Thanks. So its not a multiplier, its the couple of added frames for some way of dealing with pipeline stalls.MMAL_PARAMETER_CAMERA_CONFIG "num_preview_video_frames" field is for sizing the internal image pool. For high frame rates there is benefit in adding a few extra images (which will inherently be at a lower resolution) to smooth over any slight stalls in the pipeline. At 30fps triple buffering is sufficient.
Good to know, so I will set them always with a signle variable. I remember looking at RaspiVid.c they were different, but maybe I forgot.Framerate of the preview and video ports need to be the same - there's no mechanism for resampling the video. If they do differ and both ports are active, then the video port takes precedence.
I have measured complete image transfer from DMA start to DMA end callback - I got number close to 46 milliseconds.How long is your blit to the SPI display taking? Are servicing the queue of buffers in a sensible manner eg drop all but the latest buffer every time you want to draw. If you stall the end of the queue then it applies backpressure up the pipeline.
Queue servicing now is as simle as possible. I have a thread that parses VCHIQ messages, if it sees BUFFER_FROM_HOST, it stores them in a list "ready_to_process" buffers and signals the other thread. The other thread (BUFFER_PROCESSOR), wakes up upon signal and processes all buffers in the list. This would be a mixture of preview and encoder buffers in our case.
For a preview buffer - it is processed by only requesting pre-configured DMA SPI to transfer this buffer to display. BUFFER_PROCESSOR thread does not wait for completion and does not release this current buffer - it just goes to the next buffer, ready to be processed. Upon completion of DMA transaction DMA IRQ is being fired, and from this point buffer gets released.
For an encoder buffer (h264 compressed), I memcpy each incoming buffer to special SECTOR RAM, which is 8192 x 512 bytes size. When it is completely filled - it goes to SD card writing thread and SECTOR RAM buffer is swapped with the same but other buffer.
Around 10.How many buffers do you have on the output of the ISP/resize? Both can store up a couple of input frames if you're starving it of output buffers.
Ok, I have implemented MMAL_PARAMETER_SYSTEM_TIME and now can log times, but not sure how much time this whole command requires.Raspivid sets up the pipeline in effectively the same manner, so I can be fairly confident there isn't an inherent issueIt doesn't have the resize / isp, but I'd be surprised if dropping it into the pipeline caused an issue, at least not in the region of seconds.Code:
camera[0] -> video_render camera[1] -> video_encode
I'd suggest disabling the render temporarily, and logging the buffer callbacks with pts values. You can also get MMAL_PARAMETER_SYSTEM_TIME to compare the delta between the pts and real time.
Currently according to the logs, without drawing and any buffer processing at 25 fps, I only see around 10 buffers of one type (encoder or preview) with delay from system time of around 0.5 second maximum, but also may be less
Below is the filtered log of how I receive the buffers
first field , numeric is amount of milliseconds got at the moment of starting to process the buffer.
preview:c00000XXX - is unique handle to buffer
pts - is capture time.
So in line 1 we shows that the shot was done at 10816us, and now is 11116us, 11116 - 10861 = 255ms delay
Code:
[mmal INFO] 11116.772:preview:c0000013,'FRAME_END', pts:10.816305 [mmal INFO] 11156.780:encoder:c0000004,'FRAME_END|KEYFRAME|NAL_END', pts:10.936249[mmal INFO] 11216.802:preview:c0000013,'FRAME_END', pts:10.856286 [mmal INFO] 11276.850:encoder:c0000005,'FRAME_END|NAL_END', pts:11.16212 [mmal INFO] 11316.874:preview:c0000013,'FRAME_END', pts:10.896269 [mmal INFO] 11356.893:encoder:c0000006,'FRAME_END|NAL_END', pts:11.136157 [mmal INFO] 11416.915:preview:c0000013,'FRAME_END', pts:10.936249 [mmal INFO] 11476.960:encoder:c0000007,'FRAME_END|NAL_END', pts:11.216119 [mmal INFO] 11516.993:preview:c0000013,'FRAME_END', pts:11.16212 [mmal INFO] 11557.11:encoder:c0000008,'FRAME_END|NAL_END', pts:11.336064 [mmal INFO] 11617.34:preview:c0000013,'FRAME_END', pts:11.136157 [mmal INFO] 11677.73:encoder:c0000009,'FRAME_END|NAL_END', pts:11.416026 [mmal INFO] 11717.96:preview:c0000013,'FRAME_END', pts:11.216119 [mmal INFO] 11757.120:encoder:c000000a,'FRAME_END|NAL_END', pts:11.535971 [mmal INFO] 11817.156:preview:c0000013,'FRAME_END', pts:11.336064 [mmal INFO] 11877.182:encoder:c000000b,'FRAME_END|NAL_END', pts:11.615933 [mmal INFO] 11917.216:preview:c0000013,'FRAME_END', pts:11.416026 [mmal INFO] 11977.241:encoder:c000000c,'FRAME_END|NAL_END', pts:11.735878 [mmal INFO] 12017.260:preview:c0000013,'FRAME_END', pts:11.535971 [mmal INFO] 12077.299:encoder:c000000d,'FRAME_END|NAL_END', pts:11.815840 [mmal INFO] 12117.336:preview:c0000013,'FRAME_END', pts:11.615933 [mmal INFO] 12177.355:encoder:c000000e,'FRAME_END|NAL_END', pts:11.935785 [mmal INFO] 12217.376:preview:c0000013,'FRAME_END', pts:11.735878 [mmal INFO] 12277.414:encoder:c000000f,'FRAME_END|NAL_END', pts:12.15747 [mmal INFO] 12317.455:preview:c0000013,'FRAME_END', pts:11.815840 [mmal INFO] 12357.475:encoder:c0000010,'FRAME_END|NAL_END', pts:12.135694 [mmal INFO] 12417.491:preview:c0000013,'FRAME_END', pts:11.935785 [mmal INFO] 12517.599:preview:c0000013,'FRAME_END', pts:12.15747 [mmal INFO] 12557.578:encoder:c0000012,'FRAME_END|NAL_END', pts:12.335599 [mmal INFO] 12617.605:preview:c0000013,'FRAME_END', pts:12.135694 [mmal INFO] 12677.644:encoder:c0000003,'FRAME_END|NAL_END', pts:12.415561 [mmal INFO] 12717.683:preview:c0000013,'FRAME_END', pts:12.215655 [mmal INFO] 12757.690:encoder:c0000004,'FRAME_END|NAL_END', pts:12.535506 [mmal INFO] 12817.729:preview:c0000013,'FRAME_END', pts:12.335599 [mmal INFO] 12877.756:encoder:c0000005,'FRAME_END|NAL_END', pts:12.615468 [mmal INFO] 12917.801:preview:c0000013,'FRAME_END', pts:12.415561 [mmal INFO] 12957.807:encoder:c0000006,'FRAME_END|NAL_END', pts:12.735413 [mmal INFO] 13017.830:preview:c0000013,'FRAME_END', pts:12.535506 [mmal INFO] 13077.866:encoder:c0000007,'FRAME_END|NAL_END', pts:12.815375 [mmal INFO] 13117.913:preview:c0000013,'FRAME_END', pts:12.615468 [mmal INFO] 13157.917:encoder:c0000008,'FRAME_END|NAL_END', pts:12.935320 [mmal INFO] 13217.941:preview:c0000013,'FRAME_END', pts:12.735413 [mmal INFO] 13277.979:encoder:c0000009,'FRAME_END|NAL_END', pts:13.15282 [mmal INFO] 13318.26:preview:c0000013,'FRAME_END', pts:12.815375 [mmal INFO] 13358.28:encoder:c000000a,'FRAME_END|NAL_END', pts:13.135228 [mmal INFO] 13418.55:preview:c0000013,'FRAME_END', pts:12.935320 [mmal INFO] 13459.612:encoder:c000000b,'FRAME_END|NAL_END', pts:13.215189 [mmal INFO] 13518.136:preview:c0000013,'FRAME_END', pts:13.15282 [mmal INFO] 13547.294:encoder:c000000c,'FRAME_END|NAL_END', pts:13.335134 [mmal INFO] 13638.259:preview:c0000013,'FRAME_END', pts:13.135228 [mmal INFO] 13678.212:encoder:c000000d,'FRAME_END|NAL_END', pts:13.415096 [mmal INFO] 13738.261:preview:c0000013,'FRAME_END', pts:13.215189 [mmal INFO] 13798.307:encoder:c000000e,'FRAME_END|NAL_END', pts:13.535041 [mmal INFO] 13838.299:preview:c0000013,'FRAME_END', pts:13.335134 [mmal INFO] 13878.327:encoder:c000000f,'FRAME_END|NAL_END', pts:13.654986 [mmal INFO] 13938.369:preview:c0000013,'FRAME_END', pts:13.415096 [mmal INFO] 13979.397:encoder:c0000010,'FRAME_END|NAL_END', pts:13.734948 [mmal INFO] 14038.441:preview:c0000013,'FRAME_END', pts:13.535041 [mmal INFO] 14067.675:encoder:c0000011,'FRAME_END|NAL_END', pts:13.854893 [mmal INFO] 14158.504:preview:c0000013,'FRAME_END', pts:13.654986 [mmal INFO] 14198.508:encoder:c0000012,'FRAME_END|NAL_END', pts:13.934855 [mmal INFO] 14258.542:preview:c0000013,'FRAME_END', pts:13.734948 [mmal INFO] 14299.248:encoder:c0000003,'FRAME_END|NAL_END', pts:14.54800 Statistics: Posted by valc — Thu May 23, 2024 10:31 pm