[Rd] Faster Blas Library

[Uwe Ligges]

Prof Brian Ripley wrote:
> I presume this is on Windows (you did not actually say).
> 
> The section http://cran.r-project.org/bin/windows/contrib/ is managed by 
> Uwe Ligges, and you could send him the Rblas.dll, the sources, a 
> description file and a license (the last being rather important if this 
> is to be hosted on CRAN).


Yes, please do so. We are really interested.

Note: I doubt that double prec. floating point ops are faster than 
single prec. (according to your slides).
Moreover, I'd like to see speed comparisons related to matrix sizes on 
some logarithmic scale in order to compare the quite relevant gains for 
small sized vectors.

Best wishes,
Uwe


> On Thu, 19 Feb 2009, sachin1234 wrote:
> 
>> Hi everyone,
>>
>> I have made a faster BLAS library thanks to Nvidia CUBLAS library. I was
>> wondering how I could upload this new Rblas.dll. I've included a 
>> powerpoint
>> presentation I made on the project.
>>
>> Highlights include upto 2000% improvement in matrix multiplication 
>> timings.
> 
>> Unfortunately the link included in the presentation is only accessible by
>> CSIRO employees only. I will gladly include the source code as well.
>>
>> However for the moment could someone simply tell me where and how to 
>> upload
>> these files
>>
>> Thanks
>> Sachin http://www.nabble.com/file/p22114095/R3.ppt R3.ppt
> 
>> -- 
>> View this message in context: 
>> http://www.nabble.com/Faster-Blas-Library-tp22114095p22114095.html
>> Sent from the R devel mailing list archive at Nabble.com.
>>
>> ______________________________________________
>> R-devel at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-devel
>>
>