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

[Jacob Lifshay]

On Thu, Aug 8, 2019 at 9:16 PM lkcl <luke.leighton at gmail.com> wrote:
> Hmmmm, if speed or power consumption of an implementation is compromised by that, it would be bad (and also Khronos nonconformant, see below).

I know that for Vulkan/OpenGL compliance, it's based on if it produces
the correct results, not how long it takes to compute.
A good example of that kind of thing is that OpenGL 4.x is supported
(assuming they've finished by now) on older AMD gpus by using
soft-float for fp64:
https://www.phoronix.com/scan.php?page=news_item&px=Airlie-Preps-Soft-FP64-Mesa
If I recall correctly, soft-float is also used for fp64 on Intel Haswell GPUs.

> If those APIs "lie" about those characteristics, not only will the end users bitch like mad (justifiably), the ASIC will *FAIL* Khronos conformance and compliance and will not be permitted to be sold with the Vulkan and OpenGL badge on it (they're Trademarks).

According to Khronos's website, any Vulkan implementation is declared
conformant from a technical standpoint once it has been verified by
Khronos that it passes the Vulkan conformance tests, which is just a
suite of unit tests.
https://www.khronos.org/vulkan/adopters/
Vulkan/OpenGL/OpenGL ES conformance tests:
https://github.com/KhronosGroup/VK-GL-CTS

Jacob Lifshay