From: Neil Jerram <neil@ossau.uklinux.net>
To: Andy Wingo <wingo@pobox.com>
Cc: guile-devel <guile-devel@gnu.org>
Subject: Re: statprof working with guile-vm
Date: Thu, 05 Feb 2009 21:05:39 +0000 [thread overview]
Message-ID: <8763joiogs.fsf@arudy.ossau.uklinux.net> (raw)
In-Reply-To: <m3skmv59h5.fsf@pobox.com> (Andy Wingo's message of "Wed\, 04 Feb 2009 01\:31\:02 +0100")
Andy Wingo <wingo@pobox.com> writes:
> No. Statprof reports procedures by identity (i.e. `eq?'), not by name;
> but it discards procedures that have no names on the grounds that you
> don't really know what those procedures are.
So why were there so many entries for `retrans', then? (I assume
there's really one `retrans' procedure in the system?)
> That commit was to give names to procedures bound like (letrec ((foo
> (lambda ...)))).
OK.
> But come to think of it, we should be able to do better, reporting based
> on identity of source location (as determined for example by eq? on
> program-objcode, or on interpreted closure code) instead of on procedure
> (closure) identity.
Yes, that could be nice.
>> Where is statprof at the moment?
>
> Guile-lib
>
>> If it's now unlikely to change dramatically, it feels to me that it
>> should move into Guile core. What do you think?
>
> I agree, though there are some documentation things to take care of. It
> was only written by Rob Browning, and modified a bit by myself.
I'll add it to my list too, then. (But it's a long list...)
> Statprof on the vm is slightly different due to tail recursion -- the
> debugging evaluator keeps some of those frames on the debug stack.
> Surprising.
That the evaluator keeps debug info for those frames, you mean?
> Also here's the current GOOPS profile:
>
> scheme@(guile-user)> (use-modules (statprof))
> scheme@(guile-user)> (with-statprof #:hz 1000 (resolve-module '(oop goops)))
> % cumulative self
> time seconds seconds name
> 18.18 0.10 0.04 for-each
> 7.58 0.03 0.02 byte-length
> 7.58 0.02 0.02 record-predicate
> 4.55 0.06 0.01 glil->assembly
> 3.03 0.21 0.01 load-compiled/vm
> 3.03 0.08 0.01 write-bytecode
> 3.03 0.01 0.01 lookup-transformer
> 3.03 0.01 0.01 ghil-env-add!
> 3.03 0.01 0.01 list-index
> 3.03 0.01 0.01 cache-try-hash!
> 1.52 0.05 0.00 map
> 1.52 0.01 0.00 lp
> 1.52 0.01 0.00 make-glil-program
> 1.52 0.01 0.00 eqv?
> 1.52 0.01 0.00 %init-goops-builtins
> [...]
> 0.00 0.21 0.00 dynamic-wind
> 0.00 0.20 0.00 memoize-method!
> 0.00 0.19 0.00 compute-entry-with-cmethod
> 0.00 0.19 0.00 compile-fold
> 0.00 0.19 0.00 compile-method/vm
> 0.00 0.18 0.00 make-instance
> 0.00 0.18 0.00 initialize
> 0.00 0.08 0.00 compute-cmethod
> 0.00 0.08 0.00 %goops-loaded
> 0.00 0.08 0.00 make-extended-generic
> 0.00 0.08 0.00 make-next-method
> 0.00 0.08 0.00 compile-bytecode
> 0.00 0.06 0.00 ensure-generic
> 0.00 0.06 0.00 compile-asm
> 0.00 0.06 0.00 compile-assembly
> 0.00 0.04 0.00 save-module-excursion
> 0.00 0.03 0.00 compile-glil
> 0.00 0.03 0.00 codegen
> 0.00 0.02 0.00 call-with-ghil-bindings
> 0.00 0.02 0.00 load-file
> 0.00 0.02 0.00 add-method!
> 0.00 0.02 0.00 translate-1
> 0.00 0.01 0.00 record-accessor
> [...]
> Sample count: 66
> Total time: 0.21 seconds (2/25 seconds in GC)
>
> I think the salient point here is that out of 0.21 seconds of loading
> GOOPS, 0.19 of it is spent in the compiler (see `compile-fold').
Do you mean that this is the VM compiler compiling closures that GOOPS
creates dynamically?
(As is probably clear, I don't think I understand yet!)
Regards,
Neil
prev parent reply other threads:[~2009-02-05 21:05 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-02-02 22:45 statprof working with guile-vm Andy Wingo
2009-02-03 22:57 ` Neil Jerram
2009-02-04 0:31 ` Andy Wingo
2009-02-05 21:05 ` Neil Jerram [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
List information: https://www.gnu.org/software/guile/
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=8763joiogs.fsf@arudy.ossau.uklinux.net \
--to=neil@ossau.uklinux.net \
--cc=guile-devel@gnu.org \
--cc=wingo@pobox.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).