From 89443ff7ae81e6a439b5b67e5dd587a450090599 Mon Sep 17 00:00:00 2001
From: Justin Holewinski <justin.holewinski@gmail.com>
Date: Fri, 9 Nov 2012 23:50:24 +0000
Subject: [PATCH] [NVPTX] Use ABI alignment for parameters when alignment is
 not specified.

Affects SM 2.0+.  Fixes bug 13324.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167646 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/NVPTX/NVPTXAsmPrinter.cpp |  3 +++
 test/CodeGen/NVPTX/param-align.ll    | 25 +++++++++++++++++++++++++
 2 files changed, 28 insertions(+)
 create mode 100644 test/CodeGen/NVPTX/param-align.ll

diff --git a/lib/Target/NVPTX/NVPTXAsmPrinter.cpp b/lib/Target/NVPTX/NVPTXAsmPrinter.cpp
index d3dfb35e261..348ef720f0d 100644
--- a/lib/Target/NVPTX/NVPTXAsmPrinter.cpp
+++ b/lib/Target/NVPTX/NVPTXAsmPrinter.cpp
@@ -1525,6 +1525,9 @@ void NVPTXAsmPrinter::emitFunctionParamList(const Function *F,
       // <a> = PAL.getparamalignment
       // size = typeallocsize of element type
       unsigned align = PAL.getParamAlignment(paramIndex+1);
+      if (align == 0)
+        align = TD->getABITypeAlignment(ETy);
+
       unsigned sz = TD->getTypeAllocSize(ETy);
       O << "\t.param .align " << align
           << " .b8 ";
diff --git a/test/CodeGen/NVPTX/param-align.ll b/test/CodeGen/NVPTX/param-align.ll
new file mode 100644
index 00000000000..84ccb650d40
--- /dev/null
+++ b/test/CodeGen/NVPTX/param-align.ll
@@ -0,0 +1,25 @@
+; RUN: llc < %s -march=nvptx -mcpu=sm_20 | FileCheck %s
+
+;;; Need 4-byte alignment on float* passed byval
+define ptx_device void @t1(float* byval %x) {
+; CHECK: .func t1
+; CHECK: .param .align 4 .b8 t1_param_0[4]
+  ret void
+}
+
+
+;;; Need 8-byte alignment on double* passed byval
+define ptx_device void @t2(double* byval %x) {
+; CHECK: .func t2
+; CHECK: .param .align 8 .b8 t2_param_0[8]
+  ret void
+}
+
+
+;;; Need 4-byte alignment on float2* passed byval
+%struct.float2 = type { float, float }
+define ptx_device void @t3(%struct.float2* byval %x) {
+; CHECK: .func t3
+; CHECK: .param .align 4 .b8 t3_param_0[8]
+  ret void
+}
-- 
2.34.1