multiple mad istances about cpymad

[coldfix]

Hi, there is at leaste one error because your code doesn't ensure that at most one thread is using a particular Madx instance at a time. If two threads talk to the same Madx at the same time, it will gobble up the commands/responses and lead to crashes.

Can you try adding this guarantee, e.g.:

    def input(self,text):
        jobs = [self.pool.submit(mad.input, text) for mad in self.mad]
        return [job.result() for job in jobs]

    def input_map(self, texts):
        jobs = [self.pool.submit(mad.input, text) for mad, text in zip(self.mad, texts)]
        return [job.result() for job in jobs]

[coldfix]

I guess I could also investigate the possibility of adding async mode to Madx in which input() and commands return futures instead of blocking. Read access, however, would still be synchronous (because it will be significantly more work to update the entire API..).

[rdemaria]

It looks like it is working with your fix. I have also implemented a general method:

    def map(self,target,*args):
        lst=zip(self.mad,zip(*args))
        jobs = [self.pool.submit(target, mad,*arg) for mad,arg in lst]
        return [job.result() for job in jobs]

However I am seeing, while on an ipython shell, a call that takes much longer than expected with the threads with 3% CPU load. It could be some i/o locking maybe, but I need to investigate further. At the moment I am using 16 mad instances in a 16 core machine.