Hi Daniel, On Thu, Mar 07, 2024 at 04:08:14PM -0300, Daniel Almeida wrote: > Hi Mauro, Hans, > > While working on v1 for this patchset, I realized that we can go further by > converting the error-prone sections of our codec drivers to Rust. This also > does not need any bindings in order to work. > > As yet another proof-of-concept, I have converted parts of the rkvdec driver. > Refer to instructions in v1 to test this. I tested this on rk3399 rkvdec with: ./fluster.py run -d GStreamer-VP9-V4L2SL-Gst1.0 -ts VP9-TEST-VECTORS -j1 And I get the same result as on mainline (6.8-rc3) Ran 220/305 tests successfully in 913.877 secs So I can't say I understand all the Rust here but my testing didn't show any regressions in the VP9 decoding. :) Deb > > Notice how: > > 1) many problematic memcpy's go away, these become a simple assignment in Rust. > > 2) it can interop seamlessly with the code in rkvdec-vp9.c that was already > converted in v1 of this series. > > 3) it can use the Rust version of `seg_feat_enabled` directly in vp9.rs, while > also using the C APIs from the v4l2-vp9-rs library in rkvdec-vp9.c > > 4) more modern things become available for the programmer, like iterators and > their methods without a performance penalty. > > I want to propose the following: > > Let's merge a non-RFC version of this series and gate it behind some kconfigs > so that we can switch between the C and Rust implementations. Users get the C > version by default, while we continuously test the Rust components on a CI for > a few months. This will hopefully be enough time until the next Media Summit. > > My aim is to eventually deprecate the C parts once we're confident that the > Rust code is stable enough. I will keep my own tree, and send PRs to the media > tree if a rebase or fix is needed. > > I believe this will not be disruptive nor require any extra work from anyone > but me. > > -- Daniel > > > Again, applies on top of: > > commit d9c1fae3e5b225f2e45e0bca519f9a2967cd1062 > Author: Alice Ryhl <aliceryhl@xxxxxxxxxx> > Date: Fri Feb 9 11:18:22 2024 +0000 > > rust: file: add abstraction for `poll_table` > > For those looking for a branch instead: https://gitlab.collabora.com/dwlsalmeida/for-upstream/-/tree/vp9-rs-rkvdec?ref_type=heads > > Daniel Almeida (2): > v4l2-core: rewrite the VP9 library in Rust > media: rkvdec: rewrite parts of the driver in Rust > > drivers/media/platform/verisilicon/Kconfig | 2 +- > .../platform/verisilicon/hantro_g2_vp9_dec.c | 38 +- > .../media/platform/verisilicon/hantro_hw.h | 8 +- > drivers/media/v4l2-core/Kconfig | 5 + > drivers/staging/media/rkvdec/Kconfig | 3 +- > drivers/staging/media/rkvdec/Makefile | 2 +- > drivers/staging/media/rkvdec/cbindgen.toml | 36 + > drivers/staging/media/rkvdec/common.rs | 19 + > drivers/staging/media/rkvdec/regs.rs | 237 ++ > drivers/staging/media/rkvdec/rkvdec-vp9.c | 607 +---- > drivers/staging/media/rkvdec/rkvdec_rs.h | 125 + > drivers/staging/media/rkvdec/rkvdec_rs.rs | 14 + > drivers/staging/media/rkvdec/vp9.rs | 636 +++++ > include/media/v4l2-vp9-rs.h | 99 + > rust/bindings/bindings_helper.h | 1 + > rust/helpers.c | 7 + > rust/kernel/lib.rs | 2 + > rust/kernel/media.rs | 5 + > rust/kernel/media/v4l2_core.rs | 6 + > rust/kernel/media/v4l2_core/cbindgen.toml | 26 + > rust/kernel/media/v4l2_core/vp9.rs | 2053 +++++++++++++++++ > 21 files changed, 3415 insertions(+), 516 deletions(-) > create mode 100644 drivers/staging/media/rkvdec/cbindgen.toml > create mode 100644 drivers/staging/media/rkvdec/common.rs > create mode 100644 drivers/staging/media/rkvdec/regs.rs > create mode 100644 drivers/staging/media/rkvdec/rkvdec_rs.h > create mode 100644 drivers/staging/media/rkvdec/rkvdec_rs.rs > create mode 100644 drivers/staging/media/rkvdec/vp9.rs > create mode 100644 include/media/v4l2-vp9-rs.h > create mode 100644 rust/kernel/media.rs > create mode 100644 rust/kernel/media/v4l2_core.rs > create mode 100644 rust/kernel/media/v4l2_core/cbindgen.toml > create mode 100644 rust/kernel/media/v4l2_core/vp9.rs > > -- > 2.43.0 > >
