[SciPy-user] Initial support for sunperf and sun compilers (linux + solaris)

David Cournapeau david@ar.media.kyoto-u.ac...
Sun Sep 16 23:29:56 CDT 2007


Lucas Barbuto wrote:
> Hi David,
>
> Thanks for your continued interest in this.
>
> On 16/09/2007, at 9:39 PM, David Cournapeau wrote:
>   
>> Ok, I created a numpy branch to implement this, and get something  
>> working. This is still really rough, though. Please check out the  
>> numpy.sunperf branch:
>>
>> svn co http://svn.scipy.org/svn/numpy/branches/numpy.sunperf
>>     
>
> Just a quick correction, I had to remove branches/ from the above URL.
>   
Argh, I totally screwed up on this one, I created the branch at the 
wrong place.
>   
>> SUNPERF=SUNPERFROOT python setup.py build --compiler=sun -- 
>> fcompiler=sun
>>     
>
> I used your test_sunperf.sh script which completed the build without  
> errors, testing didn't go 100% smoothly, I don't know if those  
> warnings are serious, output below.
>   
There are several things missing in the current implementation: first, 
it should look for sunperf automatically, not using this SUNPERF thing, 
and also, the sun compilers do not use any flags. I am not sure if it is 
because I did something wrong to disable them, or if I need to add them 
anyway.

For the warning, the flags may be of some importance, I don't know. I 
don't remember having seen so many warnings on Linux, though. Would be 
interesting to know what you get on sparc. I should have said that for 
now, this is mainly random hacks on a platform I don't really know, so I 
would not use it without more testing first.

David


More information about the SciPy-user mailing list