Jeffrey Sarnoff

Is this the logic you see for mul with tiny [is huge messed up also?]: > @inline function mul_round(a::T, b::T, rounding::R)::T where {T

@dpsanders @JeffreySarnoff ```julia function mul_round(a::T, b::T, rounding::RoundingMode) where {T

natch

Because most cases are unexceptional, the usual extra cost is `!(iszero(hi) || isinf(hi))` and that is not nearly as costly as other approaches. So, if I do something similar with...

this fixes the first of those two issues ```julia function mul_round(a::T, b::T, rounding::RoundingMode) where {T

I may have an approach to the second. ``` julia> two_prod(0.25,0.25) (0.0625, 0.0) ``` you want `mul_round(0.25, 0.25, RoundUp) == mul_round(0.25, 0.25, RoundDown)` right? It's going to be about differentiating...

I added the newer mul_round, and your sqrt. Is it necessary to round the test squared-value up[down]. It looks as though the logic wouldwork using round to nearest, so the...

do you have a few test cases for division using different rounding modes? if so -- let me see

the new mul_round: ```julia @inline function mul_round(a::T, b::T, rounding::RoundingMode) where {T

lean on me