Previous implementation of InlineAllPass was greedy and dumb. For every callpoint, we inlined the gates in a Top-Down approach.
Now, first we inline the nodes inside the gate we would be inlining. Not only that, we also cache the "to-be-inlined" gates. So, it is 243248327x fast (depending on the program).
Previous implementation of
InlineAllPass
was greedy and dumb. For every callpoint, we inlined the gates in a Top-Down approach.Now, first we inline the nodes inside the gate we would be inlining. Not only that, we also cache the "to-be-inlined" gates. So, it is 243248327x fast (depending on the program).