diff --git a/lib/Target/NVPTX/NVPTXAsmPrinter.cpp b/lib/Target/NVPTX/NVPTXAsmPrinter.cpp
index abe117d0e5b..78fbd456187 100644
--- a/lib/Target/NVPTX/NVPTXAsmPrinter.cpp
+++ b/lib/Target/NVPTX/NVPTXAsmPrinter.cpp
@@ -1128,7 +1128,7 @@ void NVPTXAsmPrinter::printModuleLevelGV(const GlobalVariable *GVar,
   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))
@@ -1310,7 +1310,7 @@ void NVPTXAsmPrinter::emitPTXGlobalVariable(const GlobalVariable *GVar,
   else
     O << " .align " << GVar->getAlignment();
 
-  if (ETy->isSingleValueType()) {
+  if (ETy->isFloatingPointTy() || ETy->isIntegerTy() || ETy->isPointerTy()) {
     O << " .";
     O << getPTXFundamentalTypeStr(ETy);
     O << " ";
@@ -1349,17 +1349,6 @@ static unsigned int getOpenCLAlignment(const DataLayout *TD, Type *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;
diff --git a/test/CodeGen/NVPTX/nvcl-param-align.ll b/test/CodeGen/NVPTX/nvcl-param-align.ll
new file mode 100644
index 00000000000..c1a489f1fc4
--- /dev/null
+++ b/test/CodeGen/NVPTX/nvcl-param-align.ll
@@ -0,0 +1,16 @@
+; 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}
diff --git a/test/CodeGen/NVPTX/vector-global.ll b/test/CodeGen/NVPTX/vector-global.ll
new file mode 100644
index 00000000000..a463bee3a47
--- /dev/null
+++ b/test/CodeGen/NVPTX/vector-global.ll
@@ -0,0 +1,9 @@
+; 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];