[libre-riscv-dev] [isa-dev] Re: FP transcendentals (trigonometry, root/exp/log) proposal

[Hendrik Boom]

On Thu, Jan 16, 2020 at 04:22:21AM +0100, whygee at f-cpu.org wrote:
> On 2020-01-16 04:04, lkcl . wrote:
> > as in: for some low-cost algorithm implementations their accuracy is
> > reduced sufficiently that the last place is wrong a significant amount
> > of
> > the time (let alone rounding).  how do you *know* when the rounding (or
> > the
> > last bit of the mantissa) is wrong?
> 
> That was my reaction :
> if you *know* your LSB is wrong then you can correct it...

The problem arise when you do't know it's wrong, but you also don't 
know it's right.

-- hendrik

> and if the correction is minimal, triggering an exception is overkill.
> A flag would be enough...
> 
> > l.
> yg
> 
> _______________________________________________
> libre-riscv-dev mailing list
> libre-riscv-dev at lists.libre-riscv.org
> http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev