[EMBOSS] EMBOSS 6.1.0 release now available

[Hamish McWilliam]

Hi Peter,

>> Any thought on implementing some of the algorithms using CUDA when 
>> possible on GPUs?  This could speed up some programs significantly.

Given that our server systems do not have particularly powerful GPUs, 
but do have multiple CPU cores: threading, and possibly the use of on 
core vectorization (see http://en.wikipedia.org/wiki/SIMD), seem like 
more generally applicable methods for improving performance in our case.

One interesting option for Intel platforms is the Intel Compiler (icc), 
which will vectorize some code constructions as a platform specific 
optimization. Unfortunately we are running a mixture of AMD and Intel 
systems of various vintages, so this option is going to require a lot of 
testing to check it works and gives us any benefits.

> Yes indeed.
> 
> At BOSC/ISMB last month we were discussing closer collaborations with 
> the other Open Bio FOundation projects.
> 
> One of these is BioManyCores which is aiming at OpenCL programming but 
> is currently concentrating on CUDA.
> 
> When our new workstations are delivered we will be looking into CUDA.

Given that OpenCL supports both GPU and CPU vectorization, and CUDA is 
Nvidia GPU specific, it may be worth waiting for OpenCL to be adopted. 
MacOS X Snow Leopard is only a couple of months away after all ;-)

> Which applications would you most like to speed up? (current EMBOSS 
> programs, and suggestions for new ones)

At our end the bottlenecks are mainly the indexing (dbi* & dbx*) and 
reformatting tools (seqret).

All the best,

Hamish