[Mesa-dev] [PATCH] radeonsi: enable denorms for 64-bit and 16-bit floats

[Marek Olšák]

On Mon, Feb 8, 2016 at 5:16 PM, Matt Arsenault <arsenm2 at gmail.com> wrote:
>
> On Feb 8, 2016, at 08:08, Tom Stellard <tom at stellard.net> wrote:
>
> Do SI/CI support fp64 denorms?  If so, won't this hurt performance?
>
> We should tell the compiler we are enabling fp-64 denorms by adding
> +fp64-denormals to the feature string.  It would also be better to
> read the float_mode value from the config registers emitted by the
> compiler.
>
>
> Yes, the runtime here should read the value out of the binary and enable it
> in the compiler rather than the runtime hardcoding it. If you wanted to load
> a shader with different FP rules for example it should be able to switch.

This is not the runtime. This is the compiler. :) It's the middle-end
though, not the back-end.

I would use +fp64-denormals if it did something to graphics shaders.

Marek