Performance regression on RISC-V for hexxagon benchmark in llvm-test-suite
Open
backend:RISC-V
- Dominant language
- LLVM
- Stars
- 40.5k
- Forks
- 18.7k
- PR merge metrics
- PR metrics pending
Description
Between cdc6a84c14ea07d66d9e388171b43429eae3397b and 0c64017c9924573b81f0a74e6c7c6e3f48f41047, [MultiSource/Applications/hexxagon/hexxagon](https://cc-perf.igalia.com/db_default/v4/nts/172/graph?test.49=3) is 14.6% slower on the Banana Pi F3: https://cc-perf.igalia.com/db_default/v4/nts/172
I haven't looked closely but the perf profile says there's many more vector instructions being executed, so it could be a case of unprofitable vectorization.
Contributor guide
Assessment
This issue has not been assessed yet.