R600/SI: Commute instructions to enable more folding opportunities

llvm-svn: 225410
diff --git a/llvm/test/CodeGen/R600/mulhu.ll b/llvm/test/CodeGen/R600/mulhu.ll
index 146c92f..8659717 100644
--- a/llvm/test/CodeGen/R600/mulhu.ll
+++ b/llvm/test/CodeGen/R600/mulhu.ll
@@ -1,7 +1,7 @@
 ;RUN: llc < %s -march=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck %s
 
 ;CHECK: v_mov_b32_e32 v{{[0-9]+}}, 0xaaaaaaab
-;CHECK: v_mul_hi_u32 v0, {{[sv][0-9]+}}, {{v[0-9]+}}
+;CHECK: v_mul_hi_u32 v0, {{v[0-9]+}}, {{s[0-9]+}}
 ;CHECK-NEXT: v_lshrrev_b32_e32 v0, 1, v0
 
 define void @test(i32 %p) {
diff --git a/llvm/test/CodeGen/R600/sdiv.ll b/llvm/test/CodeGen/R600/sdiv.ll
index 8d4208c..c635c05 100644
--- a/llvm/test/CodeGen/R600/sdiv.ll
+++ b/llvm/test/CodeGen/R600/sdiv.ll
@@ -35,7 +35,7 @@
 ; FUNC-LABEL: {{^}}slow_sdiv_i32_3435:
 ; SI: buffer_load_dword [[VAL:v[0-9]+]],
 ; SI: v_mov_b32_e32 [[MAGIC:v[0-9]+]], 0x98a1930b
-; SI: v_mul_hi_i32 [[TMP:v[0-9]+]], [[VAL]], [[MAGIC]]
+; SI: v_mul_hi_i32 [[TMP:v[0-9]+]], [[MAGIC]], [[VAL]]
 ; SI: v_add_i32
 ; SI: v_lshrrev_b32
 ; SI: v_ashrrev_i32
diff --git a/llvm/test/CodeGen/R600/use-sgpr-multiple-times.ll b/llvm/test/CodeGen/R600/use-sgpr-multiple-times.ll
index 2c6ae1e..97d73ba 100644
--- a/llvm/test/CodeGen/R600/use-sgpr-multiple-times.ll
+++ b/llvm/test/CodeGen/R600/use-sgpr-multiple-times.ll
@@ -41,7 +41,7 @@
 ; SI: s_load_dword [[SGPR0:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0xb
 ; SI: s_load_dword [[SGPR1:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0xc
 ; SI: v_mov_b32_e32 [[VGPR1:v[0-9]+]], [[SGPR1]]
-; SI: v_fma_f32 [[RESULT:v[0-9]+]], [[SGPR0]], [[VGPR1]], [[SGPR0]]
+; SI: v_fma_f32 [[RESULT:v[0-9]+]], [[VGPR1]], [[SGPR0]], [[SGPR0]]
 ; SI: buffer_store_dword [[RESULT]]
 define void @test_sgpr_use_twice_ternary_op_a_b_a(float addrspace(1)* %out, float %a, float %b) #0 {
   %fma = call float @llvm.fma.f32(float %a, float %b, float %a) #1
@@ -53,7 +53,7 @@
 ; SI: s_load_dword [[SGPR0:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0xb
 ; SI: s_load_dword [[SGPR1:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0xc
 ; SI: v_mov_b32_e32 [[VGPR1:v[0-9]+]], [[SGPR1]]
-; SI: v_fma_f32 [[RESULT:v[0-9]+]], [[VGPR1]], [[SGPR0]], [[SGPR0]]
+; SI: v_fma_f32 [[RESULT:v[0-9]+]], [[SGPR0]], [[VGPR1]], [[SGPR0]]
 ; SI: buffer_store_dword [[RESULT]]
 define void @test_sgpr_use_twice_ternary_op_b_a_a(float addrspace(1)* %out, float %a, float %b) #0 {
   %fma = call float @llvm.fma.f32(float %b, float %a, float %a) #1