Re: Speed/safety trade-offs and types vs. system classes
- From: chthon <jurgen.defurne@xxxxxxxxxx>
- Date: Fri, 20 Feb 2009 04:07:27 -0800 (PST)
On Feb 20, 11:28 am, p...@xxxxxxxxxxxxxxxxx (Pascal J. Bourguignon)
wrote:
There are chances that on an implementation running on a 32-bit
computer, (make-array dimensions :element-type '(unsigned-byte 32)
:initial-element 0) be optimized to use one 32-bit word per array
slot. But this doesn't matter at all. What matter, is that your
array will work on any CL implementation, be it running on a 17-bit
eletronic computer, or on banana skins.
--
__Pascal Bourguignon__
Well, I depend on this feature, mostly developing my code on CLISP for
cross-platform development between Linux and Windows, and then testing
my code (with possible optimisations) on SBCL and ECL.
Bu this is the real beauty of CL : that I can effectively develop
something working without regards to data types, but that I can
afterward compile it with SBCL for almost free 10-fold performance (if
one doesn't mind the memory that SBCL takes up in contrast with
CLISP).
Regards,
Jurgen
.
- References:
- Speed/safety trade-offs and types vs. system classes
- From: jurgen_defurne
- Re: Speed/safety trade-offs and types vs. system classes
- From: Pascal J. Bourguignon
- Speed/safety trade-offs and types vs. system classes
- Prev by Date: Re: Trying to make some sense of these HW questions... (Here is what I have so far)
- Next by Date: Re: Speed/safety trade-offs and types vs. system classes
- Previous by thread: Re: Speed/safety trade-offs and types vs. system classes
- Next by thread: Re: Speed/safety trade-offs and types vs. system classes
- Index(es):
Relevant Pages
|