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

[lkcl]

On Friday, August 9, 2019 at 1:47:18 PM UTC+8, Allen Baum wrote:

> But unless it becomes a ratified standard, the RV Compliance Framework and Test Suite just won't deal with it.

I see the logic behind that. It costs money and time.

> If the customer is big enough, and the market is big enough, maybe that custom extension becomes a standard - at which point it all works.

Indeed. The doenside: that unfortunately would require a wing and a prayer that the opcode space doesn't get used up for other [official] purposes in between those two events.

It's an extremely risky approach, as the implementor will sure as hellfire exists not want their product - one that cost them tens to hundreds of millions to develop and market - to be "relegated" to nonstandard status by an *incompatible* post-silicon after-sales Standardisation effort, in the [highly likely] event of an opcode clash.

You can guarantee they'll fight that one, to the detriment of the entire RISCV community [cf: Altivec SSE nightmare]


On another note: I think, *deep breath*, sad to say it, the RISCV Foundation looks at our team, operating from the outside, excluded from participation due to the Membership Agreement being an NDA, and considers us to be a bit of a joke.

Our input - including warnings - can therefore be "safely ignored", just like Open Source contributors ideas and input can and have been ignored, for many years, now.

Because we don't come with a billion dollar Corporate cheque book automatically attached, our input and perspective cannot possibly have any impact, because how can such unrealistically stupid and deluded idealistic time wasters possibly get the money to actually deliver silicon, right?

Things will get a lot easier when that perspective changes. I hope and trust that that change occurs before it is too late.

L.