[llvm] r264364 - AMDGPU: TTI: Make insertelement free.

[Matt Arsenault via llvm-commits]

Author: arsenm
Date: Thu Mar 24 19:14:11 2016
New Revision: 264364

URL: http://llvm.org/viewvc/llvm-project?rev=264364&view=rev
Log:
AMDGPU: TTI: Make insertelement free.

We don't want to have a cost to scalarizing operations.

Added:
    llvm/trunk/test/Analysis/CostModel/AMDGPU/insertelement.ll
Modified:
    llvm/trunk/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp

Modified: llvm/trunk/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp?rev=264364&r1=264363&r2=264364&view=diff
==============================================================================
--- llvm/trunk/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp (original)
+++ llvm/trunk/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp Thu Mar 24 19:14:11 2016
@@ -99,6 +99,11 @@ int AMDGPUTTIImpl::getVectorInstrCost(un
                                       unsigned Index) {
   switch (Opcode) {
   case Instruction::ExtractElement:
+  case Instruction::InsertElement:
+    // Extracts are just reads of a subregister, so are free. Inserts are
+    // considered free because we don't want to have any cost for scalarizing
+    // operations, and we don't have to copy into a different register class.
+
     // Dynamic indexing isn't free and is best avoided.
     return Index == ~0u ? 2 : 0;
   default:

Added: llvm/trunk/test/Analysis/CostModel/AMDGPU/insertelement.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/CostModel/AMDGPU/insertelement.ll?rev=264364&view=auto
==============================================================================
--- llvm/trunk/test/Analysis/CostModel/AMDGPU/insertelement.ll (added)
+++ llvm/trunk/test/Analysis/CostModel/AMDGPU/insertelement.ll Thu Mar 24 19:14:11 2016
@@ -0,0 +1,37 @@
+; RUN: opt -cost-model -analyze -mtriple=amdgcn-unknown-amdhsa < %s | FileCheck %s
+
+; CHECK: 'insertelement_v2i32'
+; CHECK: estimated cost of 0 for {{.*}} insertelement <2 x i32>
+define void @insertelement_v2i32(<2 x i32> addrspace(1)* %out, <2 x i32> addrspace(1)* %vaddr) {
+  %vec = load <2 x i32>, <2 x i32> addrspace(1)* %vaddr
+  %insert = insertelement <2 x i32> %vec, i32 1, i32 123
+  store <2 x i32> %insert, <2 x i32> addrspace(1)* %out
+  ret void
+}
+
+; CHECK: 'insertelement_v2i64'
+; CHECK: estimated cost of 0 for {{.*}} insertelement <2 x i64>
+define void @insertelement_v2i64(<2 x i64> addrspace(1)* %out, <2 x i64> addrspace(1)* %vaddr) {
+  %vec = load <2 x i64>, <2 x i64> addrspace(1)* %vaddr
+  %insert = insertelement <2 x i64> %vec, i64 1, i64 123
+  store <2 x i64> %insert, <2 x i64> addrspace(1)* %out
+  ret void
+}
+
+; CHECK: 'insertelement_v2i16'
+; CHECK: estimated cost of 0 for {{.*}} insertelement <2 x i16>
+define void @insertelement_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %vaddr) {
+  %vec = load <2 x i16>, <2 x i16> addrspace(1)* %vaddr
+  %insert = insertelement <2 x i16> %vec, i16 1, i16 123
+  store <2 x i16> %insert, <2 x i16> addrspace(1)* %out
+  ret void
+}
+
+; CHECK: 'insertelement_v2i8'
+; CHECK: estimated cost of 0 for {{.*}} insertelement <2 x i8>
+define void @insertelement_v2i8(<2 x i8> addrspace(1)* %out, <2 x i8> addrspace(1)* %vaddr) {
+  %vec = load <2 x i8>, <2 x i8> addrspace(1)* %vaddr
+  %insert = insertelement <2 x i8> %vec, i8 1, i8 123
+  store <2 x i8> %insert, <2 x i8> addrspace(1)* %out
+  ret void
+}