Archives Discussions

[rchoolun]

Hello all. I am currently looking to compare the Radeon Cypress architecture with the competitors GF100 in "typical" OpenCL compute calculations for my MSc thesis. I do need help defining "typical" OpenCL algorithms and am open to suggestions.

However I am new to this so would  really appreciate any pointers that the AMD GPU experts can give me.I know that the Radeon architecture as measured in FLOPS is superior to the competition and would like to leverage this power if possible.

I am familiar with Rahul Garg's MSc thesis on a compiler for Python on RV770 and am aware that Micah Villmow and Michael Chu of AMD assisted him. I would love any pointers that these leaders in their field could give me!

I would also be happy to submit a copy of my thesis to AMD for their review at some point. The thesis is due for September 2010 so I have some time yet to formulate my plans.

Kind Regards

[noxnet]

Rahul Garg's MSc thesis sounds interesting

Is it online available?

Where can i find it?

[mlxd]

Currently have applied for an MSc myself for something similar, would be curious to know if any assistance would be issued too.

[MicahVillmow]

mlxd,
Myself and other engineers monitor these forums along with beyond3D and a couple other forums, so asking your questions here will be a good place to get assistance from AMD and from the community.

[ryta1203]

rchoolun,

  I would look at some already known benchmarks for parallel computing (not necessarily GPU computing either), some standards, and start there.

  Also, from my point of view, if you are trying to compare architectures, I think it's paramount to tune the benchmarks to the specific architectures.

JMO.

EDIT: There is already a TON of info in these forums regarding AMD GPU architectures.