3 ms·
If the goal is to save memory, you can go farther than this. You could store the JPEG in RAM, compressed, and have the GPU decompress & shade. Each of the DCT
by rahimiali 5y ago
If the goal is to save memory, you can go farther than this. You could store the JPEG in RAM, compressed, and have the GPU decompress & shade. Each of the DCT blocks can be converted massively parallely on GPU.
- astrange 5y agoYou can't do image decompression in GPGPU because the last step (Huffman coding) is not parallelizable - in fact if any compression is parallelizable it's not compressed enough. But you can do it in dedicated hardware.
- rahimiali 5y agoOk, maybe not the arithmetic code (or huffman), but I was thinking you'd at least invert the DCT on GPU.