building via gcc 4.2.x on SPARC/Solaris is very slow

[mm]

../../compiler/ghc-inplace -H16m -O -package-name  haskell-src-1.0.1 -O -Rghc-timing -fgenerics  
-package base -fgenerics -split-objs -hisuf p_hi -hcsuf p_hc -osuf p_o -prof   -c Language/Haskell/Syntax.hs -o Language/Haskell/Syntax.p_o  -ohi Language/Haskell/Syntax.p_hi
^C
<<ghc: 1692957264 bytes, 2394 GCs, 12868709/19559532 avg/max bytes residency (12 samples), 50M in use, 0.00 INIT (0.00 elapsed), 9.28 MUT (9.52 elapsed), 3.22 GC (3.50 elapsed) :ghc>>
make[3]: *** [Language/Haskell/Syntax.p_o] Fehler 1
make[2]: *** [all] Unterbrechung
make[1]: *** [all] Unterbrechung
make: *** [stage1] Unterbrechung

[igloo]

Are you sure it's hung? That file will take a comparatively long time to compile, especially as it's being compiled for profiling. If you are low on memory then it is also likely that that file is causing your machine to swap, and thus taking even longer.

What does the output of free and top -n 1 look like when it has hung?

[mm]

Thank you,

compiling it now again worked well. Maybe the system really started to swap. Probably this is a non-bug.

mm

[simonmar]

Submitter confirms we can close.

[maeder]

compiling this file takes really rather long on my poor machine.

Hardware Architecture: sparc Operating System: solaris10 Model: Sun Fire 280R Memory Size (MB): 3072 Processor Speed (MHz): 900

load averages:  1.07,  1.07,  1.06                                 15:03:32
57 processes:  54 sleeping, 2 running, 1 on cpu
CPU states:  0.0% idle, 93.2% user,  6.8% kernel,  0.0% iowait,  0.0% swap
Memory: 3072M real, 1617M free, 845M swap in use, 4450M swap free

  PID USERNAME THR PRI NICE  SIZE   RES STATE   TIME    CPU COMMAND
 8273 maeder     1   0   10  554M  541M run   148:17 96.66% cc1

Without profiling it takes already long (about 1 hour)

../../compiler/stage1/ghc-inplace -package-name haskell-src-1.0.1.2 -hide-all-packages -split-objs -i -idist/build/autogen -idist/build -i. -Idist/build -odir dist/build -hidir dist/build -stubdir dist/build -package base-3.0.2.0 -package pretty-1.0.0.0 -package array-0.1.0.0 -package haskell98-1.0.1.0 -O -XCPP -idist/build  -H16m -O -optc-mcpu=ultrasparc -opta-mcpu=ultrasparc -O -Rghc-timing -fgenerics -c Language/Haskell/Syntax.hs -o dist/build/Language/Haskell/Syntax.o  -ohi dist/build/Language/Haskell/Syntax.hi
<<ghc: 3161523500 bytes, 5142 GCs, 18526044/40419328 avg/max bytes residency (25 samples), 106M in use, 0.00 INIT (0.00 elapsed), 39.27 MUT (12245.73 elapsed), 22.81 GC (23.44 elapsed) :ghc>>

Can't this be changed, somehow?

[Isaac Dupree]

seeing cc1, maybe related to something I found here:  http://hackage.haskell.org/trac/ghc/ticket/2002#comment:12 , namely that GCC can't, performance-wise, deal with everything that GHC can generate. Is the time mostly spent in GHC or GCC? and are you having any trouble with amount of RAM/swap memory it uses, or just the amount of CPU-ish time spent?

[maeder]

It's just the CPU time spent by gcc.

[maeder]

The files are rather big:

       15051 Jun 13 12:30 /tmp/ghc8219_0/ghc8219_0.hscpp
    10497049 Jun 13 12:31 /tmp/ghc8219_0/ghc8219_0.p_hc
    15187968 Jun 13 15:50 /tmp/ghc8219_0/ghc8219_0.raw_s

[benl]

Current plan is to fix the SPARC native code generator and avoid GCC entirely.

[duncan]

Note also that using gcc-4.0.x or 4.1.x instead of 4.2.x avoids the massive compile times. See Building/Solaris.