Current status as of v1.19:
* Windows: no known issues
* Mac OS X: no known issues
* Linux: might use full CPU core (known bug/feature in AMD Catalyst drivers)

Requirements:
* AMD GPU (NVIDIA will be supported by our CUDA apps for the time being)
* OpenCL 1.0
* BOINC: at least 6.13.10
* Mac OS: 10.7 only


Oliver

Though, can you explain to me why the OpenCL BRP4 app runs at below normal priority?

Because it uses less than one CPU (avg_ncpus < 1).

If it would use a full CPU core (like CPU Apps do), it would run in "Idle" priority.

BM

So? Just because Lunatics built a 3rd party application for Seti only that does the OpenCL detection, does not mean that the BOINC you're running does the OpenCL detection. This project uses the OpenCL detection built into the 6.13 clients to determine all OpenCL capable co-processors.

Ok. Fine. Clear.
I'll try 6.13 some later, despite the fact that I don't like > 6.10 interface without Messages tab.

You are right.
The message tab is annoying.

I am now using ATI5850 Driver11.9 and SDK2.5
Boinc version 6.13.12 and app 1.19
Progress 38% in 4hours ; hopefully completed in the morning...
GPU load is rarely above 0% sometimes it uses 100% for half a second
Running on win Vista 32bit

Now I will sleep. Goodnite all

---

Could you mention a little more about what GPUs are suitable for the current A@H application? What I've seen so far suggests that it may require AMD/ATI GPUs, and is not ready for Nvidia GPUs yet.

For example, must the GPU have hardware support for double precision?

Also, is the current CPU application compiled from OpenCL, somewhat similar the the way the current POEM@HOME application is?

I've seen rumors that the 7.0 version of BOINC will be available soon, with at least some support for OpenCL GPU workunits, but almost nothing more on just how much OpenCL GPU support.

For example, must the GPU have hardware support for double precision?

No, the floating point speed of the GPU has nothing to do with OpenCL.
For AMD/ATI GPUs, only the GPUs at the requirements page on ATI's SDK web site, will have OpenCL capability. In numbers, from the HD4300 onwards and every GPU thereafter, plus the Firepor V3800 and everyone thereafter.

For nVidia it's easier, there any GPU capable of CUDA will be able to do OpenCL. In numbers, from the GEFORCE 8300 onwards and every GPU since.

I've seen rumors that the 7.0 version of BOINC will be available soon, with at least some support for OpenCL GPU workunits, but almost nothing more on just how much OpenCL GPU support.

BOINC 7 (being tested as BOINC 6.13) will support both ATI and nVidia GPUs, with a possibility for Intel GPUs to follow once Intel gets an API out. But really, BOINC doesn't need to support OpenCL, as it doesn't do any of the work. The science application will need to do OpenCL.

All that BOINC will do is detect if your GPU is OpenCL capable and if so, which version it is compliant to. Just as it already did detect if the nVidia GPU is CUDA capable and the ATI GPU CAL/Brook+ capable.

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

What I've seen so far suggests that it may require AMD/ATI GPUs, and is not ready for Nvidia GPUs yet.

As a starter please see this post of mine.

Apart from that, we do have a dedicated CUDA app so we don't focus on running our OpenCL app on NVIDIA devices for the time being.

Memory-wise 512 MB cards might be already too "small" for the current app. We always strive to provide apps that run on as many GPUs as possible but there are technical and algorithmic constraints that simply don't allow us to reduce the GPU memory footprint of the OpenCL app right now.

For example, must the GPU have hardware support for double precision?

No.


Cheers,
Oliver

Apart from that, we do have a dedicated CUDA app so we don't focus on running our OpenCL app on NVIDIA devices for the time being.

On that though, the whole point of OpenCL is to make a new standard in GPGPU and other processors, where presumably one code can work on many pieces of hardware. So in essence, you ought to be able to use the OpenCL app for an ATI GPU on an nVidia GPU, an Intel GPU, an AMD or Intel CPU, and presumably on any Tesla you throw it at. In a perfect world...

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

1 task completed!!!

:)

http://albert.phys.uwm.edu/results.php?hostid=1353&offset=0&show_names=0&state=2&appid=

http://albert.phys.uwm.edu/result.php?resultid=34526

---

On that though, the whole point of OpenCL is to make a new standard in GPGPU and other processors, where presumably one code can work on many pieces of hardware. So in essence, you ought to be able to use the OpenCL app for an ATI GPU on an nVidia GPU, an Intel GPU, an AMD or Intel CPU, and presumably on any Tesla you throw it at. In a perfect world...

Sorry, that's a misconception. While the OpenCL framework itself indeed facilitates the general approach one still has to take into account the differences of each architecture. OpenCL doesn't help you to overcome that out of the box. While a single codebase should run on most OpenCL capable devices, it won't do so efficiently (e.g. thread parallelism vs. data parallelism on NVIDIA and AMD GPUs). Thus it boils down to the fact that you can stick to a single framework for coding/building/deployment but you still have to write architecture-aware implementations for an efficient solution.


Best,
Oliver

Not just that. On my travels, I've seen that due to ATI's different approach to GPGPU, from Close-to-Metal through BrookGPU to OpenCL, that their drivers are mature enough to support OpenCL on their hardware.

However, for you as a project engineer to be able to make it work on all the different hardware, you need a different API, SDK, IDE, debugger and probably compiler for each different piece of hardware by vendor. So one of each for ATI, one of each for nVidia, one of each for Intel, etc.

Hence why I ended the previous comment with "In a perfect world..." :-)

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

It's not that bad after all. However, while our current OpenCL app runs on NVIDIA GPUs indeed it doesn't yet produce valid results on all their models. Most likely due to some subtle differences (optimizations) in NVIDIA's runtime OpenCL compiler... We're investigating but we'll focus on OpenCL@AMD first since we got a CUDA app anyway...

Oliver

Looks like the OpenCL app for ATI can't validate against the CPU app: http://albert.phys.uwm.edu/workunit.php?wuid=11889

Oh and thanks for making the initial replication 3, but leaving the third one unsent. ;-)

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

Separate post for this, can you stop sending ATI OpenCL work to non-6.13 clients?
I see one of my other tasks is waiting for a wingman, because two people got "Failed to get OpenCL platform/device info from BOINC (error: -161)!" (6.12.34 and 6.10.60), the third had it running on the failed CUDA app.

Since you're checking for the information from the client, you may just as well make the client mandatory.

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

Separate post for this, can you stop sending ATI OpenCL work to non-6.13 clients?
I see one of my other tasks is waiting for a wingman, because two people got "Failed to get OpenCL platform/device info from BOINC (error: -161)!" (6.12.34 and 6.10.60), the third had it running on the failed CUDA app.

Since you're checking for the information from the client, you may just as well make the client mandatory.

Guess what we do already? :-)
We're aware of that problem but there's no fix right now... Still investigating...


Oliver

Looks like the OpenCL app for ATI can't validate against the CPU app: http://albert.phys.uwm.edu/workunit.php?wuid=11889

We're still tuning the validator. That's part of this test, we need to sample the numerical stability/inaccuracies across various platforms and devices...

Oh and thanks for making the initial replication 3, but leaving the third one unsent. ;-)

Hm?

Oh and thanks for making the initial replication 3, but leaving the third one unsent. ;-)

Hm?

Sneaky... now it's sent to a third party.
It wasn't for several hours last night, before I made the comment... I just needed more patience then. ;-)

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

Guess what we do already? :-)
We're aware of that problem but there's no fix right now... Still investigating...

One further request then, decrease the deadline here? I see wingmen who have seemingly abandoned the cause, so now I'll have to wait 14 days before anything is resent. That's too much for a test project. Just make the deadline 3 to 5 days, that's time enough to crunch the work, even on a multi-project system, and send it back.

You want the results back fast, don't you? :-)

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

We're aware of that problem but there's no fix right now... Still investigating...

Should work now...

I've seen rumors that the 7.0 version of BOINC will be available soon, with at least some support for OpenCL GPU workunits, but almost nothing more on just how much OpenCL GPU support.

BOINC 7 (being tested as BOINC 6.13) will support both ATI and nVidia GPUs, with a possibility for Intel GPUs to follow once Intel gets an API out. But really, BOINC doesn't need to support OpenCL, as it doesn't do any of the work. The science application will need to do OpenCL.

All that BOINC will do is detect if your GPU is OpenCL capable and if so, which version it is compliant to. Just as it already did detect if the nVidia GPU is CUDA capable and the ATI GPU CAL/Brook+ capable.

Looks adequate for computers with only one GPU.

On other BOINC projects, I've seen one thing mentioned that BOINC really needs to do if you want to be able to run both an OpenCL GPU workunit and a non-OpenCL GPU workunit at the same time on computers with more than one GPU - provide a standard way of mapping the way BOINC identifies which GPU it has assigned to a workunit to the way OpenCL identifies those GPUs. Otherwise, it's likely that those two workunits will often attempt to use the same GPU at the same time, with both failing.

Except for this, I've seen nothing on features for OpenCL that need to be in BOINC instead of the application. Adding more features would be helpful, but not required, for BOINC projects wanting OpenCL GPU workunits.

I currently don't have any AMD/ATI GPUs, so I won't be able to much for Albert@Home soon, at least until I find time to measure the size requirements for putting such a graphics board into my newest computer.