I might have time to take a look at this, but I'll need a bit more
#14880: GHC panic: updateRole -------------------------------------+------------------------------------- Reporter: RyanGlScott | Owner: goldfire Type: bug | Status: new Priority: normal | Milestone: 8.8.1 Component: Compiler (Type | Version: 8.2.2 checker) | Resolution: | Keywords: TypeInType Operating System: Unknown/Multiple | Architecture: Type of failure: Compile-time | Unknown/Multiple crash or panic | Test Case: Blocked By: | Blocking: Related Tickets: #15076 | Differential Rev(s): Phab:D4769 Wiki Page: | -------------------------------------+------------------------------------- Comment (by tdammers): Replying to [comment:106 goldfire]: direction. You say that you might have missed something -- what two branches can I compare to assess that? I think you may be missing some context here, so here's the executive summary: - Your original patch is on `wip/T14880` - The baseline HEAD from right before your patch is on `wip/T14880-baseline` - Simon's patch (as mentioned above) is on `wip/T1448-accum` - `wip/T14880-reengineered` contains your patch (as on `wip/T14880`), with the approach from Simon's patch applied practically verbatim. Now here's the conundrum: - The baseline version performs within limits - Your patch applied as-is fails two performance tests: 5321Fun and 5631 - Simon's patch improves performance on 5631 - But Simon's approach applied to your patch doesn't
As for the one-line change: it's seems conceivable that the change could have a measurable performance impact. Have you tested the change, independent of anything else? If the fix makes performance worse, that would be an insight, because the `acc` check looks like it's only an optimization, not an essential check for correctness.
It is only an optimization, but that's the whole point of digging further here - the patch allocates more even though we would it expect it to allocate less. AFAIK there are no correctness issues anymore. The patch without the `acc` check fails the two performance tests, but so does the version with the check. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/14880#comment:107 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler