[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Gcl-devel] Re: multiple values
From: |
Camm Maguire |
Subject: |
[Gcl-devel] Re: multiple values |
Date: |
25 Feb 2006 14:12:08 -0500 |
User-agent: |
Gnus/5.09 (Gnus v5.9.0) Emacs/21.2 |
Greetings, and thanks!
OK hee is an idea -- the first value can be passed in a register
(i.e. not written to memory at all) using Schelter's scheme. If
correct, then *eventually*, GCL might capitalize on this, as gcc
allows applications to reserve auxiliary registers for special
purposes like this. That is, assuming 2 and 3 values is the dominant
case.
Take care,
Robert Boyer <address@hidden> writes:
> One possible conjecture from long ago. One day Schelter figured out how to
> make use of PROCLAIM tell the compiler to make ordinary user function calls
> go about as fast as c function calls. However, at least at that time, this
> feature did not work (at least not well) for multiple value returning user
> functions (obviously, since C does not directly support multiple value
> returns). So Schelter gave to GCL some special feature for ACL2 that worked
> faster than ordinary GCL multiple value returns and took advantage of his
> fast PROCLAIM c-function call, since the ACL2 functions were all PROCLAIMed
> to return only one value.
>
> Long ago, I think that essentially all KCL user functions took and returned
> all their arguments and all their values via the value stack, and that can be
> slow.
>
> Just dim mumbling, vague recollections, restatement of the obvious.
>
> No reply expected or requested.
>
> Bob
>
>
>
>
--
Camm Maguire address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens." -- Baha'u'llah
[Prev in Thread] |
Current Thread |
[Next in Thread] |
- [Gcl-devel] Re: multiple values,
Camm Maguire <=