-
Notifications
You must be signed in to change notification settings - Fork 17
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
Permutation selection #297
base: develop
Are you sure you want to change the base?
Conversation
The template parameters passed to CK permutation class cannot be extracted from them. Need to save these infomation in class DeviceElementwiseParams so that it is possible to query a specific CK permutation class by these template parameters.
return op_ptrs; | ||
std::unordered_map<hiptensor::Uid, std::unique_ptr<DeviceOp>> opPtrs; | ||
// clang-format off | ||
addInstance<256, 64, 64, 4, 4, ck::Sequence<0, 1>, ck::Sequence<4>, ck::Sequence<4>>(opPtrs); |
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.
Will add the instances with best performance in another PR when the performance test is done.
28702c2
to
811086e
Compare
- The registry contains only one hash table. The key is the hash code of all template parameters of permutation instance classes. - Removed Query class.
811086e
to
3eeab9e
Compare
@@ -26,11 +26,7 @@ | |||
|
|||
set(CK_PERMUTATION_INSTANCE_SOURCES | |||
${CMAKE_CURRENT_SOURCE_DIR}/hiptensor_permutation_scale_instances.hpp |
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.
Not sure why there are hpp files in source list?
m1PerThread, | ||
threadClusterArrangeOrder.first, | ||
threadClusterArrangeOrder.second, | ||
scalarPerVectorSeq, |
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.
Do both in and out have to be the same? Or is it interesting to have a combinatorial set for both in and out?
Added the permutation instance selection.
The work flow