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

On Fri, Apr 29, 2016 at 11:42 AM, Ian Romanick wrote: > On 02/09/2016 12:02 AM, Ian Romanick wrote: >> I submitted a public spec bug for this issue: >> >> https://www.khronos.org/bugzilla/show_bug.cgi?id=1460 >> >> I'm investigating whether a similar bug is needed for the

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

On 02/09/2016 12:02 AM, Ian Romanick wrote: > I submitted a public spec bug for this issue: > > https://www.khronos.org/bugzilla/show_bug.cgi?id=1460 > > I'm investigating whether a similar bug is needed for the SPIR-V > specification. > > I think an argument can be made for either the

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

> On Feb 9, 2016, at 11:23, Tom Stellard wrote: > > We should still add +fp64-denormals even if the backend doesn't do > anything with it now. This is the default, so it doesn’t really matter anyway. -Matt___ mesa-dev mailing list

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

On Mon, Feb 08, 2016 at 09:38:32PM +0100, Marek Olšák wrote: > On Mon, Feb 8, 2016 at 5:08 PM, Tom Stellard wrote: > > On Sat, Feb 06, 2016 at 01:15:42PM +0100, Marek Olšák wrote: > >> From: Marek Olšák > >> > >> This fixes FP16 conversion instructions for

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

> On Feb 8, 2016, at 08:08, Tom Stellard wrote: > > Do SI/CI support fp64 denorms? If so, won't this hurt performance? This is the only mode that should ever be used. I’m not sure why these are options. There technically are separate flush on input or flush on output

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

> On Feb 8, 2016, at 08:08, Tom Stellard 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

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

On Sat, Feb 06, 2016 at 01:15:42PM +0100, Marek Olšák wrote: > From: Marek Olšák > > This fixes FP16 conversion instructions for VI, which has 16-bit floats, > but not SI & CI, which can't disable denorms for those instructions. Do you know why this fixes FP16 conversions?

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

On Mon, Feb 8, 2016 at 1:11 PM, Roland Scheidegger wrote: > This looks good to me, albeit I know nothing about the hw. > So VI could do (just with some restrictios) even full-speed fp32 denorms > whereas SI/CI can't? Interesting, I suppose that would be intended for Yes, VI

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

This looks good to me, albeit I know nothing about the hw. So VI could do (just with some restrictios) even full-speed fp32 denorms whereas SI/CI can't? Interesting, I suppose that would be intended for compute. intel x86 can't even do that (actually, I think skylake can), though certainly other

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

> On Feb 8, 2016, at 12:38, Marek Olšák wrote: > >> >> 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.

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

On 06.02.2016 07:15, Marek Olšák wrote: From: Marek Olšák This fixes FP16 conversion instructions for VI, which has 16-bit floats, but not SI & CI, which can't disable denorms for those instructions. Reviewed-by: Nicolai Hähnle ---

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

On Mon, Feb 8, 2016 at 5:16 PM, Matt Arsenault wrote: > > On Feb 8, 2016, at 08:08, Tom Stellard 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 >

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

On Mon, Feb 8, 2016 at 5:08 PM, Tom Stellard wrote: > On Sat, Feb 06, 2016 at 01:15:42PM +0100, Marek Olšák wrote: >> From: Marek Olšák >> >> This fixes FP16 conversion instructions for VI, which has 16-bit floats, >> but not SI & CI, which can't disable

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

Am 09.02.2016 um 00:53 schrieb Ian Romanick: > On 02/08/2016 03:37 PM, Roland Scheidegger wrote: >> Am 09.02.2016 um 00:02 schrieb Ian Romanick: >>> On 02/08/2016 12:38 PM, Marek Olšák wrote: On Mon, Feb 8, 2016 at 5:08 PM, Tom Stellard wrote: > On Sat, Feb 06, 2016 at

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

FWIW, D3D requires round-toward-zero for conversion to FP16 colorbuffers. radeonsi does the same: round-toward-zero for colorbuffers and round-to-nearest-even for the GLSL packing functions. This looks like an inconsistency that may cause issues though. Marek On Tue, Feb 9, 2016 at 1:12 AM,

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

On 02/08/2016 12:38 PM, Marek Olšák wrote: > On Mon, Feb 8, 2016 at 5:08 PM, Tom Stellard wrote: >> On Sat, Feb 06, 2016 at 01:15:42PM +0100, Marek Olšák wrote: >>> From: Marek Olšák >>> >>> This fixes FP16 conversion instructions for VI, which has 16-bit

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

Am 09.02.2016 um 00:02 schrieb Ian Romanick: > On 02/08/2016 12:38 PM, Marek Olšák wrote: >> On Mon, Feb 8, 2016 at 5:08 PM, Tom Stellard wrote: >>> On Sat, Feb 06, 2016 at 01:15:42PM +0100, Marek Olšák wrote: From: Marek Olšák This fixes

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

On 02/08/2016 03:37 PM, Roland Scheidegger wrote: > Am 09.02.2016 um 00:02 schrieb Ian Romanick: >> On 02/08/2016 12:38 PM, Marek Olšák wrote: >>> On Mon, Feb 8, 2016 at 5:08 PM, Tom Stellard wrote: On Sat, Feb 06, 2016 at 01:15:42PM +0100, Marek Olšák wrote: > From:

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

From: Marek Olšák This fixes FP16 conversion instructions for VI, which has 16-bit floats, but not SI & CI, which can't disable denorms for those instructions. --- src/gallium/drivers/radeonsi/si_shader.c| 14 ++