[Mesa-dev] [PATCH] i965: Validate (and resolve) all the bound textures.

Chris Forbes chrisf at ijw.co.nz
Sun Mar 2 00:18:32 PST 2014


Will land this + the sampler key width change since they fix real
bugs, and then follow up with a new series to improve performance.

On Sun, Feb 23, 2014 at 9:25 AM, Kenneth Graunke <kenneth at whitecape.org> wrote:
> On 02/21/2014 09:09 PM, Chris Forbes wrote:
>> BRW_MAX_TEX_UNIT is the static limit on the number of textures we
>> support per-stage, not in total.
>>
>> Core's `Unit` array is sized by MAX_COMBINED_TEXTURE_IMAGE_UNITS, which
>> is significantly larger, and across the various shader stages, up to
>> ctx->Const.MaxCombinedTextureImageUnits elements of it may be actually
>> used.
>>
>> Fixes invisible bad behavior in piglit's max-samplers test (although
>> this escalated to an assertion failure on HSW with texture_view, since
>> non-immutable textures only have _Format set by validation.)
>>
>> Signed-off-by: Chris Forbes <chrisf at ijw.co.nz>
>> Cc: "9.2 10.0 10.1" <mesa-stable at lists.freedesktop.org>
>> Cc: Kenneth Graunke <kenneth at whitecape.org>
>> ---
>>  src/mesa/drivers/dri/i965/brw_draw.c | 2 +-
>>  src/mesa/drivers/dri/i965/brw_tex.c  | 2 +-
>>  2 files changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c
>> index 540834c..bc887fe 100644
>> --- a/src/mesa/drivers/dri/i965/brw_draw.c
>> +++ b/src/mesa/drivers/dri/i965/brw_draw.c
>> @@ -317,7 +317,7 @@ brw_predraw_resolve_buffers(struct brw_context *brw)
>>     /* Resolve depth buffer of each enabled depth texture, and color buffer of
>>      * each fast-clear-enabled color texture.
>>      */
>> -   for (int i = 0; i < BRW_MAX_TEX_UNIT; i++) {
>> +   for (int i = 0; i < ctx->Const.MaxCombinedTextureImageUnits; i++) {
>>        if (!ctx->Texture.Unit[i]._ReallyEnabled)
>>        continue;
>>        tex_obj = intel_texture_object(ctx->Texture.Unit[i]._Current);
>> diff --git a/src/mesa/drivers/dri/i965/brw_tex.c b/src/mesa/drivers/dri/i965/brw_tex.c
>> index 9234e3a..b1f4de0 100644
>> --- a/src/mesa/drivers/dri/i965/brw_tex.c
>> +++ b/src/mesa/drivers/dri/i965/brw_tex.c
>> @@ -47,7 +47,7 @@ void brw_validate_textures( struct brw_context *brw )
>>     struct gl_context *ctx = &brw->ctx;
>>     int i;
>>
>> -   for (i = 0; i < BRW_MAX_TEX_UNIT; i++) {
>> +   for (i = 0; i < ctx->Const.MaxCombinedTextureImageUnits; i++) {
>>        struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i];
>>
>>        if (texUnit->_ReallyEnabled) {
>>
>
> Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>
>
> I am concerned, though, that looping through all 96 texture units on
> every draw call may be bad for performance on CPU-bound workloads.
>
> I wonder if we couldn't use an approach similar to
> update_stage_texture_surfaces, i.e.:
>
> struct gl_program *progs = {
>    &brw->vertex_program, &brw->geometry_program, &brw->fragment_program
> };
>
> for (int p = 0; p < 3; p++) {
>    const gl_program *prog = progs[p];
>    unsigned num_samplers = _mesa_fls(prog->SamplersUsed);
>    for (int i = 0; i < num_samplers; i++) {
>       if (prog->SamplersUsed & (1 << i)) {
>          if (prog->SamplerUnits[i])
>             intel_finalize_mipmap_tree(brw, i);
>    }
> }
>


More information about the mesa-dev mailing list