expected synergy effect with function inline pass (which planned to be implemented in sprint 3), and it implies that out pass need to consider applying order at last phase of development.
How much proportion of allowed stack is appropriate?
What is good usage of oracle. We have two candidates.
Batch storer
Custom memory allocator
But custom memory allocator may not provide huge performance improvements. Batch storer is engaging choice but we have no idea of it is good choice.
What we're going to implement in sprint 3
Lots of pass candidates revealed as useless after test cases are opened.
Function inline and Oracle is selected with no doubt, but third sprint is not determined.
New optimization trick idea: we can remove every free() call which comes after last malloc() call.
(Additional) How much LICMPass (Loop invariant code motion) done good to performance
After applying it solely, it shows poor performance increase.
More detailed analysis of heap2stack pass
What is good usage of oracle. We have two candidates.
What we're going to implement in sprint 3
New optimization trick idea: we can remove every
free()
call which comes after lastmalloc()
call.(Additional) How much LICMPass (Loop invariant code motion) done good to performance