Skip to content
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

Missing mkl_somatcopy and relatives #1331

Closed
blueberry opened this issue Mar 11, 2023 · 7 comments
Closed

Missing mkl_somatcopy and relatives #1331

blueberry opened this issue Mar 11, 2023 · 7 comments
Labels

Comments

@blueberry
Copy link

blueberry commented Mar 11, 2023

Hi,

I'm using JavaCPP 1.5.8, but judging by the generated mkl_rt.java in 1.5.9-SNAPSHOT= it is the same in the latest master.

I noticed that =mkl_rt= is missing the mkl_?omatcopy function that is defined in MKL and that is present in mkl_rt for years (similarly to #1326 ).

There is a difference, though, because judging by the JavaCPP MKL configuration source, this function should be available!

"mkl_simatcopy", "mkl_dimatcopy", "mkl_cimatcopy", "mkl_zimatcopy", "mkl_somatcopy", "mkl_domatcopy", "mkl_comatcopy", "mkl_zomatcopy",

What is also strange is that batched variants of mkl_?omatcopy are present! (mkl_rt/mkl_somatcopy_batch etc.)

The same problem is present with mkl_?omatadd..

BTW the mentioned functions are present and vailable in vanilla MKL; I've been using it via custom C JNI bindings for years.

@saudet
Copy link
Member

saudet commented Mar 11, 2023

Their symbols do appear to be defined in mkl_rt, but their declarations is no where to be found in the header files. How did you figure out their signatures?

@saudet
Copy link
Member

saudet commented Mar 11, 2023

Ah, no, they seem to be available in "mkl_trans.h" these days. They used to have some strange macros attached, but that's now all in "mkl_trans_names.h" so as long as we don't try to parse that we should be fine. 🤔

@blueberry
Copy link
Author

I used a mkl.h that came with intel's MKL installation on linux. I was using the C API, though; wasn't aware that there could be something messed up in the C++ part...

@saudet
Copy link
Member

saudet commented Mar 12, 2023

I've removed those lines in commit 41d9d2b and the builds are not failing, so it looks like there are no problems.
Please give it a try with the snapshots: http://bytedeco.org/builds/

@blueberry
Copy link
Author

It works, as expected. Thank you!
BTW, Any estimate on the release of JavaCPP presets 1.5.9? Should I expect it in a month or two?

@saudet
Copy link
Member

saudet commented Mar 14, 2023

Probably something like that, yes.

@saudet
Copy link
Member

saudet commented Jun 6, 2023

Version 1.5.9 has been released! Thanks for reporting

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

No branches or pull requests

2 participants