Merge branch 'master' into vertex-loader-cleanup

Conflicts:
	Source/Core/Common/Src/CommonFuncs.h
	Source/Core/VideoCommon/Src/VertexLoader.cpp
This commit is contained in:
Jordan Woyak
2013-03-04 15:38:43 -06:00
289 changed files with 11758 additions and 2519 deletions

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

@ -51,7 +51,9 @@ void VideoBackend::UpdateFPSDisplay(const char *text)
}
void InitInterface()
{
#if defined(USE_EGL) && USE_EGL
#ifdef ANDROID
GLInterface = new cInterfaceBase;
#elif defined(USE_EGL) && USE_EGL
GLInterface = new cInterfaceEGL;
#elif defined(USE_WX) && USE_WX
GLInterface = new cInterfaceWX;

View File

@ -69,7 +69,7 @@ bool OpenGL_ReportFBOError(const char *function, const char *file, int line);
#define GL_REPORT_PROGRAM_ERROR() (void)0
#endif
#if defined __APPLE__ || defined __linux__ || defined _WIN32
#if (defined __APPLE__ || defined __linux__ || defined _WIN32) && !(defined _M_ARM)
#include <Cg/cg.h>
#include <Cg/cgGL.h>
#define HAVE_CG 1

View File

@ -17,7 +17,7 @@
#include "GLUtil.h"
#include "x64Emitter.h"
#include "ABI.h"
#include "x64ABI.h"
#include "MemoryUtil.h"
#include "VertexShaderGen.h"

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 (u32 i = 0; i != ARRAYSIZE(m_query_buffer); ++i)
glGenQueries(1, &m_query_buffer[i].query_id);
ResetQuery();
}
PerfQuery::~PerfQuery()
{
for (u32 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 u32 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];
u32 m_query_read_pos;
// TODO: sloppy
volatile u32 m_query_count;
volatile u32 m_results[PQG_NUM_MEMBERS];
};
} // namespace
#endif // _PERFQUERY_H_

View File

@ -264,7 +264,8 @@ bool PixelShaderCache::CompilePixelShader(FRAGMENTSHADER& ps, const char* pstrpr
static int num_failures = 0;
char szTemp[MAX_PATH];
sprintf(szTemp, "%sbad_ps_%04i.txt", File::GetUserPath(D_DUMP_IDX).c_str(), num_failures++);
std::ofstream file(szTemp);
std::ofstream file;
OpenFStream(file, szTemp, std::ios_base::out);
file << pstrprogram;
file.close();

View File

@ -24,6 +24,7 @@
#include <cstdio>
#include "GLUtil.h"
#include "WxUtils.h"
#include "FileUtil.h"
@ -956,20 +957,23 @@ void Renderer::SetBlendMode(bool forceUpdate)
s_blendMode = newval;
}
void DumpFrame(const std::vector<u8>& data, int w, int h)
{
#if defined(HAVE_LIBAV) || defined(_WIN32)
if (g_ActiveConfig.bDumpFrames && !data.empty())
{
AVIDump::AddFrame(&data[0], w, h);
}
#endif
}
// This function has the final picture. We adjust the aspect ratio here.
void Renderer::Swap(u32 xfbAddr, FieldType field, u32 fbWidth, u32 fbHeight,const EFBRectangle& rc,float Gamma)
{
static int w = 0, h = 0;
if (g_bSkipCurrentFrame || (!XFBWrited && !g_ActiveConfig.RealXFBEnabled()) || !fbWidth || !fbHeight)
{
if (g_ActiveConfig.bDumpFrames && frame_data)
{
#ifdef _WIN32
AVIDump::AddFrame(frame_data);
#elif defined HAVE_LIBAV
AVIDump::AddFrame((u8*)frame_data, w, h);
#endif
}
DumpFrame(frame_data, w, h);
Core::Callback_VideoCopiedToXFB(false);
return;
}
@ -979,14 +983,7 @@ void Renderer::Swap(u32 xfbAddr, FieldType field, u32 fbWidth, u32 fbHeight,cons
const XFBSourceBase* const* xfbSourceList = FramebufferManager::GetXFBSource(xfbAddr, fbWidth, fbHeight, xfbCount);
if (g_ActiveConfig.VirtualXFBEnabled() && (!xfbSourceList || xfbCount == 0))
{
if (g_ActiveConfig.bDumpFrames && frame_data)
{
#ifdef _WIN32
AVIDump::AddFrame(frame_data);
#elif defined HAVE_LIBAV
AVIDump::AddFrame((u8*)frame_data, w, h);
#endif
}
DumpFrame(frame_data, w, h);
Core::Callback_VideoCopiedToXFB(false);
return;
}
@ -1149,16 +1146,15 @@ void Renderer::Swap(u32 xfbAddr, FieldType field, u32 fbWidth, u32 fbHeight,cons
if (g_ActiveConfig.bDumpFrames)
{
std::lock_guard<std::mutex> lk(s_criticalScreenshot);
if (!frame_data || w != flipped_trc.GetWidth() ||
if (frame_data.empty() || w != flipped_trc.GetWidth() ||
h != flipped_trc.GetHeight())
{
if (frame_data) delete[] frame_data;
w = flipped_trc.GetWidth();
h = flipped_trc.GetHeight();
frame_data = new char[3 * w * h];
frame_data.resize(3 * w * h);
}
glPixelStorei(GL_PACK_ALIGNMENT, 1);
glReadPixels(flipped_trc.left, flipped_trc.bottom, w, h, GL_BGR, GL_UNSIGNED_BYTE, frame_data);
glReadPixels(flipped_trc.left, flipped_trc.bottom, w, h, GL_BGR, GL_UNSIGNED_BYTE, &frame_data[0]);
if (GL_REPORT_ERROR() == GL_NO_ERROR && w > 0 && h > 0)
{
if (!bLastFrameDumped)
@ -1179,12 +1175,11 @@ void Renderer::Swap(u32 xfbAddr, FieldType field, u32 fbWidth, u32 fbHeight,cons
}
if (bAVIDumping)
{
#ifdef _WIN32
AVIDump::AddFrame(frame_data);
#else
FlipImageData((u8*)frame_data, w, h);
AVIDump::AddFrame((u8*)frame_data, w, h);
#ifndef _WIN32
FlipImageData(&frame_data[0], w, h);
#endif
AVIDump::AddFrame(&frame_data[0], w, h);
}
bLastFrameDumped = true;
@ -1196,12 +1191,8 @@ void Renderer::Swap(u32 xfbAddr, FieldType field, u32 fbWidth, u32 fbHeight,cons
{
if (bLastFrameDumped && bAVIDumping)
{
if (frame_data)
{
delete[] frame_data;
frame_data = NULL;
w = h = 0;
}
std::vector<u8>().swap(frame_data);
w = h = 0;
AVIDump::Stop();
bAVIDumping = false;
OSD::AddMessage("Stop dumping frames", 2000);
@ -1215,9 +1206,9 @@ void Renderer::Swap(u32 xfbAddr, FieldType field, u32 fbWidth, u32 fbHeight,cons
std::string movie_file_name;
w = GetTargetRectangle().GetWidth();
h = GetTargetRectangle().GetHeight();
frame_data = new char[3 * w * h];
frame_data.resize(3 * w * h);
glPixelStorei(GL_PACK_ALIGNMENT, 1);
glReadPixels(GetTargetRectangle().left, GetTargetRectangle().bottom, w, h, GL_BGR, GL_UNSIGNED_BYTE, frame_data);
glReadPixels(GetTargetRectangle().left, GetTargetRectangle().bottom, w, h, GL_BGR, GL_UNSIGNED_BYTE, &frame_data[0]);
if (GL_REPORT_ERROR() == GL_NO_ERROR)
{
if (!bLastFrameDumped)
@ -1228,21 +1219,17 @@ void Renderer::Swap(u32 xfbAddr, FieldType field, u32 fbWidth, u32 fbHeight,cons
OSD::AddMessage("Error opening framedump.raw for writing.", 2000);
else
{
char msg [255];
sprintf(msg, "Dumping Frames to \"%s\" (%dx%d RGB24)", movie_file_name.c_str(), w, h);
OSD::AddMessage(msg, 2000);
OSD::AddMessage(StringFromFormat("Dumping Frames to \"%s\" (%dx%d RGB24)", movie_file_name.c_str(), w, h).c_str(), 2000);
}
}
if (pFrameDump)
{
FlipImageData((u8*)frame_data, w, h);
pFrameDump.WriteBytes(frame_data, w * 3 * h);
FlipImageData(&frame_data[0], w, h);
pFrameDump.WriteBytes(&frame_data[0], w * 3 * h);
pFrameDump.Flush();
}
bLastFrameDumped = true;
}
delete[] frame_data;
}
else
{
@ -1433,6 +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);
}
@ -1552,7 +1540,7 @@ void TakeScreenshot(ScrStrct* threadStruct)
// Save the screenshot and finally kill the wxImage object
// This is really expensive when saving to PNG, but not at all when using BMP
threadStruct->img->SaveFile(wxString::FromAscii(threadStruct->filename.c_str()),
threadStruct->img->SaveFile(StrToWxStr(threadStruct->filename),
wxBITMAP_TYPE_PNG);
threadStruct->img->Destroy();

View File

@ -40,6 +40,7 @@
#include "OpcodeDecoding.h"
#include "FileUtil.h"
#include "Debugger.h"
#include "PerfQueryBase.h"
#include "main.h"
@ -213,7 +214,10 @@ 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)
@ -243,10 +247,12 @@ void VertexManager::vFlush()
// save the shaders
char strfile[255];
sprintf(strfile, "%sps%.3d.txt", File::GetUserPath(D_DUMPFRAMES_IDX).c_str(), g_ActiveConfig.iSaveTargetId);
std::ofstream fps(strfile);
std::ofstream fps;
OpenFStream(fps, strfile, std::ios_base::out);
fps << ps->strprog.c_str();
sprintf(strfile, "%svs%.3d.txt", File::GetUserPath(D_DUMPFRAMES_IDX).c_str(), g_ActiveConfig.iSaveTargetId);
std::ofstream fvs(strfile);
std::ofstream fvs;
OpenFStream(fvs, strfile, std::ios_base::out);
fvs << vs->strprog.c_str();
}

View File

@ -144,7 +144,8 @@ bool VertexShaderCache::CompileVertexShader(VERTEXSHADER& vs, const char* pstrpr
static int num_failures = 0;
char szTemp[MAX_PATH];
sprintf(szTemp, "%sbad_vs_%04i.txt", File::GetUserPath(D_DUMP_IDX).c_str(), num_failures++);
std::ofstream file(szTemp);
std::ofstream file;
OpenFStream(file, szTemp, std::ios_base::out);
file << pstrprogram;
file.close();

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();
@ -204,7 +206,9 @@ void VideoBackend::Video_Prepare()
GL_REPORT_ERRORD();
VertexLoaderManager::Init();
TextureConverter::Init();
#ifndef _M_GENERIC
DLCache::Init();
#endif
// Notify the core that the video backend is ready
Host_Message(WM_USER_CREATE);
@ -219,7 +223,9 @@ void VideoBackend::Shutdown()
s_efbAccessRequested = false;
s_FifoShuttingDown = false;
s_swapRequested = false;
#ifndef _M_GENERIC
DLCache::Shutdown();
#endif
Fifo_Shutdown();
PostProcessing::Shutdown();