There was an announcement a year or so ago that the CUDA version under development didn't work very well, and offered no performance advantages over the OpenCL version. So don't hold your breath.
What they really need is a GPU (preferably CUDA) version of the gravity wave app. Everything else is secondary for me.
There was an announcement a year or so ago that the CUDA version under development didn't work very well, and offered no performance advantages over the OpenCL version. So don't hold your breath.
What they really need is a GPU (preferably CUDA) version of the gravity wave app. Everything else is secondary for me.
I hope for the same. While a CUDA app may not offer performance advantages over OpenCL, they may find it offers participation advantages. Unless you refer specifically to the high CPU usage issue.
I hope for the same. While a CUDA app may not offer performance advantages over OpenCL, they may find it offers participation advantages. Unless you refer specifically to the high CPU usage issue.
Actually, CUDA normally offers some performance improvement, and even more CPU improvement. It is a little strange that it doesn't here, though that depends a lot on the science they are trying to do. I hope they keep trying.
There has not been a new GPU app since Feb 2017, over a year. To me this is odd. Isn't the apps (and especially the GPU apps) the primary tool of this project? I'd think they would constantly tinker with the apps and improve them.
But I have no idea how the projects work backstage. Maybe the staff have more interesting jobs and just run this on spare time? :)
Given that the work went from Several hours (to days) for CPU gamma ray to a few minutes (couple of hours on low end GPUS) on GPU Gamma rays, they probably don't see a need to make cuda app.
Given that the work went from Several hours (to days) for CPU gamma ray to a few minutes (couple of hours on low end GPUS) on GPU Gamma rays, they probably don't see a need to make cuda app.
That is what I am wondering too. They may be trying to keep our CPUs busy. I would rather use a GPU for it, and free up the CPU for other uses. That would be a more efficient use of hardware insofar as I am concerned.
Given that the work went from Several hours (to days) for CPU gamma ray to a few minutes (couple of hours on low end GPUS) on GPU Gamma rays, they probably don't see a need to make cuda app.
That is what I am wondering too. They may be trying to keep our CPUs busy. I would rather use a GPU for it, and free up the CPU for other uses. That would be a more efficient use of hardware insofar as I am concerned.
Besides the motto was " lets you donate your idle computer time to science projects " . Right now it is "Lets you donate your idle computer time to idle cycling time"
There was an announcement a
)
There was an announcement a year or so ago that the CUDA version under development didn't work very well, and offered no performance advantages over the OpenCL version. So don't hold your breath.
What they really need is a GPU (preferably CUDA) version of the gravity wave app. Everything else is secondary for me.
Jim1348 wrote:There was an
)
I hope for the same. While a CUDA app may not offer performance advantages over OpenCL, they may find it offers participation advantages. Unless you refer specifically to the high CPU usage issue.
Skivelitis2 wrote:I hope for
)
Actually, CUDA normally offers some performance improvement, and even more CPU improvement. It is a little strange that it doesn't here, though that depends a lot on the science they are trying to do. I hope they keep trying.
Jim1348 wrote:I hope they
)
So do I.
There has not been a new GPU app since Feb 2017, over a year. To me this is odd. Isn't the apps (and especially the GPU apps) the primary tool of this project? I'd think they would constantly tinker with the apps and improve them.
But I have no idea how the projects work backstage. Maybe the staff have more interesting jobs and just run this on spare time? :)
Maybe the project has a near
)
Maybe the project has a near infinite amount of data to look through for binary pulsars.
Given that the work went from
)
Given that the work went from Several hours (to days) for CPU gamma ray to a few minutes (couple of hours on low end GPUS) on GPU Gamma rays, they probably don't see a need to make cuda app.
Zalster wrote:Given that the
)
That is what I am wondering too. They may be trying to keep our CPUs busy. I would rather use a GPU for it, and free up the CPU for other uses. That would be a more efficient use of hardware insofar as I am concerned.
Jim1348 wrote:Zalster
)
Besides the motto was " lets you donate your idle computer time to science projects " . Right now it is "Lets you donate your idle computer time to idle cycling time"