Add big endian support - #3057
Conversation
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
|
❌ failed run, report available here. |
fe7fe86 to
73b43f2
Compare
|
❌ failed run, report available here. |
4 similar comments
|
❌ failed run, report available here. |
|
❌ failed run, report available here. |
|
❌ failed run, report available here. |
|
❌ failed run, report available here. |
f7fc2ba to
0832b33
Compare
nazar-codethink
changed the title
7a2a4c5 to
e6d1e98
Compare
|
❌ failed run, report available here. |
1 similar comment
|
❌ failed run, report available here. |
|
❌ failed run, report available here. |
2 similar comments
|
❌ failed run, report available here. |
|
❌ failed run, report available here. |
e6d1e98 to
48503c3
Compare
a240cd2 to
4eb149b
Compare
|
❌ failed run, report available here. |
2 similar comments
|
❌ failed run, report available here. |
|
❌ failed run, report available here. |
71f4809 to
11eb038
Compare
|
Or ask maintainers to merge your modification into master... |
If we merge the modifications into master, they'll be untested. (As I've tested only CVA6 version, which is ~6 years behind). As an alternative, could you maybe fork the |
|
In that case, I have a solution: @zchamski could you synchronize with @nazar-codethink to vendorize the axi-riscv-atomics inside CVA6 ? In that way @nazar-codethink could update some files from the vendorized project. Cheers |
|
Hi @nazar-codethink, I'm in the process of creating a PR with a vendorized version of PULP's Once that PR is granted, the BE mod from https://github.com/CodethinkLabs/axi_riscv_atomics/commits/be_amo/ can be applied locally on the CVA6 tree without affecting the upstream PULP repo. |
|
Ok, sounds good! 👍 |
This PR prepares the ground for the addition of CVA6 big-endian support (PR #3057) by adding the content of PULP axi_riscv_atomics repo as a vendorized sub-tree of CVA6 under vendor/pulp-platform. This sub-tree replaces the submodule originally instantiated under corev_apu/src.
|
Hi @nazar-codethink, the PR that vendorizes |
8bac6ca to
b46c5e3
Compare
|
I rebased and added the |
|
❌ failed run, report available here. |
|
We need to merge your PR ! |
Co-authored-by: Nazar Kazakov <nazar.kazakov@codethink.co.uk>
Forward the MBE flag from the CSR_REGFILE to the load and store units.
Co-authored-by: Joseph Baker <joseph.baker@codethink.co.uk>
Co-authored-by: Joseph Baker <joseph.baker@codethink.co.uk>
b46c5e3 to
72798c5
Compare
|
I'll try to compile and run Linux there |
|
I successfully booted 32-bit Linux as before, so not sure what can fail CI |
|
❌ failed run, report available here. |
1 similar comment
|
❌ failed run, report available here. |
|
Great @nazar-codethink. Thanks for this good contribution. I merge right now. |
5 participants
Add ability to work in big endian mode.
This adds big endian support for Load/Store unit, MMU, and atomic operations. For atomic operations we had to change axi_riscv_atomics submodule, the changes are in this branch https://github.com/CodethinkLabs/axi_riscv_atomics/commits/be_amo/
With these changes we can boot big endian Linux.
What do you think?