d119505289
Integer num_format fetches now get their scale CPU-side instead of trying to guess in the shader. This is authoritative; no cvar. Both translators now use texture_integer_scale_bits after signs/gamma and before exponent bias. This alone fixes black screens and a bunch of rendering bugs across at least several dozen titles. This new information lives in the updated FormatInfo table, including fixed component widths. Resolve also has two new fixes. 8_8_8_8_GAMMA EDRAM sources can now decode through the PWL curve while still in linear space, before MSAA resolve or format conversion, then re-encoded for gamma destinations. This is also now default enabled via gamma_decode_pwl_resolve and has improved blowout in at least 4 titles, with no obvious regressions thus far. Full resolve can also now pack fixed destinations according to copy_dest_number.
65 lines
2.6 KiB
Plaintext
65 lines
2.6 KiB
Plaintext
/**
|
|
******************************************************************************
|
|
* Xenia : Xbox 360 Emulator Research Project *
|
|
******************************************************************************
|
|
* Copyright 2022 Ben Vanik. All rights reserved. *
|
|
* Released under the BSD license - see LICENSE in the root for more details. *
|
|
******************************************************************************
|
|
*/
|
|
|
|
#include "endian.xesli"
|
|
#include "pixel_formats.xesli"
|
|
#define XE_RESOLVE_COPY_EDRAM_BYTE_BUFFER_ALIGNMENT 4
|
|
#include "resolve.xesli"
|
|
|
|
byte_buffer_align16_wo_declare_xe(xe_resolve_dest, set=1, binding=0, u0, space0)
|
|
#define LOCAL_SIZE_X_XE 8
|
|
#define LOCAL_SIZE_Y_XE 8
|
|
#define LOCAL_SIZE_Z_XE 1
|
|
entry_bindings_begin_compute_xe
|
|
XE_RESOLVE_PUSH_CONST_BINDING
|
|
entry_binding_next_xe
|
|
byte_buffer_wo_binding_xe(xe_resolve_dest, buffer(1))
|
|
entry_binding_next_xe
|
|
XE_RESOLVE_COPY_EDRAM_BINDING
|
|
entry_bindings_end_inputs_begin_compute_xe
|
|
entry_in_global_thread_id_xe
|
|
entry_inputs_end_code_begin_compute_xe
|
|
{
|
|
// 1 thread = 4 host pixels.
|
|
// 1 pixel per thread is 160% as slow on Nvidia Pascal, 8 pixels per thread is
|
|
// 115% as slow.
|
|
XeResolveInfo resolve_info = XeResolveGetInfo(pass_push_consts_xe);
|
|
uint2_xe pixel_index = in_global_thread_id_xe.xy << uint2_xe(2u, 0u);
|
|
// Group height can't cross resolve granularity, Y overflow check not needed.
|
|
dont_flatten_xe
|
|
if (pixel_index.x >= resolve_info.width_div_8_scaled << 3u) {
|
|
return;
|
|
}
|
|
float4_xe pixel_0, pixel_1, pixel_2, pixel_3;
|
|
XeResolveLoad4RGBAColors(
|
|
pass_byte_buffer_xe(xe_resolve_edram) pass_next_after_byte_buffer_xe
|
|
resolve_info,
|
|
XeResolveColorCopySourcePixelAddressBytesYHalfPixelOffsetFilling(
|
|
resolve_info, pixel_index),
|
|
pixel_0, pixel_1, pixel_2, pixel_3);
|
|
uint4_xe packed = XePack32bpp4Pixels(pixel_0, pixel_1, pixel_2, pixel_3,
|
|
resolve_info.dest_format,
|
|
resolve_info.dest_num_format);
|
|
dont_flatten_xe
|
|
if (pixel_index.x == 0u &&
|
|
resolve_info.half_pixel_offset_fill_source.x != 0u) {
|
|
if (resolve_info.half_pixel_offset_fill_source.x >= 2u) {
|
|
if (resolve_info.half_pixel_offset_fill_source.x >= 3u) {
|
|
packed.z = packed.w;
|
|
}
|
|
packed.y = packed.z;
|
|
}
|
|
packed.x = packed.y;
|
|
}
|
|
byte_buffer_align16_store16_xe(
|
|
xe_resolve_dest, XeResolveDestPixelAddress(resolve_info, pixel_index, 2u),
|
|
XeEndianSwap32(packed, resolve_info.dest_endian_128));
|
|
}
|
|
entry_code_end_compute_xe
|