Timeline for Multithreaded Monte-Carlo Integration
Current License: CC BY-SA 3.0
5 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jan 14, 2018 at 16:06 | comment | added | user14717 | You're right: boost.org/doc/libs/1_50_0/boost/accumulators/statistics/… | |
| Jan 14, 2018 at 8:06 | comment | added | Roland Illig |
You could also mark the kahan class internal (maybe by putting it in a namespace called internal); there are surely people who did something like this before.
|
|
| Jan 14, 2018 at 7:57 | comment | added | Roland Illig |
Since nobody expects std::vector to be threadsafe, why should the kahan class be?
|
|
| Jan 13, 2018 at 15:21 | comment | added | user14717 |
Great suggestion. I worry, however, that if I put a Kahan class into boost there will be an expectation that operator+= will be threadsafe. In that case, there's no point in each individual thread computing it's own average . . . well, yeah that might be way better, but there might be a lot of contention on the atomic variable. Need to perf test that one.
|
|
| Jan 13, 2018 at 6:24 | history | answered | Roland Illig | CC BY-SA 3.0 |