When using algorithm of video processing, a card, how to do? How to do to speed up processing speed? And at this time the CPU utilization rate is very low

CodePudding user response:

General to release model open

Such as SSE accelerated (essentially open and openCL acceleration, etc.)

CodePudding user response:

Should use the gpu is ok, change graphics must clear

CodePudding user response:

Caton is definitely your processing time longer than a single frame, the two aspects, low CPU usage that your algorithm is low, multi-threaded use can improve from this aspect, the other image algorithm can borrow CUDA GPU calls that there will be a significant boost

CodePudding user response:

Our 4320 * 2160 double fisheye on joining together for panoramic video, video do is real-time, BuKa, suggested that the fully use graphics performance

CodePudding user response:

Algorithm and the handler is no problem, will consider memory chips and graphics performance, but in fact very card is what meaning, how long will it take? Don't the thread processing, then the processing time required?