Re: Re: [Bug-gsl] Improving gsl_blas_dsyrk routine

[王一]

The attachment is contains my source code and a 100*100 example of my matrix 
format.
to make the comparison significant you should use large matrix eg 600*6000 
following my matrix format.




王一
2008-04-16



发件人： Brian Gough
发送时间： 2008-04-16 03:52:32
收件人： 王一
抄送： bug-gsl
主题： Re: [Bug-gsl] Improving gsl_blas_dsyrk routine

At Sat, 12 Apr 2008 16:27:26 +0800,
=?GB2312?B?zfXSuw==?= wrote:
> 
> [1  <text/plain; gb2312 (base64)>]
> Dear GSL developers:
> I benifits a lot recently by using your powerful routines. Thus I became 
> interested in the detail of your routines. 
> When I look through the BLAS source code, I found  gsl_blas_dsyrk routine can 
> be significantly improved by a change of looping order. 
> As I'm just a Ph.D. student of Statistical Genetics, I'm not sure my finding 
> is generally true.
> I just wish to help the lovely GSL! Please let me know the result if possible.
> 
Thanks for your email, could you send some compilable programs (rather
than code fragments) so that we can reproduce the problem.
-- 
Brian Gough

syrk_test.rar
Description: Binary data