-
Notifications
You must be signed in to change notification settings - Fork 552
Max parameter length fetched from device #3032
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
|
This is interesting. I didn't know this parameter existed! I will test it on other devices and get back to you. Please, update the formatting of the code added and if possible please rebase your changes on top of master and removing the merge commit. Would also be nice if you could make the gitignore commit message more descriptive. Thanks |
682fa65 to
08d780b
Compare
Yes, for whatever reason, this isn't listed on online documentation of |
9prady9
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks good.
src/backend/opencl/Array.cpp
Outdated
| int heightCheckLimit = | ||
| isIntel && getDeviceType() == CL_DEVICE_TYPE_GPU ? 3 : 6; | ||
| isIntel && device.getInfo<CL_DEVICE_TYPE>() == CL_DEVICE_TYPE_GPU ? 3 | ||
| : 6; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This and the earlier line as essentially identical as device returned
Values for opencl parameter maximum length were hardcoded. The maximum is now requested at the device, so that the correct value for all devices is used.
a219ab9 to
4e2b5ff
Compare
9prady9
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks good and works for NVIDIA and AMD(Kaveri APU) for all tests except one which failed on both devices.
Random number quality test is failing for both my GTX 1060 with CUDA 11.1 and AMD API Kaveri APU. Not sure if it is a problem from PR or master yet. @umar456 Can you please confirm if this is a known issue or new one.
One minor change, can you please remove the isNvidia and isAmd variables that aren't used anymore. They appear in warnings now, they are declared in the lines 303, 305.
Thank you.
* Max parameter length is now fetched from device. Values for opencl parameter maximum length were hardcoded. The maximum is now requested at the device, so that the correct value for all devices is used. * Removed isAmd & isNvidia, since they are no longer used. (cherry picked from commit 0493478)
* Max parameter length is now fetched from device. Values for opencl parameter maximum length were hardcoded. The maximum is now requested at the device, so that the correct value for all devices is used. * Removed isAmd & isNvidia, since they are no longer used. (cherry picked from commit 0493478)
On some devices, the hard-coded maxima for the parameter length did not correspond. This is solved by asking it to the device and using the returned value.
Description
Changes to Users
Checklist