-
Notifications
You must be signed in to change notification settings - Fork 730
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
Feature/split x86 dependencies #272
Conversation
… when using mm_set1_* internally
This reverts commit 6581aae.
@markos Maybe using https://github.com/simd-everywhere/simde would help the port? It would also enable other architectures as well |
@mr-c This would require substantial refactoring of the internals of hyperscan, esp the algorithm implementations, more drastic than just adding NEON support, and it would also take more time. I'm not personally against it but something like that would have to be agreed by the Intel maintainers upfront. I do have a similar idea however, but I need to discuss it further. |
@mr-c for the time being, the focus is on getting NEON support in there. That does not exclude another approach in the future however. |
@xiangwang1 Do you see any issues with this approach? |
It has been communicated that Intel is not interested in supporting multiple architectures in this project. Closing the PR. If interested in other architectures support, please check our fork at https://github.com/VectorCamp/hyperscan. |
Hello,
This is the first part of an ongoing effort to port hyperscan to ARM, in particular ARM NEON and Advanced SIMD support. Before we add actual ARM support, it was necessary to make the source code properly platform agnostic, with platform specific implementations in a separate folder, so that it does not make code maintainance more difficult and for aesthetic reasons (not too many ifdef ARCH in the same function).
For that reason, we did the following:
So far, the changes made don't break compilation or performance or unit tests.
We would appreciate any feedback on this PR.
UPDATE: popcount abstraction was reverted, the tests were failing for no apparent reason, will need to investigate.