claude-noether/libva-v4l2-request-fourier
1
0
Fork 0
forked from marfrit/libva-v4l2-request-fourier
Code Pull Requests Activity

iter39 Option B: drop Hi10P + Main10 from RequestQueryConfigProfiles

Browse Source 
Per Phase 7 close + user-directed Option B trigger (web research /
rockchip-mpp showed Hi10P is effectively impossible on the current
stack). Cross-test on ampere RK3588 confirmed the SAME failure mode
as fresnel RK3399 — both produce all-zero output via libva; kdirect
fails with EINVAL on both. The blocker is in ffmpeg-v4l2-request
userspace plumbing for the new uAPI controls Karlman's kernel patches
introduced, NOT in our backend or the kernel.

Sources confirming kernel + HW capable but userspace pending:
  - lwn.net/Articles/950434: "to fully runtime test... you may need
    upstream DRM commits, FFmpeg patches"
  - patchwork.kernel.org Karlman v6 → v10 series on linux-media
  - Rockchip RK3399 + RK3588 datasheets list 10-bit H.264 support

Stop enumerating Hi10P + Main10 so VAAPI consumers don't try the
broken path. The backend infrastructure (codec.c profile cases,
context.c NV15 CAPTURE + synthetic SPS bit_depth=2 + video_format
invalidation, image.c P010 reporting + NV15→P010 unpack, surface.c
RT_FORMAT_YUV420_10 guard + NV15 PRIME fourcc, nv15.c + nv15.h
unpack primitive, request.h is_10bit flag) is RETAINED — just
re-add the two profiles[index++] lines and bump the H264 guard
back to (-6) when upstream ffmpeg-vaapi V4L2 hwaccel learns 10-bit.

Memory: feedback_rk3399_h264_hi10p_advertised_not_functional.md
captures the empirical evidence for future iterations.

vainfo after this commit: 10 profiles (was 12), matches the iter38
baseline. iter38 5/5 PASS preserved (no other codec touched).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This commit is contained in:
Claude (noether)
2026-05-17 16:43:44 +00:00
parent 63fed87bc5
commit 6bc12fe7e4
1 changed files with 38 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/config.c
+38 -6
View File
@@ -192,21 +192,53 @@ VAStatus RequestQueryConfigProfiles(VADriverContextP context,
}
found = any_fd_supports_output_format(driver_data, V4L2_PIX_FMT_H264_SLICE);
if (found && index < (V4L2_REQUEST_MAX_PROFILES - 6)) {
if (found && index < (V4L2_REQUEST_MAX_PROFILES - 5)) {
profiles[index++] = VAProfileH264Main;
profiles[index++] = VAProfileH264High;
profiles[index++] = VAProfileH264ConstrainedBaseline;
profiles[index++] = VAProfileH264MultiviewHigh;
profiles[index++] = VAProfileH264StereoHigh;
/* iter39: Hi10P on rkvdec (NV15 CAPTURE, bit_depth=10). */
profiles[index++] = VAProfileH264High10;
/*
* iter39 Phase 7 close (Option B): VAProfileH264High10
* DELIBERATELY NOT ENUMERATED.
*
* Hi10P on Rockchip V4L2 stateless decoders requires:
* - HW: ✓ both RK3399 + RK3588 capable (per Rockchip
* datasheets — 4K 10-bit H.264 line items)
* - Kernel: ✓ Karlman v6→v10 series merged in
* mmind v7.0 (rkvdec_h264_decoded_fmts[] has
* NV15/NV20; ctrl cfg.max=HIGH_422_INTRA;
* bit_depth_luma_minus8==2 path live in
* rkvdec-h264-common.c:196)
* - Userspace ffmpeg: ✗ ffmpeg-v4l2-request-fourier
* lacks the userspace plumbing for Hi10P;
* kdirect path fails with EINVAL, libva path
* returns CAPTURE buffer all-zero.
*
* Empirically verified on both fresnel (RK3399) and ampere
* (RK3588) 2026-05-17 — same all-zero / EINVAL failure
* mode on both. The backend infrastructure (codec.c,
* context.c, image.c, surface.c, nv15.c) is RETAINED for
* when the upstream ffmpeg gap closes — just re-add the
* profiles[index++] line and bump the (-5) guard back to
* (-6). See memory feedback_rk3399_h264_hi10p_advertised_not_functional
* for the empirical evidence.
*/
}
found = any_fd_supports_output_format(driver_data, V4L2_PIX_FMT_HEVC_SLICE);
if (found && index < (V4L2_REQUEST_MAX_PROFILES - 2)) {
if (found && index < (V4L2_REQUEST_MAX_PROFILES - 1)) {
profiles[index++] = VAProfileHEVCMain;
/* iter39: Main10 on rkvdec (NV15 CAPTURE). */
profiles[index++] = VAProfileHEVCMain10;
/*
* iter39 Phase 7 close (Option B): VAProfileHEVCMain10
* DELIBERATELY NOT ENUMERATED. Same reasoning as
* VAProfileH264High10 above — kernel + HW ready,
* userspace ffmpeg V4L2 hwaccel plumbing not. Untested
* specifically due to no Main10 fixture (system x265
* is 8-bit-only on Arch ARM), but same kernel/HW/
* userspace stack so same gap likely applies. Re-enable
* when ffmpeg-vaapi → V4L2 hwaccel adds 10-bit HEVC.
*/
}
found = any_fd_supports_output_format(driver_data, V4L2_PIX_FMT_VP8_FRAME);