Re: [Mesa-dev] [PATCH 15/18] radeonsi: upload constants into VRAM instead of GTT

2017-02-17 Thread Nicolai Hähnle
On 16.02.2017 23:36, Marek Olšák wrote: On Thu, Feb 16, 2017 at 4:21 PM, Nicolai Hähnle wrote: On 16.02.2017 13:53, Marek Olšák wrote: From: Marek Olšák This lowers lgkm wait cycles by 30% on VI and normal conditions. The might be a measurable improvement when CE is disabled (radeon) or und

Re: [Mesa-dev] [PATCH 15/18] radeonsi: upload constants into VRAM instead of GTT

2017-02-16 Thread Marek Olšák
On Thu, Feb 16, 2017 at 4:21 PM, Nicolai Hähnle wrote: > On 16.02.2017 13:53, Marek Olšák wrote: >> >> From: Marek Olšák >> >> This lowers lgkm wait cycles by 30% on VI and normal conditions. >> The might be a measurable improvement when CE is disabled (radeon) >> or under L2 thrashing. > > > Goo

Re: [Mesa-dev] [PATCH 15/18] radeonsi: upload constants into VRAM instead of GTT

2017-02-16 Thread Nicolai Hähnle
On 16.02.2017 13:53, Marek Olšák wrote: From: Marek Olšák This lowers lgkm wait cycles by 30% on VI and normal conditions. The might be a measurable improvement when CE is disabled (radeon) or under L2 thrashing. Good idea. I'm just wondering if all the users of const upload end up as stream

[Mesa-dev] [PATCH 15/18] radeonsi: upload constants into VRAM instead of GTT

2017-02-16 Thread Marek Olšák
From: Marek Olšák This lowers lgkm wait cycles by 30% on VI and normal conditions. The might be a measurable improvement when CE is disabled (radeon) or under L2 thrashing. --- src/gallium/drivers/radeon/r600_pipe_common.c | 11 --- src/gallium/drivers/radeonsi/si_compute.c | 4 ++--