6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2025-01-30 19:35:54 +00:00
Code Issues Projects Releases Wiki Activity

Resurrect r235688

Browse Source 
We should skip vector types which are not SCEVable.

test/CodeGen/NVPTX/sched2.ll passes


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235695 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Jingyue Wu 2015-04-24 04:22:39 +00:00
parent f42450abb6
commit 728ad0157c
4 changed files with 53 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
lib/Target/NVPTX/NVPTXTargetMachine.cpp
View File

@ -179,6 +179,8 @@ void NVPTXPassConfig::addIRPasses() {
addPass(createGVNPass());
else
addPass(createEarlyCSEPass());
// Run NaryReassociate after EarlyCSE/GVN to be more effective.
addPass(createNaryReassociatePass());
}
bool NVPTXPassConfig::addInstSelector() {

3
lib/Transforms/Scalar/NaryReassociate.cpp
View File

@ -181,7 +181,8 @@ bool NaryReassociate::doOneIteration(Function &F) {
Node != GraphTraits<DominatorTree *>::nodes_end(DT); ++Node) {
BasicBlock *BB = Node->getBlock();
for (auto I = BB->begin(); I != BB->end(); ++I) {
if (I->getOpcode() == Instruction::Add) {
// Skip vector types which are not SCEVable.
if (I->getOpcode() == Instruction::Add && !I->getType()->isVectorTy()) {
if (Instruction *NewI = tryReassociateAdd(I)) {
Changed = true;
SE->forgetValue(I);

2
test/Transforms/NaryReassociate/NVPTX/lit.local.cfg Normal file
View File

@ -0,0 +1,2 @@
if not 'NVPTX' in config.root.targets:
config.unsupported = True

47
test/Transforms/NaryReassociate/NVPTX/nary-slsr.ll Normal file
View File

@ -0,0 +1,47 @@
; RUN: opt < %s -slsr -nary-reassociate -S | FileCheck %s
; RUN: llc < %s -march=nvptx64 -mcpu=sm_20 | FileCheck %s --check-prefix=PTX
target datalayout = "e-i64:64-v16:16-v32:32-n16:32:64"
; foo((a + b) + c);
; foo((a + b * 2) + c);
; foo((a + b * 3) + c);
; =>
; abc = (a + b) + c;
; foo(abc);
; ab2c = abc + b;
; foo(ab2c);
; ab3c = ab2c + b;
; foo(ab3c);
define void @nary_reassociate_after_slsr(i32 %a, i32 %b, i32 %c) {
; CHECK-LABEL: @nary_reassociate_after_slsr(
; PTX-LABEL: .visible .func nary_reassociate_after_slsr(
; PTX: ld.param.u32 [[b:%r[0-9]+]], [nary_reassociate_after_slsr_param_1];
%ab = add i32 %a, %b
%abc = add i32 %ab, %c
call void @foo(i32 %abc)
; CHECK: call void @foo(i32 %abc)
; PTX: st.param.b32 [param0+0], [[abc:%r[0-9]+]];
%b2 = shl i32 %b, 1
%ab2 = add i32 %a, %b2
%ab2c = add i32 %ab2, %c
; CHECK-NEXT: %ab2c = add i32 %abc, %b
; PTX: add.s32 [[ab2c:%r[0-9]+]], [[abc]], [[b]]
call void @foo(i32 %ab2c)
; CHECK-NEXT: call void @foo(i32 %ab2c)
; PTX: st.param.b32 [param0+0], [[ab2c]];
%b3 = mul i32 %b, 3
%ab3 = add i32 %a, %b3
%ab3c = add i32 %ab3, %c
; CHECK-NEXT: %ab3c = add i32 %ab2c, %b
; PTX: add.s32 [[ab3c:%r[0-9]+]], [[ab2c]], [[b]]
call void @foo(i32 %ab3c)
; CHECK-NEXT: call void @foo(i32 %ab3c)
; PTX: st.param.b32 [param0+0], [[ab3c]];
ret void
}
declare void @foo(i32)