Dear All:
Recently we are using DMA (/dev/dm350mmap) to split a 5MP image into 4 smaller images,
using IPIPE to process them and using DMA to merge them into one big image. It works OK.
However, the DMA operations used to merged the sub-images are very very slow.
The major differences are the dstPitch and srcPitch.
For splitting, I set srcPitch as 2592, dstPitch as 1296, the width as 1296 and the height as 972.
For merging, I set srcPitch as 1296, dstPitch as 2592, the width as 1296 and the height as 972.
In this experiment, I did not set any overlap area.
The result is a little bit surprising.
The splitting DMA took about 0.02-0.03 secs, but the merging DMA took about 0.3 secs.
I do not know why these two operations have such a big difference in the performance.
Is there anyway I can improve it?
Any help will be appreciated.
Best regards,
Ray