[MLton-user] Tracking down allocations

[Joe Hurd]

Hi Matthew,

> If tuple_32 is passed as an argument to
> another SSA function, then it may be a case of the input programs being
> just different enough that the flattening optimization behaves differently
> on them.

tuple_32 is indeed passed into another function. To make things
crystal clear to the compiler I've stuck with the manually
lambda-lifted version.

One last question. I have a big function with 81 patterns like this:

datatype t = A | B | C;

fun f A A A A = ...
  | f A A A B = ...
  | ...
  | f C C C C = ...

where each of the right hand sides are just a single function call. It
is called precisely once, and the call-site is accounting for 50% of
the execution time of the whole program. Am I right in assuming that
this is reflecting the cost of the pattern match, and is there
anything that can be done to speed it up?

Thanks for your help,

Joe