atomic load/store
Try to partition off the work so that each thread can load a "piece" of the problem into its cache, work on it, and then arrive at a result. This may mean that there is a copy of sdata per-thread....
View Articleatomic load/store
That sounds right, thanks Wyck. Since there's no issue with local variables I assumed incorrectly global variables would behave the same. This was just a test program but I think I can fix the real...
View Articleatomic load/store
Your program's execution time is limited by the speed of system memory. Or as I would say, it is memory bound. (Your system almost certainly has a slower memory speed than processor speed.)I'm...
View Articleatomic load/store
If I remove the sdata global and just accumulate the i values it will complete as expected in 3s in both cases. If it had smth to do w context switching that version would be slow as well right? I also...
View Articleatomic load/store
Context switching itself is expensive. Try to make one thread a higher priority than the other. That should reduce context switching, as Windows will not switch to a low-priority thread as along as a...
View Articleatomic load/store
I get the same result without the std::cout, 6s when I use a single thread (trd1), 14s when I use both threads (trd1 + trd2). Just tried smth similar with a C# program, it's even worse there, 3s for...
View Articleatomic load/store
HiIn the following code running the task5 routine from 2 thread simultaneously on a multiprocessor machine takes approx twice as long as running it from a single thread. The atomic loads/stores seem to...
View Article