|
|
Ublas : |
Subject: Re: [ublas] ATLAS binding: error when multiplying a complex matrix by a real matrix
From: Michael Lehn (michael.lehn_at_[hidden])
Date: 2012-09-28 12:18:55
Am 28.09.2012 um 18:05 schrieb Florent Teichteil:
> Thank you very much for your response.
>
> Actually, ublas does support it, at least multiplication of complex matrix by a real one (the other way gives a compilation error):
Ok, but compared to using an optimized BLAS implementation will beat
the generic implementation of ublas by far. For big matrices the
performance gain by using ATLAS will most certainly compensate the
extra copying (assuming you have enough memory and allocate the extra
matrix only once). I think this is at least worth trying.
Cheers,
Michael
>
> ublas::axpy_prod(C, R, O) -> works
> ublas::axpy_prod(R, C, O) -> does not work
>
> where C is complex, R is real, and O is complex
>
> Cheers
> Florent
>
> Le 28/09/2012 17:42, Michael Lehn a écrit :
>> Am 28.09.2012 um 17:22 schrieb Florent Teichteil:
>>
>>> Hi all,
>>>
>>> I am trying to multiply a complex matrix by a real one using gemm:
>>>
>>> atlas::gemm(m1, m2, r);
>>>
>>
>> This operation is not supported by ATLAS (or any other standard BLAS implementation).
>>
>> cheers,
>>
>> Michael
>>
>> _______________________________________________
>> ublas mailing list
>> ublas_at_[hidden]
>> http://lists.boost.org/mailman/listinfo.cgi/ublas
>> Sent to: florent.teichteil_at_[hidden]
>>
> _______________________________________________
> ublas mailing list
> ublas_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/ublas
> Sent to: michael.lehn_at_[hidden]
>
>