On 9/12/2011 3:32 PM, Jerry Quinn wrote:
The other thing I think I have to do is sequence the initializations.
Otherwise the disk/network will thrash trying to load N copies simultaneously.
I imagine this can be done with a shared variable inside the engine.
This can easily be done by creating null engines in the
WorkerLocalStorage and then initializing them inside a synchronized block:
auto wlEngines = taskPool.workerLocalStorage(Engine.init);
auto wlIsInitialized = taskPool.workerLocalStorage(false);
// This gets executed on the pool in a Task:
void taskFunction() {
if(!wlIsInitialized.get) {
// Do initialization of thread-local engine in worker thread.
synchronized wlEngines.get = new Engine(ctorArgs);
wlIsInitialized.get = true;
}
// Do main processing.
}
I prefer to do all the initializations before work starts so that timing can be
more accurate. These things take a lot to get fully loaded and ready to run.
Unfortunately I don't have a good answer for this constraint. If you
can offer a concrete enhancement request for std.parallelism, I'll
certainly consider it, but I can't think of a good design for this kind
of per-thread initialization routine off the top of my head, either in
terms of API or implementation.
