Conditional jumps with fall-through not always ordered properly by postorder_dfs about hoopl HOT 4 CLOSED

I don't understand the problem you are trying to solve. What's wrong with the order that postorder_dfs comes up with?

Perhaps you have in mind that after postorder_dfs then you can pretty-print as a sequence of machine instructions with a fall-through for the conditional jump? But you can still get a fall-through by negating the condition can't you? And none of this matters if generating LLVM.

from hoopl.

I want a sequence of instructions that have the order as they should appear in the final output. Conditional branches need to be able to fall-through to the next block.

However, it turns out that it is easy to rectify this issue in my case as I need a branch optimizer that works on the flattened output anyway and correcting the potential out of order sequence there is trivial.

So any post processing, be it the way I do it or (apparently) LLVM can easily correct the issue.

I will close this.

from hoopl.

from hoopl.

Thank you! That is very useful information to me, I will keep it in mind for the future as I expect to get back and do more elaborate branch optimizations.

from hoopl.