2018-03-20 02:00:29 +03:00
|
|
|
// Copyright 2018 yuzu Emulator Project
|
|
|
|
// Licensed under GPLv2 or any later version
|
|
|
|
// Refer to the license.txt file included.
|
|
|
|
|
2018-07-18 07:15:20 +03:00
|
|
|
#include <cmath>
|
2018-03-22 23:17:10 +03:00
|
|
|
#include <cstring>
|
2018-10-12 21:21:53 +03:00
|
|
|
#include "common/alignment.h"
|
2018-03-22 23:17:10 +03:00
|
|
|
#include "common/assert.h"
|
2018-07-02 20:42:04 +03:00
|
|
|
#include "video_core/gpu.h"
|
2018-03-20 02:00:29 +03:00
|
|
|
#include "video_core/textures/decoders.h"
|
2018-03-22 23:17:10 +03:00
|
|
|
#include "video_core/textures/texture.h"
|
2018-03-20 02:00:29 +03:00
|
|
|
|
2018-07-21 01:14:17 +03:00
|
|
|
namespace Tegra::Texture {
|
2018-03-20 02:00:29 +03:00
|
|
|
|
2018-03-22 23:17:10 +03:00
|
|
|
/**
|
2018-09-21 18:29:54 +03:00
|
|
|
* This table represents the internal swizzle of a gob,
|
|
|
|
* in format 16 bytes x 2 sector packing.
|
2018-03-22 23:17:10 +03:00
|
|
|
* Calculates the offset of an (x, y) position within a swizzled texture.
|
2018-09-21 06:48:24 +03:00
|
|
|
* Taken from the Tegra X1 Technical Reference Manual. pages 1187-1188
|
2018-03-22 23:17:10 +03:00
|
|
|
*/
|
2018-09-21 18:29:54 +03:00
|
|
|
template <std::size_t N, std::size_t M, u32 Align>
|
|
|
|
struct alignas(64) SwizzleTable {
|
|
|
|
static_assert(M * Align == 64, "Swizzle Table does not align to GOB");
|
|
|
|
constexpr SwizzleTable() {
|
|
|
|
for (u32 y = 0; y < N; ++y) {
|
|
|
|
for (u32 x = 0; x < M; ++x) {
|
|
|
|
const u32 x2 = x * Align;
|
|
|
|
values[y][x] = static_cast<u16>(((x2 % 64) / 32) * 256 + ((y % 8) / 2) * 64 +
|
|
|
|
((x2 % 32) / 16) * 32 + (y % 2) * 16 + (x2 % 16));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
const std::array<u16, M>& operator[](std::size_t index) const {
|
|
|
|
return values[index];
|
|
|
|
}
|
|
|
|
std::array<std::array<u16, M>, N> values{};
|
|
|
|
};
|
2018-03-22 23:17:10 +03:00
|
|
|
|
2019-05-10 10:17:48 +03:00
|
|
|
constexpr u32 gob_size_x_shift = 6;
|
|
|
|
constexpr u32 gob_size_y_shift = 3;
|
|
|
|
constexpr u32 gob_size_z_shift = 0;
|
|
|
|
constexpr u32 gob_size_shift = gob_size_x_shift + gob_size_y_shift + gob_size_z_shift;
|
|
|
|
|
|
|
|
constexpr u32 gob_size_x = 1U << gob_size_x_shift;
|
|
|
|
constexpr u32 gob_size_y = 1U << gob_size_y_shift;
|
|
|
|
constexpr u32 gob_size_z = 1U << gob_size_z_shift;
|
|
|
|
constexpr u32 gob_size = 1U << gob_size_shift;
|
|
|
|
|
2018-11-17 03:55:28 +03:00
|
|
|
constexpr u32 fast_swizzle_align = 16;
|
|
|
|
|
|
|
|
constexpr auto legacy_swizzle_table = SwizzleTable<gob_size_y, gob_size_x, gob_size_z>();
|
|
|
|
constexpr auto fast_swizzle_table = SwizzleTable<gob_size_y, 4, fast_swizzle_align>();
|
2018-03-22 23:17:10 +03:00
|
|
|
|
2018-10-12 00:08:48 +03:00
|
|
|
/**
|
|
|
|
* This function manages ALL the GOBs(Group of Bytes) Inside a single block.
|
|
|
|
* Instead of going gob by gob, we map the coordinates inside a block and manage from
|
|
|
|
* those. Block_Width is assumed to be 1.
|
|
|
|
*/
|
2018-11-15 23:04:17 +03:00
|
|
|
void PreciseProcessBlock(u8* const swizzled_data, u8* const unswizzled_data, const bool unswizzle,
|
2018-10-14 03:42:38 +03:00
|
|
|
const u32 x_start, const u32 y_start, const u32 z_start, const u32 x_end,
|
|
|
|
const u32 y_end, const u32 z_end, const u32 tile_offset,
|
|
|
|
const u32 xy_block_size, const u32 layer_z, const u32 stride_x,
|
|
|
|
const u32 bytes_per_pixel, const u32 out_bytes_per_pixel) {
|
2018-10-11 21:33:50 +03:00
|
|
|
std::array<u8*, 2> data_ptrs;
|
2018-10-12 02:11:47 +03:00
|
|
|
u32 z_address = tile_offset;
|
2018-11-17 03:55:28 +03:00
|
|
|
|
2018-10-11 21:33:50 +03:00
|
|
|
for (u32 z = z_start; z < z_end; z++) {
|
2018-10-12 02:11:47 +03:00
|
|
|
u32 y_address = z_address;
|
2018-10-11 21:33:50 +03:00
|
|
|
u32 pixel_base = layer_z * z + y_start * stride_x;
|
|
|
|
for (u32 y = y_start; y < y_end; y++) {
|
2018-10-12 00:08:48 +03:00
|
|
|
const auto& table = legacy_swizzle_table[y % gob_size_y];
|
2018-10-11 21:33:50 +03:00
|
|
|
for (u32 x = x_start; x < x_end; x++) {
|
2018-10-12 02:11:47 +03:00
|
|
|
const u32 swizzle_offset{y_address + table[x * bytes_per_pixel % gob_size_x]};
|
2018-10-11 21:33:50 +03:00
|
|
|
const u32 pixel_index{x * out_bytes_per_pixel + pixel_base};
|
|
|
|
data_ptrs[unswizzle] = swizzled_data + swizzle_offset;
|
|
|
|
data_ptrs[!unswizzle] = unswizzled_data + pixel_index;
|
|
|
|
std::memcpy(data_ptrs[0], data_ptrs[1], bytes_per_pixel);
|
|
|
|
}
|
|
|
|
pixel_base += stride_x;
|
2018-10-12 00:08:48 +03:00
|
|
|
if ((y + 1) % gob_size_y == 0)
|
2018-10-12 02:11:47 +03:00
|
|
|
y_address += gob_size;
|
2018-10-11 21:33:50 +03:00
|
|
|
}
|
2018-10-12 02:11:47 +03:00
|
|
|
z_address += xy_block_size;
|
2018-10-11 21:33:50 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-10-12 00:08:48 +03:00
|
|
|
/**
|
|
|
|
* This function manages ALL the GOBs(Group of Bytes) Inside a single block.
|
|
|
|
* Instead of going gob by gob, we map the coordinates inside a block and manage from
|
|
|
|
* those. Block_Width is assumed to be 1.
|
|
|
|
*/
|
2018-11-15 23:04:17 +03:00
|
|
|
void FastProcessBlock(u8* const swizzled_data, u8* const unswizzled_data, const bool unswizzle,
|
2018-10-14 03:42:38 +03:00
|
|
|
const u32 x_start, const u32 y_start, const u32 z_start, const u32 x_end,
|
|
|
|
const u32 y_end, const u32 z_end, const u32 tile_offset,
|
|
|
|
const u32 xy_block_size, const u32 layer_z, const u32 stride_x,
|
|
|
|
const u32 bytes_per_pixel, const u32 out_bytes_per_pixel) {
|
2018-10-11 05:03:38 +03:00
|
|
|
std::array<u8*, 2> data_ptrs;
|
2018-10-12 02:11:47 +03:00
|
|
|
u32 z_address = tile_offset;
|
2018-10-11 05:03:38 +03:00
|
|
|
const u32 x_startb = x_start * bytes_per_pixel;
|
|
|
|
const u32 x_endb = x_end * bytes_per_pixel;
|
2018-11-17 03:55:28 +03:00
|
|
|
|
2018-10-11 05:03:38 +03:00
|
|
|
for (u32 z = z_start; z < z_end; z++) {
|
2018-10-12 02:11:47 +03:00
|
|
|
u32 y_address = z_address;
|
2018-10-11 05:03:38 +03:00
|
|
|
u32 pixel_base = layer_z * z + y_start * stride_x;
|
|
|
|
for (u32 y = y_start; y < y_end; y++) {
|
2018-10-12 00:08:48 +03:00
|
|
|
const auto& table = fast_swizzle_table[y % gob_size_y];
|
2018-11-17 03:55:28 +03:00
|
|
|
for (u32 xb = x_startb; xb < x_endb; xb += fast_swizzle_align) {
|
|
|
|
const u32 swizzle_offset{y_address + table[(xb / fast_swizzle_align) % 4]};
|
2018-10-11 05:03:38 +03:00
|
|
|
const u32 out_x = xb * out_bytes_per_pixel / bytes_per_pixel;
|
|
|
|
const u32 pixel_index{out_x + pixel_base};
|
2019-02-24 03:22:54 +03:00
|
|
|
data_ptrs[unswizzle ? 1 : 0] = swizzled_data + swizzle_offset;
|
|
|
|
data_ptrs[unswizzle ? 0 : 1] = unswizzled_data + pixel_index;
|
2018-11-17 03:55:28 +03:00
|
|
|
std::memcpy(data_ptrs[0], data_ptrs[1], fast_swizzle_align);
|
2018-10-11 05:03:38 +03:00
|
|
|
}
|
|
|
|
pixel_base += stride_x;
|
2018-10-12 00:08:48 +03:00
|
|
|
if ((y + 1) % gob_size_y == 0)
|
2018-10-12 02:11:47 +03:00
|
|
|
y_address += gob_size;
|
2018-10-11 05:03:38 +03:00
|
|
|
}
|
2018-10-12 02:11:47 +03:00
|
|
|
z_address += xy_block_size;
|
2018-10-11 05:03:38 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-10-12 00:08:48 +03:00
|
|
|
/**
|
|
|
|
* This function unswizzles or swizzles a texture by mapping Linear to BlockLinear Textue.
|
|
|
|
* The body of this function takes care of splitting the swizzled texture into blocks,
|
|
|
|
* and managing the extents of it. Once all the parameters of a single block are obtained,
|
2018-10-14 03:42:38 +03:00
|
|
|
* the function calls 'ProcessBlock' to process that particular Block.
|
2018-10-12 00:08:48 +03:00
|
|
|
*
|
|
|
|
* Documentation for the memory layout and decoding can be found at:
|
|
|
|
* https://envytools.readthedocs.io/en/latest/hw/memory/g80-surface.html#blocklinear-surfaces
|
|
|
|
*/
|
2018-10-14 03:42:38 +03:00
|
|
|
template <bool fast>
|
2018-11-15 23:04:17 +03:00
|
|
|
void SwizzledData(u8* const swizzled_data, u8* const unswizzled_data, const bool unswizzle,
|
|
|
|
const u32 width, const u32 height, const u32 depth, const u32 bytes_per_pixel,
|
2018-11-16 20:01:54 +03:00
|
|
|
const u32 out_bytes_per_pixel, const u32 block_height, const u32 block_depth,
|
|
|
|
const u32 width_spacing) {
|
2018-10-12 00:08:48 +03:00
|
|
|
auto div_ceil = [](const u32 x, const u32 y) { return ((x + y - 1) / y); };
|
2018-10-11 05:03:38 +03:00
|
|
|
const u32 stride_x = width * out_bytes_per_pixel;
|
|
|
|
const u32 layer_z = height * stride_x;
|
2018-11-17 03:55:28 +03:00
|
|
|
const u32 gob_elements_x = gob_size_x / bytes_per_pixel;
|
|
|
|
constexpr u32 gob_elements_y = gob_size_y;
|
|
|
|
constexpr u32 gob_elements_z = gob_size_z;
|
2018-10-11 05:03:38 +03:00
|
|
|
const u32 block_x_elements = gob_elements_x;
|
|
|
|
const u32 block_y_elements = gob_elements_y * block_height;
|
|
|
|
const u32 block_z_elements = gob_elements_z * block_depth;
|
2018-11-16 20:01:54 +03:00
|
|
|
const u32 aligned_width = Common::AlignUp(width, gob_elements_x * width_spacing);
|
|
|
|
const u32 blocks_on_x = div_ceil(aligned_width, block_x_elements);
|
2018-10-11 05:03:38 +03:00
|
|
|
const u32 blocks_on_y = div_ceil(height, block_y_elements);
|
|
|
|
const u32 blocks_on_z = div_ceil(depth, block_z_elements);
|
|
|
|
const u32 xy_block_size = gob_size * block_height;
|
|
|
|
const u32 block_size = xy_block_size * block_depth;
|
|
|
|
u32 tile_offset = 0;
|
|
|
|
for (u32 zb = 0; zb < blocks_on_z; zb++) {
|
|
|
|
const u32 z_start = zb * block_z_elements;
|
|
|
|
const u32 z_end = std::min(depth, z_start + block_z_elements);
|
|
|
|
for (u32 yb = 0; yb < blocks_on_y; yb++) {
|
|
|
|
const u32 y_start = yb * block_y_elements;
|
|
|
|
const u32 y_end = std::min(height, y_start + block_y_elements);
|
|
|
|
for (u32 xb = 0; xb < blocks_on_x; xb++) {
|
|
|
|
const u32 x_start = xb * block_x_elements;
|
|
|
|
const u32 x_end = std::min(width, x_start + block_x_elements);
|
2019-02-24 04:45:47 +03:00
|
|
|
if constexpr (fast) {
|
2018-10-14 03:42:38 +03:00
|
|
|
FastProcessBlock(swizzled_data, unswizzled_data, unswizzle, x_start, y_start,
|
|
|
|
z_start, x_end, y_end, z_end, tile_offset, xy_block_size,
|
|
|
|
layer_z, stride_x, bytes_per_pixel, out_bytes_per_pixel);
|
|
|
|
} else {
|
|
|
|
PreciseProcessBlock(swizzled_data, unswizzled_data, unswizzle, x_start, y_start,
|
|
|
|
z_start, x_end, y_end, z_end, tile_offset, xy_block_size,
|
|
|
|
layer_z, stride_x, bytes_per_pixel, out_bytes_per_pixel);
|
|
|
|
}
|
2018-10-11 05:03:38 +03:00
|
|
|
tile_offset += block_size;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-10-12 02:11:47 +03:00
|
|
|
void CopySwizzledData(u32 width, u32 height, u32 depth, u32 bytes_per_pixel,
|
2018-11-15 23:04:17 +03:00
|
|
|
u32 out_bytes_per_pixel, u8* const swizzled_data, u8* const unswizzled_data,
|
2018-11-16 20:01:54 +03:00
|
|
|
bool unswizzle, u32 block_height, u32 block_depth, u32 width_spacing) {
|
2019-05-10 10:17:48 +03:00
|
|
|
const u32 block_height_size{1U << block_height};
|
|
|
|
const u32 block_depth_size{1U << block_depth};
|
2018-11-17 03:55:28 +03:00
|
|
|
if (bytes_per_pixel % 3 != 0 && (width * bytes_per_pixel) % fast_swizzle_align == 0) {
|
2018-10-14 03:42:38 +03:00
|
|
|
SwizzledData<true>(swizzled_data, unswizzled_data, unswizzle, width, height, depth,
|
2019-05-10 10:17:48 +03:00
|
|
|
bytes_per_pixel, out_bytes_per_pixel, block_height_size,
|
|
|
|
block_depth_size, width_spacing);
|
2018-09-21 18:42:34 +03:00
|
|
|
} else {
|
2018-10-14 03:42:38 +03:00
|
|
|
SwizzledData<false>(swizzled_data, unswizzled_data, unswizzle, width, height, depth,
|
2019-05-10 10:17:48 +03:00
|
|
|
bytes_per_pixel, out_bytes_per_pixel, block_height_size,
|
|
|
|
block_depth_size, width_spacing);
|
2018-09-21 18:42:34 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-03-22 23:17:10 +03:00
|
|
|
u32 BytesPerPixel(TextureFormat format) {
|
|
|
|
switch (format) {
|
|
|
|
case TextureFormat::DXT1:
|
2018-06-02 21:17:09 +03:00
|
|
|
case TextureFormat::DXN1:
|
2018-03-22 23:17:10 +03:00
|
|
|
// In this case a 'pixel' actually refers to a 4x4 tile.
|
|
|
|
return 8;
|
2018-04-19 04:48:53 +03:00
|
|
|
case TextureFormat::DXT23:
|
|
|
|
case TextureFormat::DXT45:
|
2018-08-09 19:57:13 +03:00
|
|
|
case TextureFormat::DXN2:
|
2018-07-05 01:22:48 +03:00
|
|
|
case TextureFormat::BC7U:
|
2018-08-31 19:11:19 +03:00
|
|
|
case TextureFormat::BC6H_UF16:
|
|
|
|
case TextureFormat::BC6H_SF16:
|
2018-04-19 04:48:53 +03:00
|
|
|
// In this case a 'pixel' actually refers to a 4x4 tile.
|
|
|
|
return 16;
|
2018-08-02 21:56:38 +03:00
|
|
|
case TextureFormat::R32_G32_B32:
|
|
|
|
return 12;
|
2018-06-18 06:50:44 +03:00
|
|
|
case TextureFormat::ASTC_2D_4X4:
|
2018-10-13 17:10:26 +03:00
|
|
|
case TextureFormat::ASTC_2D_5X4:
|
2018-09-16 17:47:02 +03:00
|
|
|
case TextureFormat::ASTC_2D_8X8:
|
2018-10-13 17:10:26 +03:00
|
|
|
case TextureFormat::ASTC_2D_8X5:
|
2018-11-13 05:34:54 +03:00
|
|
|
case TextureFormat::ASTC_2D_10X8:
|
|
|
|
case TextureFormat::ASTC_2D_5X5:
|
2018-03-23 00:40:11 +03:00
|
|
|
case TextureFormat::A8R8G8B8:
|
2018-04-22 01:32:25 +03:00
|
|
|
case TextureFormat::A2B10G10R10:
|
2018-06-06 05:57:16 +03:00
|
|
|
case TextureFormat::BF10GF11RF11:
|
2018-07-24 06:21:31 +03:00
|
|
|
case TextureFormat::R32:
|
2018-07-26 03:01:29 +03:00
|
|
|
case TextureFormat::R16_G16:
|
2018-03-23 00:40:11 +03:00
|
|
|
return 4;
|
2018-05-27 17:02:05 +03:00
|
|
|
case TextureFormat::A1B5G5R5:
|
2018-04-19 02:11:14 +03:00
|
|
|
case TextureFormat::B5G6R5:
|
2018-07-15 08:33:42 +03:00
|
|
|
case TextureFormat::G8R8:
|
2018-07-24 21:39:16 +03:00
|
|
|
case TextureFormat::R16:
|
2018-04-19 02:11:14 +03:00
|
|
|
return 2;
|
2018-05-30 05:49:37 +03:00
|
|
|
case TextureFormat::R8:
|
|
|
|
return 1;
|
2018-05-31 05:24:07 +03:00
|
|
|
case TextureFormat::R16_G16_B16_A16:
|
|
|
|
return 8;
|
2018-06-30 22:23:13 +03:00
|
|
|
case TextureFormat::R32_G32_B32_A32:
|
|
|
|
return 16;
|
2018-07-24 02:10:00 +03:00
|
|
|
case TextureFormat::R32_G32:
|
|
|
|
return 8;
|
2018-03-22 23:17:10 +03:00
|
|
|
default:
|
|
|
|
UNIMPLEMENTED_MSG("Format not implemented");
|
2018-12-19 04:52:32 +03:00
|
|
|
return 1;
|
2018-03-22 23:17:10 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-02-24 08:15:35 +03:00
|
|
|
void UnswizzleTexture(u8* const unswizzled_data, u8* address, u32 tile_size_x, u32 tile_size_y,
|
2018-11-19 06:53:03 +03:00
|
|
|
u32 bytes_per_pixel, u32 width, u32 height, u32 depth, u32 block_height,
|
2018-11-16 20:01:54 +03:00
|
|
|
u32 block_depth, u32 width_spacing) {
|
2018-11-19 06:53:03 +03:00
|
|
|
CopySwizzledData((width + tile_size_x - 1) / tile_size_x,
|
|
|
|
(height + tile_size_y - 1) / tile_size_y, depth, bytes_per_pixel,
|
2019-02-24 08:15:35 +03:00
|
|
|
bytes_per_pixel, address, unswizzled_data, true, block_height, block_depth,
|
|
|
|
width_spacing);
|
2018-11-19 06:53:03 +03:00
|
|
|
}
|
|
|
|
|
2019-02-24 08:15:35 +03:00
|
|
|
std::vector<u8> UnswizzleTexture(u8* address, u32 tile_size_x, u32 tile_size_y, u32 bytes_per_pixel,
|
|
|
|
u32 width, u32 height, u32 depth, u32 block_height,
|
|
|
|
u32 block_depth, u32 width_spacing) {
|
2018-10-13 23:11:11 +03:00
|
|
|
std::vector<u8> unswizzled_data(width * height * depth * bytes_per_pixel);
|
2018-11-19 06:53:03 +03:00
|
|
|
UnswizzleTexture(unswizzled_data.data(), address, tile_size_x, tile_size_y, bytes_per_pixel,
|
2018-11-16 20:01:54 +03:00
|
|
|
width, height, depth, block_height, block_depth, width_spacing);
|
2018-03-22 23:17:10 +03:00
|
|
|
return unswizzled_data;
|
|
|
|
}
|
|
|
|
|
2018-10-18 04:26:40 +03:00
|
|
|
void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width,
|
2019-06-13 23:41:16 +03:00
|
|
|
u32 bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, u32 block_height_bit) {
|
|
|
|
const u32 block_height = 1U << block_height_bit;
|
2018-11-17 03:55:28 +03:00
|
|
|
const u32 image_width_in_gobs{(swizzled_width * bytes_per_pixel + (gob_size_x - 1)) /
|
|
|
|
gob_size_x};
|
2018-10-18 04:26:40 +03:00
|
|
|
for (u32 line = 0; line < subrect_height; ++line) {
|
|
|
|
const u32 gob_address_y =
|
2019-06-13 23:41:16 +03:00
|
|
|
(line / (gob_size_y * block_height)) * gob_size * block_height * image_width_in_gobs +
|
|
|
|
((line % (gob_size_y * block_height)) / gob_size_y) * gob_size;
|
2018-11-17 03:55:28 +03:00
|
|
|
const auto& table = legacy_swizzle_table[line % gob_size_y];
|
2018-10-18 04:26:40 +03:00
|
|
|
for (u32 x = 0; x < subrect_width; ++x) {
|
2018-11-17 03:55:28 +03:00
|
|
|
const u32 gob_address =
|
2019-06-13 23:41:16 +03:00
|
|
|
gob_address_y + (x * bytes_per_pixel / gob_size_x) * gob_size * block_height;
|
2018-11-17 03:55:28 +03:00
|
|
|
const u32 swizzled_offset = gob_address + table[(x * bytes_per_pixel) % gob_size_x];
|
2019-02-24 08:15:35 +03:00
|
|
|
u8* source_line = unswizzled_data + line * source_pitch + x * bytes_per_pixel;
|
|
|
|
u8* dest_addr = swizzled_data + swizzled_offset;
|
2018-10-18 04:26:40 +03:00
|
|
|
|
2019-02-24 08:15:35 +03:00
|
|
|
std::memcpy(dest_addr, source_line, bytes_per_pixel);
|
2018-10-18 04:26:40 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void UnswizzleSubrect(u32 subrect_width, u32 subrect_height, u32 dest_pitch, u32 swizzled_width,
|
2019-06-13 23:41:16 +03:00
|
|
|
u32 bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, u32 block_height_bit,
|
2019-02-24 08:15:35 +03:00
|
|
|
u32 offset_x, u32 offset_y) {
|
2019-06-13 23:41:16 +03:00
|
|
|
const u32 block_height = 1U << block_height_bit;
|
2018-10-18 04:26:40 +03:00
|
|
|
for (u32 line = 0; line < subrect_height; ++line) {
|
|
|
|
const u32 y2 = line + offset_y;
|
2019-06-13 23:41:16 +03:00
|
|
|
const u32 gob_address_y = (y2 / (gob_size_y * block_height)) * gob_size * block_height +
|
|
|
|
((y2 % (gob_size_y * block_height)) / gob_size_y) * gob_size;
|
2018-11-17 03:55:28 +03:00
|
|
|
const auto& table = legacy_swizzle_table[y2 % gob_size_y];
|
2018-10-18 04:26:40 +03:00
|
|
|
for (u32 x = 0; x < subrect_width; ++x) {
|
|
|
|
const u32 x2 = (x + offset_x) * bytes_per_pixel;
|
2019-06-13 23:41:16 +03:00
|
|
|
const u32 gob_address = gob_address_y + (x2 / gob_size_x) * gob_size * block_height;
|
2018-11-17 03:55:28 +03:00
|
|
|
const u32 swizzled_offset = gob_address + table[x2 % gob_size_x];
|
2019-02-24 08:15:35 +03:00
|
|
|
u8* dest_line = unswizzled_data + line * dest_pitch + x * bytes_per_pixel;
|
|
|
|
u8* source_addr = swizzled_data + swizzled_offset;
|
2018-10-18 04:26:40 +03:00
|
|
|
|
2019-02-24 08:15:35 +03:00
|
|
|
std::memcpy(dest_line, source_addr, bytes_per_pixel);
|
2018-10-18 04:26:40 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-04-16 19:00:46 +03:00
|
|
|
void SwizzleKepler(const u32 width, const u32 height, const u32 dst_x, const u32 dst_y,
|
2019-06-13 23:41:16 +03:00
|
|
|
const u32 block_height_bit, const std::size_t copy_size, const u8* source_data,
|
2019-04-16 19:00:46 +03:00
|
|
|
u8* swizzle_data) {
|
2019-06-13 23:41:16 +03:00
|
|
|
const u32 block_height = 1U << block_height_bit;
|
2019-04-16 04:06:04 +03:00
|
|
|
const u32 image_width_in_gobs{(width + gob_size_x - 1) / gob_size_x};
|
|
|
|
std::size_t count = 0;
|
2019-04-16 19:00:46 +03:00
|
|
|
for (std::size_t y = dst_y; y < height && count < copy_size; ++y) {
|
|
|
|
const std::size_t gob_address_y =
|
2019-06-13 23:41:16 +03:00
|
|
|
(y / (gob_size_y * block_height)) * gob_size * block_height * image_width_in_gobs +
|
|
|
|
((y % (gob_size_y * block_height)) / gob_size_y) * gob_size;
|
2019-04-16 04:06:04 +03:00
|
|
|
const auto& table = legacy_swizzle_table[y % gob_size_y];
|
2019-04-16 19:00:46 +03:00
|
|
|
for (std::size_t x = dst_x; x < width && count < copy_size; ++x) {
|
|
|
|
const std::size_t gob_address =
|
2019-06-13 23:41:16 +03:00
|
|
|
gob_address_y + (x / gob_size_x) * gob_size * block_height;
|
2019-04-16 19:00:46 +03:00
|
|
|
const std::size_t swizzled_offset = gob_address + table[x % gob_size_x];
|
2019-04-16 04:06:04 +03:00
|
|
|
const u8* source_line = source_data + count;
|
|
|
|
u8* dest_addr = swizzle_data + swizzled_offset;
|
|
|
|
count++;
|
|
|
|
|
|
|
|
std::memcpy(dest_addr, source_line, 1);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-03-22 23:17:10 +03:00
|
|
|
std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat format, u32 width,
|
|
|
|
u32 height) {
|
|
|
|
std::vector<u8> rgba_data;
|
|
|
|
|
|
|
|
// TODO(Subv): Implement.
|
|
|
|
switch (format) {
|
2018-03-23 00:40:11 +03:00
|
|
|
case TextureFormat::DXT1:
|
2018-04-19 04:48:53 +03:00
|
|
|
case TextureFormat::DXT23:
|
|
|
|
case TextureFormat::DXT45:
|
2018-06-02 21:17:09 +03:00
|
|
|
case TextureFormat::DXN1:
|
2018-08-09 19:57:13 +03:00
|
|
|
case TextureFormat::DXN2:
|
2018-07-05 01:22:48 +03:00
|
|
|
case TextureFormat::BC7U:
|
2018-08-31 19:11:19 +03:00
|
|
|
case TextureFormat::BC6H_UF16:
|
|
|
|
case TextureFormat::BC6H_SF16:
|
2018-06-18 06:50:44 +03:00
|
|
|
case TextureFormat::ASTC_2D_4X4:
|
2018-09-16 17:47:02 +03:00
|
|
|
case TextureFormat::ASTC_2D_8X8:
|
2018-11-13 05:34:54 +03:00
|
|
|
case TextureFormat::ASTC_2D_5X5:
|
|
|
|
case TextureFormat::ASTC_2D_10X8:
|
2018-03-23 00:40:11 +03:00
|
|
|
case TextureFormat::A8R8G8B8:
|
2018-04-22 01:32:25 +03:00
|
|
|
case TextureFormat::A2B10G10R10:
|
2018-05-27 17:02:05 +03:00
|
|
|
case TextureFormat::A1B5G5R5:
|
2018-04-19 02:11:14 +03:00
|
|
|
case TextureFormat::B5G6R5:
|
2018-05-30 05:49:37 +03:00
|
|
|
case TextureFormat::R8:
|
2018-07-15 08:33:42 +03:00
|
|
|
case TextureFormat::G8R8:
|
2018-06-06 05:57:16 +03:00
|
|
|
case TextureFormat::BF10GF11RF11:
|
2018-06-30 22:23:13 +03:00
|
|
|
case TextureFormat::R32_G32_B32_A32:
|
2018-07-24 02:10:00 +03:00
|
|
|
case TextureFormat::R32_G32:
|
2018-07-24 06:21:31 +03:00
|
|
|
case TextureFormat::R32:
|
2018-07-24 21:39:16 +03:00
|
|
|
case TextureFormat::R16:
|
2018-07-26 03:01:29 +03:00
|
|
|
case TextureFormat::R16_G16:
|
2018-08-02 21:56:38 +03:00
|
|
|
case TextureFormat::R32_G32_B32:
|
2018-03-23 00:40:11 +03:00
|
|
|
// TODO(Subv): For the time being just forward the same data without any decoding.
|
|
|
|
rgba_data = texture_data;
|
|
|
|
break;
|
2018-03-22 23:17:10 +03:00
|
|
|
default:
|
|
|
|
UNIMPLEMENTED_MSG("Format not implemented");
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
return rgba_data;
|
|
|
|
}
|
|
|
|
|
2018-10-12 21:21:53 +03:00
|
|
|
std::size_t CalculateSize(bool tiled, u32 bytes_per_pixel, u32 width, u32 height, u32 depth,
|
|
|
|
u32 block_height, u32 block_depth) {
|
|
|
|
if (tiled) {
|
2019-05-10 10:17:48 +03:00
|
|
|
const u32 aligned_width = Common::AlignBits(width * bytes_per_pixel, gob_size_x_shift);
|
|
|
|
const u32 aligned_height = Common::AlignBits(height, gob_size_y_shift + block_height);
|
|
|
|
const u32 aligned_depth = Common::AlignBits(depth, gob_size_z_shift + block_depth);
|
2019-04-14 07:44:16 +03:00
|
|
|
return aligned_width * aligned_height * aligned_depth;
|
2018-10-12 21:21:53 +03:00
|
|
|
} else {
|
|
|
|
return width * height * depth * bytes_per_pixel;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-07-21 01:14:17 +03:00
|
|
|
} // namespace Tegra::Texture
|