Video acceleration with Radeon HD 3650

[rumpumpel1]

I'm on Windows 7 32bit and I have a Radeon HD 3650.
in vlc-1.1.7 the option for "use GPU acceleration" is checked
and directx_Jun2010 is installed.
I tried three different drivers:
8.56.1.16 2009-04-25
8.771.0.0 2010-08-25
8.821.0.0 2011-01-26
None of them gives any H264 acceleration.

DXVA checker says for all three:
ModeMPEG2_IDCT DXVA2 NV12 720x480 / 1280x720 / 1920x1080
ModeMPEG2_A DXVA1 NV12 720x480 / 1280x720 / 1920x1080
ModeMPEG2_C DXVA1 NV12 720x480 / 1280x720 / 1920x1080
ModeH264_VLD_NoFGT DXVA2 NV12 720x480 / 1280x720 / 1920x1080
ModeVC1_VLD DXVA2 NV12 720x480 / 1280x720 / 1920x1080
5B23D46D-FA5F-4FDC-B78A-7EB2787942EC DXVA2 NV12 720x480 / 1280x720 / 1920x1080

I have another PC with almost the same setup, except that it has a Radeon HD 3450.
Here the GPU acceleration works perfect.

So what could be the reason it's not working with the Radeon HD 3650 ?

[Jean-Baptiste Kempf]

ModeH264_VLD_NoFGT DXVA2 NV12 720x480 / 1280x720 / 1920x1080

Should be enough.

Please provide logs.

[rumpumpel1]

I tried to enable logging in the preferences: Advanced -> Logging
Whatever filename and verbosity I specified, there is no logfile.
Did I miss something ?

[Jean-Baptiste Kempf]

Yes, Tools->Messages, verbosity to 2.

[rumpumpel1]

here is the log.
By the way: without GPU acceleration the video is choppy.
Setting "skip-filter for H264" to all improves a lot, but it's not perfect.
With GPU acceleration there is only a still picture visible, probably
the first frame in the movie.

Log:
qt4 debug: Saving the advanced preferences
main debug: opening config file (C:\Users\admin\AppData\Roaming\vlc\vlcrc)
main debug: adding item `gaga.m2ts' ( file:///E:/Filme/gaga.m2ts )
qt4 debug: Adding a new MRL to recent ones: E:\Filme\gaga.m2ts
main debug: rebuilding array of current - root Wiedergabeliste
main debug: rebuild done - 1 items, index -1
main debug: processing request item gaga.m2ts node null skip 0
main debug: resyncing on gaga.m2ts
main debug: gaga.m2ts is at 0
main debug: starting new item
main debug: creating new input thread
main debug: Creating an input for 'gaga.m2ts'
main debug: thread (input) created at priority 1 (../.././src/input/input.c:214)
main debug: thread started
main debug: using timeshift granularity of 50 MiB
main debug: using timeshift path 'C:\Users\admin\AppData\Local\Temp'
main debug: `file:///E:/Filme/gaga.m2ts' gives access `file' demux `' path `E:\Filme\gaga.m2ts'
main debug: creating demux: access='file' demux='' path='E:\Filme\gaga.m2ts'
main debug: looking for access_demux module: 2 candidates
main debug: no access_demux module matching "file" could be loaded
main debug: TIMER module_need() : 0.000 ms - Total 0.000 ms / 1 intvls (Avg 0.000 ms)
main debug: creating access 'file' path='E:\Filme\gaga.m2ts'
main debug: looking for access module: 2 candidates
filesystem debug: opening file `E:\Filme\gaga.m2ts'
main debug: using access module "filesystem"
main debug: TIMER module_need() : 0.000 ms - Total 0.000 ms / 1 intvls (Avg 0.000 ms)
main debug: Using AStream*Stream
main debug: pre buffering
main debug: received first data after 0 ms
main debug: pre-buffering done 1024 bytes in 0s - 1000000 KiB/s
main debug: looking for stream_filter module: 2 candidates
main debug: no stream_filter module matching "any" could be loaded
main debug: TIMER module_need() : 0.000 ms - Total 0.000 ms / 1 intvls (Avg 0.000 ms)
main debug: looking for stream_filter module: 1 candidate
main debug: using stream_filter module "stream_filter_record"
main debug: TIMER module_need() : 0.000 ms - Total 0.000 ms / 1 intvls (Avg 0.000 ms)
main debug: creating demux: access='file' demux='' path='E:\Filme\gaga.m2ts'
main debug: looking for demux module: 52 candidates
ts warning: lost synchro
ts debug: skipping 4 bytes of garbage
ts debug: pid[256] unknown
ts debug: pid[31] unknown
ts debug: pid[4097] unknown
ts debug: pid[4113] unknown
ts debug: pid[4352] unknown
ts debug: pid[4353] unknown
qt4 debug: IM: Setting an input
ts debug: PATCallBack called
ts debug: new PAT ts_id=1 version=0 current_next=1
ts debug: * number=0 pid=31
ts debug: * number=1 pid=256
ts debug: PMTCallBack called
ts debug: new PMT program number=1 version=0 pid_pcr=4097
ts debug: * descriptor : registration HDMV
ts debug: * descriptor : unknown (0x88)
ts debug: * es pid=4113 type=27 dr->i_tag=0x28
ts debug: * es pid=4113 type=27 fcc=h264
main debug: selecting program id=1
ts debug: * es pid=4352 type=130 dr->i_tag=0xa
ts debug: found language: deu
ts debug: * es pid=4352 type=130 fcc=dts
ts debug: * es pid=4353 type=130 dr->i_tag=0xa
ts debug: found language: eng
ts debug: * es pid=4353 type=130 fcc=dts
ts debug: * es pid=4608 type=144 dr->i_tag=0xa
ts debug: found language: deu
ts debug: * es pid=4608 type=144 fcc=bdpg
ts debug: * es pid=4609 type=144 dr->i_tag=0xa
ts debug: found language: eng
ts debug: * es pid=4609 type=144 fcc=bdpg
filesystem warning: unimplemented query 4097 in control
main debug: using demux module "ts"
main debug: TIMER module_need() : 23.000 ms - Total 23.000 ms / 1 intvls (Avg 23.000 ms)
main debug: looking for a subtitle file in E:\Filme\
main debug: looking for decoder module: 34 candidates
avcodec debug: libavcodec initialized (interface 0x346c00)
avcodec debug: trying to use direct rendering
avcodec debug: ffmpeg codec (H264 - MPEG-4 AVC (part 10)) started
main debug: using decoder module "avcodec"
main debug: TIMER module_need() : 20.000 ms - Total 20.000 ms / 1 intvls (Avg 20.000 ms)
main debug: looking for packetizer module: 21 candidates
main debug: using packetizer module "packetizer_h264"
main debug: TIMER module_need() : 4.000 ms - Total 4.000 ms / 1 intvls (Avg 4.000 ms)
main debug: Creating an input for 'gaga.m2ts'
main debug: no fetch required for (null) (art currently (null))
main debug: thread (decoder) created at priority 0 (../.././src/input/decoder.c:301)
main debug: thread started
main debug: looking for decoder module: 34 candidates
main debug: using decoder module "dts"
main debug: TIMER module_need() : 0.000 ms - Total 0.000 ms / 1 intvls (Avg 0.000 ms)
main debug: thread (decoder) created at priority 2 (../.././src/input/decoder.c:301)
main debug: thread started
ts debug: DEMUX_SET_GROUP 0 00000000
main debug: looking for meta reader module: 2 candidates
lua debug: Trying Lua scripts in C:\Users\admin\AppData\Roaming\vlc\lua\meta\reader
lua debug: Trying Lua scripts in C:\Program Files\VideoLAN\VLC\lua\meta\reader
lua debug: Trying Lua playlist script C:\Program Files\VideoLAN\VLC\lua\meta\reader\filename.lua
main debug: no meta reader module matching "any" could be loaded
main debug: TIMER module_need() : 0.000 ms - Total 0.000 ms / 1 intvls (Avg 0.000 ms)
main debug: `file:///E:/Filme/gaga.m2ts' successfully opened
ts warning: first packet for pid=4353 cc=0x5
ts warning: first packet for pid=4113 cc=0xe
ts warning: first packet for pid=4352 cc=0x1
dts info: DTS channels:6 samplerate:48000 bitrate:768000
main debug: creating aout
main debug: looking for audio output module: 4 candidates
aout_directx debug: Opening DirectSound Audio Output
packetizer_h264 warning: waiting for SPS/PPS
packetizer_h264 warning: waiting for SPS/PPS
main debug: Buffering 0%
packetizer_h264 warning: waiting for SPS/PPS
packetizer_h264 warning: waiting for SPS/PPS
main debug: Buffering 32%
packetizer_h264 warning: waiting for SPS/PPS
packetizer_h264 warning: waiting for SPS/PPS
main debug: Buffering 64%
packetizer_h264 warning: waiting for SPS/PPS
packetizer_h264 warning: waiting for SPS/PPS
main debug: Buffering 94%
packetizer_h264 warning: waiting for SPS/PPS
packetizer_h264 warning: waiting for SPS/PPS
packetizer_h264 warning: waiting for SPS/PPS
main debug: Stream buffering done (373 ms in 0 ms)
aout_directx debug: found device: Primärer Soundtreiber
aout_directx debug: found device: Lautsprecher (Realtek AC'97 Audio)
aout_directx debug: found device: Realtek Digital Output (Realtek AC'97 Audio)
aout_directx debug: device supports 5.1 channels
aout_directx debug: device supports 4 channels
aout_directx debug: device supports 2 channels
aout_directx debug: device supports 1 channel
aout_directx debug: Windows says your SpeakerConfig is stereo
aout_directx debug: creating DirectSoundThread
main debug: thread (DirectSound Notification Thread) created at priority 15 (../../.././modules/audio_output/directx.c:321)
main debug: using audio output module "aout_directx"
main debug: thread started
main debug: TIMER module_need() : 687.000 ms - Total 687.000 ms / 1 intvls (Avg 687.000 ms)
aout_directx debug: DirectSoundThread ready
main debug: output 's16l' 48000 Hz Stereo frame=1 samples/4 bytes
main debug: mixer 'f32l' 48000 Hz Stereo frame=1 samples/8 bytes
main debug: filter(s) 'f32l'->'s16l' 48000 Hz->48000 Hz Stereo->Stereo
main debug: looking for audio filter module: 14 candidates
audio_format debug: f32l->s16l, bits per sample: 32->16
main debug: using audio filter module "audio_format"
main debug: TIMER module_need() : 8.000 ms - Total 8.000 ms / 1 intvls (Avg 8.000 ms)
main debug: found a filter for the whole conversion
main debug: looking for audio mixer module: 3 candidates
main debug: using audio mixer module "float32_mixer"
main debug: TIMER module_need() : 0.000 ms - Total 0.000 ms / 1 intvls (Avg 0.000 ms)
main debug: input 'dts ' 48000 Hz 3F2R/LFE frame=512 samples/2048 bytes
main debug: looking for audio filter module: 1 candidate
scaletempo debug: format: 48000 rate, 2 nch, 4 bps, fl32
scaletempo debug: params: 30 stride, 0.200 overlap, 14 search
scaletempo debug: 1.000 scale, 1440.000 stride_in, 1440 stride_out, 1152 standing, 288 overlap, 672 search, 2400 queue, fl32 mode
main debug: using audio filter module "scaletempo"
main debug: TIMER module_need() : 0.000 ms - Total 0.000 ms / 1 intvls (Avg 0.000 ms)
main debug: filter(s) 'dts '->'f32l' 48000 Hz->48000 Hz 3F2R/LFE->Stereo
main debug: looking for audio filter module: 14 candidates
main debug: using audio filter module "dtstofloat32"
main debug: TIMER module_need() : 0.000 ms - Total 0.000 ms / 1 intvls (Avg 0.000 ms)
main debug: found a filter for the whole conversion
main debug: filter(s) 'f32l'->'f32l' 52800 Hz->48000 Hz Stereo->Stereo
main debug: looking for audio filter module: 14 candidates
bandlimited_resampler debug: f32l/52800KHz/2->f32l/48000KHz/2
main debug: using audio filter module "bandlimited_resampler"
main debug: TIMER module_need() : 0.000 ms - Total 0.000 ms / 1 intvls (Avg 0.000 ms)
main debug: found a filter for the whole conversion
main debug: End of audio preroll
main debug: Decoder buffering done in 695 ms
packetizer_h264 warning: waiting for SPS/PPS
...
packetizer_h264 warning: waiting for SPS/PPS
main debug: audio output is too slow (113044), trashing 50000us
main debug: audio output is too slow (63044), trashing 50000us
main warning: output date isn't PTS date, requesting resampling (63044)
main warning: buffer is 63043 late, triggering upsampling
packetizer_h264 warning: waiting for SPS/PPS
...
packetizer_h264 warning: waiting for SPS/PPS
main warning: resampling stopped after 8075000 usec (drift: 290)
packetizer_h264 warning: waiting for SPS/PPS
...
packetizer_h264 warning: waiting for SPS/PPS
packetizer_h264 debug: found NAL_SPS (sps_id=0)
packetizer_h264 debug: found NAL_PPS (pps_id=0 sps_id=0)
avcodec debug: Available decoder output format 61 (PIX_FMT_DXVA2_VLD)
avcodec debug: Trying DXVA2
avcodec debug: DLLs loaded
avcodec debug: D3dCreateDevice succeed
avcodec debug: OurDirect3DCreateDeviceManager9 Success!
avcodec info: obtained IDirect3DDeviceManager9
avcodec info: DXVA2CreateVideoService Success!
avcodec debug: - 'DXVA2_ModeMPEG2_IDCT' is supported by hardware
avcodec debug: - 'H.264 VLD, no FGT' is supported by hardware
avcodec debug: - 'VC-1 VLD' is supported by hardware
avcodec warning: - Unknown GUID = 5B23D46D-fa5f-4fdc-XXXX
avcodec debug: Trying to use 'H.264 VLD, no FGT' as input
avcodec debug: NV12 is supported for output
avcodec debug: Using 'H.264 VLD, no FGT' to decode to 'NV12'
avcodec debug: DxCreateVideoDecoder id 28 1920x1080
avcodec debug: IDirectXVideoAccelerationService_CreateSurface succeed with 17 surfaces (1920x1080)
avcodec debug: we got 2 decoder configurations
avcodec debug: configuration[0] ConfigBitstreamRaw 2
avcodec debug: configuration[1] ConfigBitstreamRaw 2
avcodec debug: IDirectXVideoDecoderService_CreateVideoDecoder succeed
avcodec info: Using DXVA2 (ATI Radeon HD 3600 Series , vendor 4098(ATI), device 38296, revision 0) for hardware decoding.
main debug: no usable vout present, spawning one
main debug: looking for text renderer module: 2 candidates
freetype debug: Building font databases.
freetype debug: Took 16000 microseconds
freetype debug: Using Arial as font from file C:/Windows/fonts/arial.ttf
freetype debug: using fontsize: 2
main debug: using text renderer module "freetype"
main debug: TIMER module_need() : 26.000 ms - Total 26.000 ms / 1 intvls (Avg 26.000 ms)
main debug: looking for video filter2 module: 18 candidates
swscale debug: 32x32 chroma: YUVA -> 16x16 chroma: YUVA with scaling using Bicubic (good quality)
main debug: using video filter2 module "swscale"
main debug: TIMER module_need() : 5.000 ms - Total 5.000 ms / 1 intvls (Avg 5.000 ms)
main debug: looking for video filter2 module: 18 candidates
yuvp debug: YUVP to YUVA converter
main debug: using video filter2 module "yuvp"
main debug: TIMER module_need() : 10.000 ms - Total 10.000 ms / 1 intvls (Avg 10.000 ms)
main debug: window size: 1920x1080
main debug: Deinterlacing available
main debug: deinterlace 0, mode blend, is_needed 0
main debug: looking for video output module: 1 candidate
vout_wrapper debug: Opening vout display wrapper
main debug: looking for vout display module: 8 candidates
direct3d debug: creating Vout EventThread
direct3d debug: DirectXCreateWindow
main debug: looking for vout window hwnd module: 3 candidates
qt4 debug: requesting video...
qt4 debug: Video was requested 0, 0
main debug: using vout window hwnd module "qt4"
main debug: TIMER module_need() : 42.000 ms - Total 42.000 ms / 1 intvls (Avg 42.000 ms)
direct3d debug: created video sub-window
direct3d debug: Vout EventThread running
direct3d debug: disabling screen saver
main debug: VoutDisplayEvent 'resize' 1936x1118 window
direct3d warning: trying surface pixel format: YV12
direct3d debug: selected surface pixel format is YV12
direct3d debug: Direct3D scene created successfully
direct3d debug: Direct3D device adapter successfully initialized
main debug: using vout display module "direct3d"
main debug: TIMER module_need() : 196.000 ms - Total 196.000 ms / 1 intvls (Avg 196.000 ms)
main debug: using video output module "vout_wrapper"
main debug: TIMER module_need() : 198.000 ms - Total 198.000 ms / 1 intvls (Avg 198.000 ms)
main debug: got 1 direct buffer(s)
main debug: pic render sz 1920x1080, of (0,0), vsz 1920x1080, 4cc YV12, sar 1:1, msk r0x0 g0x0 b0x0
main debug: pic in sz 1920x1080, of (0,0), vsz 1920x1080, 4cc YV12, sar 1:1, msk r0x0 g0x0 b0x0
main debug: pic out sz 1920x1080, of (0,0), vsz 1920x1080, 4cc YV12, sar 1:1, msk r0x0 g0x0 b0x0
main debug: direct render, mapping render pictures 0-23 to system pictures 1-24
main debug: VoutDisplayEvent 'resize' 1920x1080 window
main debug: End of video preroll
freetype debug: using fontsize: 67
main debug: looking for video blending module: 1 candidate
blend debug: chroma: YUVA -> YV12
main debug: using video blending module "blend"
main debug: TIMER module_need() : 11.000 ms - Total 11.000 ms / 1 intvls (Avg 11.000 ms)
main debug: Post-processing available
main warning: vlc_object_find_name(postproc) is not safe!
main debug: auto hidding mouse
main warning: late picture skipped (2218256 > -11067)
main warning: late picture skipped (3037544 > -10824)
main debug: auto hidding mouse
main warning: late picture skipped (3875833 > -10914)
avcodec error: more than 5 seconds of late video -> dropping frame (computer too slow ?)
main warning: late picture skipped (2300833 > -12772)
main warning: late picture skipped (2668044 > -6411)
main warning: late picture skipped (3418633 > -6778)
main warning: late picture skipped (4196922 > -7560)
main warning: late picture skipped (4973211 > -6718)
avcodec error: more than 5 seconds of late video -> dropping frame (computer too slow ?)
main warning: late picture skipped (2152000 > -6518)
main warning: late picture skipped (2498211 > -6889)
main warning: late picture skipped (3246800 > -7060)
main warning: late picture skipped (4032089 > -7348)
main warning: late picture skipped (4818378 > -6669)
avcodec error: more than 5 seconds of late video -> dropping frame (computer too slow ?)
main debug: auto hidding mouse
main debug: incoming request - stopping current input
main debug: dying input
main debug: control type=0
main debug: control: stopping input
main debug: dying input
main debug: thread times: real 0m31.984375s, kernel 0m0.296875s, user 0m12.203125s
avcodec debug: ffmpeg codec (H264 - MPEG-4 AVC (part 10)) stopped
main debug: removing module "avcodec"
main debug: killing decoder fourcc `h264', 0 PES in FIFO
main debug: [0] 0 0
...
main debug: [23] 0 0
main debug: saving a free vout
main debug: removing module "packetizer_h264"
main debug: thread times: real 0m32.063476s, kernel 0m0.031250s, user 0m1.125000s
main debug: removing module "dts"
main debug: killing decoder fourcc `dts ', 0 PES in FIFO
main debug: removing module "dtstofloat32"
main debug: removing module "scaletempo"
main debug: removing module "bandlimited_resampler"
aout_directx debug: closing audio device
aout_directx debug: DirectSoundThread exiting
main debug: thread ended
main debug: thread times: real 0m31.383789s, kernel 0m0.000000s, user 0m0.015625s
main debug: removing module "aout_directx"
main debug: removing module "audio_format"
main debug: removing module "float32_mixer"
main debug: releasing aout
ts debug: pid list:
ts debug: - pid[0] seen
ts debug: - pid[31] seen
ts debug: - pid[256] seen
ts debug: - pid[4097] seen
ts debug: - pid[4113] seen
ts debug: - pid[4352] seen
ts debug: - pid[4353] seen
main debug: Program doesn't contain anymore ES
ts debug: - pid[8191] seen
main debug: removing module "ts"
main debug: removing module "stream_filter_record"
main debug: removing module "filesystem"
main debug: thread ended
main debug: dead input
main debug: thread times: real 0m32.231445s, kernel 0m0.187500s, user 0m0.375000s
main debug: destroying useless vout
direct3d debug: Direct3D scene released successfully
direct3d debug: DirectXEventThread terminating
direct3d debug: DirectXCloseWindow
direct3d debug: WinProc WM_DESTROY
qt4 debug: releasing video...
qt4 debug: IM: Deleting the input
main debug: TIMER input launching for 'gaga.m2ts' : 125.000 ms - Total 125.000 ms / 1 intvls (Avg 125.000 ms)
qt4 debug: Video is not needed anymore
main debug: removing module "qt4"
main debug: removing module "direct3d"
main debug: removing module "vout_wrapper"
main debug: removing module "blend"
main debug: removing module "freetype"
main debug: removing module "yuvp"
main debug: removing module "swscale"

[Jean-Baptiste Kempf]

According to your logs, the acceleration takes place, but this card is too slow to send the data back to VLC's core...

[rumpumpel1]

Hmmm, sending back to the core ? For post-proccessing ?
Well the PC is an older shuttle with an AGP graphics card.
It was supposed to become recycled as an HTPC. I dont' t
know the bandwidth of AGP but for for sure, it's far less
than with PCIe.

Is it possible with an reasonable amount of effort to
implement an option to bypass post-processing in order
to send the data directly to the screen ?

Is this kind of architecture specific to all media-players ?
In other words, would I have more success when trying
another media-player, eg MPC Home Cinema ?

[Jean-Baptiste Kempf]

Yes, the AGP is the issue.

The right architecture isn't done now in VLC. Someone might do it...

MPC-HC should be bettter, indeed.

[rumpumpel1]

Yes, the AGP is the issue.

The right architecture isn't done now in VLC. Someone might do it...

MPC-HC should be bettter, indeed.

confirmed, with MPC-HC is works perfect.