int fib(int fibMe)
{
if(fibMe<2){
return fibMe;
}
int fib1m=cilk_spawn(fib(fibMe-1));
int fib2m=cilk_spawn(fib(fibMe-2));
cilk_sync;
int retMe=fib1m+fib2m;
return retMe;
}
void RT_RayTracer::runTasksCilk() {
for (size_t i = 0, e = taskManager.tasks.size(); i < e; ++i) {
#ifdef __cilk
cilk_spawn(taskManager.tasks[i]->run());
#else
taskManager.tasks[i]->run();
#endif
}
#ifdef __cilk
// added implicitly by the compiler at the end of a spawning function
cilk_sync;
#endif
}
void RT_RayTracer::renderFrameETC() {
uint32_t* fb1 = (uint32_t*)frameBuffer->getFrameBuffer();
uint32_t* fb2 = (uint32_t*)blockFB->getFrameBuffer();
const int widthFB1 = frameBuffer->getSizeX(); // width of full frame buffer, e.g. 1280 for 1280x720
int widthFB2 = widthFB1;
if (Engine::rectMode) {
widthFB2 = Engine::rectSizeX; // width of rect, e.g. 640 if only one half of the full frame buffer should be rendered
}
look();
taskManager.deleteAllTasks();
createRenderingTasks();
if (Engine::server) {
auto etc1_fun = [&] (size_t i) {
// render the tile to get RGBA data into the framebuffer
taskManager.tasks[i]->run();
auto src = fb1 + widthFB1 * Engine::RENDERLINE_SIZE * i;
if (Engine::rectMode) {
src += Engine::rectBottom * widthFB1 + Engine::rectLeft; // offset the RGBA access until the bottom (lower part) of the rect starts and on the left side
}
auto dst = fb2 + widthFB2 * Engine::RENDERLINE_SIZE * i;
// copy RGBA data from fb1 into the blockwise-oriented fb2
for (int blockY = 0; blockY < Engine::RENDERLINE_SIZE/4; blockY++) { // if RENDERLINE_SIZE=4 then this loop will only be executed once
for (int blockX = 0; blockX < widthFB2 / 4; blockX++) {
for (int x = 0; x < 4; x++) {
for (int y = 0; y < 4; y++) {
*dst++ = *src;
src += widthFB1;
}
src -= widthFB1 * 4 - 1;
}
}
src += widthFB1 * 3; // if RENDERLINE_SIZE=4 then this is irrelevant as src will not be used later
}
auto etc = ((uint64_t*)etcdata) + i * widthFB2 / 4;
auto etcsrc = ((uint8_t*)fb2) + widthFB2 * Engine::RENDERLINE_SIZE * i * 4;
// loop through the blockwise-oriented fb2 and compress RGBA into ETC1 and store this in etc/etcdata.
for (size_t i = 0; i < widthFB2*Engine::RENDERLINE_SIZE / 16; ++i) {
#if 0
Dither( etcsrc );
#endif
*etc++ = ProcessRGB( etcsrc );
etcsrc += 4*4*4;
}
};
auto ompf_dispatch = [&] () {
size_t e = taskManager.tasks.size();
#pragma omp parallel for
for (size_t i = 0; i < e; ++i) {
etc1_fun(i);
}
};
auto ompt_dispatch = [&] () {
#pragma omp parallel
#pragma omp single
for (size_t i = 0, e = taskManager.tasks.size(); i < e; ++i) {
#pragma omp task
etc1_fun(i);
}
};
auto cilk_dispatch = [&] () {
for (size_t i = 0, e = taskManager.tasks.size(); i < e; ++i) {
#ifdef __cilk
cilk_spawn(etc1_fun(i));
#else
etc1_fun(i);
#endif
}
#ifdef __cilk
cilk_sync;
#endif
};
auto task_dispatch = [&] () {
for (size_t i = 0, e = taskManager.tasks.size(); i < e; ++i) {
TaskDispatch::Queue([&etc1_fun, i] { etc1_fun(i); });
}
TaskDispatch::Sync();
};
switch (Engine::methodToMultiThread) {
case MultiThreadMethods::TASKDISPATCH: task_dispatch(); break;
case MultiThreadMethods::OPENMP: ompf_dispatch(); break;
case MultiThreadMethods::OPENMPT: ompt_dispatch(); break;
case MultiThreadMethods::CILK: cilk_dispatch(); break;
}
} else {
renderScene();
}
}
