C++ :: How To Prevent The Main Thread From Blocking

I have an issue where I iterate through devices and make driver API calls. Unfortunately, these calls take too much time and they are interrupting my real-time scheduler. I have 12 cores, of which one is 100% and the others are < 1%.

I'd like to multi-thread this thing. So far, I've replaced:

for (DeviceIterator d = devices.begin(); d != devices.end(); ++d) {
d->Write(words, numwords);
}

with

void* ThreadedWrite(void* p) {
WritePackage* wp = (WritePackage*)p;
wp->device->Write(wp->words, wp->numwords);
delete[] wp->words;
delete wp;

[Code]...

My problem is that this didn't improve performance at all. The main thread still takes too long to execute. Is there something I need to do to prevent the main thread from blocking?