << " .align " << GVar->getAlignment();
 
-  if (ETy->isSingleValueType()) {
+  if (ETy->isFloatingPointTy() || ETy->isIntegerTy() || ETy->isPointerTy()) {
     O << " .";
     // Special case: ABI requires that we use .u8 for predicates
     if (ETy->isIntegerTy(1))
       O << "u8";
     else
       O << getPTXFundamentalTypeStr(ETy, false);
     O << " ";
     O << *getSymbol(GVar);
void NVPTXAsmPrinter::emitPTXGlobalVariable(const GlobalVariable *GVar,
 
   O << ".";
   emitPTXAddressSpace(PTy->getAddressSpace(), O);
   if (GVar->getAlignment() == 0)
     O << " .align " << (int) TD->getPrefTypeAlignment(ETy);
   else
     O << " .align " << GVar->getAlignment();
 
-  if (ETy->isSingleValueType()) {
+  if (ETy->isFloatingPointTy() || ETy->isIntegerTy() || ETy->isPointerTy()) {
     O << " .";
     O << getPTXFundamentalTypeStr(ETy);
     O << " ";
     O << *getSymbol(GVar);
     return;
   }
 
   int64_t ElementSize = 0;

 static unsigned int getOpenCLAlignment(const DataLayout *TD, Type *Ty) {
   if (Ty->isSingleValueType())
     return TD->getPrefTypeAlignment(Ty);
 
   const ArrayType *ATy = dyn_cast<ArrayType>(Ty);
   if (ATy)
     return getOpenCLAlignment(TD, ATy->getElementType());
 
-  const VectorType *VTy = dyn_cast<VectorType>(Ty);
-  if (VTy) {
-    Type *ETy = VTy->getElementType();
-    unsigned int numE = VTy->getNumElements();
-    unsigned int alignE = TD->getPrefTypeAlignment(ETy);
-    if (numE == 3)
-      return 4 * alignE;
-    else
-      return numE * alignE;
-  }
-
   const StructType *STy = dyn_cast<StructType>(Ty);
   if (STy) {
     unsigned int alignStruct = 1;
     // Go through each element of the struct and find the
     // largest alignment.
     for (unsigned i = 0, e = STy->getNumElements(); i != e; i++) {
       Type *ETy = STy->getElementType(i);
       unsigned int align = getOpenCLAlignment(TD, ETy);

test/CodeGen/NVPTX/nvcl-param-align.ll

+; RUN: llc < %s -march=nvptx -mcpu=sm_20 | FileCheck %s
+
+target triple = "nvptx-unknown-nvcl"
+
+; CHECK-LABEL: .entry foo(
+define void @foo(i64 %img, i64 %sampler, <5 x float>* %v) {
+; The parameter alignment should be the next power of 2 of 5xsizeof(float),
+; which is 32.
+; CHECK: .param .u32 .ptr .align 32 foo_param_2
+  ret void
+}
+
+!nvvm.annotations = !{!1, !2, !3}
+!1 = !{void (i64, i64, <5 x float>*)* @foo, !"kernel", i32 1}
+!2 = !{void (i64, i64, <5 x float>*)* @foo, !"rdoimage", i32 0}
+!3 = !{void (i64, i64, <5 x float>*)* @foo, !"sampler", i32 1}

test/CodeGen/NVPTX/vector-global.ll

+; RUN: llc < %s -march=nvptx64 -mcpu=sm_20 | FileCheck %s
+
+target datalayout = "e-i64:64-v16:16-v32:32-n16:32:64"
+target triple = "nvptx64-nvidia-cuda"
+
+@g1 = external global <4 x i32> ; external global variable
+; CHECK: .extern .global .align 16 .b8 g1[16];
+@g2 = global <4 x i32> zeroinitializer ; module-level global variable
+; CHECK: .visible .global .align 16 .b8 g2[16];