Re: [Axiom-developer] Re: lisp

[Camm Maguire]

Greetings!

root <address@hidden> writes:

> The second issue, which I raised, is that Axiom is faster after
> the second system build due to the .fn files. The build cycle
> can (and used to) deal with this issue. It will again in the
> future when time permits.
> 

Thanks for bringing this to my attention. So I take it the build so
far does not take advantage of any proclamation information?  The
addition of this information can make a big difference to
performance.  I'm interested in the mechanism AXIOM used in the past
to make use of the .fn files.  What we do in GCL proper (on occasion,
not with every build, yet), is to emit all fn files in a given
directory, then make-all-proclaims, producing one big
sys-proclaims.lisp, which is then loaded automatically if it exists by
a direction in a gcl_cmpinit.lsp file in the same directory.  I can't
really think of a better way of doing it given that functions calling
each other can be in different files.  Indeed more, they can easily be
in different directories or even 'layers'.  We've recently added an
improvement in loading the sys-proclaims.lisp file for every compiled
module directory right before the save-system, so that this
information is preserved in the lisp image when moving on to the next
layer.  Still, if anyone can think of a more transparent way to do
this I'd be most interested.  I'd also be interested in looking at
what needs to be done to restore AXIOM's older ability to inhale the
.fn files.

Take care,

> The reason I raised the second issue is that Camm has made some
> performance improvements.
> 
> Tim
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah