Passing AVX 256-bit structures in Win64 was wrong.

Fixed Win64 calling conventions.

llvm-svn: 149494

lib/Target/X86/X86CallingConv.td

@@ -198,6 +198,10 @@ def CC_X86_Win64_C : CallingConv<[
   // 128 bit vectors are passed by pointer
   CCIfType<[v16i8, v8i16, v4i32, v2i64, v4f32, v2f64], CCPassIndirect<i64>>,
 
+
+  // 256 bit vectors are passed by pointer
+  CCIfType<[v32i8, v16i16, v8i32, v4i64, v8f32, v4f64], CCPassIndirect<i64>>,
+
   // The first 4 MMX vector arguments are passed in GPRs.
   CCIfType<[x86mmx], CCBitConvertToType<i64>>,
 

test/CodeGen/X86/avx-win64-args.ll

@@ -0,0 +1,18 @@
+; RUN: llc < %s -mcpu=corei7-avx -mattr=+avx | FileCheck %s
+target triple = "x86_64-pc-win32"
+
+declare <8 x float> @foo(<8 x float>, i32)
+
+define <8 x float> @test1(<8 x float> %x, <8 x float> %y) nounwind uwtable readnone ssp {
+entry:
+; CHECK: test1
+; CHECK: leaq {{.*}}, %rcx
+; CHECK: movl {{.*}}, %edx
+; CHECK: call
+; CHECK: ret
+  %x1 = fadd  <8 x float>  %x, %y
+  %call = call  <8 x float> @foo(<8 x float> %x1, i32 1) nounwind
+  %y1 = fsub  <8 x float>  %call, %y
+  ret <8 x float> %y1
+}
+

test/CodeGen/X86/avx-win64.ll

@@ -9,7 +9,6 @@ target triple = "x86_64-pc-win32"
 
 ; CHECK: f___vyf
 ; CHECK: pushq %rbp
-; CHECK-NOT: vmovaps{{.*}}(%r
 ; CHECK: vmovmsk
 ; CHECK: vmovaps %ymm{{.*}}(%r
 ; CHECK: vmovaps %ymm{{.*}}(%r