#13739: Very slow linking of profiled executables -------------------------------------+------------------------------------- Reporter: duog | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.2.1-rc2 Resolution: | Keywords: Operating System: Linux | Architecture: x86_64 Type of failure: Compile-time | (amd64) performance bug | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by RyanGlScott): Thanks for the script, duog! It'll help me figure out what commit(s) introduced this regression. I did some quick testing on my Ubuntu machine with `gcc-4.8.4`. As baselines, I tested GHC 8.0.2: {{{ real 0m0.299s user 0m0.256s sys 0m0.028s }}} and GHC 8.2.1-rc2: {{{ (Without -fwhole-archive-hs-libs) real 0m12.946s user 0m12.816s sys 0m0.136s (With -fwhole-archive-hs-libs) real 0m2.676s user 0m2.436s sys 0m0.220s }}} I noticed a significant difference around commit b207b536ded40156f9adb168565ca78e1eef2c74 (Generalize kind of the (->) tycon). Before that commit, we have: {{{ Commit efeaf9e436109cb35b491e08b5407c0598108186 (Bump nofib submodule) ------ (Without -fwhole-archive-hs-libs) real 0m4.686s user 0m4.600s sys 0m0.068s (With -fwhole-archive-hs-libs) real 0m0.055s user 0m0.040s sys 0m0.008s }}} And after: {{{ Commit b207b536ded40156f9adb168565ca78e1eef2c74 (Generalize kind of the (->) tycon) ------ (Without -fwhole-archive-hs-libs) real 0m13.047s user 0m12.948s sys 0m0.104s (With -fwhole-archive-hs-libs) real 0m2.696s user 0m2.480s sys 0m0.196s }}} However, the times for commit efeaf9e436109cb35b491e08b5407c0598108186 are still not at parity with GHC 8.0.2, so an earlier commit must have contributed as well. I'll try to track down which one(s). -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/13739#comment:9 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler