Compress with facebook zstd, decompress with nvComp

Is it possible to realize any performance gains in the case where the entire buffer was compressed with facebook zstd (no chunking) by decompressing with nvComp or does nvComp completely rely on parallel chunk decompression and therefore no gains can be expected?

Update:
I was able to implement the path to successfully decompress cpu-compressed buffers with nvComp, but performance is at best approaching that of CPU decompress. In a multi-threaded test performance drop is quite significant.
I have not, however, implemented any resource sharing optimizations at this point to optimize allocations and object creation between competing threads or within the same threads, which, I'm guessing, should help.

But the main question remains, is it possible to improve performance decompressing CPU-compressed not chunked data? Single or multi-threaded.

Thanks!