[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Gcl-devel] address@hidden: Re: lstat]
From: |
Matt Kaufmann |
Subject: |
Re: [Gcl-devel] address@hidden: Re: lstat] |
Date: |
Thu, 31 Oct 2013 11:41:55 -0500 |
P.S. By the way, the function WILD-PATHNAME-P doesn't appear to be
defined in GCL 2.6.10pre, even the ANSI version, even though it's in
the Common Lisp standard. Maybe this isn't terribly important, though
it appears to be at least one reason that one of the ACL2 books isn't
certified using GCL, which in turn makes it impossible to build the
ACL2 manual. I can probably work around this by excluding some books
from GCL-based regressions.
-- Matt
Date: Thu, 31 Oct 2013 11:34:37 -0500
From: Matt Kaufmann <address@hidden>
Cc: address@hidden, address@hidden
Hi, Camm --
>> Passed without failure.
Excellent! Thanks for the great news.
Below are some recent results I have lying around, which are
encouraging. I got these from ACL2 regressions done using a
development copy of ACL2 (hence, post-6.3), running "make" with option
"-j 8" on an unloaded 64-bit linux machine. The comparisons aren't
perfect because the books change daily, but I think they're reasonable
numbers to compare. In short, the time is looking good -- some of the
slightly slower wall-clock time compared to CCL seems to be from the
"s" time, which could easily be from gcc, as you mentioned.
; Last night with ACL2 built on /p/bin/gcl-2.6.10pre (which you
; created on Oct. 16):
27806.329u 1106.597s 1:06:37.05 723.3% 0+0k 156680+6732192io 0pf+0w
; On Oct. 29 with ACL2 built on CCL Version 1.10-dev-r15915M-trunk:
26024.566u 415.489s 1:01:01.46 722.1% 0+0k 2768+1816744io 0pf+0w
; On Oct. 14 with ACL2 built on GCL 2.6.10pre that I built on Oct. 2
; (/projects/acl2/lisps/gcl/2.6.10pre/gcl/gcl/bin/gcl):
28960.009u 1056.846s 1:18:07.39 640.3% 0+0k 147264+6462792io 2pf+0w
-- Matt
From: Camm Maguire <address@hidden>
Cc: address@hidden, address@hidden
Date: Thu, 31 Oct 2013 12:19:07 -0400
Greetings!
Matt Kaufmann <address@hidden> writes:
> Hi, Camm --
>
> That's really great! (Thanks, Harsh!)
>
> I'll keep my fingers crossed on the regression....
>
Passed without failure.
So this leaves only one possible item before 2.6.10 release. On ia64,
sgc causes fread to hang. I might look at this briefly, or just disable
sgc on this machine and be done with it.
In any case, I'm wondering if you ever happened to do a timing after the
last batch of improvements. gcc is still the bottleneck in many cases,
but in spite of this I think you should see some improvement.
Take care,
--
Camm Maguire address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens." -- Baha'u'llah
_______________________________________________
Gcl-devel mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/gcl-devel