It seems that Jaguar's implementation is the only one that achieved the 100TPS. Is this true? No offense to the others,but would there be a reason for people to use the slower versions?
I might have misunderstood the benchmarks in that thread as I didn't find any benchmarks of all four on the same computer
What you don't see in the scores, is that there was both rivalry and cooperation.
I think hoax was the first one to have fastest implementation.
Jaguar's initial trials were quite slow, and his approach changed after tips from both hoax (drop long) and BloodyRookie (switch to 16bit math).
Also BloodyRookie probably nailed down the problem in math behind
Curve.sign, which probably deserves
separte credit itself:
https://bitcointalk.org/index.php?topic=397183.msg4645132#msg4645132I was bit late t the party, I joined when I've seen info in "flaws" thread:
https://bitcointalk.org/index.php?topic=397183.1080The only credit I think I can give to myself is "hinting", what's up with magic
is_negative...
Still, we all have spent fair amount of time working on this