[GHC] #10108: Dramatic slowdown with -O2 bytestream and list streams combined.
GHC
ghc-devs at haskell.org
Wed Apr 15 00:31:06 UTC 2015
#10108: Dramatic slowdown with -O2 bytestream and list streams combined.
-------------------------------------+-------------------------------------
Reporter: Truman | Owner:
Type: bug | Status: new
Priority: normal | Milestone: 7.10.2
Component: Compiler | Version: 7.8.4
Resolution: | Keywords:
Operating System: Linux | Architecture:
Type of failure: Runtime | Unknown/Multiple
performance bug | Test Case:
Blocked By: | break_in.hs
Related Tickets: | Blocking:
| Differential Revisions:
-------------------------------------+-------------------------------------
Comment (by Truman):
Sorry about not including the library information. I should have known
better.
Here is a log showing compiling non-optimized, running, compiling
optimized, and running again, that shows the runtime difference. I
shortened the input file (described above) to 5000 lines to speed things
up, but the problem is still clear.
The versions of the two seemingly relevant imports are:
bytestring-0.10.4.0
stream-fusion 0.1.2.5
The log files detail lots of other information.
I'm running GHC 7.8.4 under Arch Linux 3.19.2-1-ARCH, and I have tested
and confirmed this on both 32 and 64 bit machines.
I will attach the two verbose compilation log files.
CPC~/ghc_prob> touch break_ins.hs
CPC~/ghc_prob> ghc -v5 break_ins.hs >ghc_nopt.log 2>&1
CPC~/ghc_prob> time break_ins <break_in.in >bi_nopt.out
real 0m0.077s
user 0m0.077s
sys 0m0.003s
CPC~/ghc_prob> touch break_ins.hs
CPC~/ghc_prob> ghc -v5 -O2 break_ins.hs >ghc_opt.log 2>&1
CPC~/ghc_prob> time break_ins <break_in.in >bi_opt.out
real 0m3.110s
user 0m3.090s
sys 0m0.017s
CPC~/ghc_prob>
--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10108#comment:2>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
More information about the ghc-tickets
mailing list