[GHC] #10083: ghc: panic! (the 'impossible' happened)
GHC
ghc-devs at haskell.org
Tue Jul 21 16:09:00 UTC 2015
#10083: ghc: panic! (the 'impossible' happened)
-------------------------------------+-------------------------------------
Reporter: hedayaty | Owner:
Type: bug | Status: infoneeded
Priority: normal | Milestone:
Component: Compiler | Version: 7.8.4
Resolution: | Keywords:
Operating System: Unknown/Multiple | Architecture:
Type of failure: Compile-time | Unknown/Multiple
crash | Test Case:
Blocked By: | Blocking:
Related Tickets: | Differential Revisions:
-------------------------------------+-------------------------------------
Comment (by simonpj):
Actually comment:17 does '''not''' fix #10083, because I found that the
fix made the compiler go slower:
{{{
Unexpected stat failures:
perf/compiler T1969 [stat not good enough] (normal)
perf/compiler T9872a [stat not good enough] (normal)
perf/compiler T9872b [stat not good enough] (normal)
perf/compiler T9872c [stat not good enough] (normal)
perf/compiler T9872d [stat not good enough] (normal)
bytes allocated value is too high:
Deviation T1969(normal) bytes allocated: 7.5 %
Deviation T9872d(normal) bytes allocated: 9.3 %
Deviation T9872c(normal) bytes allocated: 8.4 %
}}}
by about 8% allocation overall. So the patch has all the code, but the key
bit is commented out.
See `Note [Inlining and hs-boot files]` in `TcBinds`. Presumably
disabling inlining of functions exported by a `hs-boot` file really kills
GHC's performance somewhere, implausible though it sounds. I suppose the
next thing to do is to find out why, but I just don't have time to do that
now, so I'm parking the whole thing. Sigh. Does anyone want to help?
--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10083#comment:18>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
More information about the ghc-tickets
mailing list