java versus C or C++ for number crunching

[Mark Thornton]

To which of Java's several garbage collectors does your comment apply?

Young generation collections in Java are very fast, influenced only by
the number of live objects; dead ones do not add to the GC time.

What seems to be forgotten is even if there are very few live objects
the GC will still take some minimum time. Thus there will be some cost
for every time you fill the young generation. The more frequently you
fill it, the higher the cost. This cost can be reduced by giving the
process a lot of memory, and in particular configuring a very large
young generation.

Mark Thornton

 

[northerntechie]

I have not heard of any conversation regarding 'actual' process cycle
time. Most tests I have seen compare start and end clocks, nothing is
accounted for in the kernel context switching, hardware interrupts,
and all that other jazz in the background. There are a couple general
purpose hardware counters around (with a little kernel, possibly user
mode code manipulation) that can track actual cycle times. All these
numbers are irrelevant without contextual and focused tests,
optimization aside. Throw in a SMP based JVM GC, operating within my
process timeslice whenever my objects are all used up and my resources
are drained, and I would feel a little slow as well.

I think too much in the micro-controller world where multi-tasking is
rare, and if you run across it, it is only implemented by the hardy
few.

Todd Saharchuk, AScT.