Merge branch 'master' into windows-unicode

This commit is contained in:
Jordan Woyak
2013-03-02 15:33:32 -06:00
64 changed files with 1083 additions and 183 deletions

View File

@ -199,6 +199,7 @@
<ClCompile Include="Src\LineGeometryShader.cpp" />
<ClCompile Include="Src\main.cpp" />
<ClCompile Include="Src\NativeVertexFormat.cpp" />
<ClCompile Include="Src\PerfQuery.cpp" />
<ClCompile Include="Src\PixelShaderCache.cpp" />
<ClCompile Include="Src\PointGeometryShader.cpp" />
<ClCompile Include="Src\PSTextureEncoder.cpp" />
@ -228,6 +229,7 @@
<ClInclude Include="Src\Globals.h" />
<ClInclude Include="Src\LineGeometryShader.h" />
<ClInclude Include="Src\main.h" />
<ClInclude Include="Src\PerfQuery.h" />
<ClInclude Include="Src\PixelShaderCache.h" />
<ClInclude Include="Src\PointGeometryShader.h" />
<ClInclude Include="Src\PSTextureEncoder.h" />

View File

@ -57,6 +57,9 @@
<ClCompile Include="Src\PointGeometryShader.cpp">
<Filter>Render</Filter>
</ClCompile>
<ClCompile Include="Src\PerfQuery.cpp">
<Filter>Render</Filter>
</ClCompile>
</ItemGroup>
<ItemGroup>
<ClInclude Include="Src\Globals.h" />
@ -117,6 +120,9 @@
<ClInclude Include="Src\PointGeometryShader.h">
<Filter>Render</Filter>
</ClInclude>
<ClInclude Include="Src\PerfQuery.h">
<Filter>Render</Filter>
</ClInclude>
</ItemGroup>
<ItemGroup>
<Filter Include="D3D">

View File

@ -0,0 +1,150 @@
#include "RenderBase.h"
#include "D3DBase.h"
#include "PerfQuery.h"
namespace DX11 {
PerfQuery::PerfQuery()
: m_query_read_pos()
, m_query_count()
{
for (int i = 0; i != ARRAYSIZE(m_query_buffer); ++i)
{
D3D11_QUERY_DESC qdesc = CD3D11_QUERY_DESC(D3D11_QUERY_OCCLUSION, 0);
D3D::device->CreateQuery(&qdesc, &m_query_buffer[i].query);
}
ResetQuery();
}
PerfQuery::~PerfQuery()
{
for (int i = 0; i != ARRAYSIZE(m_query_buffer); ++i)
{
// TODO: EndQuery?
m_query_buffer[i].query->Release();
}
}
void PerfQuery::EnableQuery(PerfQueryGroup type)
{
// Is this sane?
if (m_query_count > ARRAYSIZE(m_query_buffer) / 2)
WeakFlush();
if (ARRAYSIZE(m_query_buffer) == m_query_count)
{
// TODO
FlushOne();
ERROR_LOG(VIDEO, "flushed query buffer early!");
}
// start query
if (type == PQG_ZCOMP_ZCOMPLOC || type == PQG_ZCOMP)
{
auto& entry = m_query_buffer[(m_query_read_pos + m_query_count) % ARRAYSIZE(m_query_buffer)];
D3D::context->Begin(entry.query);
entry.query_type = type;
++m_query_count;
}
}
void PerfQuery::DisableQuery(PerfQueryGroup type)
{
// stop query
if (type == PQG_ZCOMP_ZCOMPLOC || type == PQG_ZCOMP)
{
auto& entry = m_query_buffer[(m_query_read_pos + m_query_count + ARRAYSIZE(m_query_buffer)-1) % ARRAYSIZE(m_query_buffer)];
D3D::context->End(entry.query);
}
}
void PerfQuery::ResetQuery()
{
m_query_count = 0;
std::fill_n(m_results, ARRAYSIZE(m_results), 0);
}
u32 PerfQuery::GetQueryResult(PerfQueryType type)
{
u32 result = 0;
if (type == PQ_ZCOMP_INPUT_ZCOMPLOC || type == PQ_ZCOMP_OUTPUT_ZCOMPLOC)
{
result = m_results[PQG_ZCOMP_ZCOMPLOC];
}
else if (type == PQ_ZCOMP_INPUT || type == PQ_ZCOMP_OUTPUT)
{
result = m_results[PQG_ZCOMP];
}
else if (type == PQ_BLEND_INPUT)
{
result = m_results[PQG_ZCOMP] + m_results[PQG_ZCOMP_ZCOMPLOC];
}
else if (type == PQ_EFB_COPY_CLOCKS)
{
result = m_results[PQG_EFB_COPY_CLOCKS];
}
return result / 4;
}
void PerfQuery::FlushOne()
{
auto& entry = m_query_buffer[m_query_read_pos];
UINT64 result = 0;
HRESULT hr = S_FALSE;
while (hr != S_OK)
{
// TODO: Might cause us to be stuck in an infinite loop!
hr = D3D::context->GetData(entry.query, &result, sizeof(result), 0);
}
// NOTE: Reported pixel metrics should be referenced to native resolution
m_results[entry.query_type] += (u64)result * EFB_WIDTH / g_renderer->GetTargetWidth() * EFB_HEIGHT / g_renderer->GetTargetHeight();
m_query_read_pos = (m_query_read_pos + 1) % ARRAYSIZE(m_query_buffer);
--m_query_count;
}
// TODO: could selectively flush things, but I don't think that will do much
void PerfQuery::FlushResults()
{
while (!IsFlushed())
FlushOne();
}
void PerfQuery::WeakFlush()
{
while (!IsFlushed())
{
auto& entry = m_query_buffer[m_query_read_pos];
UINT64 result = 0;
HRESULT hr = D3D::context->GetData(entry.query, &result, sizeof(result), D3D11_ASYNC_GETDATA_DONOTFLUSH);
if (hr == S_OK)
{
// NOTE: Reported pixel metrics should be referenced to native resolution
m_results[entry.query_type] += (u64)result * EFB_WIDTH / g_renderer->GetTargetWidth() * EFB_HEIGHT / g_renderer->GetTargetHeight();
m_query_read_pos = (m_query_read_pos + 1) % ARRAYSIZE(m_query_buffer);
--m_query_count;
}
else
{
break;
}
}
}
bool PerfQuery::IsFlushed() const
{
return 0 == m_query_count;
}
} // namespace

View File

@ -0,0 +1,46 @@
#ifndef _PERFQUERY_H_
#define _PERFQUERY_H_
#include "PerfQueryBase.h"
namespace DX11 {
class PerfQuery : public PerfQueryBase
{
public:
PerfQuery();
~PerfQuery();
void EnableQuery(PerfQueryGroup type);
void DisableQuery(PerfQueryGroup type);
void ResetQuery();
u32 GetQueryResult(PerfQueryType type);
void FlushResults();
bool IsFlushed() const;
private:
struct ActiveQuery
{
ID3D11Query* query;
PerfQueryGroup query_type;
};
void WeakFlush();
// Only use when non-empty
void FlushOne();
// when testing in SMS: 64 was too small, 128 was ok
static const int PERF_QUERY_BUFFER_SIZE = 512;
ActiveQuery m_query_buffer[PERF_QUERY_BUFFER_SIZE];
int m_query_read_pos;
// TODO: sloppy
volatile int m_query_count;
volatile u32 m_results[PQG_NUM_MEMBERS];
};
} // namespace
#endif // _PERFQUERY_H_

View File

@ -65,6 +65,7 @@ ID3D11RasterizerState* resetraststate = NULL;
static ID3D11Texture2D* s_screenshot_texture = NULL;
// GX pipeline state
struct
{

View File

@ -208,7 +208,6 @@ void VertexManager::Draw(UINT stride)
if (IndexGenerator::GetNumLines() > 0 || IndexGenerator::GetNumPoints() > 0)
((DX11::Renderer*)g_renderer)->RestoreCull();
}
void VertexManager::vFlush()
{
if (LocalVBuffer == s_pCurBufferPointer) return;
@ -274,8 +273,10 @@ void VertexManager::vFlush()
unsigned int stride = g_nativeVertexFmt->GetVertexStride();
g_nativeVertexFmt->SetupVertexPointers();
g_renderer->ApplyState(useDstAlpha);
g_perf_query->EnableQuery(bpmem.zcontrol.early_ztest ? PQG_ZCOMP_ZCOMPLOC : PQG_ZCOMP);
Draw(stride);
g_perf_query->DisableQuery(bpmem.zcontrol.early_ztest ? PQG_ZCOMP_ZCOMPLOC : PQG_ZCOMP);
GFX_DEBUGGER_PAUSE_AT(NEXT_FLUSH, true);

View File

@ -42,6 +42,7 @@
#include "D3DUtil.h"
#include "D3DBase.h"
#include "PerfQuery.h"
#include "PixelShaderCache.h"
#include "TextureCache.h"
#include "VertexManager.h"
@ -183,6 +184,7 @@ void VideoBackend::Video_Prepare()
g_renderer = new Renderer;
g_texture_cache = new TextureCache;
g_vertex_manager = new VertexManager;
g_perf_query = new PerfQuery;
VertexShaderCache::Init();
PixelShaderCache::Init();
D3D::InitUtils();
@ -225,6 +227,7 @@ void VideoBackend::Shutdown()
D3D::ShutdownUtils();
PixelShaderCache::Shutdown();
VertexShaderCache::Shutdown();
delete g_perf_query;
delete g_vertex_manager;
delete g_texture_cache;
delete g_renderer;

View File

@ -57,6 +57,7 @@
#include "ConfigManager.h"
#include "VideoBackend.h"
#include "PerfQueryBase.h"
namespace DX9
{
@ -97,8 +98,6 @@ void InitBackendInfo()
g_Config.backend_info.bSupports3DVision = true;
g_Config.backend_info.bSupportsDualSourceBlend = false;
g_Config.backend_info.bSupportsFormatReinterpretation = true;
g_Config.backend_info.bSupportsPixelLighting = C_PLIGHTS + 40 <= maxConstants && C_PMATERIALS + 4 <= maxConstants;
// adapters
@ -172,6 +171,7 @@ void VideoBackend::Video_Prepare()
g_vertex_manager = new VertexManager;
g_renderer = new Renderer;
g_texture_cache = new TextureCache;
g_perf_query = new PerfQueryBase;
// VideoCommon
BPInit();
Fifo_Init();
@ -209,6 +209,7 @@ void VideoBackend::Shutdown()
// internal interfaces
PixelShaderCache::Shutdown();
VertexShaderCache::Shutdown();
delete g_perf_query;
delete g_texture_cache;
delete g_renderer;
delete g_vertex_manager;

View File

@ -2,6 +2,7 @@ set(SRCS Src/FramebufferManager.cpp
Src/GLUtil.cpp
Src/main.cpp
Src/NativeVertexFormat.cpp
Src/PerfQuery.cpp
Src/PixelShaderCache.cpp
Src/PostProcessing.cpp
Src/RasterFont.cpp

View File

@ -200,6 +200,7 @@
<ClCompile Include="Src\GLUtil.cpp" />
<ClCompile Include="Src\main.cpp" />
<ClCompile Include="Src\NativeVertexFormat.cpp" />
<ClCompile Include="Src\PerfQuery.cpp" />
<ClCompile Include="Src\PixelShaderCache.cpp" />
<ClCompile Include="Src\PostProcessing.cpp" />
<ClCompile Include="Src\RasterFont.cpp" />
@ -222,6 +223,7 @@
<ClInclude Include="Src\Globals.h" />
<ClInclude Include="Src\GLUtil.h" />
<ClInclude Include="Src\main.h" />
<ClInclude Include="Src\PerfQuery.h" />
<ClInclude Include="Src\PixelShaderCache.h" />
<ClInclude Include="Src\PostProcessing.h" />
<ClInclude Include="Src\RasterFont.h" />

View File

@ -36,6 +36,9 @@
<ClCompile Include="Src\VertexShaderCache.cpp">
<Filter>Render</Filter>
</ClCompile>
<ClCompile Include="Src\PerfQuery.cpp">
<Filter>Render</Filter>
</ClCompile>
</ItemGroup>
<ItemGroup>
<ClInclude Include="Src\Globals.h" />
@ -72,6 +75,9 @@
<ClInclude Include="Src\VertexShaderCache.h">
<Filter>Render</Filter>
</ClInclude>
<ClInclude Include="Src\PerfQuery.h">
<Filter>Render</Filter>
</ClInclude>
</ItemGroup>
<ItemGroup>
<None Include="CMakeLists.txt" />

View File

@ -0,0 +1,133 @@
#include "RenderBase.h"
#include "GLUtil.h"
#include "PerfQuery.h"
namespace OGL
{
PerfQuery::PerfQuery()
: m_query_read_pos()
, m_query_count()
{
for (int i = 0; i != ARRAYSIZE(m_query_buffer); ++i)
glGenQueries(1, &m_query_buffer[i].query_id);
ResetQuery();
}
PerfQuery::~PerfQuery()
{
for (int i = 0; i != ARRAYSIZE(m_query_buffer); ++i)
glDeleteQueries(1, &m_query_buffer[i].query_id);
}
void PerfQuery::EnableQuery(PerfQueryGroup type)
{
// Is this sane?
if (m_query_count > ARRAYSIZE(m_query_buffer) / 2)
WeakFlush();
if (ARRAYSIZE(m_query_buffer) == m_query_count)
{
FlushOne();
//ERROR_LOG(VIDEO, "flushed query buffer early!");
}
// start query
if (type == PQG_ZCOMP_ZCOMPLOC || type == PQG_ZCOMP)
{
auto& entry = m_query_buffer[(m_query_read_pos + m_query_count) % ARRAYSIZE(m_query_buffer)];
glBeginQuery(GL_SAMPLES_PASSED, entry.query_id);
entry.query_type = type;
++m_query_count;
}
}
void PerfQuery::DisableQuery(PerfQueryGroup type)
{
// stop query
if (type == PQG_ZCOMP_ZCOMPLOC || type == PQG_ZCOMP)
{
glEndQuery(GL_SAMPLES_PASSED);
}
}
bool PerfQuery::IsFlushed() const
{
return 0 == m_query_count;
}
void PerfQuery::FlushOne()
{
auto& entry = m_query_buffer[m_query_read_pos];
GLuint result = 0;
glGetQueryObjectuiv(entry.query_id, GL_QUERY_RESULT, &result);
// NOTE: Reported pixel metrics should be referenced to native resolution
m_results[entry.query_type] += (u64)result * EFB_WIDTH / g_renderer->GetTargetWidth() * EFB_HEIGHT / g_renderer->GetTargetHeight();
m_query_read_pos = (m_query_read_pos + 1) % ARRAYSIZE(m_query_buffer);
--m_query_count;
}
// TODO: could selectively flush things, but I don't think that will do much
void PerfQuery::FlushResults()
{
while (!IsFlushed())
FlushOne();
}
void PerfQuery::WeakFlush()
{
while (!IsFlushed())
{
auto& entry = m_query_buffer[m_query_read_pos];
GLuint result = GL_FALSE;
glGetQueryObjectuiv(entry.query_id, GL_QUERY_RESULT_AVAILABLE, &result);
if (GL_TRUE == result)
{
FlushOne();
}
else
{
break;
}
}
}
void PerfQuery::ResetQuery()
{
m_query_count = 0;
std::fill_n(m_results, ARRAYSIZE(m_results), 0);
}
u32 PerfQuery::GetQueryResult(PerfQueryType type)
{
u32 result = 0;
if (type == PQ_ZCOMP_INPUT_ZCOMPLOC || type == PQ_ZCOMP_OUTPUT_ZCOMPLOC)
{
result = m_results[PQG_ZCOMP_ZCOMPLOC];
}
else if (type == PQ_ZCOMP_INPUT || type == PQ_ZCOMP_OUTPUT)
{
result = m_results[PQG_ZCOMP];
}
else if (type == PQ_BLEND_INPUT)
{
result = m_results[PQG_ZCOMP] + m_results[PQG_ZCOMP_ZCOMPLOC];
}
else if (type == PQ_EFB_COPY_CLOCKS)
{
result = m_results[PQG_EFB_COPY_CLOCKS];
}
return result / 4;
}
} // namespace

View File

@ -0,0 +1,46 @@
#ifndef _PERFQUERY_H_
#define _PERFQUERY_H_
#include "PerfQueryBase.h"
namespace OGL {
class PerfQuery : public PerfQueryBase
{
public:
PerfQuery();
~PerfQuery();
void EnableQuery(PerfQueryGroup type);
void DisableQuery(PerfQueryGroup type);
void ResetQuery();
u32 GetQueryResult(PerfQueryType type);
void FlushResults();
bool IsFlushed() const;
private:
struct ActiveQuery
{
GLuint query_id;
PerfQueryGroup query_type;
};
// when testing in SMS: 64 was too small, 128 was ok
static const int PERF_QUERY_BUFFER_SIZE = 512;
void WeakFlush();
// Only use when non-empty
void FlushOne();
// This contains gl query objects with unretrieved results.
ActiveQuery m_query_buffer[PERF_QUERY_BUFFER_SIZE];
int m_query_read_pos;
// TODO: sloppy
volatile int m_query_count;
volatile u32 m_results[PQG_NUM_MEMBERS];
};
} // namespace
#endif // _PERFQUERY_H_

View File

@ -1420,6 +1420,7 @@ void Renderer::SetDepthMode()
else
{
// if the test is disabled write is disabled too
// TODO: When PE performance metrics are being emulated via occlusion queries, we should (probably?) enable depth test with depth function ALWAYS here
glDisable(GL_DEPTH_TEST);
glDepthMask(GL_FALSE);
}

View File

@ -40,6 +40,7 @@
#include "OpcodeDecoding.h"
#include "FileUtil.h"
#include "Debugger.h"
#include "PerfQueryBase.h"
#include "main.h"
@ -217,7 +218,10 @@ void VertexManager::vFlush()
if (ps) PixelShaderCache::SetCurrentShader(ps->glprogid); // Lego Star Wars crashes here.
if (vs) VertexShaderCache::SetCurrentShader(vs->glprogid);
g_perf_query->EnableQuery(bpmem.zcontrol.early_ztest ? PQG_ZCOMP_ZCOMPLOC : PQG_ZCOMP);
Draw();
g_perf_query->DisableQuery(bpmem.zcontrol.early_ztest ? PQG_ZCOMP_ZCOMPLOC : PQG_ZCOMP);
//ERROR_LOG(VIDEO, "PerfQuery result: %d", g_perf_query->GetQueryResult(bpmem.zcontrol.early_ztest ? PQ_ZCOMP_OUTPUT_ZCOMPLOC : PQ_ZCOMP_OUTPUT));
// run through vertex groups again to set alpha
if (useDstAlpha && !dualSourcePossible)

View File

@ -92,6 +92,7 @@ Make AA apply instantly during gameplay if possible
#include "FramebufferManager.h"
#include "Core.h"
#include "Host.h"
#include "PerfQuery.h"
#include "VideoState.h"
#include "VideoBackend.h"
@ -194,6 +195,7 @@ void VideoBackend::Video_Prepare()
BPInit();
g_vertex_manager = new VertexManager;
g_perf_query = new PerfQuery;
Fifo_Init(); // must be done before OpcodeDecoder_Init()
OpcodeDecoder_Init();
VertexShaderCache::Init();

View File

@ -90,6 +90,21 @@ void SWBPWritten(int address, int newvalue)
SWPixelEngine::pereg.boxBottom = newvalue >> 10;
SWPixelEngine::pereg.boxTop = newvalue & 0x3ff;
break;
case BPMEM_CLEAR_PIXEL_PERF:
// TODO: I didn't test if the value written to this register affects the amount of cleared registers
SWPixelEngine::pereg.perfZcompInputZcomplocLo = 0;
SWPixelEngine::pereg.perfZcompInputZcomplocHi = 0;
SWPixelEngine::pereg.perfZcompOutputZcomplocLo = 0;
SWPixelEngine::pereg.perfZcompOutputZcomplocHi = 0;
SWPixelEngine::pereg.perfZcompInputLo = 0;
SWPixelEngine::pereg.perfZcompInputHi = 0;
SWPixelEngine::pereg.perfZcompOutputLo = 0;
SWPixelEngine::pereg.perfZcompOutputHi = 0;
SWPixelEngine::pereg.perfBlendInputLo = 0;
SWPixelEngine::pereg.perfBlendInputHi = 0;
SWPixelEngine::pereg.perfEfbCopyClocksLo = 0;
SWPixelEngine::pereg.perfEfbCopyClocksHi = 0;
break;
case BPMEM_LOADTLUT0: // This one updates bpmem.tlutXferSrc, no need to do anything here.
break;
case BPMEM_LOADTLUT1: // Load a Texture Look Up Table

View File

@ -65,6 +65,13 @@ namespace Clipper
OutputVertexData ClippedVertices[NUM_CLIPPED_VERTICES];
OutputVertexData *Vertices[NUM_INDICES];
void DoState(PointerWrap &p)
{
p.DoArray(m_ViewOffset,2);
for (int i = 0; i< NUM_CLIPPED_VERTICES; ++i)
ClippedVertices[i].DoState(p);
}
void Init()
{
for (int i = 0; i < NUM_CLIPPED_VERTICES; ++i)

View File

@ -21,6 +21,7 @@
#include "Common.h"
#include "NativeVertexFormat.h"
#include "ChunkFile.h"
namespace Clipper
@ -36,6 +37,8 @@ namespace Clipper
bool CullTest(OutputVertexData *v0, OutputVertexData *v1, OutputVertexData *v2, bool &backface);
void PerspectiveDivide(OutputVertexData *vertex);
void DoState(PointerWrap &p);
}

View File

@ -28,6 +28,7 @@ u8 efb[EFB_WIDTH*EFB_HEIGHT*6];
namespace EfbInterface
{
u8 efbColorTexture[EFB_WIDTH*EFB_HEIGHT*4];
inline u32 GetColorOffset(u16 x, u16 y)
@ -40,6 +41,12 @@ namespace EfbInterface
return (x + y * EFB_WIDTH) * 3 + DEPTH_BUFFER_START;
}
void DoState(PointerWrap &p)
{
p.DoArray(efb, EFB_WIDTH*EFB_HEIGHT*6);
p.DoArray(efbColorTexture, EFB_WIDTH*EFB_HEIGHT*4);
}
void SetPixelAlphaOnly(u32 offset, u8 a)
{
switch (bpmem.zcontrol.pixel_format)

View File

@ -47,6 +47,7 @@ namespace EfbInterface
void UpdateColorTexture();
extern u8 efbColorTexture[EFB_WIDTH*EFB_HEIGHT*4]; // RGBA format
void DoState(PointerWrap &p);
}
#endif

View File

@ -19,6 +19,7 @@
#define _NATIVEVERTEXFORMAT_H
#include "Vec3.h"
#include "ChunkFile.h"
#ifdef WIN32
#define LOADERDECL __cdecl
@ -92,6 +93,18 @@ struct OutputVertexData
#undef LINTERP
#undef LINTERP_INT
}
void DoState(PointerWrap &p)
{
mvPosition.DoState(p);
p.Do(projectedPosition);
screenPosition.DoState(p);
for (int i = 0; i < 3;++i)
normal[i].DoState(p);
p.DoArray(color, sizeof color);
for (int i = 0; i < 8;++i)
texCoords[i].DoState(p);
}
};
#endif

View File

@ -35,7 +35,6 @@ typedef void (*DecodingFunction)(u32);
namespace OpcodeDecoder
{
static DecodingFunction currentFunction = NULL;
static u32 minCommandSize;
static u16 streamSize;
@ -46,6 +45,20 @@ static bool inObjectStream;
static u8 lastPrimCmd;
void DoState(PointerWrap &p)
{
p.Do(minCommandSize);
// Not sure what is wrong with this. Something(s) in here is causing dolphin to crash/hang when loading states saved from another run of dolphin. Doesn't seem too important anyway...
//vertexLoader.DoState(p);
p.Do(readOpcode);
p.Do(inObjectStream);
p.Do(lastPrimCmd);
p.Do(streamSize);
p.Do(streamAddress);
if (p.GetMode() == PointerWrap::MODE_READ)
ResetDecoding();
}
void DecodePrimitiveStream(u32 iBufferSize)
{
u32 vertexSize = vertexLoader.GetVertexSize();

View File

@ -20,6 +20,7 @@
#define _OPCODEDECODER_H_
#include "CommonTypes.h"
#include "ChunkFile.h"
namespace OpcodeDecoder
{
@ -57,6 +58,8 @@ namespace OpcodeDecoder
bool CommandRunnable(u32 iBufferSize);
void Run(u32 iBufferSize);
void DoState(PointerWrap &p);
}
#endif

View File

@ -23,6 +23,7 @@
#include "BPMemLoader.h"
#include "XFMemLoader.h"
#include "Tev.h"
#include "SWPixelEngine.h"
#include "SWStatistics.h"
#include "SWVideoConfig.h"
@ -62,6 +63,28 @@ s32 scissorBottom = 0;
Tev tev;
RasterBlock rasterBlock;
void DoState(PointerWrap &p)
{
ZSlope.DoState(p);
WSlope.DoState(p);
for (int i=0;i<2;++i)
for (int n=0; n<4; ++n)
ColorSlopes[i][n].DoState(p);
for (int i=0;i<8;++i)
for (int n=0; n<3; ++n)
TexSlopes[i][n].DoState(p);
p.Do(vertex0X);
p.Do(vertex0Y);
p.Do(vertexOffsetX);
p.Do(vertexOffsetY);
p.Do(scissorLeft);
p.Do(scissorTop);
p.Do(scissorRight);
p.Do(scissorBottom);
tev.DoState(p);
p.Do(rasterBlock);
}
void Init()
{
tev.Init();
@ -127,9 +150,15 @@ inline void Draw(s32 x, s32 y, s32 xi, s32 yi)
if (bpmem.zcontrol.early_ztest && bpmem.zmode.testenable && g_SWVideoConfig.bZComploc)
{
// early z
if (!EfbInterface::ZCompare(x, y, z))
return;
// TODO: Test if perf regs are incremented even if test is disabled
SWPixelEngine::pereg.IncZInputQuadCount(true);
if (bpmem.zmode.testenable)
{
// early z
if (!EfbInterface::ZCompare(x, y, z))
return;
}
SWPixelEngine::pereg.IncZOutputQuadCount(true);
}
RasterBlockPixel& pixel = rasterBlock.Pixel[xi][yi];

View File

@ -19,6 +19,7 @@
#define _RASTERIZER_H_
#include "NativeVertexFormat.h"
#include "ChunkFile.h"
namespace Rasterizer
{
@ -37,6 +38,12 @@ namespace Rasterizer
float f0;
float GetValue(float dx, float dy) { return f0 + (dfdx * dx) + (dfdy * dy); }
void DoState(PointerWrap &p)
{
p.Do(dfdx);
p.Do(dfdy);
p.Do(f0);
}
};
struct RasterBlockPixel
@ -53,6 +60,8 @@ namespace Rasterizer
s32 TextureLod[16];
bool TextureLinear[16];
};
void DoState(PointerWrap &p);
}

View File

@ -57,6 +57,15 @@ CPReg cpreg; // shared between gfx and emulator thread
void DoState(PointerWrap &p)
{
p.Do(cpreg);
p.DoArray(commandBuffer, commandBufferSize);
p.Do(readPos);
p.Do(writePos);
p.Do(et_UpdateInterrupts);
p.Do(interruptSet);
p.Do(interruptWaiting);
// Is this right?
p.DoArray(g_pVideoData,writePos);
}
// does it matter that there is no synchronization between threads during writes?

View File

@ -53,6 +53,8 @@ void DoState(PointerWrap &p)
p.Do(pereg);
p.Do(g_bSignalTokenInterrupt);
p.Do(g_bSignalFinishInterrupt);
p.Do(et_SetTokenOnMainThread);
p.Do(et_SetFinishOnMainThread);
}
void UpdateInterrupts();
@ -77,7 +79,7 @@ void Read16(u16& _uReturnValue, const u32 _iAddress)
u16 address = _iAddress & 0xFFF;
if (address <= 0x16)
if (address <= 0x2e)
_uReturnValue = ((u16*)&pereg)[address >> 1];
}
@ -109,7 +111,7 @@ void Write16(const u16 _iValue, const u32 _iAddress)
}
break;
default:
if (address <= 0x16)
if (address <= 0x2e)
((u16*)&pereg)[address >> 1] = _iValue;
break;
}

View File

@ -38,6 +38,21 @@ namespace SWPixelEngine
PE_BBOX_RIGHT = 0x012, // Flip Right
PE_BBOX_TOP = 0x014, // Flip Top
PE_BBOX_BOTTOM = 0x016, // Flip Bottom
// NOTE: Order not verified
// These indicate the number of quads that are being used as input/output for each particular stage
PE_PERF_ZCOMP_INPUT_ZCOMPLOC_L = 0x18,
PE_PERF_ZCOMP_INPUT_ZCOMPLOC_H = 0x1a,
PE_PERF_ZCOMP_OUTPUT_ZCOMPLOC_L = 0x1c,
PE_PERF_ZCOMP_OUTPUT_ZCOMPLOC_H = 0x1e,
PE_PERF_ZCOMP_INPUT_L = 0x20,
PE_PERF_ZCOMP_INPUT_H = 0x22,
PE_PERF_ZCOMP_OUTPUT_L = 0x24,
PE_PERF_ZCOMP_OUTPUT_H = 0x26,
PE_PERF_BLEND_INPUT_L = 0x28,
PE_PERF_BLEND_INPUT_H = 0x2a,
PE_PERF_EFB_COPY_CLOCKS_L = 0x2c,
PE_PERF_EFB_COPY_CLOCKS_H = 0x2e,
};
union UPEZConfReg
@ -125,10 +140,72 @@ namespace SWPixelEngine
UPECtrlReg ctrl;
u16 unk0;
u16 token;
u16 boxLeft;
u16 boxRight;
u16 boxTop;
u16 boxBottom;
u16 perfZcompInputZcomplocLo;
u16 perfZcompInputZcomplocHi;
u16 perfZcompOutputZcomplocLo;
u16 perfZcompOutputZcomplocHi;
u16 perfZcompInputLo;
u16 perfZcompInputHi;
u16 perfZcompOutputLo;
u16 perfZcompOutputHi;
u16 perfBlendInputLo;
u16 perfBlendInputHi;
u16 perfEfbCopyClocksLo;
u16 perfEfbCopyClocksHi;
// NOTE: hardware doesn't process individual pixels but quads instead. Current software renderer architecture works on pixels though, so we have this "quad" hack here to only increment the registers on every fourth rendered pixel
void IncZInputQuadCount(bool early_ztest)
{
static int quad = 0;
if (++quad != 3)
return;
quad = 0;
if (early_ztest)
{
if (++perfZcompInputZcomplocLo == 0)
perfZcompInputZcomplocHi++;
}
else
{
if (++perfZcompInputLo == 0)
perfZcompInputHi++;
}
}
void IncZOutputQuadCount(bool early_ztest)
{
static int quad = 0;
if (++quad != 3)
return;
quad = 0;
if (early_ztest)
{
if (++perfZcompOutputZcomplocLo == 0)
perfZcompOutputZcomplocHi++;
}
else
{
if (++perfZcompOutputLo == 0)
perfZcompOutputHi++;
}
}
void IncBlendInputQuadCount()
{
static int quad = 0;
if (++quad != 3)
return;
quad = 0;
if (++perfBlendInputLo == 0)
perfBlendInputHi++;
}
};
extern PEReg pereg;

View File

@ -328,4 +328,15 @@ void SWVertexLoader::LoadTexCoord(SWVertexLoader *vertexLoader, InputVertexData
vertexLoader->m_texCoordLoader[index]();
}
void SWVertexLoader::DoState(PointerWrap &p)
{
p.DoArray(m_AttributeLoaders, sizeof m_AttributeLoaders);
p.Do(m_VertexSize);
p.Do(*m_CurrentVat);
p.Do(m_positionLoader);
p.Do(m_normalLoader);
p.DoArray(m_colorLoader, sizeof m_colorLoader);
p.Do(m_NumAttributeLoaders);
m_SetupUnit->DoState(p);
p.Do(m_TexGenSpecialCase);
}

View File

@ -22,6 +22,7 @@
#include "NativeVertexFormat.h"
#include "CPMemLoader.h"
#include "ChunkFile.h"
class SetupUnit;
@ -69,7 +70,7 @@ public:
u32 GetVertexSize() { return m_VertexSize; }
void LoadVertex();
void DoState(PointerWrap &p);
};
#endif

View File

@ -39,6 +39,9 @@
#include "FileUtil.h"
#include "VideoBackend.h"
#include "Core.h"
#include "OpcodeDecoder.h"
#include "SWVertexLoader.h"
#include "SWStatistics.h"
#define VSYNC_ENABLED 0
@ -93,9 +96,33 @@ bool VideoSoftware::Initialize(void *&window_handle)
return true;
}
void VideoSoftware::DoState(PointerWrap&)
void VideoSoftware::DoState(PointerWrap& p)
{
// NYI
bool software = true;
p.Do(software);
if (p.GetMode() == PointerWrap::MODE_READ && software == false)
// change mode to abort load of incompatible save state.
p.SetMode(PointerWrap::MODE_VERIFY);
// TODO: incomplete?
SWCommandProcessor::DoState(p);
SWPixelEngine::DoState(p);
EfbInterface::DoState(p);
OpcodeDecoder::DoState(p);
Clipper::DoState(p);
p.Do(swxfregs);
p.Do(bpmem);
p.Do(swstats);
// CP Memory
p.DoArray(arraybases, 16);
p.DoArray(arraystrides, 16);
p.Do(MatrixIndexA);
p.Do(MatrixIndexB);
p.Do(g_VtxDesc.Hex);
p.DoArray(g_VtxAttr, 8);
p.DoMarker("CP Memory");
}
void VideoSoftware::CheckInvalidState()
@ -198,6 +225,12 @@ u32 VideoSoftware::Video_AccessEFB(EFBAccessType type, u32 x, u32 y, u32 InputDa
return value;
}
u32 VideoSoftware::Video_GetQueryResult(PerfQueryType type)
{
// TODO:
return 0;
}
bool VideoSoftware::Video_Screenshot(const char *_szFilename)
{
return false;

View File

@ -25,13 +25,13 @@
void SetupUnit::Init(u8 primitiveType)
{
m_PrimType = primitiveType;
m_PrimType = primitiveType;
m_VertexCounter = 0;
m_VertPointer[0] = &m_Vertices[0];
m_VertPointer[1] = &m_Vertices[1];
m_VertPointer[2] = &m_Vertices[2];
m_VertWritePointer = m_VertPointer[0];
m_VertexCounter = 0;
m_VertPointer[0] = &m_Vertices[0];
m_VertPointer[1] = &m_Vertices[1];
m_VertPointer[2] = &m_Vertices[2];
m_VertWritePointer = m_VertPointer[0];
}
void SetupUnit::SetupVertex()
@ -169,3 +169,21 @@ void SetupUnit::SetupLineStrip()
void SetupUnit::SetupPoint()
{}
void SetupUnit::DoState(PointerWrap &p)
{
// TODO: some or all of this is making the save states stop working once dolphin is closed...sometimes (usually)
// I have no idea what specifically is wrong, or if this is even important. Disabling it doesn't seem to make any noticible difference...
/* p.Do(m_PrimType);
p.Do(m_VertexCounter);
for (int i = 0; i < 3; ++i)
m_Vertices[i].DoState(p);
if (p.GetMode() == PointerWrap::MODE_READ)
{
m_VertPointer[0] = &m_Vertices[0];
m_VertPointer[1] = &m_Vertices[1];
m_VertPointer[2] = &m_Vertices[2];
m_VertWritePointer = m_VertPointer[0];
}*/
}

View File

@ -21,6 +21,7 @@
#include "Common.h"
#include "NativeVertexFormat.h"
#include "ChunkFile.h"
class SetupUnit
{
@ -45,6 +46,7 @@ public:
OutputVertexData* GetVertex() { return m_VertWritePointer; }
void SetupVertex();
void DoState(PointerWrap &p);
};
#endif

View File

@ -20,6 +20,7 @@
#include "Tev.h"
#include "EfbInterface.h"
#include "TextureSampler.h"
#include "SWPixelEngine.h"
#include "SWStatistics.h"
#include "SWVideoConfig.h"
#include "DebugUtil.h"
@ -787,8 +788,13 @@ void Tev::Draw()
bool late_ztest = !bpmem.zcontrol.early_ztest || !g_SWVideoConfig.bZComploc;
if (late_ztest && bpmem.zmode.testenable)
{
if (!EfbInterface::ZCompare(Position[0], Position[1], Position[2]))
return;
// TODO: Check against hw if these values get incremented even if depth testing is disabled
SWPixelEngine::pereg.IncZInputQuadCount(false);
if (!EfbInterface::ZCompare(Position[0], Position[1], Position[2]))
return;
SWPixelEngine::pereg.IncZOutputQuadCount(false);
}
#if ALLOW_TEV_DUMPS
@ -812,6 +818,7 @@ void Tev::Draw()
#endif
INCSTAT(swstats.thisFrame.tevPixelsOut);
SWPixelEngine::pereg.IncBlendInputQuadCount();
EfbInterface::BlendTev(Position[0], Position[1], output);
}
@ -827,3 +834,31 @@ void Tev::SetRegColor(int reg, int comp, bool konst, s16 color)
Reg[reg][comp] = color;
}
}
void Tev::DoState(PointerWrap &p)
{
p.DoArray(Reg, sizeof(Reg));
p.DoArray(KonstantColors, sizeof(KonstantColors));
p.DoArray(TexColor,4);
p.DoArray(RasColor,4);
p.DoArray(StageKonst,4);
p.DoArray(Zero16,4);
p.DoArray(FixedConstants,9);
p.Do(AlphaBump);
p.DoArray(IndirectTex, sizeof(IndirectTex));
p.Do(TexCoord);
p.DoArray(m_BiasLUT,4);
p.DoArray(m_ScaleLShiftLUT,4);
p.DoArray(m_ScaleRShiftLUT,4);
p.DoArray(Position,3);
p.DoArray(Color, sizeof(Color));
p.DoArray(Uv, 8);
p.DoArray(IndirectLod,4);
p.DoArray(IndirectLinear,4);
p.DoArray(TextureLod,16);
p.DoArray(TextureLinear,16);
}

View File

@ -19,6 +19,7 @@
#define _TEV_H_
#include "BPMemLoader.h"
#include "ChunkFile.h"
class Tev
{
@ -96,6 +97,8 @@ public:
void SetRegColor(int reg, int comp, bool konst, s16 color);
enum { ALP_C, BLU_C, GRN_C, RED_C };
void DoState(PointerWrap &p);
};
#endif

View File

@ -20,6 +20,7 @@
#include <stdlib.h>
#include <math.h>
#include "ChunkFile.h"
class Vec3
{
@ -111,6 +112,12 @@ public:
{
memset((void *)this,0,sizeof(float)*3);
}
void DoState(PointerWrap &p)
{
p.Do(x);
p.Do(y);
p.Do(z);
}
};
#endif

View File

@ -26,7 +26,9 @@ class VideoSoftware : public VideoBackend
void Video_ExitLoop();
void Video_BeginField(u32, FieldType, u32, u32);
void Video_EndField();
u32 Video_AccessEFB(EFBAccessType, u32, u32, u32);
u32 Video_GetQueryResult(PerfQueryType type);
void Video_AddMessage(const char* pstr, unsigned int milliseconds);
void Video_ClearMessages();