gaspersic/triton
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[CODEGEN/DRIVER] Tweaks for performance optimization (#193)

Browse Source
This commit is contained in:
Philippe Tillet
2021-08-07 16:41:44 -07:00
committed by GitHub
parent 6cd1ec3955
commit 298da78058
8 changed files with 69 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
lib/codegen/analysis/align.cc
View File

@@ -331,6 +331,11 @@ std::vector<unsigned> align::populate_max_contiguous_cast(ir::cast_inst* v){
std::vector<unsigned> align::populate_max_contiguous(ir::value *v){
if(max_contiguous_.find(v) != max_contiguous_.end())
return max_contiguous_.at(v);
if(auto *x = dynamic_cast<ir::instruction*>(v)){
unsigned max_contiguous = x->get_metadata(ir::metadata::max_contiguous);
if(max_contiguous > 0)
return add_to_cache(x, {max_contiguous}, max_contiguous_);
}
if(auto *x = dynamic_cast<ir::cast_inst*>(v))
return populate_max_contiguous_cast(x);
if(auto *x = dynamic_cast<ir::splat_inst*>(v))