[Mesa-dev] [PATCH 01/13] radeonsi: reduce the size of si_pm4_state

[Marek Olšák]

Sure. I've reverted it.

Marek

On Wed, Jan 7, 2015 at 9:03 PM, Tom Stellard <tom at stellard.net> wrote:
> On Mon, Jan 05, 2015 at 12:18:40AM +0100, Marek Olšák wrote:
>> From: Marek Olšák <marek.olsak at amd.com>
>>
>> - the relocs array is unused, remove it
>> - ndw is at most 115 (init), set 140 as the maximum
>> - compute needs 4 buffers per state, graphics only needs 1; set 4 as the maximum
>
> The number of buffers per state is dependent on the input arguments to the kernel,
> so it can be way more than 4.
>
> OpenCL requires that the input buffer be at least 256 bytes, which means at minimum,
> we must be able to support:
>
> (256 / 8) + 3 (internal buffers) = 35 buffers
>
> This patch breaks most of the OpenCL tests, can we revert it for now?
>
> -Tom
>
>> ---
>>  src/gallium/drivers/radeonsi/si_pm4.c | 6 +-----
>>  src/gallium/drivers/radeonsi/si_pm4.h | 9 ++-------
>>  2 files changed, 3 insertions(+), 12 deletions(-)
>>
>> diff --git a/src/gallium/drivers/radeonsi/si_pm4.c b/src/gallium/drivers/radeonsi/si_pm4.c
>> index 954eb6e..21ab9f2 100644
>> --- a/src/gallium/drivers/radeonsi/si_pm4.c
>> +++ b/src/gallium/drivers/radeonsi/si_pm4.c
>> @@ -145,17 +145,13 @@ unsigned si_pm4_dirty_dw(struct si_context *sctx)
>>  void si_pm4_emit(struct si_context *sctx, struct si_pm4_state *state)
>>  {
>>       struct radeon_winsys_cs *cs = sctx->b.rings.gfx.cs;
>> +
>>       for (int i = 0; i < state->nbo; ++i) {
>>               r600_context_bo_reloc(&sctx->b, &sctx->b.rings.gfx, state->bo[i],
>>                                     state->bo_usage[i], state->bo_priority[i]);
>>       }
>>
>>       memcpy(&cs->buf[cs->cdw], state->pm4, state->ndw * 4);
>> -
>> -     for (int i = 0; i < state->nrelocs; ++i) {
>> -             cs->buf[cs->cdw + state->relocs[i]] += cs->cdw << 2;
>> -     }
>> -
>>       cs->cdw += state->ndw;
>>
>>  #if SI_TRACE_CS
>> diff --git a/src/gallium/drivers/radeonsi/si_pm4.h b/src/gallium/drivers/radeonsi/si_pm4.h
>> index 8680a9e..388bb4b 100644
>> --- a/src/gallium/drivers/radeonsi/si_pm4.h
>> +++ b/src/gallium/drivers/radeonsi/si_pm4.h
>> @@ -29,9 +29,8 @@
>>
>>  #include "radeon/drm/radeon_winsys.h"
>>
>> -#define SI_PM4_MAX_DW                256
>> -#define SI_PM4_MAX_BO                32
>> -#define SI_PM4_MAX_RELOCS    4
>> +#define SI_PM4_MAX_DW                140
>> +#define SI_PM4_MAX_BO                4
>>
>>  // forward defines
>>  struct si_context;
>> @@ -54,10 +53,6 @@ struct si_pm4_state
>>       enum radeon_bo_usage    bo_usage[SI_PM4_MAX_BO];
>>       enum radeon_bo_priority bo_priority[SI_PM4_MAX_BO];
>>
>> -     /* relocs for shader data */
>> -     unsigned        nrelocs;
>> -     unsigned        relocs[SI_PM4_MAX_RELOCS];
>> -
>>       bool compute_pkt;
>>  };
>>
>> --
>> 2.1.0
>>
>> _______________________________________________
>> mesa-dev mailing list
>> mesa-dev at lists.freedesktop.org
>> http://lists.freedesktop.org/mailman/listinfo/mesa-dev