ReplaceJumpWithRet #1159
ReplaceJumpWithRet #1159
Conversation
Hecate2
force-pushed
the
delete-jmp-to-ret
branch
from
480e590
to
93a5924
Compare
|
This is the first time I add new instructions to the optimized contract, and the added RET is not binded to source codes with debuginfo. For now this may not be a big problem, but in the future we may need to bind new instructions to source codes. |
Hecate2
force-pushed
the
delete-jmp-to-ret
branch
from
84b11b3
to
21c640f
Compare
|
This make harder the method detection, currently we only have one ret for each method |
do we have limitation on using ret? i dont know that. |
Are we now really restricting the count of RET and detecting methods using the only RET? I can write a robust method detection if needed. I can also abandon this PR and (slowly) write something else to ensure only a single RET exists in a method. |
|
Go compiler can generate as many RETs for a method as needed and typical Go style even encourages that. Method boundaries are usually tracked with manifests, it's just more reliable. |
| @@ -22,7 +22,7 @@ namespace Neo.Optimizer | |||
| { | |||
| static class Reachability | |||
| { | |||
| [Strategy(Priority = int.MaxValue)] | |||
| [Strategy(Priority = int.MaxValue - 16)] | |||
There was a problem hiding this comment.
a brief comment to explain what 16 means here would be great
There was a problem hiding this comment.
It means nothing. Just reserving spaces for other optimizers. We are not even using the Strategy attribute now. In the future we may sort all the [Strategy] methods and execute them (cyclically) ordered by Priority.
@shargon Hey shargon, any further detail about this? |
|
It complicate the reverse engineering but maybe we should take care more avoid optimization instead of reversing |
If a JMP or JMP_L jumps to a RET, replace the JMP with RET