int main(int argc, char *argv[])
{
int i;
matrix_t mat[2]; /* A, B*/
matrix_t *C;
int m;
for (i = 0; i < 2; i++) {
str2mat(argv[i+1], &(mat[i]), ":", ",");
}
C = pmat(&(mat[0]), &(mat[1]));
for (i = 0; i < 2; i++) {
if (i == 0) {
printf("A=\n");
} else if (i == 1){
printf("B=\n");
}
for (m = 0; m < mat[i].row; m++) {
pdarray((mat[i].value)[m], mat[i].column);
}
}
printf("A*B=\n");
for (m = 0; m < C->row; m++) {
pdarray((C->value)[m], C->column);
}
return 0;
}
void Ssao::runMain(const RenderingContext& ctx, RenderPassWorkContext& rgraphCtx)
{
CommandBufferPtr& cmdb = rgraphCtx.m_commandBuffer;
cmdb->bindShaderProgram(m_main.m_grProg);
// Bind resources
cmdb->bindSampler(0, 0, m_r->getSamplers().m_trilinearClamp);
cmdb->bindSampler(0, 1, m_r->getSamplers().m_trilinearRepeat);
rgraphCtx.bindTexture(0, 2, m_r->getDepthDownscale().getHiZRt(), HIZ_HALF_DEPTH);
cmdb->bindTexture(0, 3, m_main.m_noiseTex->getGrTextureView(), TextureUsageBit::SAMPLED_FRAGMENT);
if(m_useNormal)
{
rgraphCtx.bindColorTexture(0, 4, m_r->getGBuffer().getColorRt(2));
}
struct Unis
{
Vec4 m_unprojectionParams;
Vec4 m_projectionMat;
Mat3x4 m_viewRotMat;
} unis;
const Mat4& pmat = ctx.m_renderQueue->m_projectionMatrix;
unis.m_unprojectionParams = ctx.m_unprojParams;
unis.m_projectionMat = Vec4(pmat(0, 0), pmat(1, 1), pmat(2, 2), pmat(2, 3));
unis.m_viewRotMat = Mat3x4(ctx.m_renderQueue->m_viewMatrix.getRotationPart());
cmdb->setPushConstants(&unis, sizeof(unis));
if(m_useCompute)
{
rgraphCtx.bindImage(0, 5, m_runCtx.m_rts[0], TextureSubresourceInfo());
const U sizeX = (m_width + m_workgroupSize[0] - 1) / m_workgroupSize[0];
const U sizeY = (m_height + m_workgroupSize[1] - 1) / m_workgroupSize[1];
cmdb->dispatchCompute(sizeX, sizeY, 1);
}
else
{
cmdb->setViewport(0, 0, m_width, m_height);
drawQuad(cmdb);
}
}
void f()
{ int i,j;
double A[M*N];
Assignment result;
for( i=0; i<M*N; i++) A[i] = fabs( ((int)rand())/1024./1024./1024. );
//printf("*************\n");
//pmat( A, M,N );
result = match( A, M, N );
printf("----\n");
pmat( A, M,N );
for( i=0; i<result.n; i++ )
printf( "%d --> %g\n", i, result.assignment[i] );
printf("Cost: %g\n", result.cost);
}
//#define BUG 0
void fdjac2(void (*fcn)(int, int, double[], double[], int *, void *),
int m,
int n,
double x[],
double fvec[],
double fjac[],
int ldfjac,
int *iflag,
double epsfcn,
double wa[],
void *data)
{
int i;
int j;
int ij;
double eps;
double h;
double temp;
//static double zero = 0.0;
//dkb - changed to equal TvdB solver code, use constant delta = 1e-5
// temp = dmax1(epsfcn,MACHEP);
// eps = sqrt(temp);
JACOBIAN = 1;
eps = 1e-5;
#ifdef BUG
printf( "fdjac2\n" );
#endif
ij = 0;
for (j=0; j<n; j++)
{
temp = x[j];
// h = eps * fabs(temp);
// if (h == zero)
// h = eps;
h = eps; // added dkb
x[j] = temp + h;
fcn(m,n,x,wa,iflag, data);
x[j] = temp;
if (*iflag < 0)
return;
for (i=0; i<m; i++)
{
fjac[ij] = (wa[i] - fvec[i])/h;
ij += 1; /* fjac[i+m*j] */
}
}
#ifdef BUG
printf("jacobian:\n");
pmat( m, n, fjac );
for (i = 0; i < m*n; i++)
printf("%6e ", fjac[i]);
#endif
JACOBIAN = 0;
}
void GSRenderer::VSync(int field)
{
GSPerfMonAutoTimer pmat(&m_perfmon);
m_perfmon.Put(GSPerfMon::Frame);
Flush();
if(s_dump && s_n >= s_saven)
{
m_regs->Dump(root_sw + format("%05d_f%lld_gs_reg.txt", s_n, m_perfmon.GetFrame()));
}
if(!m_dev->IsLost(true))
{
if(!Merge(field ? 1 : 0))
{
return;
}
}
else
{
ResetDevice();
}
m_dev->AgePool();
// osd
if((m_perfmon.GetFrame() & 0x1f) == 0)
{
m_perfmon.Update();
double fps = 1000.0f / m_perfmon.Get(GSPerfMon::Frame);
string s;
#ifdef GSTITLEINFO_API_FORCE_VERBOSE
if(1)//force verbose reply
#else
if(m_wnd->IsManaged())
#endif
{
//GSdx owns the window's title, be verbose.
string s2 = m_regs->SMODE2.INT ? (string("Interlaced ") + (m_regs->SMODE2.FFMD ? "(frame)" : "(field)")) : "Progressive";
s = format(
"%lld | %d x %d | %.2f fps (%d%%) | %s - %s | %s | %d S/%d P/%d D | %d%% CPU | %.2f | %.2f",
m_perfmon.GetFrame(), GetInternalResolution().x, GetInternalResolution().y, fps, (int)(100.0 * fps / GetTvRefreshRate()),
s2.c_str(),
theApp.m_gs_interlace[m_interlace].name.c_str(),
theApp.m_gs_aspectratio[m_aspectratio].name.c_str(),
(int)m_perfmon.Get(GSPerfMon::SyncPoint),
(int)m_perfmon.Get(GSPerfMon::Prim),
(int)m_perfmon.Get(GSPerfMon::Draw),
m_perfmon.CPU(),
m_perfmon.Get(GSPerfMon::Swizzle) / 1024,
m_perfmon.Get(GSPerfMon::Unswizzle) / 1024
);
double fillrate = m_perfmon.Get(GSPerfMon::Fillrate);
if(fillrate > 0)
{
s += format(" | %.2f mpps", fps * fillrate / (1024 * 1024));
int sum = 0;
for(int i = 0; i < 16; i++)
{
sum += m_perfmon.CPU(GSPerfMon::WorkerDraw0 + i);
}
s += format(" | %d%% CPU", sum);
}
}
else
{
// Satisfy PCSX2's request for title info: minimal verbosity due to more external title text
s = format("%dx%d | %s", GetInternalResolution().x, GetInternalResolution().y, theApp.m_gs_interlace[m_interlace].name.c_str());
}
if(m_capture.IsCapturing())
{
s += " | Recording...";
}
if(m_wnd->IsManaged())
{
m_wnd->SetWindowText(s.c_str());
}
else
{
// note: do not use TryEnterCriticalSection. It is unnecessary code complication in
// an area that absolutely does not matter (even if it were 100 times slower, it wouldn't
// be noticeable). Besides, these locks are extremely short -- overhead of conditional
// is way more expensive than just waiting for the CriticalSection in 1 of 10,000,000 tries. --air
std::lock_guard<std::mutex> lock(m_pGSsetTitle_Crit);
strncpy(m_GStitleInfoBuffer, s.c_str(), countof(m_GStitleInfoBuffer) - 1);
m_GStitleInfoBuffer[sizeof(m_GStitleInfoBuffer) - 1] = 0; // make sure null terminated even if text overflows
}
}
else
{
// [TODO]
// We don't have window title rights, or the window has no title,
// so let's use actual OSD!
}
if(m_frameskip)
{
return;
}
// present
#if 0
// This will scale the OSD to the PS2's output resolution.
// Will be affected by 2x, 4x, etc scaling.
m_dev->m_osd.m_real_size = m_real_size
#elif 0
// This will scale the OSD to the window's size.
// Will maintiain the font size no matter what size the window is.
GSVector4i window_size = m_wnd->GetClientRect();
m_dev->m_osd.m_real_size.x = window_size.v[2];
m_dev->m_osd.m_real_size.y = window_size.v[3];
#else
// This will scale the OSD to the native resolution.
// Will size font relative to the window's size.
// TODO this should probably be done with native calls
m_dev->m_osd.m_real_size.x = 1024;
m_dev->m_osd.m_real_size.y = 768;
#endif
m_dev->Present(m_wnd->GetClientRect().fit(m_aspectratio), m_shader);
// snapshot
if(!m_snapshot.empty())
{
if(!m_dump && m_shift_key)
{
GSFreezeData fd = {0, nullptr};
Freeze(&fd, true);
fd.data = new uint8[fd.size];
Freeze(&fd, false);
#ifdef LZMA_SUPPORTED
if (m_control_key)
m_dump = std::unique_ptr<GSDumpBase>(new GSDump(m_snapshot, m_crc, fd, m_regs));
else
m_dump = std::unique_ptr<GSDumpBase>(new GSDumpXz(m_snapshot, m_crc, fd, m_regs));
#else
m_dump = std::unique_ptr<GSDumpBase>(new GSDump(m_snapshot, m_crc, fd, m_regs));
#endif
delete [] fd.data;
}
if(GSTexture* t = m_dev->GetCurrent())
{
t->Save(m_snapshot + ".bmp");
}
m_snapshot.clear();
}
else if(m_dump)
{
if(m_dump->VSync(field, !m_control_key, m_regs))
m_dump.reset();
}
// capture
if(m_capture.IsCapturing())
{
if(GSTexture* current = m_dev->GetCurrent())
{
GSVector2i size = m_capture.GetSize();
if(GSTexture* offscreen = m_dev->CopyOffscreen(current, GSVector4(0, 0, 1, 1), size.x, size.y))
{
GSTexture::GSMap m;
if(offscreen->Map(m))
{
m_capture.DeliverFrame(m.bits, m.pitch, !m_dev->IsRBSwapped());
offscreen->Unmap();
}
m_dev->Recycle(offscreen);
}
}
}
}
void GSRenderer::VSync(int field)
{
GSPerfMonAutoTimer pmat(&m_perfmon);
m_perfmon.Put(GSPerfMon::Frame);
Flush();
if(!m_dev->IsLost(true))
{
if(!Merge(field ? 1 : 0))
{
return;
}
}
else
{
ResetDevice();
}
m_dev->AgePool();
// osd
if((m_perfmon.GetFrame() & 0x1f) == 0)
{
m_perfmon.Update();
double fps = 1000.0f / m_perfmon.Get(GSPerfMon::Frame);
string s;
#ifdef GSTITLEINFO_API_FORCE_VERBOSE
if(1)//force verbose reply
#else
if(m_wnd->IsManaged())
#endif
{
//GSdx owns the window's title, be verbose.
string s2 = m_regs->SMODE2.INT ? (string("Interlaced ") + (m_regs->SMODE2.FFMD ? "(frame)" : "(field)")) : "Progressive";
s = format(
"%lld | %d x %d | %.2f fps (%d%%) | %s - %s | %s | %d S/%d P/%d D | %d%% CPU | %.2f | %.2f",
m_perfmon.GetFrame(), GetInternalResolution().x, GetInternalResolution().y, fps, (int)(100.0 * fps / GetTvRefreshRate()),
s2.c_str(),
theApp.m_gs_interlace[m_interlace].name.c_str(),
theApp.m_gs_aspectratio[m_aspectratio].name.c_str(),
(int)m_perfmon.Get(GSPerfMon::SyncPoint),
(int)m_perfmon.Get(GSPerfMon::Prim),
(int)m_perfmon.Get(GSPerfMon::Draw),
m_perfmon.CPU(),
m_perfmon.Get(GSPerfMon::Swizzle) / 1024,
m_perfmon.Get(GSPerfMon::Unswizzle) / 1024
);
double fillrate = m_perfmon.Get(GSPerfMon::Fillrate);
if(fillrate > 0)
{
s += format(" | %.2f mpps", fps * fillrate / (1024 * 1024));
int sum = 0;
for(int i = 0; i < 16; i++)
{
sum += m_perfmon.CPU(GSPerfMon::WorkerDraw0 + i);
}
s += format(" | %d%% CPU", sum);
}
}
else
{
// Satisfy PCSX2's request for title info: minimal verbosity due to more external title text
s = format("%dx%d | %s", GetInternalResolution().x, GetInternalResolution().y, theApp.m_gs_interlace[m_interlace].name.c_str());
}
if(m_capture.IsCapturing())
{
s += " | Recording...";
}
if(m_wnd->IsManaged())
{
m_wnd->SetWindowText(s.c_str());
}
else
{
// note: do not use TryEnterCriticalSection. It is unnecessary code complication in
// an area that absolutely does not matter (even if it were 100 times slower, it wouldn't
// be noticeable). Besides, these locks are extremely short -- overhead of conditional
// is way more expensive than just waiting for the CriticalSection in 1 of 10,000,000 tries. --air
std::lock_guard<std::mutex> lock(m_pGSsetTitle_Crit);
strncpy(m_GStitleInfoBuffer, s.c_str(), countof(m_GStitleInfoBuffer) - 1);
m_GStitleInfoBuffer[sizeof(m_GStitleInfoBuffer) - 1] = 0; // make sure null terminated even if text overflows
}
}
else
{
// [TODO]
// We don't have window title rights, or the window has no title,
// so let's use actual OSD!
}
if(m_frameskip)
{
return;
}
// present
m_dev->Present(m_wnd->GetClientRect().fit(m_aspectratio), m_shader);
// snapshot
if(!m_snapshot.empty())
{
bool shift = false;
#ifdef _WIN32
shift = !!(::GetAsyncKeyState(VK_SHIFT) & 0x8000);
#else
shift = m_shift_key;
#endif
if(!m_dump && shift)
{
GSFreezeData fd;
fd.size = 0;
fd.data = NULL;
Freeze(&fd, true);
fd.data = new uint8[fd.size];
Freeze(&fd, false);
m_dump.Open(m_snapshot, m_crc, fd, m_regs);
delete [] fd.data;
}
if(GSTexture* t = m_dev->GetCurrent())
{
t->Save(m_snapshot + ".bmp");
}
m_snapshot.clear();
}
else
{
if(m_dump)
{
bool control = false;
#ifdef _WIN32
control = !!(::GetAsyncKeyState(VK_CONTROL) & 0x8000);
#else
control = m_control_key;
#endif
m_dump.VSync(field, !control, m_regs);
}
}
// capture
if(m_capture.IsCapturing())
{
if(GSTexture* current = m_dev->GetCurrent())
{
GSVector2i size = m_capture.GetSize();
if(GSTexture* offscreen = m_dev->CopyOffscreen(current, GSVector4(0, 0, 1, 1), size.x, size.y))
{
GSTexture::GSMap m;
if(offscreen->Map(m))
{
m_capture.DeliverFrame(m.bits, m.pitch, !m_dev->IsRBSwapped());
offscreen->Unmap();
}
m_dev->Recycle(offscreen);
}
}
}
}
void DisplayDeviceOpenGL::render(const Renderable* r) const
{
if(!r->isEnabled()) {
// Renderable item not enabled then early return.
return;
}
StencilScopePtr stencil_scope;
if(r->hasClipSettings()) {
ModelManager2D mm(r->getPosition().x, r->getPosition().y);
auto clip_shape = r->getStencilMask();
bool cam_set = false;
if(clip_shape->getCamera() == nullptr && r->getCamera() != nullptr) {
cam_set = true;
clip_shape->setCamera(r->getCamera());
}
stencil_scope.reset(new StencilScopeOGL(r->getStencilSettings()));
glColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE);
glDepthMask(GL_FALSE);
glClear(GL_STENCIL_BUFFER_BIT);
render(clip_shape.get());
stencil_scope->applyNewSettings(keep_stencil_settings);
glColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
glDepthMask(GL_TRUE);
if(cam_set) {
clip_shape->setCamera(nullptr);
}
}
auto shader = r->getShader();
shader->makeActive();
BlendEquationScopeOGL be_scope(*r);
BlendModeScopeOGL bm_scope(*r);
// apply lighting/depth check/depth write here.
bool use_lighting = r->isLightingStateSet() ? r->useLighting() : false;
// Set the depth enable.
if(r->isDepthEnableStateSet()) {
if(get_current_depth_enable() != r->isDepthEnabled()) {
if(r->isDepthEnabled()) {
glEnable(GL_DEPTH_TEST);
} else {
glDisable(GL_DEPTH_TEST);
}
get_current_depth_enable() = r->isDepthEnabled();
}
} else {
// We assume that depth is disabled if not specified.
if(get_current_depth_enable() == true) {
glDisable(GL_DEPTH_TEST);
get_current_depth_enable() = false;
}
}
glm::mat4 pmat(1.0f);
glm::mat4 vmat(1.0f);
if(r->getCamera()) {
// set camera here.
pmat = r->getCamera()->getProjectionMat();
vmat = r->getCamera()->getViewMat();
} else if(get_default_camera() != nullptr) {
pmat = get_default_camera()->getProjectionMat();
vmat = get_default_camera()->getViewMat();
}
if(use_lighting) {
for(auto lp : r->getLights()) {
/// xxx need to set lights here.
}
}
if(r->getRenderTarget()) {
r->getRenderTarget()->apply();
}
if(shader->getPUniform() != ShaderProgram::INVALID_UNIFORM) {
shader->setUniformValue(shader->getPUniform(), glm::value_ptr(pmat));
}
if(shader->getMvUniform() != ShaderProgram::INVALID_UNIFORM) {
glm::mat4 mvmat = vmat;
if(is_global_model_matrix_valid() && !r->ignoreGlobalModelMatrix()) {
mvmat *= get_global_model_matrix() * r->getModelMatrix();
} else {
mvmat *= r->getModelMatrix();
}
shader->setUniformValue(shader->getMvUniform(), glm::value_ptr(mvmat));
}
if(shader->getMvpUniform() != ShaderProgram::INVALID_UNIFORM) {
glm::mat4 pvmat(1.0f);
if(is_global_model_matrix_valid() && !r->ignoreGlobalModelMatrix()) {
pvmat = pmat * vmat * get_global_model_matrix() * r->getModelMatrix();
} else {
pvmat = pmat * vmat * r->getModelMatrix();
}
shader->setUniformValue(shader->getMvpUniform(), glm::value_ptr(pvmat));
}
if(shader->getColorUniform() != ShaderProgram::INVALID_UNIFORM) {
if(r->isColorSet()) {
shader->setUniformValue(shader->getColorUniform(), r->getColor().asFloatVector());
} else {
shader->setUniformValue(shader->getColorUniform(), ColorScope::getCurrentColor().asFloatVector());
}
}
shader->setUniformsForTexture(r->getTexture());
// XXX we should make this either or with setting the mvp/color uniforms above.
auto uniform_draw_fn = shader->getUniformDrawFunction();
if(uniform_draw_fn) {
uniform_draw_fn(shader);
}
// Loop through uniform render variables and set them.
/*for(auto& urv : r->UniformRenderVariables()) {
for(auto& rvd : urv->VariableDescritionList()) {
auto rvdd = std::dynamic_pointer_cast<RenderVariableDeviceData>(rvd->GetDisplayData());
ASSERT_LOG(rvdd != nullptr, "Unable to cast DeviceData to RenderVariableDeviceData.");
shader->SetUniformValue(rvdd->GetActiveMapIterator(), urv->Value());
}
}*/
// Need to figure the interaction with shaders.
/// XXX Need to create a mapping between attributes and the index value below.
for(auto as : r->getAttributeSet()) {
//ASSERT_LOG(as->getCount() > 0, "No (or negative) number of vertices in attribute set. " << as->getCount());
if((!as->isMultiDrawEnabled() && as->getCount() <= 0) || (as->isMultiDrawEnabled() && as->getMultiDrawCount() <= 0)) {
//LOG_WARN("No (or negative) number of vertices in attribute set. " << as->getCount());
continue;
}
GLenum draw_mode = convert_drawing_mode(as->getDrawMode());
// apply blend, if any, from attribute set.
BlendEquationScopeOGL be_scope(*as);
BlendModeScopeOGL bm_scope(*as);
if(shader->getColorUniform() != ShaderProgram::INVALID_UNIFORM && as->isColorSet()) {
shader->setUniformValue(shader->getColorUniform(), as->getColor().asFloatVector());
}
for(auto& attr : as->getAttributes()) {
if(attr->isEnabled()) {
shader->applyAttribute(attr);
}
}
if(as->isInstanced()) {
if(as->isIndexed()) {
as->bindIndex();
// XXX as->GetIndexArray() should be as->GetIndexArray()+as->GetOffset()
glDrawElementsInstanced(draw_mode, static_cast<GLsizei>(as->getCount()), convert_index_type(as->getIndexType()), as->getIndexArray(), as->getInstanceCount());
as->unbindIndex();
} else {
glDrawArraysInstanced(draw_mode, static_cast<GLint>(as->getOffset()), static_cast<GLsizei>(as->getCount()), as->getInstanceCount());
}
} else {
if(as->isIndexed()) {
as->bindIndex();
// XXX as->GetIndexArray() should be as->GetIndexArray()+as->GetOffset()
glDrawElements(draw_mode, static_cast<GLsizei>(as->getCount()), convert_index_type(as->getIndexType()), as->getIndexArray());
as->unbindIndex();
} else {
if(as->isMultiDrawEnabled()) {
glMultiDrawArrays(draw_mode, as->getMultiOffsetArray().data(), as->getMultiCountArray().data(), as->getMultiDrawCount());
} else {
glDrawArrays(draw_mode, static_cast<GLint>(as->getOffset()), static_cast<GLsizei>(as->getCount()));
}
}
}
shader->cleanUpAfterDraw();
glBindBuffer(GL_ARRAY_BUFFER, 0);
}
if(r->getRenderTarget()) {
r->getRenderTarget()->unapply();
}
}
