Merging two independently compressed deflate chunks in a PNG compatible zstream

I am working on compression bitmap images in png. The input data are quite large, therefore I need to divide the image in chunks, after that I use the deflate compression in nvcomp. Given 2 independently compressed chunks with deflate, I want to connect them. In zlib a similar feature is achieved handling correctly the padding in the last byte, and adding an uncompressed block of size 0 (0x00 0x00 0xFF 0xFF). Is there a similar functionality in nvcomp? At least, can I retrieve the exact number of bits (not bytes) without padding of the compressed block, because in this way I can handle myself the BFINAL|BTYPE header.

The idea is to join in this way multiple compressed chunks, using both GPU speed, avoiding the 64KB limit for the chunk size, and creating a valid zstream that I can store myself in a valid png. I need to work on native Windows (therefore I cannot use DALI, because it is for Windows).