fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2025-12-20 11:40:10 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

freedreno/lrz: Add gen8 lrz layout support

Browse source 
Signed-off-by: Rob Clark <rob.clark@oss.qualcomm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38930>
This commit is contained in:
Rob Clark 2025-12-12 06:28:47 -08:00 • committed by Marge Bot
parent 81c465372a
commit 14a23e8b3e
2 changed files with 66 additions and 10 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
src/freedreno/common/freedreno_lrz.h
View file

@ -69,7 +69,7 @@ template <chip CHIP>
struct PACKED fd_lrzfc_layout<chip_range(CHIP >= A7XX)> { struct PACKED fd_lrzfc_layout<chip_range(CHIP >= A7XX)> {
static const bool HAS_BIDIR = true; static const bool HAS_BIDIR = true;
static const bool HAS_CB = true; static const bool HAS_CB = true;
static const size_t FC_SIZE = 1024; static const size_t FC_SIZE = (CHIP < A8XX) ? 1024 : 2048;
union { union {
struct { struct {

68
src/freedreno/fdl/freedreno_lrz_layout.h
View file

@ -15,6 +15,7 @@ BEGINC;
struct fdl_lrz_layout { struct fdl_lrz_layout {
uint32_t lrz_offset; uint32_t lrz_offset;
uint32_t lrz_pitch; uint32_t lrz_pitch;
uint32_t lrz_slice_pitch; /* gen8+ */
uint32_t lrz_height; uint32_t lrz_height;
uint32_t lrz_layer_size; uint32_t lrz_layer_size;
uint32_t lrz_buffer_size; uint32_t lrz_buffer_size;
@ -120,17 +121,72 @@ fdl6_lrz_layout_init(struct fdl_lrz_layout *lrz_layout,
const struct fd_dev_info *dev_info, uint32_t lrz_offset, const struct fd_dev_info *dev_info, uint32_t lrz_offset,
uint32_t array_layers) uint32_t array_layers)
{ {
if (CHIP >= A8XX) {
static const struct lrz_block {
uint16_t width;
uint16_t height;
} lrz_block_sizes[4][4] = {
[0] = {
[MSAA_ONE] = { 64, 128 },
[MSAA_TWO] = { 64, 64 },
[MSAA_FOUR] = { 32, 64 },
[MSAA_EIGHT] = { 32, 32 },
},
[1] = {
[MSAA_ONE] = { 128, 128 },
[MSAA_TWO] = { 128, 64 },
[MSAA_FOUR] = { 64, 64 },
[MSAA_EIGHT] = { 64, 32 },
},
[2] = {
[MSAA_ONE] = { 192, 128 },
[MSAA_TWO] = { 192, 64 },
[MSAA_FOUR] = { 96, 64 },
[MSAA_EIGHT] = { 96, 32 },
},
[3] = {
[MSAA_ONE] = { 128, 256 },
[MSAA_TWO] = { 128, 128 },
[MSAA_FOUR] = { 64, 128 },
[MSAA_EIGHT] = { 64, 64 },
},
}, *lrz_block = &lrz_block_sizes[dev_info->num_slices - 1][ffs(layout->nr_samples) - 1];
const unsigned per_slice_block_width_in_tiles = 8;
const unsigned per_slice_block_height_in_tiles = 16;
const unsigned surface_width_in_blocks =
DIV_ROUND_UP(layout->width0 + extra_width, lrz_block->width);
const unsigned surface_height_in_blocks =
DIV_ROUND_UP(layout->height0 + extra_height, lrz_block->height);
lrz_layout->lrz_pitch =
align(surface_width_in_blocks * per_slice_block_width_in_tiles, 64);
/* Construct a "fake" height to use for fallback lrz clear on
* the blitter. Since we use lrz_pitch as the width, this is
* just lrz_layer_size / lrz_pitch
*/
lrz_layout->lrz_height = dev_info->num_slices *
surface_height_in_blocks * per_slice_block_height_in_tiles;
lrz_layout->lrz_slice_pitch = surface_height_in_blocks *
per_slice_block_height_in_tiles * lrz_layout->lrz_pitch * 2;
lrz_layout->lrz_layer_size = lrz_layout->lrz_slice_pitch * dev_info->num_slices;
} else {
unsigned width = layout->width0 + extra_width; unsigned width = layout->width0 + extra_width;
unsigned height = layout->height0 + extra_height; unsigned height = layout->height0 + extra_height;
fdl6_lrz_get_super_sampled_size(&width, &height, layout->nr_samples); fdl6_lrz_get_super_sampled_size(&width, &height, layout->nr_samples);
unsigned lrz_pitch = align(DIV_ROUND_UP(width, 8), 32); lrz_layout->lrz_pitch = align(DIV_ROUND_UP(width, 8), 32);
unsigned lrz_height = align(DIV_ROUND_UP(height, 8), 32); lrz_layout->lrz_height = align(DIV_ROUND_UP(height, 8), 32);
lrz_layout->lrz_layer_size =
lrz_layout->lrz_pitch * lrz_layout->lrz_height * sizeof(uint16_t);
}
lrz_layout->lrz_offset = lrz_offset; lrz_layout->lrz_offset = lrz_offset;
lrz_layout->lrz_height = lrz_height;
lrz_layout->lrz_pitch = lrz_pitch;
lrz_layout->lrz_layer_size = lrz_pitch * lrz_height * sizeof(uint16_t);
lrz_layout->lrz_buffer_size = lrz_layout->lrz_layer_size * array_layers; lrz_layout->lrz_buffer_size = lrz_layout->lrz_layer_size * array_layers;
/* Fast-clear buffer is 1bit/block */ /* Fast-clear buffer is 1bit/block */
@ -144,7 +200,7 @@ fdl6_lrz_layout_init(struct fdl_lrz_layout *lrz_layout,
/* Allocate 2 LRZ buffers for double-buffering on a7xx. */ /* Allocate 2 LRZ buffers for double-buffering on a7xx. */
uint32_t lrz_size = lrz_layout->lrz_buffer_size * uint32_t lrz_size = lrz_layout->lrz_buffer_size *
(dev_info->chip >= 7 ? 2 : 1); (CHIP >= A7XX ? 2 : 1);
if (dev_info->props.enable_lrz_fast_clear || if (dev_info->props.enable_lrz_fast_clear ||
dev_info->props.has_lrz_dir_tracking) { dev_info->props.has_lrz_dir_tracking) {