Applications page...

Well someone appears to be using it, there's reported gigaflops. But is it just in the lab? I cannot find an address to connect to the test.

Applications page...

Well someone appears to be using it, there's reported gigaflops. But is it just in the lab? I cannot find an address to connect to the test.

Likely what you're seeing is in-house testing. It appears at this point that it would only be available for ATI GPUs. In your preferences, you would have to enable "run test applications" to receive any work that may become available.

---

Applications page...

Well someone appears to be using it, there's reported gigaflops. But is it just in the lab? I cannot find an address to connect to the test.

Likely what you're seeing is in-house testing. It appears at this point that it would only be available for ATI GPUs. In your preferences, you would have to enable "run test applications" to receive any work that may become available.

That's on, but I haven't seen any.

Server status page shows no work available.

I have some progress (but small) in porting application to GPU. Now, I will porting main routine that computes simulation. I am not using mentioned clrungen (generator) due to many bugs and limitations in that utility and that an application's code requires some hand-made optimization for GPU.

I have some progress (but small) in porting application to GPU. Now, I will porting main routine that computes simulation. I am not using mentioned clrungen (generator) due to many bugs and limitations in that utility and that an application's code requires some hand-made optimization for GPU.

Do you plan a public test?

I have more progress in porting application. Now, I am successfully write bit more code into OpenCL. Many routines in code is called needlessly many times. I found some way to reducing this obsolete execution to minimum or even zero. Very likely, the GPU application will be much better optimized than current CPU application, thanks a mentioned optimizations and making a code more parallel.

Any chance in the future to have a CUDA GPU app along with OpenCL app ?

We have no plans for CUDA GPU app. Altough, Our OpenCL application is designed to be portable between various modern GPU architectures. After the first tests on Radeon GPU's, we will be porting and testing this same application on NVIDIA GPU's.

Next progress in porting code to GPU. The almost part of subroutines of main calculating routine (singl) has been ported to GPU code. Now, I will be joining this code to make that routine. Many duplicated executions will be simplified to single calls and some functions will be replaced with single subroutines that doing this same thing (reusing this same code). Very likely, these optimizations will be performed while constructing big parts of code (including that main routine) of application.

Your hard work is much appreciated.

After the first tests on Radeon GPU's, we will be porting and testing this same application on NVIDIA GPU's.

Am I reading this correctly as you will only send this out only to identified Radeon GPUs and then later it will move to nVidia GPUs too

Thanks,

~Y

If OpenCL 2.0 would do you any good at this point, you will be able to use it with the Nvidia cards too.
https://www.phoronix.com/scan.php?page=news_item&px=OpenCL-2.0-NVIDIA-Preps

And I would always suggest optimizing a project (especially a new project) for the newest video cards.

I am writing next tool to porting code to GPU. New tool will be much automatic and more smart and it will generates same OpenCL code, tests. Tool is still in development and design stage. More info about progress will be later.

Here an suggestion if you need help to test your application on different GPU's.

universe.test@home

Here an suggestion if you need help to test your application on different GPU's.

universe.test@home

What do you mean? Is there a test version (what's the full address?) or are you suggesting he should make one?

Hi. My experiments with my new unfinished tools gave some considerations and experience. I don't know, what I will make decision. Maybe, I will go back to older tools and I will extend them (this is the most likely choice) to make them more suitable to this project. Nevertheless, works under GPU app will be continued.

Here an suggestion if you need help to test your application on different GPU's.

universe.test@home

What do you mean? Is there a test version (what's the full address?) or are you suggesting he should make one?

Was only an suggestion from me. ;-)

I resumed works on GPU app with older tools. Older tools has been extended to use more complex kernel call's chains and they allow to easier testing. I have hope that changes and improvements will help me while my work.
Now, I will be porting main routine of app.

Now, I will be porting main routine of app.

The work on new cpu code is slowing down the work on gpu??
Any test soon???