On 5/31/10 8:38 AM, Olive wrote:
> On Mon, 31 May 2010 07:39:30 -0400
> Raymond Toy <toy.raymond at gmail.com> wrote:
>
>
>> On 5/31/10 1:17 AM, Olive wrote:
>>
>>> On Sun, 30 May 2010 09:45:24 -0400
>>> Raymond Toy <toy.raymond at gmail.com> wrote:
>>>
>>>
>>>
>>>> On 5/30/10 1:45 AM, Olive wrote:
>>>>
>>>>
>>>>> I'm trying to use maxima (5.21.1) with cmucl. I've tried 20a and
>>>>> 19f (using binaries from the cmucl page) but although maxima
>>>>> builds, it does not pass the test suite (I do not remember the
>>>>> errors, it varies according to the lisp version, but I could
>>>>> reproduce it if necessary). Are these errors known? Does someone
>>>>> know if there is a specific version I have to use?
>>>>>
>>>>>
>>>>>
>>>> I have regularly used 19f and 20a (but not recently). AFAIK, cmucl
>>>> works very well with maxima. I currently use the 2010-05 snapshot
>>>> of cmucl with maxima. If you could send information about the
>>>> failed tests and the version, I'm sure we can figure out what is
>>>> going on.
>>>>
>>> Here is the output of the tests with cmucl 20a (using the official
>>> binary cmucl-20a-non-unicode-x86-linux.tar.bz2 and
>>> cmucl-20a-non-unicode-x86-linux.extra.tar.bz2) and maxima 5.21.1
>>> (compiled from source) (I had less errors with cmucl 19f but there
>>> was).
>>>
>> Thanks for the test results. Some of the failures in rtest14 look
>> strange. I don't remember seeing anything like that. I'll have to
>> try this with 20a and 5.21.1, and the non-unicode version. (I
>> normally use the unicode version.)
>>
>>
>> The ones in rtest_gamma look somewhat familiar. Can you tell me if
>> you were running the tests using the sse2 version of cmucl? (Just
>> start up cmucl and look to see what core is being used. If it's
>> lisp-sse2.core, then you're using the sse2 version). If you are,
>> there are known bugs in arithmetic in the sse2 version. These were
>> fixed in the 2009-11 and 2010-04 snapshots. If you don't mind, could
>> you try again with the latest (2010-06) snapshot from common-lisp.net?
>>
> I was indeed using the lisp-sse2 core. I have rebuild maxima using the
> lisp-x87.core (setting the CMUCLCORE environment variable) and there are
> no errors anymore.
>
Fantastic! I think the sse2 version is somewhat faster to much faster
on floating point code, but that's not usually an issue in maxima.
> I am unsure how to get the 2009-11 or 2010 snapshots but I am happy to
> use the x87.core (If you really want me to test the 2010-04 could you
> tell me in a line which command I have to type to get it, using svn?
> cvs?).
>
No, you don't need to test the snapshots. I regularly build/test maxima
with them myself, but if you want a snapshot, go to
<http://common-lisp.net/project/cmucl/mirror/download.html>; to get a
snapshot and see
<http://common-lisp.net/project/cmucl/mirror/install.html>; for
installation instructions.
Ray