From 4781d314b7b0bd239dc3986b4157726e80a270ba Mon Sep 17 00:00:00 2001
From: Tom Stellard <thomas.stellard@amd.com>
Date: Fri, 16 Aug 2013 01:12:00 +0000
Subject: [PATCH] R600: Add support for v4i32 stores on Cayman

Tested-by: Aaron Watry <awatry@gmail.com>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188518 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/R600/R600Instructions.td        |  1 +
 test/CodeGen/R600/store.ll                 | 14 ++++++++++++++
 test/CodeGen/R600/vertex-fetch-encoding.ll |  2 +-
 3 files changed, 16 insertions(+), 1 deletion(-)

diff --git a/lib/Target/R600/R600Instructions.td b/lib/Target/R600/R600Instructions.td
index bacedfc0a61..06886ce0560 100644
--- a/lib/Target/R600/R600Instructions.td
+++ b/lib/Target/R600/R600Instructions.td
@@ -1808,6 +1808,7 @@ class RAT_STORE_DWORD <RegisterClass rc, ValueType vt, bits<4> mask> :
 
 def RAT_STORE_DWORD32 : RAT_STORE_DWORD <R600_TReg32_X, i32, 0x1>;
 def RAT_STORE_DWORD64 : RAT_STORE_DWORD <R600_Reg64, v2i32, 0x3>;
+def RAT_STORE_DWORD128 : RAT_STORE_DWORD <R600_Reg128, v4i32, 0xf>;
 
 class VTX_READ_cm <string name, bits<8> buffer_id, dag outs, list<dag> pattern>
     : VTX_WORD0_cm, VTX_READ<name, buffer_id, outs, pattern> {
diff --git a/test/CodeGen/R600/store.ll b/test/CodeGen/R600/store.ll
index 5dc0a84bbe5..f2a8dd7d265 100644
--- a/test/CodeGen/R600/store.ll
+++ b/test/CodeGen/R600/store.ll
@@ -31,6 +31,20 @@ entry:
   ret void
 }
 
+; EG-CHECK: @store_v4i32
+; EG-CHECK: MEM_RAT_CACHELESS STORE_RAW
+; EG-CHECK-NOT: MEM_RAT_CACHELESS STORE_RAW
+; CM-CHECK: @store_v4i32
+; CM-CHECK: MEM_RAT_CACHELESS STORE_DWORD
+; CM-CHECK-NOT: MEM_RAT_CACHELESS STORE_DWORD
+; SI-CHECK: @store_v4i32
+; SI-CHECK: BUFFER_STORE_DWORDX4
+define void @store_v4i32(<4 x i32> addrspace(1)* %out, <4 x i32> %in) {
+entry:
+  store <4 x i32> %in, <4 x i32> addrspace(1)* %out
+  ret void
+}
+
 ; The stores in this function are combined by the optimizer to create a
 ; 64-bit store with 32-bit alignment.  This is legal for SI and the legalizer
 ; should not try to split the 64-bit store back into 2 32-bit stores.
diff --git a/test/CodeGen/R600/vertex-fetch-encoding.ll b/test/CodeGen/R600/vertex-fetch-encoding.ll
index d892229f49d..7ea7a5c079c 100644
--- a/test/CodeGen/R600/vertex-fetch-encoding.ll
+++ b/test/CodeGen/R600/vertex-fetch-encoding.ll
@@ -1,5 +1,5 @@
 ; RUN: llc < %s -march=r600 -show-mc-encoding -mcpu=barts | FileCheck --check-prefix=NI-CHECK %s
-; RUN: not llc < %s -march=r600 -show-mc-encoding -mcpu=cayman | FileCheck --check-prefix=CM-CHECK %s
+; RUN: llc < %s -march=r600 -show-mc-encoding -mcpu=cayman | FileCheck --check-prefix=CM-CHECK %s
 
 ; NI-CHECK: @vtx_fetch32
 ; NI-CHECK: VTX_READ_32 T[[GPR:[0-9]]].X, T[[GPR]].X, 0 ; encoding: [0x40,0x01,0x0[[GPR]],0x10,0x0[[GPR]],0xf0,0x5f,0x13,0x00,0x00,0x08,0x00