Feature: Automated CI/CD Pipeline and Repository Templates#7
Feature: Automated CI/CD Pipeline and Repository Templates#7GiGiKoneti wants to merge 4 commits into
Conversation
…lation pipeline - Added rv_sparse.h with cache-aligned CSR/CSC definitions and 32-bit indices for optimized register packing. - Implemented vsetvl-driven strip-mined SpMV utilizing vloxei32 gathers. - Implemented contiguous vector SpMM utilizing dense column strip-mining. - Configured CMake cross-compilation toolchain targeting rv64gcv.
|
Update: I've significantly expanded the scope of this PR to include full RISC-V Hardware Emulation and a unit testing framework. New Features Added:
The pipeline now successfully boots QEMU and passes the test. This completely solves the "how do we verify RVV math w/o hardware?" problem and sets a highly robust standard for all future algorithm PRs! |
|
Hi @5hayanB and @Talha-Ahmed-1 , Following up on the QEMU integration, I wanted to share a quick note on my broader vision and commitment for this PR for following month and the rv-sparse repository, I'll keep on refining this PR. When looking at the LFX mentorship challenge, it is clear that the ultimate goal isn't just to write vector intrinsics, but to build an accelerator-independent, production-ready library that can actually be trusted in modern machine learning and scientific computing workloads. My vision right now is to establish a bulletproof, hardware-aligned foundation so that building the core rv-sparse library becomes a seamless experience for everyone involved. By defining the repository templates, configuring the bare-metal CMake linking rules, and providing automated QEMU emulation in the CI pipeline, we now have a mathematically rigorous environment. We can safely introduce highly complex RISC-V Vector kernels (like SpMM) and CSC formats in future PRs, knowing that the infrastructure will automatically catch any memory alignment faults or precision errors before they are ever merged. I am deeply committed to making sure the architecture surrounding this library is as perfect and optimized as the mathematics inside of it. Let me know if the CI setup looks good to you, and I look forward to contributing the core sparse algorithms next! |
1 participant
Resolves #6. This PR introduces a complete GitHub Actions workflow to verify cross-compilation of the RVV static library via the RISC-V GNU toolchain. Additionally, it establishes standard Issue and Pull Request templates to standardize future LFX mentorship contributions.