Add linux-riscv64 binary

[luhenry]

No description provided.

[yawkat]

Thank you! I'll probably make a minor release when #44 is ready.

[yawkat]

@luhenry I manually triggered a workflow run and it passed, but would you mind testing the final jar on riscv? It's available here, hopefully you have access: https://github.com/yawkat/lz4-java/actions/runs/24154088159/artifacts/6335517206

[luhenry]

@luhenry I manually triggered a workflow run and it passed, but would you mind testing the final jar on riscv? It's available here, hopefully you have access: https://github.com/yawkat/lz4-java/actions/runs/24154088159/artifacts/6335517206

Will do! I’m thinking of using some of these GitHub riscv64 runners on my fork to run the testing there. Would you generally be happy to integrate such change?

[yawkat]

Hmm maybe. The macos and windows builds already use runners and I think they run tests. The Linux builds don't. Maybe that could change, but s390x and x86 still have to be supported so some cross compilation will remain. There's also actions credit cost to consider, but I think Linux runners are fairly cheap on all platforms.

[luhenry]

Testing at https://github.com/luhenry/lz4-java/actions/workflows/test.yml. Running into some missing dependencies. Debugging and updating my branch.

It's a lot slower on riscv64 for two reasons:

• hardware isn't on par with the Github-hosted x86_64 and arm runners
• xxhash 0.83.2 is not optimized for riscv64. It's been optimized in Add RVV Optimizations for RISC-V Architecture Platforms Cyan4973/xxHash#1043, refactor riscv-v implementation. Cyan4973/xxHash#1069, enhance riscv-v implementation performance. Cyan4973/xxHash#1070; it hasn't been released yet.

From the early testing (https://github.com/luhenry/lz4-java/actions/runs/24181872888/job/70576636625):

[INFO] Tests run: 1, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 0.343 s -- in net.jpountz.lz4.LZ4FactoryTest
[INFO] Tests run: 78, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 60.12 s -- in net.jpountz.lz4.LZ4BlockStreamingTest
[INFO] Tests run: 272, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 163.2 s -- in net.jpountz.lz4.LZ4FrameIOStreamTest
[INFO] Tests run: 16, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 0.085 s -- in net.jpountz.lz4.LZ4FastResetTest
[INFO] Tests run: 79, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 521.6 s -- in net.jpountz.lz4.LZ4Test
[INFO] Tests run: 63, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 104.1 s -- in net.jpountz.xxhash.XXHash64Test
[INFO] Tests run: 1, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 0.004 s -- in net.jpountz.xxhash.XXHashFactoryTest
[INFO] Tests run: 63, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 85.75 s -- in net.jpountz.xxhash.XXHash32Test
[INFO] Tests run: 1085, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 10.06 s -- in net.jpountz.lz4.OutOfBoundsTest
[INFO] Tests run: 7, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 13.74 s -- in net.jpountz.lz4.UtilsTest

So it's mostly working. Some failures still, reproducing locally and figuring out what's missing.

[luhenry]

The failures are because jazzer is not currently supported on linux-riscv64: https://github.com/CodeIntelligenceTesting/jazzer/blob/main/README.md?plain=1#L29. I'll submit a PR to try to get them to enable it 🤞

In the meantime I'll disable fuzzing on riscv64 specifically.