Version 4 of the patches is upstream.
-
Upstream fixes for the latest feedback.
- Pushed a new version of the patches (v4).
- V5 might be needed to find an alternative way to generate atomic memory operations.
- No issues found with patch 1 of the patch set (small vectors).
-
Start optimization for whole register load/store operations through tcg op generation.
- nothing to report yet.
- we're looking into adding resources to speed up the work.
-
Investigate the SIGILL QEMU error in the SiFive library functions
- the error might be due a bad combination of LMUL and VLEN with some instructions (thanks Nathan for suggesting).
- Jeremy has been improving the scripts and aims at having the issue resolved by the end of the month.
-
Actions:
- Paolo to update the slides for the RISC-V Summit.
- COMPLETE.
- Paolo to update the slides for the RISC-V Summit.
- WP2
- Address any new comments on patch 2 (it might not need new implementations, just a better description).
- Continue the optimization work for whole register load/store operations through tcg op generation.
- Ultimate the scripts for the SiFive tests.
- Paolo Savini will be on vacation 8 November and 13-18 November
- MS0 and MS1 paid
- Pending agreed deliverables and RISE approval:
- MS2 Oct-31, 2024
- MS3 Dec-16, 2024