From 5b3fad49661525e4053acaa13f8ac5bde5328462 Mon Sep 17 00:00:00 2001
From: Daniil Fukalov <daniil.fukalov@amd.com>
Date: Mon, 30 Aug 2021 19:05:55 +0300
Subject: [PATCH] [AMDGPU][CostModel] Update shuffle instruction tests. NFC.

New tests ported over from test/Analysis/CostModel/AArch64/shuffle-other.ll.
---
 .../Analysis/CostModel/AMDGPU/shufflevector.ll     | 382 ++++++++++++++++++---
 1 file changed, 330 insertions(+), 52 deletions(-)

diff --git a/llvm/test/Analysis/CostModel/AMDGPU/shufflevector.ll b/llvm/test/Analysis/CostModel/AMDGPU/shufflevector.ll
index 99467ca..7fe1ceb 100644
--- a/llvm/test/Analysis/CostModel/AMDGPU/shufflevector.ll
+++ b/llvm/test/Analysis/CostModel/AMDGPU/shufflevector.ll
@@ -1,64 +1,342 @@
-; RUN: opt -cost-model -analyze -mtriple=amdgcn-unknown-amdhsa -mcpu=gfx900 %s | FileCheck -check-prefixes=GFX9,GCN,TPT %s
-; RUN: opt -cost-model -analyze -mtriple=amdgcn-unknown-amdhsa -mcpu=fiji %s | FileCheck -check-prefixes=VI,GCN,TPT %s
-; RUN: opt -cost-model -cost-kind=code-size -analyze -mtriple=amdgcn-unknown-amdhsa -mcpu=gfx900 %s | FileCheck -check-prefixes=GFX9,GCN,CS %s
-; RUN: opt -cost-model -cost-kind=code-size -analyze -mtriple=amdgcn-unknown-amdhsa -mcpu=fiji %s | FileCheck -check-prefixes=VI,GCN,CS %s
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
+; RUN: opt < %s -cost-model -analyze -mtriple=amdgcn-unknown-amdhsa -mcpu=gfx1010 -S | FileCheck -check-prefixes=GFX10 %s
+; RUN: opt < %s -cost-model -analyze -mtriple=amdgcn-unknown-amdhsa -mcpu=gfx1010 -cost-kind=code-size -S | FileCheck -check-prefixes=GFX10-CS %s
+; RUN: opt < %s -cost-model -analyze -mtriple=amdgcn-unknown-amdhsa -mcpu=gfx900 -S | FileCheck -check-prefixes=GFX9 %s
+; RUN: opt < %s -cost-model -analyze -mtriple=amdgcn-unknown-amdhsa -mcpu=gfx900 -cost-kind=code-size -S | FileCheck -check-prefixes=GFX9-CS %s
+; RUN: opt < %s -cost-model -analyze -mtriple=amdgcn-unknown-amdhsa -mcpu=fiji -S | FileCheck -check-prefixes=VI %s
+; RUN: opt < %s -cost-model -analyze -mtriple=amdgcn-unknown-amdhsa -mcpu=fiji -cost-kind=code-size -S | FileCheck -check-prefixes=VI-CS %s
 
-; GCN-LABEL: 'shufflevector_00_v2i16'
-; GFX9: estimated cost of 0 for {{.*}} shufflevector <2 x i16> %vec, <2 x i16> undef, <2 x i32> zeroinitializer
-; VI: estimated cost of 1 for {{.*}} shufflevector <2 x i16> %vec, <2 x i16> undef, <2 x i32> zeroinitializer
-define amdgpu_kernel void @shufflevector_00_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %vaddr) {
-  %vec = load <2 x i16>, <2 x i16> addrspace(1)* %vaddr
-  %shuf = shufflevector <2 x i16> %vec, <2 x i16> undef, <2 x i32> zeroinitializer
-  store <2 x i16> %shuf, <2 x i16> addrspace(1)* %out
+define amdgpu_kernel void @shufflevector_00_v2i16(<2 x i16> %vec0, <2 x i16> %vec1) {
+; GFX10-LABEL: 'shufflevector_00_v2i16'
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> zeroinitializer
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 0, i32 1>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 2, i32 0>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
+;
+; GFX10-CS-LABEL: 'shufflevector_00_v2i16'
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> zeroinitializer
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 0, i32 1>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 2, i32 0>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; GFX9-LABEL: 'shufflevector_00_v2i16'
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> zeroinitializer
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 0, i32 1>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 2, i32 0>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
+;
+; GFX9-CS-LABEL: 'shufflevector_00_v2i16'
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> zeroinitializer
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 0, i32 1>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 2, i32 0>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; VI-LABEL: 'shufflevector_00_v2i16'
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf00 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> zeroinitializer
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf10 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf11 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf22 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
+;
+; VI-CS-LABEL: 'shufflevector_00_v2i16'
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf00 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> zeroinitializer
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 0, i32 1>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf10 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf11 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 2, i32 0>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf22 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  %shuf00 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> zeroinitializer
+  %shuf01 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 0, i32 1>
+  %shuf10 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+  %shuf11 = shufflevector <2 x i16> %vec0, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+  %shuf02 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
+  %shuf20 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 2, i32 0>
+  %shuf22 = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
   ret void
 }
 
-; GCN-LABEL: 'shufflevector_01_v2i16'
-; GCN: estimated cost of 0 for {{.*}} shufflevector <2 x i16> %vec, <2 x i16> undef, <2 x i32> <i32 0, i32 1>
-define amdgpu_kernel void @shufflevector_01_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %vaddr) {
-  %vec = load <2 x i16>, <2 x i16> addrspace(1)* %vaddr
-  %shuf = shufflevector <2 x i16> %vec, <2 x i16> undef, <2 x i32> <i32 0, i32 1>
-  store <2 x i16> %shuf, <2 x i16> addrspace(1)* %out
-  ret void
-}
-
-; GCN-LABEL: 'shufflevector_10_v2i16'
-; GFX9: estimated cost of 0 for {{.*}} shufflevector <2 x i16> %vec, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; VI: estimated cost of 2 for {{.*}} shufflevector <2 x i16> %vec, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-define amdgpu_kernel void @shufflevector_10_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %vaddr) {
-  %vec = load <2 x i16>, <2 x i16> addrspace(1)* %vaddr
-  %shuf = shufflevector <2 x i16> %vec, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-  store <2 x i16> %shuf, <2 x i16> addrspace(1)* %out
-  ret void
-}
-
-; GCN-LABEL: 'shufflevector_11_v2i16'
-; GFX9: estimated cost of 0 for {{.*}} shufflevector <2 x i16> %vec, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; VI: estimated cost of 2 for {{.*}} shufflevector <2 x i16> %vec, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-define amdgpu_kernel void @shufflevector_11_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %vaddr) {
-  %vec = load <2 x i16>, <2 x i16> addrspace(1)* %vaddr
-  %shuf = shufflevector <2 x i16> %vec, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-  store <2 x i16> %shuf, <2 x i16> addrspace(1)* %out
+; Should not assert
+define amdgpu_kernel void @shufflevector_xxx(<2 x i8> %vec0) {
+; GFX10-LABEL: 'shufflevector_xxx'
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01uu = shufflevector <2 x i8> %vec0, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10uu = shufflevector <2 x i8> %vec0, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
+;
+; GFX10-CS-LABEL: 'shufflevector_xxx'
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01uu = shufflevector <2 x i8> %vec0, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10uu = shufflevector <2 x i8> %vec0, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; GFX9-LABEL: 'shufflevector_xxx'
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01uu = shufflevector <2 x i8> %vec0, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10uu = shufflevector <2 x i8> %vec0, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
+;
+; GFX9-CS-LABEL: 'shufflevector_xxx'
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01uu = shufflevector <2 x i8> %vec0, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10uu = shufflevector <2 x i8> %vec0, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; VI-LABEL: 'shufflevector_xxx'
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01uu = shufflevector <2 x i8> %vec0, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10uu = shufflevector <2 x i8> %vec0, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+; VI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
+;
+; VI-CS-LABEL: 'shufflevector_xxx'
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01uu = shufflevector <2 x i8> %vec0, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10uu = shufflevector <2 x i8> %vec0, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  %shuf01uu = shufflevector <2 x i8> %vec0, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+  %shuf10uu = shufflevector <2 x i8> %vec0, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
   ret void
 }
 
-; GCN-LABEL: 'shufflevector_02_v2i16'
-; GCN: estimated cost of 2 for {{.*}} shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
-define amdgpu_kernel void @shufflevector_02_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %vaddr0, <2 x i16> addrspace(1)* %vaddr1) {
-  %vec0 = load <2 x i16>, <2 x i16> addrspace(1)* %vaddr0
-  %vec1 = load <2 x i16>, <2 x i16> addrspace(1)* %vaddr1
-  %shuf = shufflevector <2 x i16> %vec0, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
-  store <2 x i16> %shuf, <2 x i16> addrspace(1)* %out
+; Other shuffle cases
+define void @shuffle() {
+; GFX10-LABEL: 'shuffle'
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v7 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v9 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v10 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v11 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v12 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v13 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v14 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v15 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16 = shufflevector <2 x float> undef, <2 x float> undef, <2 x i32> <i32 1, i32 0>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v17 = shufflevector <4 x float> undef, <4 x float> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
+;
+; GFX10-CS-LABEL: 'shuffle'
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v7 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v9 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v10 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v11 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v12 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v13 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v14 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v15 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16 = shufflevector <2 x float> undef, <2 x float> undef, <2 x i32> <i32 1, i32 0>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v17 = shufflevector <4 x float> undef, <4 x float> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; GFX9-LABEL: 'shuffle'
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v7 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v9 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v10 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v11 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v12 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v13 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v14 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v15 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16 = shufflevector <2 x float> undef, <2 x float> undef, <2 x i32> <i32 1, i32 0>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v17 = shufflevector <4 x float> undef, <4 x float> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
+;
+; GFX9-CS-LABEL: 'shuffle'
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v7 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v9 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v10 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v11 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v12 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v13 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v14 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v15 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16 = shufflevector <2 x float> undef, <2 x float> undef, <2 x i32> <i32 1, i32 0>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v17 = shufflevector <4 x float> undef, <4 x float> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; VI-LABEL: 'shuffle'
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v7 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v9 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v10 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v11 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v12 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v13 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v14 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v15 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16 = shufflevector <2 x float> undef, <2 x float> undef, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v17 = shufflevector <4 x float> undef, <4 x float> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
+;
+; VI-CS-LABEL: 'shuffle'
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v7 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v9 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v10 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v11 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v12 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v13 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v14 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v15 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16 = shufflevector <2 x float> undef, <2 x float> undef, <2 x i32> <i32 1, i32 0>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v17 = shufflevector <4 x float> undef, <4 x float> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  %v7 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
+  %v8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+  %v9 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+  %v10 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+  %v11 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+  %v12 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+  %v13 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+  %v14 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
+  %v15 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+  %v16 = shufflevector <2 x float> undef, <2 x float> undef, <2 x i32> <i32 1, i32 0>
+  %v17 = shufflevector <4 x float> undef, <4 x float> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
   ret void
 }
 
-; GCN-LABEL: 'shufflevector_xxx'
-; TPT: estimated cost of 0 for {{.*}} shufflevector <2 x i8> %vec, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
-; CS: estimated cost of 0 for {{.*}} shufflevector <2 x i8> %vec, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
-; Should not assert
-define amdgpu_kernel void @shufflevector_xxx(<4 x i8> addrspace(1)* %out, <2 x i8> addrspace(1)* %vaddr) {
-  %vec = load <2 x i8>, <2 x i8> addrspace(1)* %vaddr
-  %shuf = shufflevector <2 x i8> %vec, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
-  store <4 x i8> %shuf, <4 x i8> addrspace(1)* %out
+define void @concat() {
+; GFX10-LABEL: 'concat'
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i32 = shufflevector <4 x i32> undef, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i64 = shufflevector <2 x i64> undef, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4f16 = shufflevector <2 x half> undef, <2 x half> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8f16 = shufflevector <4 x half> undef, <4 x half> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16f16 = shufflevector <8 x half> undef, <8 x half> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32 = shufflevector <2 x float> undef, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f32 = shufflevector <4 x float> undef, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f64 = shufflevector <2 x double> undef, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
+;
+; GFX10-CS-LABEL: 'concat'
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i32 = shufflevector <4 x i32> undef, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i64 = shufflevector <2 x i64> undef, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4f16 = shufflevector <2 x half> undef, <2 x half> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8f16 = shufflevector <4 x half> undef, <4 x half> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16f16 = shufflevector <8 x half> undef, <8 x half> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32 = shufflevector <2 x float> undef, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f32 = shufflevector <4 x float> undef, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f64 = shufflevector <2 x double> undef, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX10-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; GFX9-LABEL: 'concat'
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i32 = shufflevector <4 x i32> undef, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i64 = shufflevector <2 x i64> undef, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4f16 = shufflevector <2 x half> undef, <2 x half> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8f16 = shufflevector <4 x half> undef, <4 x half> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16f16 = shufflevector <8 x half> undef, <8 x half> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32 = shufflevector <2 x float> undef, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f32 = shufflevector <4 x float> undef, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f64 = shufflevector <2 x double> undef, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
+;
+; GFX9-CS-LABEL: 'concat'
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i32 = shufflevector <4 x i32> undef, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i64 = shufflevector <2 x i64> undef, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4f16 = shufflevector <2 x half> undef, <2 x half> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8f16 = shufflevector <4 x half> undef, <4 x half> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16f16 = shufflevector <8 x half> undef, <8 x half> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32 = shufflevector <2 x float> undef, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f32 = shufflevector <4 x float> undef, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f64 = shufflevector <2 x double> undef, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX9-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; VI-LABEL: 'concat'
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; VI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; VI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; VI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i32 = shufflevector <4 x i32> undef, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i64 = shufflevector <2 x i64> undef, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4f16 = shufflevector <2 x half> undef, <2 x half> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8f16 = shufflevector <4 x half> undef, <4 x half> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; VI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16f16 = shufflevector <8 x half> undef, <8 x half> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32 = shufflevector <2 x float> undef, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f32 = shufflevector <4 x float> undef, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f64 = shufflevector <2 x double> undef, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
+;
+; VI-CS-LABEL: 'concat'
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i32 = shufflevector <4 x i32> undef, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i64 = shufflevector <2 x i64> undef, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4f16 = shufflevector <2 x half> undef, <2 x half> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8f16 = shufflevector <4 x half> undef, <4 x half> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16f16 = shufflevector <8 x half> undef, <8 x half> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32 = shufflevector <2 x float> undef, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f32 = shufflevector <4 x float> undef, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f64 = shufflevector <2 x double> undef, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; VI-CS-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  %v4i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v8i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %v16i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  %v4i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v8i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %v16i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  %v4i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v8i32 = shufflevector <4 x i32> undef, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %v4i64 = shufflevector <2 x i64> undef, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v4f16 = shufflevector <2 x half> undef, <2 x half> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v8f16 = shufflevector <4 x half> undef, <4 x half> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %v16f16 = shufflevector <8 x half> undef, <8 x half> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  %v4f32 = shufflevector <2 x float> undef, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v8f32 = shufflevector <4 x float> undef, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %v4f64 = shufflevector <2 x double> undef, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   ret void
 }
-- 
2.7.4