-
Notifications
You must be signed in to change notification settings - Fork 160
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
[DOCS]Update new backend guide to support the latest develop #593
base: develop
Are you sure you want to change the base?
Conversation
I am curious, have you tried to add a dummy backend and compile the files generated by the python scripts?
|
@@ -124,6 +138,7 @@ Below you can see structure of oneMKL top-level include directory: | |||
<other backends>/ | |||
<other domains>/ | |||
|
|||
Note: The actual structure may be different than what is shown here. To ensure your addition of a new backend is correct, verify that the above scripts correctly generate sample header files from the new files you have added. | |||
|
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.
MIght as well use the rST note role:
.. note::
The actual structure…
…
I did try that with BLAS, as mentioned in the document and saw some issues during the CMake configure steps. I believe that part is dependent on the backend library being used/ported, so the existing documentation may not be very useful beyond the generation of the CMake config file. There is room for improvement here.
The problem I see here is that each of the domains like BLAS, DFT etc have different ways of implementing the headers and the API declarations. This requires a different layout, which the scripts cannot manage at this time to the extent of being helpful for other domains than BLAS. I agree with your POV that the My recommendation is that we update the document as per this PR for now (it still works), and then have a community poll to decide if we want to keep these generation scripts or not. If the consensus is to keep them, we may try to tailor them to support the specific differences of each domain as described in the oneMKL Specification, else remove them in a separate PR. What do you think? |
That's fine with me.
I agree, this is part of the issue. The way I see it we have 3 different kinds of domains today:
|
The poll is now active: #598 |
Description
It was found that the create_new_backend.rst document is outdated and the instructions do not work with the current develop branch.
Solution/Changes
.hpp
->.hxx
) for actual API declarations to be picked up by the scripts.generate_ct_templates.py
usage.Test
All script (
scripts/*.py
) outputs now match the expected output shown in the docs.