On Mon, 19 Mar 2018 14:43:05 +0000
Joachim Hein <[email protected]> wrote:

> I am wondering how do we want to organise us in future?  Do we want to 
> continue with the goolfc idea or do we go for a “core” cuda and cuDNN?  I 
> feel this needs standardising soonish.

On this topic, does anyone have any opinions about making use of Magma instead 
of standard blas in whatever "cuda" toolchain we come up with?

Even further, is anyone willing to explore this magic:
http://spcl.inf.ethz.ch/Research/Parallel_Programming/Polly-ACC/
There was a talk at ISC claiming they achieve 60% of performance on unmodified 
C code compared to manually ported OpenACC code. 


-- 

  Jurij Pečar
  HPC Engineer, IT Operations, IT Services
  EMBL Heidelberg, Meyerhofstraße 1, 69117, Heidelberg, Germany
  Room 13-401

Reply via email to