Merge pull request #8511 from lioncash/if-constexpr

VideoCommon/OpcodeDecoding: Make use of if constexpr
This commit is contained in:
Anthony
2019-12-07 18:40:15 -08:00
committed by GitHub
6 changed files with 132 additions and 113 deletions

View File

@ -25,6 +25,7 @@
#include "VideoCommon/Fifo.h" #include "VideoCommon/Fifo.h"
#include "VideoCommon/FramebufferManager.h" #include "VideoCommon/FramebufferManager.h"
#include "VideoCommon/GeometryShaderManager.h" #include "VideoCommon/GeometryShaderManager.h"
#include "VideoCommon/OpcodeDecoding.h"
#include "VideoCommon/PerfQueryBase.h" #include "VideoCommon/PerfQueryBase.h"
#include "VideoCommon/PixelEngine.h" #include "VideoCommon/PixelEngine.h"
#include "VideoCommon/PixelShaderManager.h" #include "VideoCommon/PixelShaderManager.h"
@ -341,7 +342,7 @@ static void BPWritten(const BPCmd& bp)
Memory::CopyFromEmu(texMem + tlutTMemAddr, addr, tlutXferCount); Memory::CopyFromEmu(texMem + tlutTMemAddr, addr, tlutXferCount);
if (g_bRecordFifoData) if (OpcodeDecoder::g_record_fifo_data)
FifoRecorder::GetInstance().UseMemory(addr, tlutXferCount, MemoryUpdate::TMEM); FifoRecorder::GetInstance().UseMemory(addr, tlutXferCount, MemoryUpdate::TMEM);
TextureCacheBase::InvalidateAllBindPoints(); TextureCacheBase::InvalidateAllBindPoints();
@ -563,7 +564,7 @@ static void BPWritten(const BPCmd& bp)
} }
} }
if (g_bRecordFifoData) if (OpcodeDecoder::g_record_fifo_data)
FifoRecorder::GetInstance().UseMemory(src_addr, bytes_read, MemoryUpdate::TMEM); FifoRecorder::GetInstance().UseMemory(src_addr, bytes_read, MemoryUpdate::TMEM);
TextureCacheBase::InvalidateAllBindPoints(); TextureCacheBase::InvalidateAllBindPoints();

View File

@ -9,15 +9,14 @@
// Super Mario Galaxy has nearly all geometry and more than half of the state in DLs (great!) // Super Mario Galaxy has nearly all geometry and more than half of the state in DLs (great!)
// Note that it IS NOT GENERALLY POSSIBLE to precompile display lists! You can compile them as they // Note that it IS NOT GENERALLY POSSIBLE to precompile display lists! You can compile them as they
// are // are while interpreting them, and hope that the vertex format doesn't change, though, if you do
// while interpreting them, and hope that the vertex format doesn't change, though, if you do it // it right when they are called. The reason is that the vertex format affects the sizes of the
// right // vertices.
// when they are called. The reason is that the vertex format affects the sizes of the vertices.
#include "VideoCommon/OpcodeDecoding.h" #include "VideoCommon/OpcodeDecoding.h"
#include "Common/CommonTypes.h" #include "Common/CommonTypes.h"
#include "Common/Logging/Log.h" #include "Common/Logging/Log.h"
#include "Common/MsgHandler.h"
#include "Core/FifoPlayer/FifoRecorder.h" #include "Core/FifoPlayer/FifoRecorder.h"
#include "Core/HW/Memmap.h" #include "Core/HW/Memmap.h"
#include "VideoCommon/BPMemory.h" #include "VideoCommon/BPMemory.h"
@ -27,33 +26,32 @@
#include "VideoCommon/Fifo.h" #include "VideoCommon/Fifo.h"
#include "VideoCommon/Statistics.h" #include "VideoCommon/Statistics.h"
#include "VideoCommon/VertexLoaderManager.h" #include "VideoCommon/VertexLoaderManager.h"
#include "VideoCommon/VideoCommon.h"
#include "VideoCommon/XFMemory.h" #include "VideoCommon/XFMemory.h"
bool g_bRecordFifoData = false;
namespace OpcodeDecoder namespace OpcodeDecoder
{ {
static bool s_bFifoErrorSeen = false; namespace
static u32 InterpretDisplayList(u32 address, u32 size)
{ {
u8* startAddress; bool s_is_fifo_error_seen = false;
u32 InterpretDisplayList(u32 address, u32 size)
{
u8* start_address;
if (Fifo::UseDeterministicGPUThread()) if (Fifo::UseDeterministicGPUThread())
startAddress = (u8*)Fifo::PopFifoAuxBuffer(size); start_address = static_cast<u8*>(Fifo::PopFifoAuxBuffer(size));
else else
startAddress = Memory::GetPointer(address); start_address = Memory::GetPointer(address);
u32 cycles = 0; u32 cycles = 0;
// Avoid the crash if Memory::GetPointer failed .. // Avoid the crash if Memory::GetPointer failed ..
if (startAddress != nullptr) if (start_address != nullptr)
{ {
// temporarily swap dl and non-dl (small "hack" for the stats) // temporarily swap dl and non-dl (small "hack" for the stats)
g_stats.SwapDL(); g_stats.SwapDL();
Run(DataReader(startAddress, startAddress + size), &cycles, true); Run(DataReader(start_address, start_address + size), &cycles, true);
INCSTAT(g_stats.this_frame.num_dlists_called); INCSTAT(g_stats.this_frame.num_dlists_called);
// un-swap // un-swap
@ -63,57 +61,70 @@ static u32 InterpretDisplayList(u32 address, u32 size)
return cycles; return cycles;
} }
static void InterpretDisplayListPreprocess(u32 address, u32 size) void InterpretDisplayListPreprocess(u32 address, u32 size)
{ {
u8* startAddress = Memory::GetPointer(address); u8* const start_address = Memory::GetPointer(address);
Fifo::PushFifoAuxBuffer(startAddress, size); Fifo::PushFifoAuxBuffer(start_address, size);
if (startAddress != nullptr) if (start_address == nullptr)
{ return;
Run<true>(DataReader(startAddress, startAddress + size), nullptr, true);
} Run<true>(DataReader(start_address, start_address + size), nullptr, true);
} }
} // Anonymous namespace
bool g_record_fifo_data = false;
void Init() void Init()
{ {
s_bFifoErrorSeen = false; s_is_fifo_error_seen = false;
} }
template <bool is_preprocess> template <bool is_preprocess>
u8* Run(DataReader src, u32* cycles, bool in_display_list) u8* Run(DataReader src, u32* cycles, bool in_display_list)
{ {
u32 totalCycles = 0; u32 total_cycles = 0;
u8* opcodeStart; u8* opcode_start = nullptr;
const auto finish_up = [cycles, &opcode_start, &total_cycles] {
if (cycles != nullptr)
{
*cycles = total_cycles;
}
return opcode_start;
};
while (true) while (true)
{ {
opcodeStart = src.GetPointer(); opcode_start = src.GetPointer();
if (!src.size()) if (!src.size())
goto end; return finish_up();
u8 cmd_byte = src.Read<u8>(); const u8 cmd_byte = src.Read<u8>();
int refarray;
switch (cmd_byte) switch (cmd_byte)
{ {
case GX_NOP: case GX_NOP:
totalCycles += 6; // Hm, this means that we scan over nop streams pretty slowly... total_cycles += 6; // Hm, this means that we scan over nop streams pretty slowly...
break; break;
case GX_UNKNOWN_RESET: case GX_UNKNOWN_RESET:
totalCycles += 6; // Datel software uses this command total_cycles += 6; // Datel software uses this command
DEBUG_LOG(VIDEO, "GX Reset?: %08x", cmd_byte); DEBUG_LOG(VIDEO, "GX Reset?: %08x", cmd_byte);
break; break;
case GX_LOAD_CP_REG: case GX_LOAD_CP_REG:
{ {
if (src.size() < 1 + 4) if (src.size() < 1 + 4)
goto end; return finish_up();
totalCycles += 12;
u8 sub_cmd = src.Read<u8>(); total_cycles += 12;
u32 value = src.Read<u32>();
const u8 sub_cmd = src.Read<u8>();
const u32 value = src.Read<u32>();
LoadCPReg(sub_cmd, value, is_preprocess); LoadCPReg(sub_cmd, value, is_preprocess);
if (!is_preprocess) if constexpr (!is_preprocess)
INCSTAT(g_stats.this_frame.num_cp_loads); INCSTAT(g_stats.this_frame.num_cp_loads);
} }
break; break;
@ -121,15 +132,18 @@ u8* Run(DataReader src, u32* cycles, bool in_display_list)
case GX_LOAD_XF_REG: case GX_LOAD_XF_REG:
{ {
if (src.size() < 4) if (src.size() < 4)
goto end; return finish_up();
u32 Cmd2 = src.Read<u32>();
int transfer_size = ((Cmd2 >> 16) & 15) + 1; const u32 cmd2 = src.Read<u32>();
const u32 transfer_size = ((cmd2 >> 16) & 15) + 1;
if (src.size() < transfer_size * sizeof(u32)) if (src.size() < transfer_size * sizeof(u32))
goto end; return finish_up();
totalCycles += 18 + 6 * transfer_size;
if (!is_preprocess) total_cycles += 18 + 6 * transfer_size;
if constexpr (!is_preprocess)
{ {
u32 xf_address = Cmd2 & 0xFFFF; const u32 xf_address = cmd2 & 0xFFFF;
LoadXFReg(transfer_size, xf_address, src); LoadXFReg(transfer_size, xf_address, src);
INCSTAT(g_stats.this_frame.num_xf_loads); INCSTAT(g_stats.this_frame.num_xf_loads);
@ -138,58 +152,61 @@ u8* Run(DataReader src, u32* cycles, bool in_display_list)
} }
break; break;
case GX_LOAD_INDX_A: // used for position matrices case GX_LOAD_INDX_A: // Used for position matrices
refarray = 0xC; case GX_LOAD_INDX_B: // Used for normal matrices
goto load_indx; case GX_LOAD_INDX_C: // Used for postmatrices
case GX_LOAD_INDX_B: // used for normal matrices case GX_LOAD_INDX_D: // Used for lights
refarray = 0xD; {
goto load_indx;
case GX_LOAD_INDX_C: // used for postmatrices
refarray = 0xE;
goto load_indx;
case GX_LOAD_INDX_D: // used for lights
refarray = 0xF;
goto load_indx;
load_indx:
if (src.size() < 4) if (src.size() < 4)
goto end; return finish_up();
totalCycles += 6;
if (is_preprocess) total_cycles += 6;
PreprocessIndexedXF(src.Read<u32>(), refarray);
// Map the command byte to its ref array.
// GX_LOAD_INDX_A (32) -> 0xC
// GX_LOAD_INDX_B (40) -> 0xD
// GX_LOAD_INDX_C (48) -> 0xE
// GX_LOAD_INDX_D (56) -> 0xF
const int ref_array = (cmd_byte / 8) + 8;
if constexpr (is_preprocess)
PreprocessIndexedXF(src.Read<u32>(), ref_array);
else else
LoadIndexedXF(src.Read<u32>(), refarray); LoadIndexedXF(src.Read<u32>(), ref_array);
break; }
break;
case GX_CMD_CALL_DL: case GX_CMD_CALL_DL:
{ {
if (src.size() < 8) if (src.size() < 8)
goto end; return finish_up();
u32 address = src.Read<u32>();
u32 count = src.Read<u32>(); const u32 address = src.Read<u32>();
const u32 count = src.Read<u32>();
if (in_display_list) if (in_display_list)
{ {
totalCycles += 6; total_cycles += 6;
INFO_LOG(VIDEO, "recursive display list detected"); INFO_LOG(VIDEO, "recursive display list detected");
} }
else else
{ {
if (is_preprocess) if constexpr (is_preprocess)
InterpretDisplayListPreprocess(address, count); InterpretDisplayListPreprocess(address, count);
else else
totalCycles += 6 + InterpretDisplayList(address, count); total_cycles += 6 + InterpretDisplayList(address, count);
} }
} }
break; break;
case GX_CMD_UNKNOWN_METRICS: // zelda 4 swords calls it and checks the metrics registers after case GX_CMD_UNKNOWN_METRICS: // zelda 4 swords calls it and checks the metrics registers after
// that // that
totalCycles += 6; total_cycles += 6;
DEBUG_LOG(VIDEO, "GX 0x44: %08x", cmd_byte); DEBUG_LOG(VIDEO, "GX 0x44: %08x", cmd_byte);
break; break;
case GX_CMD_INVL_VC: // Invalidate Vertex Cache case GX_CMD_INVL_VC: // Invalidate Vertex Cache
totalCycles += 6; total_cycles += 6;
DEBUG_LOG(VIDEO, "Invalidate (vertex cache?)"); DEBUG_LOG(VIDEO, "Invalidate (vertex cache?)");
break; break;
@ -198,10 +215,12 @@ u8* Run(DataReader src, u32* cycles, bool in_display_list)
// tokens and stuff. TODO: Call a much simplified LoadBPReg instead. // tokens and stuff. TODO: Call a much simplified LoadBPReg instead.
{ {
if (src.size() < 4) if (src.size() < 4)
goto end; return finish_up();
totalCycles += 12;
u32 bp_cmd = src.Read<u32>(); total_cycles += 12;
if (is_preprocess)
const u32 bp_cmd = src.Read<u32>();
if constexpr (is_preprocess)
{ {
LoadBPRegPreprocess(bp_cmd); LoadBPRegPreprocess(bp_cmd);
} }
@ -219,47 +238,43 @@ u8* Run(DataReader src, u32* cycles, bool in_display_list)
{ {
// load vertices // load vertices
if (src.size() < 2) if (src.size() < 2)
goto end; return finish_up();
u16 num_vertices = src.Read<u16>();
int bytes = VertexLoaderManager::RunVertices( const u16 num_vertices = src.Read<u16>();
const int bytes = VertexLoaderManager::RunVertices(
cmd_byte & GX_VAT_MASK, // Vertex loader index (0 - 7) cmd_byte & GX_VAT_MASK, // Vertex loader index (0 - 7)
(cmd_byte & GX_PRIMITIVE_MASK) >> GX_PRIMITIVE_SHIFT, num_vertices, src, is_preprocess); (cmd_byte & GX_PRIMITIVE_MASK) >> GX_PRIMITIVE_SHIFT, num_vertices, src, is_preprocess);
if (bytes < 0) if (bytes < 0)
goto end; return finish_up();
src.Skip(bytes); src.Skip(bytes);
// 4 GPU ticks per vertex, 3 CPU ticks per GPU tick // 4 GPU ticks per vertex, 3 CPU ticks per GPU tick
totalCycles += num_vertices * 4 * 3 + 6; total_cycles += num_vertices * 4 * 3 + 6;
} }
else else
{ {
if (!s_bFifoErrorSeen) if (!s_is_fifo_error_seen)
CommandProcessor::HandleUnknownOpcode(cmd_byte, opcodeStart, is_preprocess); CommandProcessor::HandleUnknownOpcode(cmd_byte, opcode_start, is_preprocess);
ERROR_LOG(VIDEO, "FIFO: Unknown Opcode(0x%02x @ %p, preprocessing = %s)", cmd_byte, ERROR_LOG(VIDEO, "FIFO: Unknown Opcode(0x%02x @ %p, preprocessing = %s)", cmd_byte,
opcodeStart, is_preprocess ? "yes" : "no"); opcode_start, is_preprocess ? "yes" : "no");
s_bFifoErrorSeen = true; s_is_fifo_error_seen = true;
totalCycles += 1; total_cycles += 1;
} }
break; break;
} }
// Display lists get added directly into the FIFO stream // Display lists get added directly into the FIFO stream
if (!is_preprocess && g_bRecordFifoData && cmd_byte != GX_CMD_CALL_DL) if constexpr (!is_preprocess)
{ {
u8* opcodeEnd; if (g_record_fifo_data && cmd_byte != GX_CMD_CALL_DL)
opcodeEnd = src.GetPointer(); {
FifoRecorder::GetInstance().WriteGPCommand(opcodeStart, u32(opcodeEnd - opcodeStart)); const u8* const opcode_end = src.GetPointer();
FifoRecorder::GetInstance().WriteGPCommand(opcode_start, u32(opcode_end - opcode_start));
}
} }
} }
end:
if (cycles)
{
*cycles = totalCycles;
}
return opcodeStart;
} }
template u8* Run<true>(DataReader src, u32* cycles, bool in_display_list); template u8* Run<true>(DataReader src, u32* cycles, bool in_display_list);

View File

@ -10,6 +10,9 @@ class DataReader;
namespace OpcodeDecoder namespace OpcodeDecoder
{ {
// Global flag to signal if FifoRecorder is active.
extern bool g_record_fifo_data;
enum enum
{ {
GX_NOP = 0x00, GX_NOP = 0x00,

View File

@ -65,6 +65,7 @@
#include "VideoCommon/NetPlayChatUI.h" #include "VideoCommon/NetPlayChatUI.h"
#include "VideoCommon/NetPlayGolfUI.h" #include "VideoCommon/NetPlayGolfUI.h"
#include "VideoCommon/OnScreenDisplay.h" #include "VideoCommon/OnScreenDisplay.h"
#include "VideoCommon/OpcodeDecoding.h"
#include "VideoCommon/PixelEngine.h" #include "VideoCommon/PixelEngine.h"
#include "VideoCommon/PixelShaderManager.h" #include "VideoCommon/PixelShaderManager.h"
#include "VideoCommon/PostProcessing.h" #include "VideoCommon/PostProcessing.h"
@ -880,19 +881,18 @@ std::tuple<int, int> Renderer::CalculateOutputDimensions(int width, int height)
void Renderer::CheckFifoRecording() void Renderer::CheckFifoRecording()
{ {
bool wasRecording = g_bRecordFifoData; const bool was_recording = OpcodeDecoder::g_record_fifo_data;
g_bRecordFifoData = FifoRecorder::GetInstance().IsRecording(); OpcodeDecoder::g_record_fifo_data = FifoRecorder::GetInstance().IsRecording();
if (g_bRecordFifoData) if (!OpcodeDecoder::g_record_fifo_data)
return;
if (!was_recording)
{ {
if (!wasRecording) RecordVideoMemory();
{
RecordVideoMemory();
}
FifoRecorder::GetInstance().EndFrame(CommandProcessor::fifo.CPBase,
CommandProcessor::fifo.CPEnd);
} }
FifoRecorder::GetInstance().EndFrame(CommandProcessor::fifo.CPBase, CommandProcessor::fifo.CPEnd);
} }
void Renderer::RecordVideoMemory() void Renderer::RecordVideoMemory()

View File

@ -38,6 +38,7 @@
#include "VideoCommon/BPMemory.h" #include "VideoCommon/BPMemory.h"
#include "VideoCommon/FramebufferManager.h" #include "VideoCommon/FramebufferManager.h"
#include "VideoCommon/HiresTextures.h" #include "VideoCommon/HiresTextures.h"
#include "VideoCommon/OpcodeDecoding.h"
#include "VideoCommon/PixelShaderManager.h" #include "VideoCommon/PixelShaderManager.h"
#include "VideoCommon/RenderBase.h" #include "VideoCommon/RenderBase.h"
#include "VideoCommon/SamplerCommon.h" #include "VideoCommon/SamplerCommon.h"
@ -1260,9 +1261,11 @@ TextureCacheBase::GetTexture(u32 address, u32 width, u32 height, const TextureFo
// If we are recording a FifoLog, keep track of what memory we read. FifoRecorder does // If we are recording a FifoLog, keep track of what memory we read. FifoRecorder does
// its own memory modification tracking independent of the texture hashing below. // its own memory modification tracking independent of the texture hashing below.
if (g_bRecordFifoData && !from_tmem) if (OpcodeDecoder::g_record_fifo_data && !from_tmem)
{
FifoRecorder::GetInstance().UseMemory(address, texture_size + additional_mips_size, FifoRecorder::GetInstance().UseMemory(address, texture_size + additional_mips_size,
MemoryUpdate::TEXTURE_MAP); MemoryUpdate::TEXTURE_MAP);
}
// TODO: This doesn't hash GB tiles for preloaded RGBA8 textures (instead, it's hashing more data // TODO: This doesn't hash GB tiles for preloaded RGBA8 textures (instead, it's hashing more data
// from the low tmem bank than it should) // from the low tmem bank than it should)
@ -2294,7 +2297,7 @@ void TextureCacheBase::CopyRenderTargetToTexture(
++iter.first; ++iter.first;
} }
if (g_bRecordFifoData) if (OpcodeDecoder::g_record_fifo_data)
{ {
// Mark the memory behind this efb copy as dynamicly generated for the Fifo log // Mark the memory behind this efb copy as dynamicly generated for the Fifo log
u32 address = dstAddr; u32 address = dstAddr;

View File

@ -6,9 +6,6 @@
#include "Common/CommonTypes.h" #include "Common/CommonTypes.h"
// Global flag to signal if FifoRecorder is active.
extern bool g_bRecordFifoData;
// These are accurate (disregarding AA modes). // These are accurate (disregarding AA modes).
constexpr u32 EFB_WIDTH = 640; constexpr u32 EFB_WIDTH = 640;
constexpr u32 EFB_HEIGHT = 528; constexpr u32 EFB_HEIGHT = 528;