-
Notifications
You must be signed in to change notification settings - Fork 1k
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
cpu: risc-v: add RISC-V defines and fix indentation #1148
Conversation
Hello @aaronfranke , Usually if test is indicated as failed in summary it prints errors during the execution above so you can find the reason of a failure there. Regards, |
If you want these changes to land in oneDNN v2.4 please backport to rls-v2.4 once the PR lands into master. |
@igorsafo Here are some instructions based on what I did. It's roughly based on https://wiki.ubuntu.com/RISC-V but note that this documentation is a bit out of date.
qemu-system-riscv64 \
-machine virt \
-nographic \
-m 8192 -smp 4 \
-bios /usr/lib/riscv64-linux-gnu/opensbi/generic/fw_jump.elf \
-kernel /usr/lib/u-boot/qemu-riscv64_smode/uboot.elf \
-device virtio-net-pci,netdev=eth0 \
-netdev user,id=eth0,hostfwd=tcp::2222-:22 \
-drive file=ubuntu.img,format=raw,if=virtio \
|
@aaronfranke Unfortunately I wasn't able to run ubuntu under qemu targeting riscv cpu. Could you please provide more details about the issues you see on riscv? |
@igorsafo The failed tests did not give me any debug information other than what I posted above. If you would like, I can re-run the tests if you provide information of what arguments/flags/etc to use to show us more debug information. That said, fixing those two tests is necessarily not a blocker to get this merged. 98% of tests passing on RISC-V is better than not having any RISC-V support. |
@aaronfranke To get more details about test issues you need to run the failing tests manually: $ DNNL_VERBOSE=1 ./build/tests/gtests/test_lrn_forward
$ DNNL_VERBOSE=1 ./build/tests/gtests/test_lrn_backward The tests will print cases that failed and a root cause. I think it is ok to promote RISC-V support if it provides value to oneDNN users even if there are minor issues related to particular implementations, which can be addressed later. |
Hmm, weird, now they say pass. I'm not sure if there is anything to do here. |
Description
This PR does 2 things:
cmake/platform.cmake
, part of the file was indented by 5 spaces instead of 4.cc1plus: error: bad value (‘rv64gc’) for ‘-march=’ switch
, which is nice.Closes #1146. I did not make an RFC for this, I'm not sure that it's necessary for such a small change.
Checklist
General
make test
andmake test_benchdnn_*
) pass locally for each commit?make test
on RISC-V and most tests pass but these two tests fail:I don't know how to debug the test. Also, the very long time is expected because this was ran in an emulator.
clang-format src/cpu/platform.hpp > src/cpu/platform.hpp2
then copied from hpp2 to hpp.New features