GNU bug report logs - #1334
23.0.60; bug of bytecomp arithmetic operations

Previous Next

Full log

Message #46 received at 1334-done <at> emacsbugs.donarmstrong.com (full text, mbox):

From: "Shigeru Fukaya" <shigeru.fukaya <at> gmail.com>
To: "Chong Yidong" <cyd <at> stupidchicken.com>
Cc: 1334-done <at> debbugs.gnu.org
Subject: Re: 23.0.60; bug of bytecomp arithmetic operations
Date: Sun, 23 Nov 2008 14:38:41 +0900

>> *** Optimize addtion/subtraction where their last operand is 1 or
>> -1 to use `1+', `1-' funcitons. This pattern often appears and
>> more effective at least in size (byte-add1 vs constant 1).
>
> Let's leave this till after the release.  It's not clear to me that
> there is a significant speed increase from it, and we should try to
> restrict the changes to bug fixes for the moment.

I myself was actually irresolute.

But, at least, for three operands I think it it worths
optimizing. For more than three operands, only expectation of
save of storage(for constant 1), and probability of the later
re-optimization induced by decresing of the number of operands.


(+ a b 1)

no optimization     optimazation
0 constant  +       varref    a
1 varref    a       varref    b
2 varref    b       plus
3 constant  1       add1
4 call      3       return
5 return


(+ a b c 1)
0 constant  +       constant  +
1 varref    a       varref    a
2 varref    b       varref    b
3 varref    c       varref    c
4 constant  1       call      3
5 call      4       add1
6 return            return

(byte-compile (lambda () (+ a b c 1)))
#[nil "\303\b\t\n\304$\207" [a b c + 1] 5]
#[nil "\303\b\t\n#T\207"    [a b c +] 4]

Anyway, please do what you think is best now.

Regards,
Shigeru

2008/11/22 Chong Yidong <cyd <at> stupidchicken.com>:
> I've reviewed your patch, and checked most of it into CVS (I took the
> liberty of rewording some comments, plus some minor edits).
>
>> *** Optimize addtion/subtraction where their last operand is 1 or
>> -1 to use `1+', `1-' funcitons. This pattern often appears and
>> more effective at least in size (byte-add1 vs constant 1).
>
> Let's leave this till after the release.  It's not clear to me that
> there is a significant speed increase from it, and we should try to
> restrict the changes to bug fixes for the moment.
>
> Thanks very much for working on this.
>

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.