/**
* Compute some pixels, and store them.
*
* This uses a "self-dispatching" parallel algorithm. Executes until
* there is no more work to be done, or is told to stop.
*
* In order to reduce the traffic through the res_worker critical
* section, a multiple pixel block may be removed from the work queue
* at once.
*
* For a general-purpose version, see LIBRT rt_shoot_many_rays()
*/
void
worker(int cpu, void *UNUSED(arg))
{
int pixel_start;
int pixelnum;
int pat_num = -1;
/* The more CPUs at work, the bigger the bites we take */
if (per_processor_chunk <= 0) per_processor_chunk = npsw;
if (cpu >= MAX_PSW) {
bu_log("rt/worker() cpu %d > MAX_PSW %d, array overrun\n", cpu, MAX_PSW);
bu_exit(EXIT_FAILURE, "rt/worker() cpu > MAX_PSW, array overrun\n");
}
RT_CK_RESOURCE(&resource[cpu]);
pat_num = -1;
if (hypersample) {
int i, ray_samples;
ray_samples = hypersample + 1;
for (i=0; pt_pats[i].num_samples != 0; i++) {
if (pt_pats[i].num_samples == ray_samples) {
pat_num = i;
goto pat_found;
}
}
}
pat_found:
if (transpose_grid) {
int tmp;
/* switch cur_pixel and last_pixel */
tmp = cur_pixel;
cur_pixel = last_pixel;
last_pixel = tmp;
while (1) {
if (stop_worker)
return;
bu_semaphore_acquire(RT_SEM_WORKER);
pixel_start = cur_pixel;
cur_pixel -= per_processor_chunk;
bu_semaphore_release(RT_SEM_WORKER);
for (pixelnum = pixel_start; pixelnum > pixel_start-per_processor_chunk; pixelnum--) {
if (pixelnum < last_pixel)
return;
do_pixel(cpu, pat_num, pixelnum);
}
}
} else if (random_mode) {
while (1) {
/* Generate a random pixel id between 0 and last_pixel
inclusive - TODO: check if there is any issue related
with multi-threaded RNG */
pixelnum = rand()*1.0/RAND_MAX*(last_pixel + 1);
if (pixelnum >= last_pixel) pixelnum = last_pixel;
do_pixel(cpu, pat_num, pixelnum);
}
} else {
while (1) {
if (stop_worker)
return;
bu_semaphore_acquire(RT_SEM_WORKER);
pixel_start = cur_pixel;
cur_pixel += per_processor_chunk;
bu_semaphore_release(RT_SEM_WORKER);
for (pixelnum = pixel_start; pixelnum < pixel_start+per_processor_chunk; pixelnum++) {
if (pixelnum > last_pixel)
return;
do_pixel(cpu, pat_num, pixelnum);
}
}
}
}
/*
* W O R K E R
*
* Compute some pixels, and store them.
* A "self-dispatching" parallel algorithm.
* Executes until there is no more work to be done, or is told to stop.
*
* In order to reduce the traffic through the res_worker critical section,
* a multiple pixel block may be removed from the work queue at once.
*
* For a general-purpose version, see LIBRT rt_shoot_many_rays()
*/
void
worker(int cpu, genptr_t arg)
{
int pixel_start;
int pixelnum;
int pat_num = -1;
/* The more CPUs at work, the bigger the bites we take */
if ( per_processor_chunk <= 0 ) per_processor_chunk = npsw;
if ( cpu >= MAX_PSW ) {
bu_log("rt/worker() cpu %d > MAX_PSW %d, array overrun\n", cpu, MAX_PSW);
bu_exit(EXIT_FAILURE, "rt/worker() cpu > MAX_PSW, array overrun\n");
}
RT_CK_RESOURCE( &resource[cpu] );
pat_num = -1;
if (hypersample) {
int i, ray_samples;
ray_samples = hypersample + 1;
for (i=0; pt_pats[i].num_samples != 0; i++) {
if (pt_pats[i].num_samples == ray_samples) {
pat_num = i;
goto pat_found;
}
}
}
pat_found:
if (transpose_grid) {
int tmp;
/* switch cur_pixel and last_pixel */
tmp = cur_pixel;
cur_pixel = last_pixel;
last_pixel = tmp;
while (1) {
if (stop_worker)
return;
bu_semaphore_acquire(RT_SEM_WORKER);
pixel_start = cur_pixel;
cur_pixel -= per_processor_chunk;
bu_semaphore_release(RT_SEM_WORKER);
for (pixelnum = pixel_start; pixelnum > pixel_start-per_processor_chunk; pixelnum--) {
if (pixelnum < last_pixel)
return;
do_pixel(cpu, pat_num, pixelnum);
}
}
} else {
while (1) {
if (stop_worker)
return;
bu_semaphore_acquire(RT_SEM_WORKER);
pixel_start = cur_pixel;
cur_pixel += per_processor_chunk;
bu_semaphore_release(RT_SEM_WORKER);
for (pixelnum = pixel_start; pixelnum < pixel_start+per_processor_chunk; pixelnum++) {
if (pixelnum > last_pixel)
return;
do_pixel(cpu, pat_num, pixelnum);
}
}
}
}
