and no full supported video hw acceleration
on am6plus not working HW decoding for many long time supported formats (no audio and black screen) which were ok on CE21NG (and CE22NO to 20.12.2024 with kernel 5.15.119).
It is possible solved this issue with no HW decoder on AM6plus ? (this issue is only from time of newer kernel)
Mar 02 16:57:05.212250 CoreELEC kernel: 0: Warning: max_dec_frame_buffering(3) is less than DPB size(9) calculated from Profile/Level.
https://paste.coreelec.org/UrsulaWinch
https://paste.coreelec.org/MeagerDress
https://paste.coreelec.org/FictionThunder
kodi.old.log (1.5 MB)
According to this article must be enabled bi-pyramid: What is DPB size? [Archive] - Doom9's Forum
I am not programator, but i saw too very different structure of code compare to aml 4.9:
static bool is_buffer_available(struct vdec_s *vdec)
{
bool buffer_available = 1;
struct vdec_h264_hw_s *hw = (struct vdec_h264_hw_s *)(vdec->private);
struct h264_dpb_stru *p_H264_Dpb = &hw->dpb;
struct DecodedPictureBuffer *p_Dpb = &p_H264_Dpb->mDPB;
if ((kfifo_len(&hw->newframe_q) <= 0) ||
((hw->config_bufmgr_done) && (!have_free_buf_spec(vdec))) ||
((p_H264_Dpb->mDPB.init_done) &&
(p_H264_Dpb->mDPB.used_size == p_H264_Dpb->mDPB.size) &&
(is_there_unused_frame_from_dpb(&p_H264_Dpb->mDPB) == 0))) {
dpb_print(DECODE_ID(hw), PRINT_FLAG_VDEC_DETAIL,
"%s, empty, newq(%d), free_spec(%d), initdon(%d), used_size(%d/%d), unused_fr_dpb(%d)\n",
__func__,
kfifo_len(&hw->newframe_q),
have_free_buf_spec(vdec),
p_H264_Dpb->mDPB.init_done,
p_H264_Dpb->mDPB.used_size, p_H264_Dpb->mDPB.size,
is_there_unused_frame_from_dpb(&p_H264_Dpb->mDPB)
);
buffer_available = 0;
if (dpb_is_debug(DECODE_ID(hw),
DEBUG_DISABLE_RUNREADY_RMBUF))
return buffer_available;
**!**one_packet_multi_frames_multi_run && - it is really command or comment ?
Once time is with “!” and next time is this line without “!”. In printscreen in blue line at left side. For last kernel are these blue and red lines missing after many years.
So different kernel have different behaviour of HW video acceleration on am6plus.
5.15.119 - everything was OK (identification of cpu dont see attribute x-j)
5.15.137 - no video, no audio (identification of cpu dont see attribute x-j)
5.15.153 - no video, audio with tearing (first time going my s922xj on 500 mhz in ondemance gouvernamnt. It has been usually on minimal 667 mhz; only CEC remote; identification of cpu dont see attribute x-j)
https://paste.coreelec.org/InitialsFished
https://paste.coreelec.org/DrillsBossy
https://paste.coreelec.org/CoffeeMammal
sorry, sample of what ? This video is impacted ,e.g.:
all video have issue. Only one free christian stream have video HW acceleration (but also too long 10 seconds of waiting before playing): http://live.tvlux.sk:1935/lux/lux.stream/playlist.m3u8
it is too different NG and NO due to 32bit vs 64 bit, but i thing also, that something is missing in last 64 bit kernel on NO in compare to NO 64 bit kernel from 20. december 2024. In linux world i have skilll that something very needed (executive file) by mistake was deleted long time in builds for ubuntu (only as example, not in relation to CE). NO 64 bit kernel from 20. december 2024 was last good kernel with HW acceleration of videos on AM6plus. **Maybe would be for stabile build would be better using of kernel 5.15.119.**No errors, no stress, only fun 
On am6plus is always bad playing of hw playing with all big bunny samples (no video, only sound).
https://paste.coreelec.org/DumboBraver
https://paste.coreelec.org/ShortJavier
https://paste.coreelec.org/OrderlyPoured
It is same with my AM6plus. Also with 2 last kernel i have this artefacts before boot of CE22NO (with older kernel 5.15.119 was not this artefact - dtb issue or something with 50 vs 60 hz ?):
kodi.log (9.4 MB)
ce-debug_info.txt (46.0 KB)
Would be help also to Am6plus ? On kernel 5.15.119 (with HW acceleration) and also newer kernel it dont have mounted vendor filesystems on CE22NO (vendor filesystem is mounted only in NG).
AM6plus video firmwares (original firmware Android 9 V. 0.5.4 for AM6 Family):
video_ucode.bin (1.6 MB)
h264_enc.bin (74.5 KB)
GDC affine (FHD, 4K):
gdc.tar (80 KB)
On AM6plus any good news. Same situation, but maybe not mounted vendor filesystem can be reason of not working your solution.
https://paste.coreelec.org/GrammyMowed
https://paste.coreelec.org/CanonTrophy
https://paste.coreelec.org/ChangeCatholic
This bug - no bug was not on build with 64bit kernel 5.15.119 on this device am6plus (g12b with x-j cpu). I understand that on this ugoos device is encrypted bootloader. But it has worked to time of changing kernel version. Maybe for this device is ok only this older version for unidentified dependencies.
What means sample ? I used all big bunny videos for testing as sample.
Ok, here are testing samples which i tried. Only mpeg2 ( bun33s.ts) was ok.
Sorry, i am out now. I send you later.
So i tested your command for am6plus, but no needed TEE in case of NO as also NG.
NO: https://paste.coreelec.org/PajamaRefuse
NG: https://paste.coreelec.org/UnloadGrownup
I have clear install of CE22NO and firmware of Android is original 0.5.4 (this firmware is last for it)
HW logs of current build on my device: Amlogic-no Discussion - #1684
I see anomaly:
Kodi compiled 2025-03-21 02:30:11 by GCC 14.2.0 for Linux ARM 64-bit version 5.15.153 vs 5.15.137-android14-11-maybe-dirty SMP from uppster command.