OccAnal loop-breaker scoring for NOINLINE things
When working on #14566 (closed) I found a case where the occurrence analyser generated
Rec { {-# NOINLINE f #-}
f = e1[g]
; g {-# LOOPBREAKER #-} = e2[f] }
That is, even though f is marked NOINLINE we chose g to be the loop breaker. Stupid! If f is marked NOINLINE it would be a much better loop breaker.
Trac metadata
| Trac field | Value |
|---|---|
| Version | 8.2.1 |
| Type | Bug |
| TypeOfFailure | OtherFailure |
| Priority | normal |
| Resolution | Unresolved |
| Component | Compiler |
| Test case | |
| Differential revisions | |
| BlockedBy | |
| Related | |
| Blocking | |
| CC | |
| Operating system | |
| Architecture |