[Mesa-dev] [PATCH 5/5] i965/miptree: Don't gtt map from map_depthstencil

Scott D Phillips scott.d.phillips at intel.com
Tue Apr 3 18:53:19 UTC 2018


Nanley Chery <nanleychery at gmail.com> writes:

> On Tue, Jan 09, 2018 at 11:17:02PM -0800, Scott D Phillips wrote:
>> Instead of gtt mapping, call out to other map functions (map_map
>> or map_tiled_memcpy) for the depth surface. Removes a place where
>> gtt mapping is used.
>> ---
>> This is a bit icky, perhaps something like mapping z_mt with
>> BRW_MAP_DIRECT_BIT could be cleaner (but in that case the
>> depthstencil mapping and the DIRECT one would fight for the map
>> slot in mt->level[level].slice[slice].map).
>> 
>>  src/mesa/drivers/dri/i965/intel_mipmap_tree.c | 48 +++++++++++++++++----------
>>  1 file changed, 30 insertions(+), 18 deletions(-)
>> 
>> diff --git a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
>> index fa4ae06399..0b9aafe205 100644
>> --- a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
>> +++ b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
>> @@ -3460,16 +3460,21 @@ intel_miptree_map_depthstencil(struct brw_context *brw,
>>      * temporary buffer back out.
>>      */
>>     if (!(map->mode & GL_MAP_INVALIDATE_RANGE_BIT)) {
>> +      struct intel_miptree_map z_mt_map = {
>> +         .mode = map->mode & ~GL_MAP_WRITE_BIT, .x = map->x, .y = map->y,
>
> The old paths were simpler in that they constants instead of map->mode.
> Why the change?

The reason is that the map/unmap_tiled_memcpy will do the tiling at both
map and unmap time depending on the mode flags. If we didn't alter the
flags then both map_depthstencil and unmap_depthstencil could do both a
read and a write, doubling the memory traffic needlessly.

>> +         .w = map->w, .h = map->h,
>> +      };
>> +      if (z_mt->surf.tiling == ISL_TILING_LINEAR)
>> +         intel_miptree_map_map(brw, z_mt, &z_mt_map, level, slice);
>> +      else
>> +         intel_miptree_map_tiled_memcpy(brw, z_mt, &z_mt_map, level, slice);
>> +      uint32_t *z_map = z_mt_map.ptr;
>>        uint32_t *packed_map = map->ptr;
>>        uint8_t *s_map = intel_miptree_map_raw(brw, s_mt, GL_MAP_READ_BIT);
>> -      uint32_t *z_map = intel_miptree_map_raw(brw, z_mt, GL_MAP_READ_BIT);
>>        unsigned int s_image_x, s_image_y;
>> -      unsigned int z_image_x, z_image_y;
>>  
>>        intel_miptree_get_image_offset(s_mt, level, slice,
>>  				     &s_image_x, &s_image_y);
>> -      intel_miptree_get_image_offset(z_mt, level, slice,
>> -				     &z_image_x, &z_image_y);
>>  
>>        for (uint32_t y = 0; y < map->h; y++) {
>>  	 for (uint32_t x = 0; x < map->w; x++) {
>> @@ -3478,9 +3483,7 @@ intel_miptree_map_depthstencil(struct brw_context *brw,
>>  						 map_x + s_image_x,
>>  						 map_y + s_image_y,
>>  						 brw->has_swizzling);
>> -	    ptrdiff_t z_offset = ((map_y + z_image_y) *
>> -                                  (z_mt->surf.row_pitch / 4) +
>> -				  (map_x + z_image_x));
>> +	    ptrdiff_t z_offset = y * (z_mt_map.stride / 4) + x;
>>  	    uint8_t s = s_map[s_offset];
>>  	    uint32_t z = z_map[z_offset];
>>  
>> @@ -3494,12 +3497,15 @@ intel_miptree_map_depthstencil(struct brw_context *brw,
>>        }
>>  
>>        intel_miptree_unmap_raw(s_mt);
>> -      intel_miptree_unmap_raw(z_mt);
>> +      if (z_mt->surf.tiling == ISL_TILING_LINEAR)
>> +         intel_miptree_unmap_map(z_mt);
>> +      else
>> +         intel_miptree_unmap_tiled_memcpy(brw, z_mt, &z_mt_map, level, slice);
>>  
>>        DBG("%s: %d,%d %dx%d from z mt %p %d,%d, s mt %p %d,%d = %p/%d\n",
>>  	  __func__,
>>  	  map->x, map->y, map->w, map->h,
>> -	  z_mt, map->x + z_image_x, map->y + z_image_y,
>> +	  z_mt, map->x, map->y,
>
> I can see this update and the similar one below leading to confusion for
> a user reading the debug output if they aren't aware of this change. The
> user may map a rectangle that's not at (level,slice) (0,0) and be
> surprised that the second x,y coordinate is unchanged from the first.
> One solution would be to instead print the level and slice for the z mt.

Good idea, I've added that to the patch.

> -Nanley
>
>>  	  s_mt, map->x + s_image_x, map->y + s_image_y,
>>  	  map->ptr, map->stride);
>>     } else {
>> @@ -3521,16 +3527,21 @@ intel_miptree_unmap_depthstencil(struct brw_context *brw,
>>     bool map_z32f_x24s8 = mt->format == MESA_FORMAT_Z_FLOAT32;
>>  
>>     if (map->mode & GL_MAP_WRITE_BIT) {
>> +      struct intel_miptree_map z_mt_map = {
>> +         .mode = map->mode | GL_MAP_INVALIDATE_RANGE_BIT, .x = map->x,
>> +         .y = map->y, .w = map->w, .h = map->h,
>> +      };
>> +      if (z_mt->surf.tiling == ISL_TILING_LINEAR)
>> +         intel_miptree_map_map(brw, z_mt, &z_mt_map, level, slice);
>> +      else
>> +         intel_miptree_map_tiled_memcpy(brw, z_mt, &z_mt_map, level, slice);
>> +      uint32_t *z_map = z_mt_map.ptr;
>>        uint32_t *packed_map = map->ptr;
>>        uint8_t *s_map = intel_miptree_map_raw(brw, s_mt, GL_MAP_WRITE_BIT);
>> -      uint32_t *z_map = intel_miptree_map_raw(brw, z_mt, GL_MAP_WRITE_BIT);
>>        unsigned int s_image_x, s_image_y;
>> -      unsigned int z_image_x, z_image_y;
>>  
>>        intel_miptree_get_image_offset(s_mt, level, slice,
>>  				     &s_image_x, &s_image_y);
>> -      intel_miptree_get_image_offset(z_mt, level, slice,
>> -				     &z_image_x, &z_image_y);
>>  
>>        for (uint32_t y = 0; y < map->h; y++) {
>>  	 for (uint32_t x = 0; x < map->w; x++) {
>> @@ -3538,9 +3549,7 @@ intel_miptree_unmap_depthstencil(struct brw_context *brw,
>>  						 x + s_image_x + map->x,
>>  						 y + s_image_y + map->y,
>>  						 brw->has_swizzling);
>> -	    ptrdiff_t z_offset = ((y + z_image_y + map->y) *
>> -                                  (z_mt->surf.row_pitch / 4) +
>> -				  (x + z_image_x + map->x));
>> +	    ptrdiff_t z_offset = y * (z_mt_map.stride / 4) + x;
>>  
>>  	    if (map_z32f_x24s8) {
>>  	       z_map[z_offset] = packed_map[(y * map->w + x) * 2 + 0];
>> @@ -3554,13 +3563,16 @@ intel_miptree_unmap_depthstencil(struct brw_context *brw,
>>        }
>>  
>>        intel_miptree_unmap_raw(s_mt);
>> -      intel_miptree_unmap_raw(z_mt);
>> +      if (z_mt->surf.tiling == ISL_TILING_LINEAR)
>> +         intel_miptree_unmap_map(z_mt);
>> +      else
>> +         intel_miptree_unmap_tiled_memcpy(brw, z_mt, &z_mt_map, level, slice);
>>  
>>        DBG("%s: %d,%d %dx%d from z mt %p (%s) %d,%d, s mt %p %d,%d = %p/%d\n",
>>  	  __func__,
>>  	  map->x, map->y, map->w, map->h,
>>  	  z_mt, _mesa_get_format_name(z_mt->format),
>> -	  map->x + z_image_x, map->y + z_image_y,
>> +	  map->x, map->y,
>>  	  s_mt, map->x + s_image_x, map->y + s_image_y,
>>  	  map->ptr, map->stride);
>>     }
>> -- 
>> 2.14.3
>> 
>> _______________________________________________
>> mesa-dev mailing list
>> mesa-dev at lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/mesa-dev
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/mesa-dev


More information about the mesa-dev mailing list