look at this CUDA-based MD5/SHA1 implementations: https://sourceforge.net/projects/storegpu/
CUDA can be used to run MD5/SHA1 on different files (or messages) in parallel, but it is useless in the most common case of processing single file.
Current implementation of MD5/SHA1 gives speed about 300 MB per second on Core 2 CPU, so the real bottleneck is filesystem I/O :)
Log in to post a comment.
CUDA can be used to run MD5/SHA1 on different files (or messages) in parallel, but it is useless in the most common case of processing single file.
Current implementation of MD5/SHA1 gives speed about 300 MB per second on Core 2 CPU, so the real bottleneck is filesystem I/O :)