tqchen commented on code in PR #16972:
URL: https://github.com/apache/tvm/pull/16972#discussion_r1593920403
##########
src/target/target_kind.cc:
##########
@@ -340,9 +340,9 @@ TVM_REGISTER_TARGET_KIND("rocm", kDLROCM)
.set_target_parser(UpdateROCmAttrs);
TVM_REGISTER_TARGET_KIND("opencl", kDLOpenCL)
- .add_attr_option<Integer>("max_threads_per_block", Integer(256))
+ .add_attr_option<Integer>("max_threads_per_block", Integer(1024))
.add_attr_option<Integer>("max_shared_memory_per_block", Integer(16384))
- .add_attr_option<Integer>("max_num_threads", Integer(256))
+ .add_attr_option<Integer>("max_num_threads", Integer(1024))
Review Comment:
see an example in
https://github.com/mlc-ai/mlc-llm/blob/main/python/mlc_llm/support/auto_target.py#L424C11-L424C42
##########
src/target/target_kind.cc:
##########
@@ -340,9 +340,9 @@ TVM_REGISTER_TARGET_KIND("rocm", kDLROCM)
.set_target_parser(UpdateROCmAttrs);
TVM_REGISTER_TARGET_KIND("opencl", kDLOpenCL)
- .add_attr_option<Integer>("max_threads_per_block", Integer(256))
+ .add_attr_option<Integer>("max_threads_per_block", Integer(1024))
.add_attr_option<Integer>("max_shared_memory_per_block", Integer(16384))
- .add_attr_option<Integer>("max_num_threads", Integer(256))
+ .add_attr_option<Integer>("max_num_threads", Integer(1024))
Review Comment:
OK, maybe we can take a different approach. When we do `--device
android:adreno`, we can translate that to target opencl and overrides the
`max_num_threads` to 1024
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
