Cache MKL plans#3101
Conversation
- Make sure DLIB_USE_MKL_SEQUENTIAL and DLIB_USE_MKL_WITH_TBB are added to the list of preprocessor definitions when appropriate - Add path to libtbb. You don't want to mix libraries in /opt/intel/oneapi/... and /usr/lib/x86...
|
@davisking Ready for an initial review. Linking to libmkl_rt.so was broken. Mainly because by default it loads libmkl_intel_thread.so at runtime using
For the BLAS stuff i'm now using cmake_dependent_option which simplifies the super messy logic in the CMakeLists.txt script. |
|
@davisking Do you mind having a look at this? Or are you wanting to merge the CUDA stuff first? I would rather get this in before i touch cmake revamping stuff again as I would be duplicating certain things in here already. |
|
@davisking hold fire on merging. I haven't got the windows jobs working correctly with openblas and MKL. |
…on Github Actions
…e, it can throw. Don't know why gcc and clang were happy with that - attempt 1 at downloading openblas
- trying to fix openblas download
… actions - why is openblas not working
…g right on windows
|
Right, i'm gonna need some help. I don't have a windows machine i can use for dev. I'm giving up trying to debug the windows builds simply by looking at github actions logs. |
|
Warning: this issue has been inactive for 35 days and will be automatically closed on 2025-10-16 if there is no further activity. If you are waiting for a response but haven't received one it's possible your question is somehow inappropriate. E.g. it is off topic, you didn't follow the issue submission instructions, or your question is easily answerable by reading the FAQ, dlib's official compilation instructions, dlib's API documentation, or a Google search. |
|
Warning: this issue has been inactive for 42 days and will be automatically closed on 2025-10-16 if there is no further activity. If you are waiting for a response but haven't received one it's possible your question is somehow inappropriate. E.g. it is off topic, you didn't follow the issue submission instructions, or your question is easily answerable by reading the FAQ, dlib's official compilation instructions, dlib's API documentation, or a Google search. |
|
Notice: this issue has been closed because it has been inactive for 45 days. You may reopen this issue if it has been closed in error. |
This PR caches MKL plans. This substantially improves performance when using MKL sequential and MKL TBB. Not benchmarked with MKL RT but I expect the same results.