Comments for GPU Software Blog http://blog.accelereyes.com/blog Helpful posts about GPU computing. Discussion of Jacket and ArrayFire. Real speedups on real code! Wed, 02 Feb 2011 15:32:00 +0000 hourly 1 http://wordpress.org/?v=3.3.1 Comment on Beam Propagation Methods – Jacket is 3.5X faster than the CPU and 2X faster than PCT by AccelerEyes http://blog.accelereyes.com/blog/2010/11/05/beam_propagation_methods/comment-page-1/#comment-323 AccelerEyes Wed, 02 Feb 2011 15:32:00 +0000 http://blog.accelereyes.com/blog/?p=848#comment-323 You're right... the file had bad permissions. It is fixed now. Enjoy! You’re right… the file had bad permissions. It is fixed now. Enjoy!

]]> Comment on Beam Propagation Methods – Jacket is 3.5X faster than the CPU and 2X faster than PCT by Vic http://blog.accelereyes.com/blog/2010/11/05/beam_propagation_methods/comment-page-1/#comment-322 Vic Tue, 01 Feb 2011 19:52:00 +0000 http://blog.accelereyes.com/blog/?p=848#comment-322 Hey! Link for "CUDA_BPM_NOV_04_2010_AccelerEyes" is not working!!! How to get the file? Hey! Link for “CUDA_BPM_NOV_04_2010_AccelerEyes” is not working!!! How to get the file?

]]> Comment on LIBJACKET on Amazon EC2 GPU Cloud Instances by Jeff Nucciarone http://blog.accelereyes.com/blog/2010/12/14/libjacket-on-amazon-ec2-gpu-cloud/comment-page-1/#comment-321 Jeff Nucciarone Tue, 14 Dec 2010 22:12:00 +0000 http://blog.accelereyes.com/blog/?p=907#comment-321 Use an elastic ip (yet another nickel and dime service, but relatively cheap) to associate your instance with a fixed address. I use one so I only have to poke holes in my local firewall for a specific address. Use an elastic ip (yet another nickel and dime service, but relatively cheap) to associate your instance with a fixed address. I use one so I only have to poke holes in my local firewall for a specific address.

]]> Comment on Median Filtering: CUDA tips and tricks by James Malcolm http://blog.accelereyes.com/blog/2010/03/04/median-filtering-cuda-tips-and-tricks/comment-page-1/#comment-320 James Malcolm Tue, 14 Dec 2010 04:07:00 +0000 http://www.accelereyes.com/blog/?p=190#comment-320 Hi Indalo-- You're right -- it gets much tougher beyond 3x3. It was tedious, but we scaled up the algorithm to sort sets and subsets until 15x15 size kernels. Do you need bigger, or were you just asking? Here's more documentation: http://wiki.accelereyes.com/wiki/index.php/MEDFILT2 Hi Indalo–
You’re right — it gets much tougher beyond 3×3. It was tedious, but we scaled up the algorithm to sort sets and subsets until 15×15 size kernels. Do you need bigger, or were you just asking? Here’s more documentation: http://wiki.accelereyes.com/wiki/index.php/MEDFILT2

]]> Comment on Median Filtering: CUDA tips and tricks by James Malcolm http://blog.accelereyes.com/blog/2010/03/04/median-filtering-cuda-tips-and-tricks/comment-page-1/#comment-319 James Malcolm Tue, 14 Dec 2010 04:06:00 +0000 http://www.accelereyes.com/blog/?p=190#comment-319 Hey Nick-- You're right -- there's likely a bug in this demo code. I just checked the production version currently in Jacket --- running in a loop, various sizes,etc. -- and it checks out with some nice speedups. Hey Nick–
You’re right — there’s likely a bug in this demo code. I just checked the production version currently in Jacket — running in a loop, various sizes,etc. — and it checks out with some nice speedups.

]]> Comment on GPU accelerated lattice Boltzmann model for shallow water flow and mass transport by Gallagher Pryor http://blog.accelereyes.com/blog/2010/12/01/lattice_boltzmann_model/comment-page-1/#comment-318 Gallagher Pryor Fri, 03 Dec 2010 18:30:00 +0000 http://blog.accelereyes.com/blog/?p=889#comment-318 Were the visualizations done in MATLAB? The axes on the plots (videos) look very crisp. Were the visualizations done in MATLAB? The axes on the plots (videos) look very crisp.

]]> Comment on GPU accelerated lattice Boltzmann model for shallow water flow and mass transport by Anonymous http://blog.accelereyes.com/blog/2010/12/01/lattice_boltzmann_model/comment-page-1/#comment-317 Anonymous Thu, 02 Dec 2010 16:13:00 +0000 http://blog.accelereyes.com/blog/?p=889#comment-317 Thanks for the comment! Two responses: * Ha, yes, you are right! I have a hard time turning it off. At least I didn't say it was super-duper-uber-fantastically-magically-splendidly good stuff :) BTW, the scientific paper is much more scientifically pure and objective! * Microsoft products: Excel for the chart and PPT for the border and green thing. MS really stepped up their game with Office 2010. Thanks for the comment!

Two responses:
* Ha, yes, you are right! I have a hard time turning it off. At least I didn’t say it was super-duper-uber-fantastically-magically-splendidly good stuff :) BTW, the scientific paper is much more scientifically pure and objective!
* Microsoft products: Excel for the chart and PPT for the border and green thing. MS really stepped up their game with Office 2010.

]]> Comment on GPU accelerated lattice Boltzmann model for shallow water flow and mass transport by tomrod http://blog.accelereyes.com/blog/2010/12/01/lattice_boltzmann_model/comment-page-1/#comment-316 tomrod Thu, 02 Dec 2010 16:01:00 +0000 http://blog.accelereyes.com/blog/?p=889#comment-316 Two things: *this reads like an advert for Jacket. Not necessarily a bad thing, but full disclosure (you are a paid spokesperson, you are not a paid spokesperson) would be nice. * How did you make that beautiful barchart? Two things:
*this reads like an advert for Jacket. Not necessarily a bad thing, but full disclosure (you are a paid spokesperson, you are not a paid spokesperson) would be nice.
* How did you make that beautiful barchart?

]]> Comment on Median Filtering: CUDA tips and tricks by Nick http://blog.accelereyes.com/blog/2010/03/04/median-filtering-cuda-tips-and-tricks/comment-page-1/#comment-315 Nick Thu, 18 Nov 2010 03:33:00 +0000 http://www.accelereyes.com/blog/?p=190#comment-315 There seems to be a bug somewhere. Try repeating the filtering over and over on the same matrix for 20 iteration. This work if the input has dimensions that are a multiple of 16 in both axis, so a 16x16 will produce correct results. Next try a 32x16, now something is broken. There seems to be a bug somewhere.

Try repeating the filtering over and over on the same matrix for 20 iteration. This work if the input has dimensions that are a multiple of 16 in both axis, so a 16×16 will produce correct results.

Next try a 32×16, now something is broken.

]]> Comment on The Jacket SDK trumps standalone MEX by Anonymous http://blog.accelereyes.com/blog/2010/10/29/jacket_sdk_trumps_mex/comment-page-1/#comment-314 Anonymous Sat, 30 Oct 2010 15:11:00 +0000 http://blog.accelereyes.com/blog/?p=825#comment-314 Thanks for the quick reply. I think these comments will be useful to clarify for others. On the PCT comparison, as far as I understand from what I've seen posted on various forums, PCT does not support GPU computation on Mac OS X. I updated the wording on the comparison page to add more clarity. If my understanding is wrong, I'm happy to correct the mistake too. Thanks for the quick reply. I think these comments will be useful to clarify for others.

On the PCT comparison, as far as I understand from what I’ve seen posted on various forums, PCT does not support GPU computation on Mac OS X. I updated the wording on the comparison page to add more clarity. If my understanding is wrong, I’m happy to correct the mistake too.

]]>