debuggerx/eden
1
0
Fork 0
mirror of https://git.eden-emu.dev/eden-emu/eden.git synced 2026-09-01 18:53:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

[vulkan] Re-adjusting ASTC resolve (#4250)

Browse Source 
This is an adjustment made due to problems caused by my ASTC changes some pr's earlier; added sanity to decoded blocks, removes the inconditional storage view for ASTC formats (verified against MaxwellToVk logic), added support to VK_EXT_astc_decode_mode which will ensure a better way to resolve RGFA16/RGFA32 on tilers, simplified ASTC HDR decoding into LDR valid formats.

This fixes some issues with SteamDeck, Fire Emblem: Engage washed colors and other problems related to ASTC.

_Special Thanks_

1.- Meowly The Smol (@Gidoly)
2.- @simply0001 for the actual idea on the redundant storage views on the transcoded ASTC path.

Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4250
This commit is contained in:
CamilleLaVey
2026-08-05 09:24:35 +02:00
committed by crueter
parent ba9130fbf9
commit 5e1d5e82dc
6 changed files with 57 additions and 56 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/video_core/host_shaders/astc_decoder.comp
+16 -16
View File
@@ -77,14 +77,8 @@ uvec4 local_buff;
uvec4 color_endpoint_data;
int color_bitsread = 0;
// Global "vector" to be pushed into when decoding
// At most will require BLOCK_WIDTH x BLOCK_HEIGHT in single plane mode
// At most will require BLOCK_WIDTH x BLOCK_HEIGHT x 2 in dual plane mode
// So the maximum would be 144 (12 x 12) elements, x 2 for two planes
#define DIVCEIL(number, divisor) (number + divisor - 1) / divisor
#define ARRAY_NUM_ELEMENTS 144
#define VECTOR_ARRAY_SIZE DIVCEIL(ARRAY_NUM_ELEMENTS * 2, 4)
uint result_vector[ARRAY_NUM_ELEMENTS * 2];
#define MAX_WEIGHT_VALUES 64
uint result_vector[MAX_WEIGHT_VALUES];
int result_index = 0;
uint result_vector_max_index;
@@ -492,7 +486,7 @@ void DecodeColorValues(uvec4 modes, uint num_partitions, uint color_data_bits, o
A = ReplicateBitTo9((bitval & 1));
switch (encoding) {
case JUST_BITS:
color_values[++out_index] = FastReplicateTo8(bitval, bitlen);
color_values[out_index++] = FastReplicateTo8(bitval, bitlen);
break;
case TRIT: {
D = QuintTritValue(val);
@@ -571,7 +565,7 @@ void DecodeColorValues(uvec4 modes, uint num_partitions, uint color_data_bits, o
uint T = (D * C) + B;
T ^= A;
T = (A & 0x80) | (T >> 2);
color_values[++out_index] = T;
color_values[out_index++] = T;
}
}
}
@@ -753,12 +747,12 @@ void ComputeEndpoints(out uvec4 ep1, out uvec4 ep2, uint color_endpoint_mode, ui
#define READ_UINT_VALUES(N) \
uvec4 V[2]; \
for (uint i = 0; i < N; i++) { \
V[i / 4][i % 4] = color_values[++colvals_index]; \
V[i / 4][i % 4] = color_values[colvals_index++]; \
}
#define READ_INT_VALUES(N) \
ivec4 V[2]; \
for (uint i = 0; i < N; i++) { \
V[i / 4][i % 4] = int(color_values[++colvals_index]); \
V[i / 4][i % 4] = int(color_values[colvals_index++]); \
}
switch (color_endpoint_mode) {
@@ -1225,6 +1219,10 @@ void DecompressBlock(ivec3 coord) {
FillError(coord);
return;
}
if (GetNumWeightValues(size_params, dual_plane) > MAX_WEIGHT_VALUES) {
FillError(coord);
return;
}
uint partition_index = 1;
uvec4 color_endpoint_mode = uvec4(0);
uint ced_pointer = 0;
@@ -1239,6 +1237,10 @@ void DecompressBlock(ivec3 coord) {
const uint base_mode = base_cem & 3;
const uint max_weight = DecodeMaxWeight(mode);
const uint weight_bits = GetPackedBitSize(size_params, dual_plane, max_weight);
if (weight_bits < 24 || weight_bits > 96) {
FillError(coord);
return;
}
uint remaining_bits = 128 - weight_bits - total_bitsread;
uint extra_cem_bits = 0;
if (base_mode > 0) {
@@ -1253,6 +1255,7 @@ void DecompressBlock(ivec3 coord) {
extra_cem_bits += 8;
break;
default:
FillError(coord);
return;
}
}
@@ -1262,6 +1265,7 @@ void DecompressBlock(ivec3 coord) {
if (remaining_bits > 128) {
// Bad data, more remaining bits than 4 bytes
// return early
FillError(coord);
return;
}
// Read color data...
@@ -1384,11 +1388,7 @@ void DecompressBlock(ivec3 coord) {
p = Cf / 65535.0f;
}
#ifdef VULKAN
imageStore(dest_image, coord + ivec3(i, j, 0), p.gbar);
#else
imageStore(dest_image, coord + ivec3(i, j, 0), clamp(p, 0.0f, 1.0f).gbar);
#endif
}
}
}