Re: [Mesa-dev] [PATCH 1/2] nir: add an isinf opcode, and an option to use it.

2017-03-16 Thread Roland Scheidegger
Am 17.03.2017 um 04:33 schrieb Roland Scheidegger: > Am 17.03.2017 um 02:29 schrieb Dave Airlie: >> On 17 March 2017 at 11:09, Jason Ekstrand wrote: >>> On March 16, 2017 5:04:37 PM Dave Airlie wrote: >>> From: Dave Airlie

Re: [Mesa-dev] [PATCH 1/2] nir: add an isinf opcode, and an option to use it.

2017-03-16 Thread Jason Ekstrand
On Thu, Mar 16, 2017 at 8:33 PM, Roland Scheidegger wrote: > Am 17.03.2017 um 02:29 schrieb Dave Airlie: > > On 17 March 2017 at 11:09, Jason Ekstrand wrote: > >> On March 16, 2017 5:04:37 PM Dave Airlie wrote: > >> > >>> From: Dave

Re: [Mesa-dev] [PATCH 1/2] nir: add an isinf opcode, and an option to use it.

2017-03-16 Thread Roland Scheidegger
Am 17.03.2017 um 02:29 schrieb Dave Airlie: > On 17 March 2017 at 11:09, Jason Ekstrand wrote: >> On March 16, 2017 5:04:37 PM Dave Airlie wrote: >> >>> From: Dave Airlie >>> >>> In order to get isinf(NaN) correct, at least radv can't

[Mesa-dev] [PATCH 1/2] nir: add an isinf opcode, and an option to use it. (v2)

2017-03-16 Thread Dave Airlie
From: Dave Airlie In order to get isinf(NaN) correct, at least radv can't use an unordered equals which feq has to be for us, this passes isinf to the backend and let's it sort it out as it pleases. This turns lowering on for i965 only as it's the only other spir-v consumer

Re: [Mesa-dev] [PATCH 1/2] nir: add an isinf opcode, and an option to use it.

2017-03-16 Thread Dave Airlie
> Another option would be to make this lower_isinf and add a quick lowering > line to nir_opt_algebraic. That's more idiomatic for nir. If I do that though won't that mean I have to set lower_isinf for all current NIR users? Dave. ___ mesa-dev mailing

Re: [Mesa-dev] [PATCH 1/2] nir: add an isinf opcode, and an option to use it.

2017-03-16 Thread Dave Airlie
On 17 March 2017 at 11:09, Jason Ekstrand wrote: > On March 16, 2017 5:04:37 PM Dave Airlie wrote: > >> From: Dave Airlie >> >> In order to get isinf(NaN) correct, at least radv can't >> use an unordered equals which feq has to be for

Re: [Mesa-dev] [PATCH 1/2] nir: add an isinf opcode, and an option to use it.

2017-03-16 Thread Jason Ekstrand
On March 16, 2017 5:04:37 PM Dave Airlie wrote: From: Dave Airlie In order to get isinf(NaN) correct, at least radv can't use an unordered equals which feq has to be for us, this passes isinf to the backend and let's it sort it out as it pleases. I

Re: [Mesa-dev] [PATCH 1/2] nir: add an isinf opcode, and an option to use it.

2017-03-16 Thread Bas Nieuwenhuizen
On Fri, Mar 17, 2017 at 1:04 AM, Dave Airlie wrote: > From: Dave Airlie > > In order to get isinf(NaN) correct, at least radv can't > use an unordered equals which feq has to be for us, this Why do we have to use an unordered equal normally? SPIR-V has

[Mesa-dev] [PATCH 1/2] nir: add an isinf opcode, and an option to use it.

2017-03-16 Thread Dave Airlie
From: Dave Airlie In order to get isinf(NaN) correct, at least radv can't use an unordered equals which feq has to be for us, this passes isinf to the backend and let's it sort it out as it pleases. Signed-off-by: Dave Airlie --- src/compiler/nir/nir.h