[plt-scheme] Profiling code

My game prototype in DrScheme is now feature complete and I would like
to optimise it to make things run faster. The inbuilt profiler is
wonderfully well integrated and has already helped me remove some
bottlenecks.

However there are a few functions in my code that get called very
often and have a very small execution time. I want to know if there is
an additional cost in terms of "function call overhead" that exists
for functions beyond their actual execution time.

On a more general note, I wanted to know how I can observe what my
code is doing at a slightly lower level of detail. Things I would be
interested in

1) How much memory is being consed.
2) When does GC get invoked.
3) What kind of assembly code is getting generated for my functions.
4) Am I hitting some inefficient execution paths with DrScheme.

Any pointers if this is possible at all ?