Re: [Sbcl-devel] Non-linear slowdown loading eql-specialized methods

[Jeremy B. <jh...@it...>]

Juho Snellman <js...@ik...> writes:

> Jeremy Brown <jh...@it...> writes:
>
>> In sbcl-0.9.18-x86-linux, there seems to be a non-linear slowdown
>> loading files with large number of eql-specialized methods.  To
>> demonstrate the problem, untar the attached tgz, cd into demo, and run
>> 
>> sbcl --eval '(load "doit.lisp")'
>> 
>> Note that the timing for loading each of the files b1 through b4 goes
>> something like 2.4s, 11.7s, 32.8s, 62.3s (your milage will vary by
>> processor) even though each is loading 200 eql-specialized functions.
>
> This came up recently on sbcl-help [*]. The quick fix is:
>
>   (SETF SB-PCL::*MAX-EMF-PRECOMPUTATION-METHODS* 100)
>
> [*] <http://groups.google.com/group/sbcl-help-archive/browse_frm/thread/d878dceb20f2fe89/b780303e982b5ce3>
>
> -- 
> Juho Snellman

Ah, thank you very much.  Since bug report 281 is gone from the BUGS
file, I assume I don't need to worry about the problem where "setting
it to a non-NIL value is likely to render the system vulnerable to a
carefully crafted test along the lines of that report" mentioned here?

http://groups.google.com/group/sbcl-help-archive/browse_frm/thread/d878dceb20f2fe89/b780303e982b5ce3

Thanks again, and sorry for the duplicate report.

Jeremy