Quote from: jl777 on January 21, 2014, 07:05:15 PM

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#msg4645132

I was bit late t the party, I joined when I've seen info in "flaws" thread: https://bitcointalk.org/index.php?topic=397183.1080
The 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