Build/performance trouble?

Build/performance trouble?

[Neil Jerram]

FYI, something seems awry in my automatic daily Guile build.

I noticed that the build machine is spending much longer churning away
than it normally does.  Today I found that this is while building the
master branch, and that it apparently spent all day trying
unsuccessfully to

  GUILEC language/tree-il/analyze.go

Also, looking at the benchmark graphs at
http://ossau.homelinux.net/~neil/bm_master_i.html, there's a bunch of
benchmarks related to latin-1 / UTF-8 that are apparently taking 100s of
times longer than they used to.  This applies both to master and
stable-2.0.

The latter may be a red herring, though, because my build log for
stable-2.0 indicates a backtrace in the middle of the benchmarking,
again related to latin-1:

(the interleaving of stdout and sterr here may not be significant)

...
("srfi-13.bm: strings: modification: comparison: hash hash-ci" 1000 user 0.24 benchmark 0.23968147277832 bench/interp 0.23968147277832 gc 0.0)
("srfi-13.bm: strings: searching: prefix-length suffix-length" 270 user 0.16 benchmark 0.159913997650146 bench/interp 0.159913997650146 gc 0.0)
("srfi-13.bm: strings: searching: prefix? suffix?" 270 user 0.16 benchmark 0.159913997650146 bench/interp 0.159913997650146 gc 0.0)
("srfi-13.bm: strings: searching: i;;; note: auto-compilation is enabled, set GUILE_AUTO_COMPILE=0
;;;       or pass the --no-auto-compile argument to disable.
;;; compiling /home/neil/guile-build-tracking/stable-2.0/benchmark-suite/guile-benchmark
;;; compiling /home/neil/guile-build-tracking/stable-2.0/benchmark-suite/lib.scm
;;; compiled /home/neil/guile-build-tracking/stable-2.0/cache/guile/ccache/2.0-LE-4-2.0/home/neil/guile-build-tracking/stable-2.0/benchmark-suite/lib.scm.go
;;; compiled /home/neil/guile-build-tracking/stable-2.0/cache/guile/ccache/2.0-LE-4-2.0/home/neil/guile-build-tracking/stable-2.0/benchmark-suite/guile-benchmark.go
;;; compiling /home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/0-reference.bm
;;; compiled /home/neil/guile-build-tracking/stable-2.0/cache/guile/ccache/2.0-LE-4-2.0/home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/0-reference.bm.go
;;; compiling /home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/arithmetic.bm
;;; compiled /home/neil/guile-build-tracking/stable-2.0/cache/guile/ccache/2.0-LE-4-2.0/home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/arithmetic.bm.go
;;; compiling /home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/bytevectors.bm
;;; compiled /home/neil/guile-build-tracking/stable-2.0/cache/guile/ccache/2.0-LE-4-2.0/home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/bytevectors.bm.go
;;; compiling /home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/chars.bm
;;; compiled /home/neil/guile-build-tracking/stable-2.0/cache/guile/ccache/2.0-LE-4-2.0/home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/chars.bm.go
;;; compiling /home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/continuations.bm
;;; compiled /home/neil/guile-build-tracking/stable-2.0/cache/guile/ccache/2.0-LE-4-2.0/home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/continuations.bm.go
;;; compiling /home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/if.bm
;;; compiled /home/neil/guile-build-tracking/stable-2.0/cache/guile/ccache/2.0-LE-4-2.0/home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/if.bm.go
;;; compiling /home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/logand.bm
;;; compiled /home/neil/guile-build-tracking/stable-2.0/cache/guile/ccache/2.0-LE-4-2.0/home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/logand.bm.go
;;; compiling /home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/ports.bm
;;; compiled /home/neil/guile-build-tracking/stable-2.0/cache/guile/ccache/2.0-LE-4-2.0/home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/ports.bm.go
;;; compiling /home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/r6rs-arithmetic.bm
;;; compiled /home/neil/guile-build-tracking/stable-2.0/cache/guile/ccache/2.0-LE-4-2.0/home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/r6rs-arithmetic.bm.go
;;; compiling /home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/read.bm
;;; compiled /home/neil/guile-build-tracking/stable-2.0/cache/guile/ccache/2.0-LE-4-2.0/home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/read.bm.go
;;; compiling /home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/srfi-1.bm
;;; compiled /home/neil/guile-build-tracking/stable-2.0/cache/guile/ccache/2.0-LE-4-2.0/home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/srfi-1.bm.go
;;; compiling /home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/srfi-13.bm
;;; compiled /home/neil/guile-build-tracking/stable-2.0/cache/guile/ccache/2.0-LE-4-2.0/home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/srfi-13.bm.go
;;; compiling /home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/strings.bm
;;; compiled /home/neil/guile-build-tracking/stable-2.0/cache/guile/ccache/2.0-LE-4-2.0/home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/strings.bm.go
Backtrace:
In ice-9/boot-9.scm:
 157: 14 [catch #t #<catch-closure 85e7e90> ...]
In unknown file:
   ?: 13 [catch-closure]
In ice-9/boot-9.scm:
  63: 12 [call-with-prompt prompt0 ...]
In ice-9/eval.scm:
 407: 11 [eval # #]
In benchmark-suite/guile-benchmark:
 211: 10 [main #]
In ice-9/boot-9.scm:
 696: 9 [for-each #<procedure 87ec200 at benchmark-suite/guile-benchmark:211:18 (benchmark)> ...]
In benchmark-suite/lib.scm:
 361: 8 [with-benchmark-prefix* "strings.bm" #<procedure 991dbf0 ()>]
In ice-9/boot-9.scm:
2189: 7 [save-module-excursion #<procedure 99a5820 at ice-9/boot-9.scm:3724:3 ()>]
3729: 6 [#<procedure 99a5820 at ice-9/boot-9.scm:3724:3 ()>]
In unknown file:
   ?: 5 [load-compiled/vm "/home/neil/guile-build-tracking/stable-2.0/cache/guile/ccache/2.0-LE-4-2.0/home/neil/guile-build-tracking/stable-2.0/benchmark-suite/benchmarks/strings.bm.go"]
In benchmark-suite/lib.scm:
 361: 4 [with-benchmark-prefix* "string ops" #<procedure a3333c0 ()>]
In benchmark-suite/benchmarks/strings.bm:
 165: 3 [#<procedure a3333c0 ()>]
In benchmark-suite/lib.scm:
 361: 2 [with-benchmark-prefix* "short Latin1" #<procedure 9fc40a0 ()>]
In benchmark-suite/benchmarks/strings.bm:
 191: 1 [#<procedure 9fc40a0 ()>]
In benchmark-suite/lib.scm:
 319: 0 [local-run-benchmark "French Latin-1 locale compare initially differing strings" ...]

benchmark-suite/lib.scm:319:14: In procedure local-run-benchmark:
benchmark-suite/lib.scm:319:14: Throw to key `unresolved' with args `()'.
ndex index-right rindex" 100000 user 2.21 benchmark 2.17814727783203 bench/interp 2.17814727783203 gc 0.0)
("srfi-13.bm: strings: searching: skip skip-right?" 100000 user 2.49 benchmark 2.45814727783203 bench/interp 2.45814727783203 gc 0.0)
("srfi-13.bm: strings: searching: count" 10000 user 15.67 benchmark 15.6668147277832 bench/interp 15.6668147277832 gc 0.0)
("srfi-13.bm: strings: searching: contains contains-ci" 34000 user 1.64 benchmark 1.62917007446289 bench/interp 1.62917007446289 gc 0.0)
("srfi-13.bm: strings: searching: upcase downcase upcase! downcase!" 600 user 0.17 benchmark 0.169808883666992 bench/interp 0.169808883666992 gc 0.0)
...

I probably won't have time to look further into this myself, but thought
you'd like to know.

      Neil

Re: Build/performance trouble?

[Neil Jerram]

Neil Jerram <neil@ossau.homelinux.net> writes:

> FYI, something seems awry in my automatic daily Guile build.
>
> I noticed that the build machine is spending much longer churning away
> than it normally does.  Today I found that this is while building the
> master branch, and that it apparently spent all day trying
> unsuccessfully to
>
>   GUILEC language/tree-il/analyze.go
>
> Also, looking at the benchmark graphs at
> http://ossau.homelinux.net/~neil/bm_master_i.html, there's a bunch of
> benchmarks related to latin-1 / UTF-8 that are apparently taking 100s of
> times longer than they used to.  This applies both to master and
> stable-2.0.
>
> The latter may be a red herring, though, because my build log for
> stable-2.0 indicates a backtrace in the middle of the benchmarking,
> again related to latin-1:

OK, I think the benchmark backtrace/performance is caused by me having
inadvertently uninstalled my French locales recently.  I've reinstalled
them now, so hopefully those benchmarks will return to normal
performance.

I doubt that's related to the problem compiling analyze.go though - I'll
see if I still see that tomorrow.

       Neil

Re: Build/performance trouble?

[Ludovic Courtès]

Hi Neil!

Thanks for keeping those benchs running!

Neil Jerram <neil@ossau.homelinux.net> skribis:

> Also, looking at the benchmark graphs at
> http://ossau.homelinux.net/~neil/bm_master_i.html, there's a bunch of
> benchmarks related to latin-1 / UTF-8 that are apparently taking 100s of
> times longer than they used to.  This applies both to master and
> stable-2.0.

Commit da35d2eaa9dbc1d3cf098c9a1c9bc62dcb2515bd changes ‘ports.bm’ in a
way that affects timings, which is what we observe on the graphs of
these specific benchmarks.

Not sure if there’s anything else alarming.

Thanks,
Ludo’.

Re: Build/performance trouble?

[Neil Jerram]

ludo@gnu.org (Ludovic Courtès) writes:

> Hi Neil!
>
> Thanks for keeping those benchs running!

No prob; it all just runs itself.

Except... that there does now seem to be a significant problem with
master's compilation time.  In practice this is now taking so long that
it has no chance of completing (on my build machine) within a day.
Hence for now I'm going to disable the master build.

> Neil Jerram <neil@ossau.homelinux.net> skribis:
>
>> Also, looking at the benchmark graphs at
>> http://ossau.homelinux.net/~neil/bm_master_i.html, there's a bunch of
>> benchmarks related to latin-1 / UTF-8 that are apparently taking 100s of
>> times longer than they used to.  This applies both to master and
>> stable-2.0.
>
> Commit da35d2eaa9dbc1d3cf098c9a1c9bc62dcb2515bd changes ‘ports.bm’ in a
> way that affects timings, which is what we observe on the graphs of
> these specific benchmarks.

Ah right, thanks for pointing that out.

     Neil

Re: Build/performance trouble?

[Andy Wingo]

On Fri 16 Mar 2012 19:44, Neil Jerram <neil@ossau.homelinux.net> writes:

> Except... that there does now seem to be a significant problem with
> master's compilation time.  In practice this is now taking so long that
> it has no chance of completing (on my build machine) within a day.
> Hence for now I'm going to disable the master build.

I have observed a bug on master whereby sometimes, not reproducibly, it
gets "stuck" doing something.  I think it's related to the
rob_from_rich() bug that Ludovic reported recently.  I hope to figure
out what it is soon.

Andy
-- 
http://wingolog.org/