[Openmcl-devel] Sudden increment in executable image size.

Kaz Kylheku kaz at kylheku.com
Fri Nov 19 02:53:53 UTC 2010


On Thu, 18 Nov 2010 18:54:20 -0700 (MST), Gary Byers <gb at clozure.com>
wrote:
> Just to take another stab in the dark: someone (possibly you) was asking
> about inline LAP a week or two ago and implied that they were defining
> LAP functions.

I gave up on that whole idea. The only low level stuff I use is a bit
of FFI, which is in another module (not the one being loaded when
that big heap usage occurs).

None of the program's code (other than macro expanders and whatever
read-time stuff there may be) is being executed, only loaded.

It's just: launching the image, doing an ASDF load operation to bring
in a bunch of fasls, and dumping the image.

The program seems to be working perfectly well, so I'm not considering
this to be a show stopper. The large size doesn't seem to have
any appreciable impact on the startup time or anything like that.
I think the program can be released in spite of the bloated heap
image. Solving the problem won't shrink the compressed installation
image, nor improve any user experience other than reduce
"sticker shock" when they look at the size. Though I will continue
to investigate the issue as an interesting problem and report any
findings.

Now that I'm familiar with rebuilding CCL, I should be able to
debug into it. I can perhaps add some instrumentation to the
fasl loading to pinpoint where the sudden heap increase occurs.




More information about the Openmcl-devel mailing list