Profiler which takes I/O into account?

[Skip Montanaro]

Is there a profiler for Python (2.7 in my case) which recognizes when
a process (or all threads but the profiler's) is blocked on I/O? I'm
using cProfile at the moment, which is fine as far as it goes, but the
program I'm profiling does a fair amount of I/O, so that dominates the
actual time the program spends calculating. My goal in this case is
minimizing latency of request handling, so while it's not CPU-bound,
it's still important to minimize the relevant code paths.

Thx,

Skip