Xargon/Duckstation
1
0
Fork 0
mirror of https://github.com/RetroDECK/Duckstation.git synced 2024-11-29 09:05:41 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

MDEC: Store scale/IDCT matrix transposed

Browse source
This commit is contained in:
Stenzek 2024-07-02 14:31:55 +10:00
parent 6d6659c85e
commit e1f90e6baf
No known key found for this signature in database
2 changed files with 44 additions and 18 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

58
src/core/mdec.cpp
View file

@ -107,6 +107,7 @@ static bool HandleDecodeMacroblockCommand();
static void HandleSetQuantTableCommand(); static void HandleSetQuantTableCommand();
static void HandleSetScaleCommand(); static void HandleSetScaleCommand();
static void SetScaleMatrix(const u16* values);
static bool DecodeMonoMacroblock(); static bool DecodeMonoMacroblock();
static bool DecodeColoredMacroblock(); static bool DecodeColoredMacroblock();
static void ScheduleBlockCopyOut(TickCount ticks); static void ScheduleBlockCopyOut(TickCount ticks);
@ -181,7 +182,18 @@ bool MDEC::DoState(StateWrapper& sw)
sw.Do(&s_remaining_halfwords); sw.Do(&s_remaining_halfwords);
sw.Do(&s_iq_uv); sw.Do(&s_iq_uv);
sw.Do(&s_iq_y); sw.Do(&s_iq_y);
if (sw.GetVersion() < 66) [[unlikely]]
{
std::array<u16, 64> old_scale_matrix;
sw.Do(&old_scale_matrix);
SetScaleMatrix(old_scale_matrix.data());
}
else
{
sw.Do(&s_scale_table); sw.Do(&s_scale_table);
}
sw.Do(&s_blocks); sw.Do(&s_blocks);
sw.Do(&s_current_block); sw.Do(&s_current_block);
sw.Do(&s_current_coefficient); sw.Do(&s_current_coefficient);
@ -817,7 +829,7 @@ void MDEC::IDCT_Old(s16* blk)
{ {
s64 sum = 0; s64 sum = 0;
for (u32 u = 0; u < 8; u++) for (u32 u = 0; u < 8; u++)
sum += s32(blk[u * 8 + x]) * s32(s_scale_table[u * 8 + y]); sum += s32(blk[u * 8 + x]) * s32(s_scale_table[y * 8 + u]);
temp_buffer[x + y * 8] = sum; temp_buffer[x + y * 8] = sum;
} }
} }
@ -827,7 +839,7 @@ void MDEC::IDCT_Old(s16* blk)
{ {
s64 sum = 0; s64 sum = 0;
for (u32 u = 0; u < 8; u++) for (u32 u = 0; u < 8; u++)
sum += s64(temp_buffer[u + y * 8]) * s32(s_scale_table[u * 8 + x]); sum += s64(temp_buffer[u + y * 8]) * s32(s_scale_table[x * 8 + u]);
blk[x + y * 8] = blk[x + y * 8] =
static_cast<s16>(std::clamp<s32>(SignExtendN<9, s32>((sum >> 32) + ((sum >> 31) & 1)), -128, 127)); static_cast<s16>(std::clamp<s32>(SignExtendN<9, s32>((sum >> 32) + ((sum >> 31) & 1)), -128, 127));
@ -923,29 +935,35 @@ bool MDEC::DecodeRLE_New(s16* blk, const u8* qt)
return false; return false;
} }
template<typename BlkType>
static s32 IDCTRow(const BlkType* blk, const s16* idct_matrix)
{
// IDCT matrix is -32768..32767, block is -16384..16383. 4 adds can happen without overflow.
const s32 sum1 = static_cast<s32>(blk[0]) * static_cast<s32>(idct_matrix[0]) +
static_cast<s32>(blk[1]) * static_cast<s32>(idct_matrix[1]) +
static_cast<s32>(blk[2]) * static_cast<s32>(idct_matrix[2]) +
static_cast<s32>(blk[3]) * static_cast<s32>(idct_matrix[3]);
const s32 sum2 = static_cast<s32>(blk[4]) * static_cast<s32>(idct_matrix[4]) +
static_cast<s32>(blk[5]) * static_cast<s32>(idct_matrix[5]) +
static_cast<s32>(blk[6]) * static_cast<s32>(idct_matrix[6]) +
static_cast<s32>(blk[7]) * static_cast<s32>(idct_matrix[7]);
return static_cast<s32>(((static_cast<s64>(sum1) + static_cast<s64>(sum2)) + 0x20000) >> 18);
}
void MDEC::IDCT_New(s16* blk) void MDEC::IDCT_New(s16* blk)
{ {
std::array<s32, 64> temp; std::array<s32, 64> temp;
for (u32 x = 0; x < 8; x++) for (u32 x = 0; x < 8; x++)
{ {
for (u32 y = 0; y < 8; y++) for (u32 y = 0; y < 8; y++)
{ temp[y * 8 + x] = IDCTRow(&blk[x * 8], &s_scale_table[y * 8]);
// TODO: We could invert scale_table to get these in row-major order,
// in which case we could do optimize this to a vector multiply.
s32 sum = 0;
for (u32 z = 0; z < 8; z++)
sum += (s32(blk[x * 8 + z]) * s32(s_scale_table[z * 8 + y])) / 8;
temp[y * 8 + x] = static_cast<s32>((sum + 0x4000) >> 15);
}
} }
for (u32 x = 0; x < 8; x++) for (u32 x = 0; x < 8; x++)
{ {
for (u32 y = 0; y < 8; y++) for (u32 y = 0; y < 8; y++)
{ {
s32 sum = 0; const s32 sum = IDCTRow(&temp[x * 8], &s_scale_table[y * 8]);
for (u32 z = 0; z < 8; z++) blk[x * 8 + y] = static_cast<s16>(std::clamp(SignExtendN<9, s32>(sum), -128, 127));
sum += (temp[x * 8 + z] * s32(s_scale_table[z * 8 + y])) / 8;
blk[x * 8 + y] = static_cast<s16>(std::clamp(SignExtendN<9, s32>((sum + 0x4000) >> 15), -128, 127));
} }
} }
} }
@ -1005,11 +1023,19 @@ void MDEC::HandleSetScaleCommand()
{ {
DebugAssert(s_remaining_halfwords == 64); DebugAssert(s_remaining_halfwords == 64);
// TODO: Remove extra copies..
std::array<u16, 64> packed_data; std::array<u16, 64> packed_data;
s_data_in_fifo.PopRange(packed_data.data(), static_cast<u32>(packed_data.size())); s_data_in_fifo.PopRange(packed_data.data(), static_cast<u32>(packed_data.size()));
s_remaining_halfwords -= 32; s_remaining_halfwords -= 32;
std::memcpy(s_scale_table.data(), packed_data.data(), s_scale_table.size() * sizeof(s16)); SetScaleMatrix(packed_data.data());
}
void MDEC::SetScaleMatrix(const u16* values)
{
for (u32 y = 0; y < 8; y++)
{
for (u32 x = 0; x < 8; x++)
s_scale_table[y * 8 + x] = values[x * 8 + y];
}
} }
void MDEC::DrawDebugStateWindow() void MDEC::DrawDebugStateWindow()

2
src/core/save_state_version.h
View file

@ -5,7 +5,7 @@
#include "types.h" #include "types.h"
static constexpr u32 SAVE_STATE_MAGIC = 0x43435544; static constexpr u32 SAVE_STATE_MAGIC = 0x43435544;
static constexpr u32 SAVE_STATE_VERSION = 65; static constexpr u32 SAVE_STATE_VERSION = 66;
static constexpr u32 SAVE_STATE_MINIMUM_VERSION = 42; static constexpr u32 SAVE_STATE_MINIMUM_VERSION = 42;
static_assert(SAVE_STATE_VERSION >= SAVE_STATE_MINIMUM_VERSION); static_assert(SAVE_STATE_VERSION >= SAVE_STATE_MINIMUM_VERSION);