...
- WG Opens/challenges
- Future topics
- Tensorflow for RISC-V
- x264 for RISC-V
- AIs from last week
Appendix
...
- Tensorflow for RISC-V (email sent to libraries-runtimes WG)
- If yes, what is the latest status?
- We are getting asked questions Questions regarding the availability of Tensorflow on RISC-V
- Tried to compile Tensorflow within Ventana and things fail due to missing support for Bazel
- Is anyone from RISE or any upstream company working on Tensorflow?
- If no one is working on Tensorflow from RISE, can we make this an RFP and get some bids as Tensorflow for RISC-V is a pretty important framework for AI/ML
- x264 for RISC-V discussion
- Bytedance team started work on x264 porting and optimization with RVV1.0
- Repo located here https://code.videolan.org/yintong.ustc/x-264-bd-ventana
- Is anyone from RISE or upstream working on x264? Bytedance would like to collaborate with other work if anyAn open discussion occured around Bazel, and RISC-V, and what is available. After the discussion it was determined that the group could continue going forward and see what the next level of errors would be.
- Working Groups Opens / Challenges
- Distro WG - RVI email thread
- [RISE-Distro-Integration-WG] [tech-profiles] Feedback on having multiple applications profiles from RISE Discussion
02 May 2024
- Michelle Martineau, The Linux Foundation
- Erich Hanke, AMD
- Tomas Evensen, AMD
- Anup Patel, Ventana
- Heinrich Schuchardt, Canonical
- Robin Randhawa, SiFive
- Gordan Markus, Canonical
- Sunil Vl, Ventana
- Myungjoo Ham, Samsung
- Daniel Barboza, Ventana
- Simon Harvey, Imagination
- Barna Ibrahim, Rivos
- David Weaver, Akean
- Kumar Sankaran, Ventana
- Paul Walmsley, SiFive
- Victor Costan, Google
- Alexey Bataev, SiFive
- Nathan Egge, Google
- Ludovic Henry, Rivos
- Ming-yu Hung, Mediatek
- Jian Zhang, BOSC
...
- LR_01_013: Enable pip on riscv64
- Necessary to run pip install on riscv64; released with 24.0
- LR_01_015: Enable cibuildwheel on riscv64
LR_01_017: Package and distribute numpy wheel
LR_01_020: Package and distribute ninja wheel
LR_01_008: Package and distribute Scikit-Learn wheel
LR_01_022: Package and distribute markupsafe wheel
LR_01_024: Package and distribute psutil wheel
LR_01_027: Package and distribute Pillow wheel- Effort to package and distribute Python packages on RISC-V
- Can’t be done by upstream project (discussed in previous TSC meetings)
- Approach approved with LFEU Legal on licensing
Go - LR_02_0015: Support for Bitmanip extensions in assembler
- LR_02_024: Use Bitmanip instruction in compiler where optimal
- LR_02_002: Support for Vector extension in Go Assembler
- LR_02_0029: Add RISC-V boards to Go upstream CI
- LR_01_013: Enable pip on riscv64
- Work Prioritization
- Please comment on mailing list for prioritization
- Anything not listed you’d like to see?
- Anything not as high on the priority list as you expected?
- Please comment on mailing list for prioritization
- Scaleway / Board farm
- Current allocation
-
Project
Boards
Adoptium / Java distribution
10 / 15
AlmaLinux
1 / 5
RockyLinux
1 / 5
Go
1 / 5
-
- Current allocation
- Python packaging
- https://gitlab.com/groups/riseproject/python/-/packages
- Available: ninja, numpy, scipy, pillow, psutil, PyYAML, optree, MarkupSafe, cmake, ninja
- Next: scikit-learn
- https://wiki.riseproject.dev/display/HOME/LR_01%3A+Python
- Python has a JIT now
- It’s doesn’t support RISC-V
- https://gitlab.com/groups/riseproject/python/-/packages
...