[OpenCL] Don't initialize OpenCL runtime on host #15745
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Contributor
Author
|
cc: @elvin-n |
elvin-n
suggested changes
Sep 14, 2023
After adding OpenCL wrapper, it is possible to build TVM with OpenCL support also on the host which doesn't have OpenCL libraries. But if you want to compile OpenCL module for a remote device on such host machine then you will see an error that OpenCL lib cannot be open. To avoid such problem, we need to call OpenCL functions only in runtime. So function for initializing OpenCL workspace was removed from OpenCLModuleNode. And a new function `IsProgramCreated` was added. The last function is necessary to prepare vectors with OpenCL programs, associated with OpenCL devices. Previously it was done during OpenCLModule initialization. So, now we create such vectors only in runtime after getting list of available OpenCL devices.
echuraev
force-pushed
the
echuraev/do_not_initialize_opencl_on_host
branch
from
f1ad8d0 to
4b8c06f
Compare
Contributor
Author
masahi
approved these changes
Sep 25, 2023
vinx13
pushed a commit
to vinx13/tvm
that referenced
this pull request
Oct 17, 2023
* [OpenCL] Don't initialize OpenCL runtime on host After adding OpenCL wrapper, it is possible to build TVM with OpenCL support also on the host which doesn't have OpenCL libraries. But if you want to compile OpenCL module for a remote device on such host machine then you will see an error that OpenCL lib cannot be open. To avoid such problem, we need to call OpenCL functions only in runtime. So function for initializing OpenCL workspace was removed from OpenCLModuleNode. And a new function `IsProgramCreated` was added. The last function is necessary to prepare vectors with OpenCL programs, associated with OpenCL devices. Previously it was done during OpenCLModule initialization. So, now we create such vectors only in runtime after getting list of available OpenCL devices. * Call workspace init function before all OpenCL API calls
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
After adding OpenCL wrapper, it is possible to build TVM with OpenCL support also on the host which doesn't have OpenCL libraries. But if you want to compile OpenCL module for a remote device on such host machine then you will see an error that OpenCL lib cannot be open.
To avoid such problem, we need to call OpenCL functions only in runtime. So function for initializing OpenCL workspace was removed from OpenCLModuleNode. And a new function
IsProgramCreatedwas added. The last function is necessary to prepare vectors with OpenCL programs, associated with OpenCL devices. Previously it was done during OpenCLModule initialization. So, now we create such vectors only in runtime after getting list of available OpenCL devices.