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

[GPU/WGF] Use ByteAddressBuffer in precompiled shaders

Browse Source 
Drivers and hardware may provide more optimal paths for untyped buffers
compared to typed.

Also, ByteAddressBuffer may be bound via a root descriptor in Direct3D 12,
greatly simplifying the binding logic.

ByteAddressBuffer also doesn't have the 128 * 2^20 element count limit that
typed and structured buffers have, and doesn't require the structure stride
to be specified at resource creation time in Direct3D 11.
This commit is contained in:
Triang3l
2026-02-09 21:29:18 +03:00
parent c7f61342d7
commit 80dad41454
211 changed files with 121690 additions and 123176 deletions
Download Patch File Download Diff File Expand all files Collapse all files

40
src/xenia/gpu/shaders/texture_load_128bpb.xesli
View File

@@ -9,19 +9,7 @@
#include "texture_load.xesli"
array_buffer_wo_declare_xe(uint4_xe, xe_texture_load_dest, set=0, binding=0, u0,
space0)
array_buffer_declare_xe(uint4_xe, xe_texture_load_source, set=1, binding=0, t0,
space0)
entry_bindings_begin_compute_xe
XE_TEXTURE_LOAD_PUSH_CONST_BINDING
entry_binding_next_xe
array_buffer_wo_binding_xe(uint4_xe, xe_texture_load_dest, buffer(1))
entry_binding_next_xe
array_buffer_binding_xe(uint4_xe, xe_texture_load_source, buffer(2))
entry_bindings_end_inputs_begin_compute_xe
entry_in_global_thread_id_xe
entry_inputs_end_code_begin_compute_xe
XE_TEXTURE_LOAD_ENTRY(16)
{
// 1 thread = 2 blocks.
XeTextureLoadInfo load_info = XeTextureLoadGetInfo(pass_push_consts_xe);
@@ -30,24 +18,24 @@ entry_inputs_end_code_begin_compute_xe
if (any(greater_than_equal_xe(block_index.xy, load_info.size_blocks.xy))) {
return;
}
uint block_offset_host = uint(
(XeTextureHostLinearOffset(int3_xe(block_index), load_info.host_pitch,
load_info.size_blocks.y, 16u) +
load_info.host_offset) >> 4u);
uint block_offset_host =
load_info.host_offset +
uint(XeTextureHostLinearOffset(int3_xe(block_index), load_info.host_pitch,
load_info.size_blocks.y, 16u));
uint block_offset_guest =
XeTextureLoadSourceAddress(load_info, block_index, 4u) >> 4u;
array_buffer_store_xe(
XeTextureLoadSourceAddress(load_info, block_index, 4u);
byte_buffer_align16_store16_xe(
xe_texture_load_dest, block_offset_host,
XeEndianSwap32(array_buffer_load_xe(xe_texture_load_source,
block_offset_guest),
XeEndianSwap32(byte_buffer_align16_load16_xe(xe_texture_load_source,
block_offset_guest),
load_info.endian_32));
++block_offset_host;
block_offset_host += 16u;
block_offset_guest +=
XeTextureLoadLocalXAddressXor(1u, 4u, load_info.is_tiled) >> 4u;
array_buffer_store_xe(
XeTextureLoadLocalXAddressXor(1u, 4u, load_info.is_tiled);
byte_buffer_align16_store16_xe(
xe_texture_load_dest, block_offset_host,
XeEndianSwap32(array_buffer_load_xe(xe_texture_load_source,
block_offset_guest),
XeEndianSwap32(byte_buffer_align16_load16_xe(xe_texture_load_source,
block_offset_guest),
load_info.endian_32));
}
entry_code_end_compute_xe