[Openmcl-devel] Hardware requirements

[Marko Kocić]

Software emulation is good for me, as long as it would be possible to
generate sse2 executable from emulated ccl.

Do you have some guidance what / where should be changed in order to
create non-sse2 version?

Regards,
Marko Kocić

2009/3/10 Gary Byers <gb at clozure.com>:
> It's probably complaining about the lack of SSE2.  (Yes, it'd
> be nice if it told you exactly what required features were missing.)
> Most (but certainly not all) x86 CPUs sold in the last several
> years offer SSE2 support, but since there are exceptions we should
> mention this requirement.  (One notable SSE2-less chip that's
> used in new hardware is the AMD Geode, which is used in the OLPC.)
>
> CCL just uses SSE2 hardware for scalar floating-point arithmetic, and
> it -might- be possible/practical to emulate the handful (maybe two
> handfuls ...) of SSE2 instructions that it actually uses.  Getting
> all of the hard parts (rounding, exceptions, ...) right is probably
> a lot of work; getting enough right so that it'd at least run and
> do incidental floating-point and get something close to the right
> answer when it does is ... well, it's probably still a lot of work,
> but might be practical and useful.
>
> I don't know how likely that (emulation) is to happen or how soon
> it'd happen if it did, but it seems more attractive than it would
> be to have to support two binary-incompatible versions (SSE2 and x87)
> for every ia32 platform.
>
> In the meantime - we should add something to the front page
> of the website that clarifies this requirement; sorry that it
> wasn't there.
>
> On Tue, 10 Mar 2009, Marko KociÄ~G wrote:
>
>> What are hardware requirements for ccl?
>
> I tried svn version on both win xp and linux on my laptop, and in both
> cases I get "this processor does not support required features"
> message when I try to start the program.
>
> I have old Athlon XP 2400 processor. Would it be possible to build it
> some other way, or there is no chance to run it on this machine.
>
> Thanks,
> Marko Kocić
> _______________________________________________
> Openmcl-devel mailing list
> Openmcl-devel at clozure.com
> http://clozure.com/mailman/listinfo/openmcl-devel