[GHC] #9961: compile-time performance regression compiling genprimcode

[GHC]

#9961: compile-time performance regression compiling genprimcode
-------------------------------------+-------------------------------------
        Reporter:  hvr               |                   Owner:  simonpj
            Type:  bug               |                  Status:  new
        Priority:  normal            |               Milestone:  7.10.1
       Component:  Compiler          |                 Version:  7.10.1-rc1
      Resolution:                    |                Keywords:
Operating System:  Unknown/Multiple  |            Architecture:
 Type of failure:  Compile-time      |  Unknown/Multiple
  performance bug                    |               Test Case:
      Blocked By:                    |                Blocking:
 Related Tickets:                    |  Differential Revisions:
-------------------------------------+-------------------------------------
Description changed by hvr:

Old description:

> compiling `utils/genprimopcode` (where compiling the Alex-generated Lexer
> takes up the most amount of time) with GHC 7.8.4:
>
> ```
> $ time /opt/ghc/7.8.4/bin/ghc  -Rghc-timing --make -O Main.hs
> [1 of 5] Compiling ParserM          ( ParserM.hs, ParserM.o )
> [2 of 5] Compiling Lexer            ( Lexer.hs, Lexer.o )
> [3 of 5] Compiling Syntax           ( Syntax.hs, Syntax.o )
> [4 of 5] Compiling Parser           ( Parser.hs, Parser.o )
> [5 of 5] Compiling Main             ( Main.hs, Main.o )
> Linking Main ...
> <<ghc: 20742311240 bytes, 2092 GCs, 60174814/140258264 avg/max bytes
> residency (33 samples), 332M in use, 0.00 INIT (0.00 elapsed), 7.84 MUT
> (8.35 elapsed), 6.93 GC (6.93 elapsed) :ghc>>
>
> real    0m15.331s
> user    0m15.173s
> sys     0m0.172s
> ```
>
> (btw, is it ok that GC takes up so much?)
>
> the same with a GHC 7.10 snapshot takes *alot* longer:
>
> ```
> $ time /opt/ghc/7.10.1/bin/ghc  -Rghc-timing --make -O Main.hs
> [1 of 5] Compiling ParserM          ( ParserM.hs, ParserM.o )
> [2 of 5] Compiling Lexer            ( Lexer.hs, Lexer.o )
> [3 of 5] Compiling Syntax           ( Syntax.hs, Syntax.o )
> [4 of 5] Compiling Parser           ( Parser.hs, Parser.o )
> [5 of 5] Compiling Main             ( Main.hs, Main.o )
> Linking Main ...
> <<ghc: 147681534568 bytes, 37824 GCs, 203338614/375314176 avg/max bytes
> residency (538 samples), 876M in use, 0.001 INIT (0.001 elapsed), 81.374
> MUT (81.888 elapsed), 235.240 GC (235.136 elapsed) :ghc>>
>
> real    5m17.061s
> user    5m16.702s
> sys     0m0.499s
> ```
>
> ...and here the GC/MUT ratio is even more off.

New description:

 compiling `utils/genprimopcode` (where compiling the Alex-generated Lexer
 takes up the most amount of time) with GHC 7.8.4:

 {{{
 $ time /opt/ghc/7.8.4/bin/ghc  -Rghc-timing --make -O Main.hs
 [1 of 5] Compiling ParserM          ( ParserM.hs, ParserM.o )
 [2 of 5] Compiling Lexer            ( Lexer.hs, Lexer.o )
 [3 of 5] Compiling Syntax           ( Syntax.hs, Syntax.o )
 [4 of 5] Compiling Parser           ( Parser.hs, Parser.o )
 [5 of 5] Compiling Main             ( Main.hs, Main.o )
 Linking Main ...
 <<ghc: 20742311240 bytes, 2092 GCs, 60174814/140258264 avg/max bytes
 residency (33 samples), 332M in use, 0.00 INIT (0.00 elapsed), 7.84 MUT
 (8.35 elapsed), 6.93 GC (6.93 elapsed) :ghc>>

 real    0m15.331s
 user    0m15.173s
 sys     0m0.172s
 }}}

 (btw, is it ok that GC takes up so much?)

 the same with a GHC 7.10 snapshot takes *alot* longer:

 {{{
 $ time /opt/ghc/7.10.1/bin/ghc  -Rghc-timing --make -O Main.hs
 [1 of 5] Compiling ParserM          ( ParserM.hs, ParserM.o )
 [2 of 5] Compiling Lexer            ( Lexer.hs, Lexer.o )
 [3 of 5] Compiling Syntax           ( Syntax.hs, Syntax.o )
 [4 of 5] Compiling Parser           ( Parser.hs, Parser.o )
 [5 of 5] Compiling Main             ( Main.hs, Main.o )
 Linking Main ...
 <<ghc: 147681534568 bytes, 37824 GCs, 203338614/375314176 avg/max bytes
 residency (538 samples), 876M in use, 0.001 INIT (0.001 elapsed), 81.374
 MUT (81.888 elapsed), 235.240 GC (235.136 elapsed) :ghc>>

 real    5m17.061s
 user    5m16.702s
 sys     0m0.499s
 }}}

 ...and here the GC/MUT ratio is even more off.

--

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9961#comment:1>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler