Rename README to README.md

disable buffer by default

note about size_t max value

use cudaMallocHost and cudaMemcpy for buffering

use queue to store buffered frame nums to delete old frame

change size of buffer to size_t to support big size

map and unmap surface in batch outside loop

check if there is enough frame to write video or not

add a way to wait for job to finish executing

pass pipeline to bus watch callback

move nvvideoconvert to buffering branch, support 720p video

guard buffering and getting buffered frames

add thread pool to write video output, use atomic for safer ref count

fix bug use wrong variable to create memory view

memory buffer using preallocate malloc memory