Hi Damian --

I know of a user who has a good start on the effort to wrap BLAS 
(specifically OpenBLAS) routines in Chapel, and we've been working to get 
him under a CLA so that he can contribute it back; if I've understood the 
status properly, it's likely that we'd want to do further Chapelification 
of the interface (similar to the recent work on wrapping FFTW).  I 
consider this a priority for version 1.12, and hope that the code will 
start showing up under GitHub within the next month or so...

-Brad



On Thu, 23 Apr 2015, Damian McGuckin wrote:

>
> Dear All,
>
> I know that some of what are Level 2 and Level 3 routines are given in the
> file 'hpl.chpl' of the benchmarks, has anybody got Routines, or a Module
> which implement BLAS for Chapel? It may just be an interface into some C
> routines, or as Chapel itself, especially given that many components of
> Level 1 BLAS can be implemented extremely cleanly, if sub-optimally as
>
>       NRM2
>
>               s = + reduce [i in x.domain] x[i] ** 2;
>
>               // The above can overflow & underflow badly, and more
>               // importantly has poor accuracy characteristics unless
>               // 'x' satisfies certain criteria and in practice, would
>               // and should be implemented in a more effective fashion.
>
>       ASUM
>
>               s = + reduce [i in x.domain] abs(x[i]); // can also overflow
>
>       AMAX
>
>               i = maxloc reduce [i in x.domain] abs(x[i]);
>
>       SWAP
>
>               x <=> y;
>
>       COPY
>
>               x = y
>
>       SCALE
>
>               x *= a
>
>
>       AXPY
>               y += a * x
>
> while for the implementations of
>
>       ROTG
>
>               Compute Given's Rotation
>
>       ROTMG
>
>               Compute Modified Given's Rotation
>
>       ROT
>
>               Apply Given's Rotation
>
>       ROTM
>
>               Apply Modified Given's Rotation
>
> need lots more code than 1-liners and
>
>       DOT
>
>               vector inner product
>
> while it can be expressed as a reduction, has lots of variants.
>
> Regards - Damian
>
> Pacific Engineering Systems International, 277-279 Broadway, Glebe NSW 2037
> Ph:+61-2-8571-0847 .. Fx:+61-2-9692-9623 | unsolicited email not wanted here
> Views & opinions here are mine and not those of any past or present employer
>
> ------------------------------------------------------------------------------
> One dashboard for servers and applications across Physical-Virtual-Cloud
> Widest out-of-the-box monitoring support with 50+ applications
> Performance metrics, stats and reports that give you Actionable Insights
> Deep dive visibility with transaction tracing using APM Insight.
> http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
> _______________________________________________
> Chapel-users mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/chapel-users
>

------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud 
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
Chapel-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/chapel-users

Reply via email to