| Issue |
181308
|
| Summary |
RISC-V regression in nbench with unprofitable SLP vectorization
|
| Labels |
backend:RISC-V,
llvm:SLPVectorizer
|
| Assignees |
|
| Reporter |
lukel97
|
LNT picked up a ~2% regression in the nbench benchmark in llvm-test-suite between 6ffbb320d7578f0bed137a75e6d50ca0eb1e5ab4 and ac1220f860a8639881e28ba7824e90ba1155d4f4: https://cc-perf.igalia.com/db_default/v4/nts/profile/203/236/232
The DivideInternalFPF function has more SLP vectorized code now which seems likely unprofitable, and there's 15% more vector instructions executed overall.
cc) @alexey-bataev
![]()
_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
