fabi/Xenia-Canary
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[D3D12] Rename XeTextureCopy to XeTextureLoad in shaders

Browse Source
This commit is contained in:
Triang3l
2018-10-05 20:02:36 +03:00
parent 9dc7dfab02
commit 34946c57b0
58 changed files with 730 additions and 730 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
src/xenia/gpu/d3d12/shaders/texture_load_dxn_rg8.cs.hlsl
View File

@@ -1,24 +1,24 @@
#include "pixel_formats.hlsli"
#include "texture_copy.hlsli"
#include "texture_load.hlsli"
[numthreads(8, 32, 1)]
void main(uint3 xe_thread_id : SV_DispatchThreadID) {
// 1 thread = 4 DXN (16bpb) blocks to 16x4 R8G8 texels.
uint3 block_index = xe_thread_id;
block_index.x <<= 2u;
[branch] if (any(block_index >= xe_texture_copy_size_blocks)) {
[branch] if (any(block_index >= xe_texture_load_size_blocks)) {
return;
}
uint4 block_offsets_guest =
XeTextureCopyGuestBlockOffsets(block_index, 16u, 4u);
uint4 block_0 = xe_texture_copy_source.Load4(block_offsets_guest.x);
uint4 block_1 = xe_texture_copy_source.Load4(block_offsets_guest.y);
uint4 block_2 = xe_texture_copy_source.Load4(block_offsets_guest.z);
uint4 block_3 = xe_texture_copy_source.Load4(block_offsets_guest.w);
block_0 = XeByteSwap(block_0, xe_texture_copy_endianness);
block_1 = XeByteSwap(block_1, xe_texture_copy_endianness);
block_2 = XeByteSwap(block_2, xe_texture_copy_endianness);
block_3 = XeByteSwap(block_3, xe_texture_copy_endianness);
XeTextureLoadGuestBlockOffsets(block_index, 16u, 4u);
uint4 block_0 = xe_texture_load_source.Load4(block_offsets_guest.x);
uint4 block_1 = xe_texture_load_source.Load4(block_offsets_guest.y);
uint4 block_2 = xe_texture_load_source.Load4(block_offsets_guest.z);
uint4 block_3 = xe_texture_load_source.Load4(block_offsets_guest.w);
block_0 = XeByteSwap(block_0, xe_texture_load_endianness);
block_1 = XeByteSwap(block_1, xe_texture_load_endianness);
block_2 = XeByteSwap(block_2, xe_texture_load_endianness);
block_3 = XeByteSwap(block_3, xe_texture_load_endianness);
uint4 r_blocks_0 = uint4(block_0.x, block_1.x, block_2.x, block_3.x);
uint4 r_blocks_1 = uint4(block_0.y, block_1.y, block_2.y, block_3.y);
uint4 g_blocks_0 = uint4(block_0.z, block_1.z, block_2.z, block_3.z);
@@ -47,8 +47,8 @@ void main(uint3 xe_thread_id : SV_DispatchThreadID) {
// Uncompress and write the rows.
uint3 texel_index_host = block_index << uint3(2u, 2u, 0u);
uint texel_offset_host = XeTextureHostLinearOffset(
texel_index_host, xe_texture_copy_size_texels.y,
xe_texture_copy_host_pitch, 2u) + xe_texture_copy_host_base;
texel_index_host, xe_texture_load_size_texels.y,
xe_texture_load_host_pitch, 2u) + xe_texture_load_host_base;
for (uint i = 0u; i < 4u; ++i) {
uint4 r_row = XeDXT5Four8StepBlocksRowToA8(
r_end_low, r_end_high,
@@ -67,16 +67,16 @@ void main(uint3 xe_thread_id : SV_DispatchThreadID) {
(((r_row.xxyy >> uint4(8u, 24u, 8u, 24u)) & 0xFFu) << 16u) |
(((g_row.xxyy >> uint4(0u, 16u, 0u, 16u)) & 0xFFu) << 8u) |
(((g_row.xxyy >> uint4(8u, 24u, 8u, 24u)) & 0xFFu) << 24u);
xe_texture_copy_dest.Store4(texel_offset_host, rg_row_half);
xe_texture_load_dest.Store4(texel_offset_host, rg_row_half);
rg_row_half =
((r_row.zzww >> uint4(0u, 16u, 0u, 16u)) & 0xFFu) |
(((r_row.zzww >> uint4(8u, 24u, 8u, 24u)) & 0xFFu) << 16u) |
(((g_row.zzww >> uint4(0u, 16u, 0u, 16u)) & 0xFFu) << 8u) |
(((g_row.zzww >> uint4(8u, 24u, 8u, 24u)) & 0xFFu) << 24u);
xe_texture_copy_dest.Store4(texel_offset_host + 16u, rg_row_half);
if (++texel_index_host.y >= xe_texture_copy_size_texels.y) {
xe_texture_load_dest.Store4(texel_offset_host + 16u, rg_row_half);
if (++texel_index_host.y >= xe_texture_load_size_texels.y) {
return;
}
texel_offset_host += xe_texture_copy_host_pitch;
texel_offset_host += xe_texture_load_host_pitch;
}
}