Plexi09/eden-miror
1
0
Fork 0
mirror of https://git.eden-emu.dev/eden-emu/eden synced 2026-05-17 08:07:01 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 4

linearize GetBitLength

Browse source
This commit is contained in:
lizzie 2026-01-29 22:04:39 +00:00 • committed by crueter
parent 5a10e95188
commit d7e3fd7fe8
1 changed files with 49 additions and 68 deletions
Download patch file Download diff file Expand all files Collapse all files

117
src/video_core/host_shaders/astc_decoder.comp
View file

@ -48,6 +48,7 @@ const uint GOB_SIZE_SHIFT = GOB_SIZE_X_SHIFT + GOB_SIZE_Y_SHIFT;
const uint BYTES_PER_BLOCK_LOG2 = 4; const uint BYTES_PER_BLOCK_LOG2 = 4;
// DO NOT CHANGE - code depends on the value of these!
const uint JUST_BITS = 0u; const uint JUST_BITS = 0u;
const uint QUINT = 1u; const uint QUINT = 1u;
const uint TRIT = 2u; const uint TRIT = 2u;
@ -55,12 +56,29 @@ const uint TRIT = 2u;
// ASTC Encodings data, sorted in ascending order based on their BitLength value // ASTC Encodings data, sorted in ascending order based on their BitLength value
// (see GetBitLength() function) // (see GetBitLength() function)
const uint encoding_values[22] = uint[]( const uint encoding_values[22] = uint[](
(JUST_BITS), (JUST_BITS | (1u << 8u)), (TRIT), (JUST_BITS | (2u << 8u)), (JUST_BITS),
(QUINT), (TRIT | (1u << 8u)), (JUST_BITS | (3u << 8u)), (QUINT | (1u << 8u)), (JUST_BITS | (1u << 8u)),
(TRIT | (2u << 8u)), (JUST_BITS | (4u << 8u)), (QUINT | (2u << 8u)), (TRIT | (3u << 8u)), (TRIT),
(JUST_BITS | (5u << 8u)), (QUINT | (3u << 8u)), (TRIT | (4u << 8u)), (JUST_BITS | (6u << 8u)), (JUST_BITS | (2u << 8u)),
(QUINT | (4u << 8u)), (TRIT | (5u << 8u)), (JUST_BITS | (7u << 8u)), (QUINT | (5u << 8u)), (QUINT),
(TRIT | (6u << 8u)), (JUST_BITS | (8u << 8u))); (TRIT | (1u << 8u)),
(JUST_BITS | (3u << 8u)),
(QUINT | (1u << 8u)),
(TRIT | (2u << 8u)),
(JUST_BITS | (4u << 8u)),
(QUINT | (2u << 8u)),
(TRIT | (3u << 8u)),
(JUST_BITS | (5u << 8u)),
(QUINT | (3u << 8u)),
(TRIT | (4u << 8u)),
(JUST_BITS | (6u << 8u)),
(QUINT | (4u << 8u)),
(TRIT | (5u << 8u)),
(JUST_BITS | (7u << 8u)),
(QUINT | (5u << 8u)),
(TRIT | (6u << 8u)),
(JUST_BITS | (8u << 8u))
);
// Input ASTC texture globals // Input ASTC texture globals
int total_bitsread = 0; int total_bitsread = 0;
@ -157,22 +175,6 @@ uint FastReplicateTo6(uint value, uint num_bits) {
return ReplicateBits(value, num_bits, 6); return ReplicateBits(value, num_bits, 6);
} }
uint Div3Floor(uint v) {
return (v * 0x5556) >> 16;
}
uint Div3Ceil(uint v) {
return Div3Floor(v + 2);
}
uint Div5Floor(uint v) {
return (v * 0x3334) >> 16;
}
uint Div5Ceil(uint v) {
return Div5Floor(v + 4);
}
uint Hash52(uint p) { uint Hash52(uint p) {
p ^= p >> 15; p ^= p >> 15;
p -= p << 17; p -= p << 17;
@ -260,15 +262,16 @@ EncodingData GetEncodingFromVector(uint index) {
// Returns the number of bits required to encode n_vals values. // Returns the number of bits required to encode n_vals values.
uint GetBitLength(uint n_vals, uint encoding_index) { uint GetBitLength(uint n_vals, uint encoding_index) {
// uint Div3Floor(uint v) { return (v * 0x5556) >> 16; }
// uint Div3Ceil(uint v) { return Div3Floor(v + 2); }
// uint Div5Floor(uint v) { return (v * 0x3334) >> 16; }
// uint Div5Ceil(uint v) { return Div5Floor(v + 4); }
const EncodingData encoding_value = EncodingData(encoding_values[encoding_index]); const EncodingData encoding_value = EncodingData(encoding_values[encoding_index]);
const uint encoding = Encoding(encoding_value); const uint encoding = Encoding(encoding_value);
uint total_bits = NumBits(encoding_value) * n_vals; const uvec3 div_constant = uvec3(0, 0x5556, 0x3334);
if (encoding == TRIT) { return NumBits(encoding_value) * n_vals
total_bits += Div5Ceil(n_vals * 8); + ((((n_vals * ((0x870 >> (encoding * 4)) & 0xf)) + ((0x420 >> (encoding * 4)) & 0xf))
} else if (encoding == QUINT) { * div_constant[encoding]) >> 16);
total_bits += Div3Ceil(n_vals * 7);
}
return total_bits;
} }
uint GetNumWeightValues(uvec2 size, bool dual_plane) { uint GetNumWeightValues(uvec2 size, bool dual_plane) {
@ -423,11 +426,10 @@ void DecodeColorValues(uvec4 modes, uint num_partitions, uint color_data_bits, o
const uint encoding = Encoding(val); const uint encoding = Encoding(val);
const uint bitlen = NumBits(val); const uint bitlen = NumBits(val);
const uint bitval = BitValue(val); const uint bitval = BitValue(val);
uint A = 0, B = 0, C = 0, D = 0; uint B = 0, C = 0, D = 0;
A = ReplicateBitTo9((bitval & 1)); uint A = ReplicateBitTo9((bitval & 1));
switch (encoding) { switch (encoding) {
case JUST_BITS: case JUST_BITS:
color_values[++out_index] = FastReplicateTo8(bitval, bitlen);
break; break;
case TRIT: { case TRIT: {
D = QuintTritValue(val); D = QuintTritValue(val);
@ -502,12 +504,12 @@ void DecodeColorValues(uvec4 modes, uint num_partitions, uint color_data_bits, o
break; break;
} }
} }
if (encoding != JUST_BITS) { uint unq = D * C + B;
uint T = (D * C) + B; unq = unq ^ A;
T ^= A; unq = (A & 0x80) | (unq >> 2);
T = (A & 0x80) | (T >> 2); color_values[++out_index] = encoding == JUST_BITS
color_values[++out_index] = T; ? FastReplicateTo8(bitval, bitlen)
} : unq;
} }
} }
@ -566,12 +568,8 @@ void ComputeEndpoints(out uvec4 ep1, out uvec4 ep2, uint color_endpoint_mode, ui
} }
case 5: { case 5: {
READ_INT_VALUES(4) READ_INT_VALUES(4)
ivec2 transferred = BitTransferSigned(V[0].y, V[0].x); V[0].yx = BitTransferSigned(V[0].y, V[0].x);
V[0].y = transferred.x; V[0].wz = BitTransferSigned(V[0].w, V[0].z);
V[0].x = transferred.y;
transferred = BitTransferSigned(V[0].w, V[0].z);
V[0].w = transferred.x;
V[0].z = transferred.y;
ep1 = ClampByte(ivec4(V[0].z, V[0].x, V[0].x, V[0].x)); ep1 = ClampByte(ivec4(V[0].z, V[0].x, V[0].x, V[0].x));
ep2 = ClampByte(ivec4(V[0].z + V[0].w, V[0].x + V[0].y, V[0].x + V[0].y, V[0].x + V[0].y)); ep2 = ClampByte(ivec4(V[0].z + V[0].w, V[0].x + V[0].y, V[0].x + V[0].y, V[0].x + V[0].y));
break; break;
@ -595,15 +593,9 @@ void ComputeEndpoints(out uvec4 ep1, out uvec4 ep2, uint color_endpoint_mode, ui
} }
case 9: { case 9: {
READ_INT_VALUES(6) READ_INT_VALUES(6)
ivec2 transferred = BitTransferSigned(V[0].y, V[0].x); V[0].yx = BitTransferSigned(V[0].y, V[0].x);
V[0].y = transferred.x; V[0].wz = BitTransferSigned(V[0].w, V[0].z);
V[0].x = transferred.y; V[1].yx = BitTransferSigned(V[1].y, V[1].x);
transferred = BitTransferSigned(V[0].w, V[0].z);
V[0].w = transferred.x;
V[0].z = transferred.y;
transferred = BitTransferSigned(V[1].y, V[1].x);
V[1].y = transferred.x;
V[1].x = transferred.y;
if ((V[0].y + V[0].w + V[1].y) >= 0) { if ((V[0].y + V[0].w + V[1].y) >= 0) {
ep1 = ClampByte(ivec4(0xFF, V[0].x, V[0].z, V[1].x)); ep1 = ClampByte(ivec4(0xFF, V[0].x, V[0].z, V[1].x));
ep2 = ClampByte(ivec4(0xFF, V[0].x + V[0].y, V[0].z + V[0].w, V[1].x + V[1].y)); ep2 = ClampByte(ivec4(0xFF, V[0].x + V[0].y, V[0].z + V[0].w, V[1].x + V[1].y));
@ -632,21 +624,10 @@ void ComputeEndpoints(out uvec4 ep1, out uvec4 ep2, uint color_endpoint_mode, ui
} }
case 13: { case 13: {
READ_INT_VALUES(8) READ_INT_VALUES(8)
ivec2 transferred = BitTransferSigned(V[0].y, V[0].x); V[0].yx = BitTransferSigned(V[0].y, V[0].x);
V[0].y = transferred.x; V[0].wz = BitTransferSigned(V[0].w, V[0].z);
V[0].x = transferred.y; V[1].yx = BitTransferSigned(V[1].y, V[1].x);
transferred = BitTransferSigned(V[0].w, V[0].z); V[1].wz = BitTransferSigned(V[1].w, V[1].z);
V[0].w = transferred.x;
V[0].z = transferred.y;
transferred = BitTransferSigned(V[1].y, V[1].x);
V[1].y = transferred.x;
V[1].x = transferred.y;
transferred = BitTransferSigned(V[1].w, V[1].z);
V[1].w = transferred.x;
V[1].z = transferred.y;
if ((V[0].y + V[0].w + V[1].y) >= 0) { if ((V[0].y + V[0].w + V[1].y) >= 0) {
ep1 = ClampByte(ivec4(V[1].z, V[0].x, V[0].z, V[1].x)); ep1 = ClampByte(ivec4(V[1].z, V[0].x, V[0].z, V[1].x));
ep2 = ClampByte(ivec4(V[1].w + V[1].z, V[0].x + V[0].y, V[0].z + V[0].w, V[1].x + V[1].y)); ep2 = ClampByte(ivec4(V[1].w + V[1].z, V[0].x + V[0].y, V[0].z + V[0].w, V[1].x + V[1].y));