[Bug tree-optimization/126376] aarch64: cunrolli unrolls small constant-trip inner loops and blocks dot-product vectorization
"konstantinos.eleftheriou at vrull dot eu via Gcc-bugs" <[email protected]>
| Newsgroups | gmane.comp.gcc.bugs |
|---|---|
| Message-ID | <[email protected]/bugzilla/> |
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126376 --- Comment #4 from Konstantinos Eleftheriou <konstantinos.eleftheriou at vrull dot eu> --- (In reply to Drea Pinski from comment #3) > (In reply to Konstantinos Eleftheriou from comment #1) > > We are preparing a patch for the suggested fix. > > except that is just a workaround. True. The problem is that after the unrolling we lose the reduction status. Teaching SLP to recognize DOT_PROD patterns in straight-line code will be a much more complicated solution with a higher regression risk. We went with this approach as a simple, low-risk workaround for now.