http://www.youtube.com/watch?v=r1sN1ELJfNo&eurl=http://www.generation-3d.com/actualite-La-premiere-demo-OpenCL-sur-nVidia,ac14322.htm

@+
*_*

---

Have you read the discussions over on Rosetta@home asking for Nvidia GPU support? The first efforts to try it showed that the minirosetta algorithms are too memory-demanding to make much use of enough GPU cores to get very much speedup on Nvidia boards with a currently reasonable amount of video memory. Therefore, such a conversion would require so much of a rewrite that it currently doesn't look reasonable.

So far, it looks like this would apply for AMD/ATI boards as well, the other kind BOINC is already able to handle.

I suspect that this may change when either Nvidia or AMD/ATI start offering compilers for the computer language is already written in that are good enough at spreading the work onto multiple GPU cores without increasing the memory requirements, and doing this even for the older GPU boards.

Since Ralph@home runs essentially the same software planned for Rosetta@home later, this should apply to Ralph@home as well.

Good news from ATi with new SDK version

ati stream

The 2.05 version seems to be so stable (over 2 mounth of test) to give us a gpu version.... :-P

New version of OpenCl
OpenCl 1.1

Now also POEM@home support OpenCL....

Now also POEM@home support OpenCL....

And also Milky....

Now also POEM@home support OpenCL....

Correct, but for CPU workunits only so far.

They appear likely to be waiting for the 6.14.* series of BOINC versions that should finally add a reasonable amount of support for OpenCL GPU workunits.


No word yet on whether the new Rosetta Mini Beta 3.* workunits use any OpenCL or not.


Milkyway@Home is offering both CPU and GPU OpenCL workunits, but watch for a few restrictions on the GPU workunits:

1. If your computer has more than one GPU, you must make Milkyway@Home the only BOINC project using any of them, so they can use their own mapping from the GPU numbers BOINC uses to the ones OpenCL uses. However, if your computer has only one GPU, the fact the BOINC has told it to use a GPU is enough to insure that any GPU it can find is the right one, even if some other BOINC projects uses it at other times.

2. You'll need to expect many changes in the GPU workunits when the 6.14.* versions of BOINC start becoming available.

3. Unclear if their AMD/ATI workunits are ready to use OpenCL yet.

4. You MUST have a GPU capable of double precision to run their GPU workunits - a few of the high-end AMD/ATI-based graphics boards from the last few generations, a graphics board based on an Nvidia Fermi chip, a GTX 2xx series Nvidia-based board from GTX 260 up, or a few other high-end graphics boards.


I would have expected more BOINC projects offering GPU workunits sooner if Nvidia and AMD/ATI had made it easier by offering ways to create programs partly in OpenCL and partly in their previous GPU computer languages, but that would have put AMD/ATI at a disadvantage since they had much less complete function libraries for their previous GPU computer language.

Correct, but for CPU workunits only so far.
They appear likely to be waiting for the 6.14.* series of BOINC versions that should finally add a reasonable amount of support for OpenCL GPU workunits.

From boinc alpha newsletter:
We've finished adding new features to the 6.13 client.
Now we need your help to finish testing it.
When it's fully tested as free of bugs as possible, we'll
release it to the public as 7.0.

The new features of 6.13 include:

- A new Simple GUI, which is translatable and accessible via the
keyboard.
- The client enforces resource shares better and does fewer scheduler
RPCs.
- The client supports OpenCL applications.
(We will follow-up with a project at a later date)

New OpenCl version is out:

The new OpenCL 1.2 specification released on November 15th 2011, provides enhanced performance and functionality in response to requests from the developer community – while retaining backwards compatibility with OpenCL 1.0 and 1.1. New features in OpenCL 1.2 include seamless sharing of media and surfaces with DirectX® 9 and 11, enhanced image support, custom devices and kernels, device partitioning and separate compilation and linking of objects.

Poem@Home realease a gpu opencl client. Part of the code run on cpu, part on gpu....

New boinc client 7.0.25 with features:

Localized and accessible Simple GUI
OpenCL Support
Improved Virtual Machine Support
REC-Based scheduler

:-)

During CASP, no ralph wus... :-(
But some news for opencl:

http://developer.amd.com/sdks/AMDAPPSDK/downloads/Pages/default.aspx

New sdk with opencl 1.2 full support

Have you read the discussions over on Rosetta@home asking for Nvidia GPU support? The first efforts to try it showed that the minirosetta algorithms are too memory-demanding to make much use of enough GPU cores to get very much speedup on Nvidia boards with a currently reasonable amount of video memory.

Why not use AMD APU?? Memory shared between cpu and gpu is not a problem...

Now HCC project of WCG uses gpu (opencl)
GPU app is 18-20 times faster than cpu!!

Some info from OpenCl's world:
1) New OpenCL 1.2 extensions
2) New AMD/ATI SDK with BOLT (open C++ template library)
3) New AMD CodeXL developer tool suite (with debugger and profiler)

Some info from OpenCl's world:
1) New OpenCL 1.2 extensions
2) New AMD/ATI SDK with BOLT (open C++ template library)
3) New AMD CodeXL developer tool suite (with debugger and profiler)

3) New AMD CodeXL developer tool suite (with debugger and profiler)

New version of CodeXL is out!
http://developer.amd.com/community/blog/amd-codexl-version-1-1-released-to-developer-amd-com/

Some news about opencl/heterogenous computing/gpgpu.

Opencl 2.0 is released in provisional mode (with a lot of improvement), 6 month to ratificate it.

HSA Foundation release the first specification

AMD HD 7990 has 6gb (!!) of GDDR5 and 8,2 Tflops in single precision...

Opencl 2.0 is released in provisional mode (with a lot of improvement), 6 month to ratificate it.

Opencl 2.0 released
With, also, a new versione of SPIR

SPIR’ stands for Standard Portable Intermediate Representation and is a portable non-source representation for OpenCL 1.2 device programs. It enables application developers to avoid shipping kernel source and to manage the proliferation of devices and drivers from multiple vendors