Re: [Mesa-dev] [PATCH 3/3] i965: Stop splitting fma() prior to optimization

2016-05-05 Thread Jason Ekstrand
On May 5, 2016 6:35 PM, "Connor Abbott" wrote: > > On Thu, May 5, 2016 at 8:51 PM, Jason Ekstrand wrote: > > According to the GLSL spec, if the user uses fma directly and you have it > > in your hardware, you shouldn't split it. For a while now, we've been > > splitting all fma's up-front and th

Re: [Mesa-dev] [PATCH 3/3] i965: Stop splitting fma() prior to optimization

2016-05-05 Thread Connor Abbott
On Thu, May 5, 2016 at 8:51 PM, Jason Ekstrand wrote: > According to the GLSL spec, if the user uses fma directly and you have it > in your hardware, you shouldn't split it. For a while now, we've been > splitting all fma's up-front and then planned to fuse them later. The only > reason why this

Re: [Mesa-dev] [PATCH 3/3] i965: Stop splitting fma() prior to optimization

2016-05-05 Thread Jason Ekstrand
On Thu, May 5, 2016 at 6:22 PM, Jason Ekstrand wrote: > > On May 5, 2016 6:13 PM, "Ilia Mirkin" wrote: > > > > On Thu, May 5, 2016 at 8:51 PM, Jason Ekstrand > wrote: > > > According to the GLSL spec, if the user uses fma directly and you have > it > > > in your hardware, you shouldn't split it

Re: [Mesa-dev] [PATCH 3/3] i965: Stop splitting fma() prior to optimization

2016-05-05 Thread Jason Ekstrand
On May 5, 2016 6:13 PM, "Ilia Mirkin" wrote: > > On Thu, May 5, 2016 at 8:51 PM, Jason Ekstrand wrote: > > According to the GLSL spec, if the user uses fma directly and you have it > > in your hardware, you shouldn't split it. For a while now, we've been > > splitting all fma's up-front and then

Re: [Mesa-dev] [PATCH 3/3] i965: Stop splitting fma() prior to optimization

2016-05-05 Thread Ilia Mirkin
On Thu, May 5, 2016 at 8:51 PM, Jason Ekstrand wrote: > According to the GLSL spec, if the user uses fma directly and you have it > in your hardware, you shouldn't split it. For a while now, we've been > splitting all fma's up-front and then planned to fuse them later. The only > reason why this

[Mesa-dev] [PATCH 3/3] i965: Stop splitting fma() prior to optimization

2016-05-05 Thread Jason Ekstrand
According to the GLSL spec, if the user uses fma directly and you have it in your hardware, you shouldn't split it. For a while now, we've been splitting all fma's up-front and then planned to fuse them later. The only reason why this possibly helped before was for ARB programs which is handled b