Releases: johguse/profanity
profanity-v1.60
profanity-v1.50
Montgomery multiplication has been replaced by improved interleaved Barret reduction which together with improved memory access has led to speedups of 14% on nVidia (GTX 1070) and 8% on AMD (RX 480).
profanity-v1.40
New major optimization implemented. 15% more performance on nVidia (GTX 1070) and 5% more performance on AMD (RX 480). On AMD (or at least RX 480) I suggest using larger than default --inverse-multiple values, for example 32768 or 65536.
profanity-v1.31
Fix contract address generation broken by comment.
profanity-v1.3
This release mainly improves performance on nVidia cards by improving the Keccak implementation that apparently was a major bottleneck on my GTX 1070. I saw a speed boost of about 4x and it's now outperforming all of my RX 480/580 cards.
Some other minor changes by me and other people since the last release is included in this binary:
- Possible to search for contract addresses
- Infinite wait on initialization step fixed
- Output improved by writing to stderr and also by clearing current line when outputting new result
- Max score of 40 is now possible
- Mask now actually possible in mode --matching as given in example in --help
profanity-v1.22
Initialization time is now printed and included in overall timer. Uppercase hex is now handled correctly.
profanity-v1.21
Reworked initialization so devices can be initialized in parallel.
profanity-v1.2
Major overhaul. Users now have more control over memory usage on the device via the inverse-size and inverse-multiple arguments. Measures to avoid exhausting resources on nVidia cards have been added. Slight speed increase overall.
profanity-v1.12
Memory footprint reduced from ~2.47 GiB to ~1 GiB. Statically linked Windows executable runtime libraries.
profanity-v1.11
Improve printed GPU index and fallback to automatic local worksize on error.