RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-23 19:23:23 +01:00
Code Issues Projects Releases Wiki Activity

[LV] Ignore more debug info.

Browse Source 
We can avoid doing some unnecessary work by skipping debug instructions
in a few loops. It also helps to ensure debug instructions do not
prevent vectorization, although I do not have any concrete test cases
for that.

Reviewers: rengolin, hsaito, dcaballe, aprantl, vsk

Reviewed By: rengolin, dcaballe

Differential Revision: https://reviews.llvm.org/D53091

llvm-svn: 344232
This commit is contained in:
Florian Hahn 2018-10-11 09:27:24 +00:00
parent 33073994e4
commit 8d61d25682
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
lib/Transforms/Vectorize/LoopVectorize.cpp
View File

@ -4655,7 +4655,7 @@ LoopVectorizationCostModel::getSmallestAndWidestTypes() {
// For each block.
for (BasicBlock *BB : TheLoop->blocks()) {
// For each instruction in the loop.
for (Instruction &I : *BB) {
for (Instruction &I : BB->instructionsWithoutDebug()) {
Type *T = I.getType();
// Skip ignored values.
@ -4893,7 +4893,7 @@ LoopVectorizationCostModel::calculateRegisterUsage(ArrayRef<unsigned> VFs) {
unsigned Index = 0;
for (BasicBlock *BB : make_range(DFS.beginRPO(), DFS.endRPO())) {
for (Instruction &I : *BB) {
for (Instruction &I : BB->instructionsWithoutDebug()) {
IdxToInstr[Index++] = &I;
// Save the end location of each USE.