JIT: revise checking for computed preds #81582
Conversation
Since pred lists now exist pervasively, change most methods that had conditional pred list updates to assert preds exist and always update. To make sure I got all uses, I renamed `fgComputePredsDone` to `fgPredsComputed`. Remove the ability to drop EH, as it doesn't update pred lists properly and so has been broken for quite a while now. Remove some of the logic for fixing up finally targets, since we now always have pred lists around and so don't need the blanket invaliation. Closes dotnet#80193.
|
Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch, @kunalspathak Issue DetailsSince pred lists now exist pervasively, change most methods that had conditional pred list updates to assert preds exist and always update. To make sure I got all uses, I renamed Remove the ability to drop EH, as it doesn't update pred lists properly and so has been broken for quite a while now. Remove some of the logic for fixing up finally targets, since we now always have pred lists around and so don't need the blanket invaliation. Closes #80193.
|
|
@EgorBo PTAL This is the final bit of revision for pred lists. I may still try and encapsulate the remove/add ref patterns that are now fairly pervasive but will likely instead start building upon the new capabilities that persistent pred lists offer. |
|
|
||
| bool fgModified; // True if the flow graph has been modified recently | ||
| bool fgComputePredsDone; // Have we computed the bbPreds list | ||
| bool fgPredsComputed; // Have we computed the bbPreds list |
There was a problem hiding this comment.
Can you educate me why we need this at all? I assume it's false only during importing?
There was a problem hiding this comment.
We actually build pred lists before importation, so fgPredsComputed is true for all phases. So it seems like something we don't need. But I kept this concept around for two reasons:
- we sometimes do a bit of flow graph modification very early before we have added preds lists, and call methods that are also called after we have pred edges; those methods need to behave differently before and after (mainly
fgEnsureFirstBBisScratchand friends - we want to make sure no other pred-list sensitive method is called early.
|
Failure is known issue. |
Since pred lists now exist pervasively, change most methods that had conditional pred list updates to assert preds exist and always update.
To make sure I got all uses, I renamed
fgComputePredsDonetofgPredsComputed.Remove the ability to drop EH, as it doesn't update pred lists properly and so has been broken for quite a while now.
Remove some of the logic for fixing up finally targets, since we now always have pred lists around and so don't need the blanket invaliation.
Closes #80193.