[clisp-list] Re: FFI variable length arrays optimization

[Sam S. <sd...@gn...>]

> * Hoehle, Joerg-Cyril <Wbret-Plevy.Ubruyr@g-flfgrzf.pbz> [2004-09-03 10:32:25 +0200]:
>
> Sam Steingold asked:
>>well, whatever the means, I want ZLIB:COMPRESS and ZLIB:UNCOMPRESS in
>>the new ZLIB module to be as efficient as possible.
>
> Well, that means writing a module instead of using the FFI.
> With the FFI, there will always the overhead of parsing the internal
> c-type description of the foreign function.

I thought other lisps could compile foreign calls to something non-consing.

> In particular, if you want compress/uncompress as part of a "core"
> CLISP and possibly take advantage of them in other parts of core CLISP
> (e.g. transparent compressed streams), then the FFI is *not* the way
> to go.

of course.
making i/o pluggable (e.g., letting OPEN handle compressed streams and
SSL sockets) would take a rewrite of stream.d anyway.

>>if they were written in C using modprep.lisp they would not 
>>cons at all.
>>is it possible to make them non-consing while still using FFI?
>
> d) Even with a module, what would you do with the in and out buffers,
> do they count as consing? Would you always stack-allocate? Would you

both COMPRESS and UNCOMPRESS should take both input and output buffers
from the outside.

> use a fill-pointer to resize the output buffer to the actual size
> (trading array-copying time with wasted buffer space)?

yes.

-- 
Sam Steingold (http://www.podval.org/~sds) running w2k
<http://www.camera.org> <http://www.iris.org.il> <http://www.memri.org/>
<http://www.mideasttruth.com/> <http://www.honestreporting.com>
The world will end in 5 minutes.  Please log out.