[libre-riscv-dev] [isa-dev] Re: FP transcendentals (trigonometry, root/exp/log) proposal
andrew at sifive.com
Thu Aug 8 15:14:42 BST 2019
On Thu, Aug 8, 2019 at 7:01 AM lkcl . <luke.leighton at gmail.com> wrote:
> On Thu, Aug 8, 2019 at 11:00 AM Andrew Waterman <andrew at sifive.com> wrote:
> >> wait... hang on: there are now *four* potential Platforms against which
> this statement has to be verified. are you saying that for a *UNIX*
> platform that correctly-rounded transcendentals are potentially undesirable?
> > The sentence you quoted began with the adjective "ISA-level". We
> happily provide correctly rounded transcendental math on Linux as-is.
> i am very confused. we seem to be talking at cross-purposes, and i
> have no idea where the confusion lies.
> it appears that you are rejecting the possibility of providing
> ISA-level support for transcendental and trigonometric operations for
> *four* possible platform scenarios just because "correctly rounded
> transcendental math is provided on linux". this has me utterly
> particularly when, even *on* one of those platforms - the standard
> UNIX Platform - Jacob pointed out that there may exist
> High-Performance Server scenarios that would want the increased
> performance - *on linux* - that such ISA-level support would provide.
> i apologise: i don't understand what is going on.
> why would even *one* argument "we provide accurate SOFTWARE math
> libraries on linux" be reasonable cause to reject HARDWARE support for
> the same?
> why would that argument be relevant for THREE OTHER completely
> different Platform Profiles?
> i don't understand.
I don’t understand the need for all those capital letters.
As I mentioned earlier, an instruction being useful is not by itself a
justification for adding it to the ISA. Where’s the quantitative case for
> You received this message because you are subscribed to the Google Groups
> "RISC-V ISA Dev" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to isa-dev+unsubscribe at groups.riscv.org.
> To view this discussion on the web visit
More information about the libre-riscv-dev