Re: Southport + Seeding FPGA application acceleration in our community
Hi Tony,
On 10/27/2011 01:14 PM, Tony Travis wrote:
> On 26/10/11 18:12, Steffen Möller wrote:
>> Heyho,
>
> During the 1980's, I commissioned a 9216 processor CLIP4R (Cellular
> Logic Image Processor) SIMD array built for the MRC by the Rutherford
> Appleton labs. This used FPGA's and custom LSI processors to run IPC,
> which was a customised version of C for image processing (not
> Inter-Process Communication).
>
> I spent three years learning to program this machine in the CLIP4R
> assembly language and IPC. I also worked with my colleagues on
> alternative loosely-coupled 680x0-based VME systems that we compared
> with the CLIP4R processor array. We concluded that it was not worth
> the huge effort of using such highly specialised systems that only a
> few people in the world, including me, knew how to program.
Fascinating.
> I know that things have moved on since then, but the principle remains
> the same. Unless there are many people using the same technology, the
> task of developing for it is extremely difficult. That said, using
> FPU's for SIMD not to mention the GPU's has come into the mainstream.
>
> My experience was that the (huge) performance advantage of the CLIP4R
> system was overwhelmed by the difficulty of actually using it to do a
> job analysing images of metaphase chromosomes...
>
> I would stick to using GPU's and CUDA/OpenCL if you want to improve
> the performance of your bioinformatics applications.
Maybe. Maybe not. This will all depend on what we can share among
ourselves and about what there is out there beyond VHDL and Verilog to
get things done. My personal interest is not so much the immediate
acceleration for a given problem. We get some additional freedom. What
exactly to do first or later with it, we will see. In your overview you
mention quite a number of concerns that underline the importance of a
community to backup any such developments. Can you join on Friday?
Best,
Steffen
Reply to: