0.9999~ = 9/9
0.8888~ = 8/9
saved you 7:20
0.8888~ = 8/9
saved you 7:20
Timeline
Post
Remote status
Replies
7
@lain but what is 0.00000.... equal to???
@bartholin it is a mystery
@cell @bartholin IEEE is the sound you make when you figure out how floating point math works (or doesn't)
Remember when DJB implemented poly1305 on the FPU ?
That's the one.
Per my recollection, it's based on 16*16->32 multiplication on 53 bit doubles, which is kind of wasteful. IMO it should have been 26*26->52 multiplication, which can do 128 bits with 5 limbs. There were other implementations of Poly1305 which used 26 bit limbs too, but they were in int32 and doing classical 32*32->64 multiplication.
Per my recollection, it's based on 16*16->32 multiplication on 53 bit doubles, which is kind of wasteful. IMO it should have been 26*26->52 multiplication, which can do 128 bits with 5 limbs. There were other implementations of Poly1305 which used 26 bit limbs too, but they were in int32 and doing classical 32*32->64 multiplication.
@lain I don't believe he's genuinely curious. I think it's an act.