void display(void)
{
pthread_mutex_lock(&display_mutex);
if (!ref_file)
{
sdkStartTimer(&timer);
simulateFluids();
}
// render points from vertex buffer
glClear(GL_COLOR_BUFFER_BIT);
glClearColor(1.f/256*172, 1.f/256*101, 1.f/256*4, 0.f);
glColor4f(1.f, 1.f, 1.f, 0.5f);
glPointSize(1);
glEnable(GL_POINT_SMOOTH);
glEnable(GL_BLEND);
glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
glEnableClientState(GL_VERTEX_ARRAY);
glDisable(GL_DEPTH_TEST);
glDisable(GL_CULL_FACE);
glBindBufferARB(GL_ARRAY_BUFFER_ARB, vbo);
#ifdef OPTIMUS
glBufferDataARB(GL_ARRAY_BUFFER_ARB, sizeof(cData) * DS,
particles, GL_DYNAMIC_DRAW_ARB);
#endif
glVertexPointer(2, GL_FLOAT, 0, NULL);
glDrawArrays(GL_POINTS, 0, DS);
glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
glDisableClientState(GL_VERTEX_ARRAY);
glDisableClientState(GL_TEXTURE_COORD_ARRAY);
glDisable(GL_TEXTURE_2D);
if (ref_file)
{
return;
}
// Finish timing before swap buffers to avoid refresh sync
sdkStopTimer(&timer);
glutSwapBuffers();
fpsCount++;
if (fpsCount == fpsLimit)
{
char fps[256];
float ifps = 1.f / (sdkGetAverageTimerValue(&timer) / 1000.f);
sprintf(fps, "Caffe Macchiato / Stable Fluids (%d x %d): %3.1f fps", DIM, DIM, ifps);
glutSetWindowTitle(fps);
fpsCount = 0;
fpsLimit = (int)MAX(ifps, 1.f);
sdkResetTimer(&timer);
}
glutPostRedisplay();
pthread_mutex_unlock(&display_mutex);
}
void autoTest(char **argv)
{
CFrameBufferObject *fbo = new CFrameBufferObject(wWidth, wHeight, 4, false, GL_TEXTURE_2D);
g_CheckRender = new CheckFBO(wWidth, wHeight, 4, fbo);
g_CheckRender->setPixelFormat(GL_RGBA);
g_CheckRender->setExecPath(argv[0]);
g_CheckRender->EnableQAReadback(true);
fbo->bindRenderPath();
reshape(wWidth, wHeight);
for (int count=0; count<g_iFrameToCompare; count++)
{
simulateFluids();
// add in a little force so the automated testing is interesing.
if (ref_file)
{
int x = wWidth/(count+1);
int y = wHeight/(count+1);
float fx = (x / (float)wWidth);
float fy = (y / (float)wHeight);
int nx = (int)(fx * DIM);
int ny = (int)(fy * DIM);
int ddx = 35;
int ddy = 35;
fx = ddx / (float)wWidth;
fy = ddy / (float)wHeight;
int spy = ny-FR;
int spx = nx-FR;
addForces(dvfield, DIM, DIM, spx, spy, FORCE * DT * fx, FORCE * DT * fy, FR);
lastx = x;
lasty = y;
}
}
display();
fbo->unbindRenderPath();
// compare to offical reference image, printing PASS or FAIL.
printf("> (Frame %d) Readback BackBuffer\n", 100);
g_CheckRender->readback(wWidth, wHeight);
g_CheckRender->savePPM("fluidsGL.ppm", true, NULL);
if (!g_CheckRender->PPMvsPPM("fluidsGL.ppm", ref_file, MAX_EPSILON_ERROR, 0.25f))
{
g_TotalErrors++;
}
}
void display(void)
{
if (!ref_file)
{
sdkStartTimer(&timer);
simulateFluids();
}
// render points from vertex buffer
glClear(GL_COLOR_BUFFER_BIT);
glColor4f(0,1,0,0.5f);
glPointSize(1);
glEnable(GL_POINT_SMOOTH);
glEnable(GL_BLEND);
glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
glEnableClientState(GL_VERTEX_ARRAY);
glDisable(GL_DEPTH_TEST);
glDisable(GL_CULL_FACE);
glBindBufferARB(GL_ARRAY_BUFFER_ARB, vbo);
glVertexPointer(2, GL_FLOAT, 0, NULL);
glDrawArrays(GL_POINTS, 0, DS);
glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
glDisableClientState(GL_VERTEX_ARRAY);
glDisableClientState(GL_TEXTURE_COORD_ARRAY);
glDisable(GL_TEXTURE_2D);
if (ref_file)
{
return;
}
// Finish timing before swap buffers to avoid refresh sync
sdkStopTimer(&timer);
glutSwapBuffers();
fpsCount++;
if (fpsCount == fpsLimit)
{
char fps[256];
float ifps = 1.f / (sdkGetAverageTimerValue(&timer) / 1000.f);
sprintf(fps, "Cuda/GL Stable Fluids (%d x %d): %3.1f fps", DIM, DIM, ifps);
glutSetWindowTitle(fps);
fpsCount = 0;
fpsLimit = (int)MAX(ifps, 1.f);
sdkResetTimer(&timer);
}
glutPostRedisplay();
}
GLFluids::GLFluids(QWidget *parent)
: QGLWidget(parent),
QGLFunctions()
{
vbo = 0;
wWidth = qMax(512, DIM);
wHeight = qMax(512, DIM);
hvfield = (float2 *)malloc(sizeof(float2) * DS);
memset(hvfield, 0, sizeof(float2) * DS);
// Allocate and initialize device data
cudaMallocPitch((void **)&dvfield, &tPitch, sizeof(float2)*DIM, DIM);
cudaMemcpy(dvfield, hvfield, sizeof(float2) * DS,
cudaMemcpyHostToDevice);
// Temporary complex velocity field data
cudaMalloc((void **)&vxfield, sizeof(float2) * PDS);
cudaMalloc((void **)&vyfield, sizeof(float2) * PDS);
setup_texture(DIM, DIM);
bind_texture();
// Create particle array
particles = (float2 *)malloc(sizeof(float2) * DS);
memset(particles, 0, sizeof(float2) * DS);
initParticles(particles, DIM, DIM);
// Create CUFFT transform plan configuration
cufftPlan2d(&planr2c, DIM, DIM, CUFFT_R2C);
cufftPlan2d(&planc2r, DIM, DIM, CUFFT_C2R);
// TODO: update kernels to use the new unpadded memory layout for perf
// rather than the old FFTW-compatible layout
cufftSetCompatibilityMode(planr2c, CUFFT_COMPATIBILITY_FFTW_PADDING);
cufftSetCompatibilityMode(planc2r, CUFFT_COMPATIBILITY_FFTW_PADDING);
QTimer *timer = new QTimer(this);
connect(timer, &QTimer::timeout, [&](){
simulateFluids();
updateGL();
});
timer->start(0);
}
