| Issue |
153813
|
| Summary |
[X86] Avoid spilling whole vectors where only a subvector will be extracted after reload
|
| Labels |
backend:X86,
llvm:regalloc,
missed-optimization
|
| Assignees |
|
| Reporter |
RKSimon
|
An initial attempt was made here: #137316 - performing a subvector partial reload of a larger vector spill store - but that could lead to STLF issues.
What we need to do is perform the extract+store at the spill stage so we're only spilling/reloading the required subvector.
![]()
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
