Fix handling of virtual exit node in PostDomTree
#53739
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The `dom_edges()` for an exit block in the CFG are empty when computing the PostDomTree, although the algorithm is supposed to treat this as being dominated by a "virtual" -1 node (which is already included in the DFSTree). As a result, we can't rely on the loop below this code to initialize the semi-dominator correctly for us.
aviatesk
approved these changes
Mar 15, 2024
There was a problem hiding this comment.
It looks like this change fixes the problem #53739 tried to fix with the augmented control-flow graph right at its root. If you're up for it, switching the base branch to #53739 could let us use the test cases from there to check this fix. Though, having more direct test cases could be even better, but not necessary if we use test cases from #53739.
aviatesk
force-pushed
the
ct/fix-postdomtree
branch
from
4ff4981
to
c860691
Compare
|
I confirm this PR passes all the test cases added in #53642. |
aviatesk
added
compiler:optimizer
aviatesk
added a commit
that referenced
this pull request
Mar 19, 2024
This commit was initially crafted to fix an issue with post-opt analysis that came to light in #53613, planning to solve it by incorporating an augmented CFG directly into the post-opt analysis. But, with the issue now sorted out by #53739, this commit shifted focus to just adding test cases. === The original commit message === post-opt: use augmented post-domtree for `visit_conditional_successors` This commit fixes the first problem that was found while digging into #53613. It turns out that the post-domtree constructed from regular `IRCode` doesn't work for visiting conditional successors for post-opt analysis in cases like: ```julia julia> let code = Any[ # block 1 GotoIfNot(Argument(2), 3), # block 2 ReturnNode(Argument(3)), # block 3 (we should visit this block) Expr(:call, throw, "potential throw"), ReturnNode(), # unreachable ] ir = make_ircode(code; slottypes=Any[Any,Bool,Bool]) visited = BitSet() @test !Core.Compiler.visit_conditional_successors(CC.LazyPostDomtree(ir), ir, #=bb=#1) do succ::Int push!(visited, succ) return false end @test 2 ∉ visited @test 3 ∈ visited end Test Failed at REPL[14]:16 Expression: 2 ∉ visited Evaluated: 2 ∉ BitSet([2]) ``` This might mean that we need to fix on the `postdominates` end, but for now, this commit tries to get around it by using the augmented post domtree in `visit_conditional_successors`, while also enforcing the augmented control flow graph (`construct_augmented_cfg`) to have a single exit node really. Since the augmented post domtree is now enforced to have a single return, we can keep using the current `postdominates` to fix the issue. However, this commit isn't enough to fix the NeuralNetworkReachability segfault as reported in #53613, and we need to tackle the second issue reported there too (#53613 (comment)).
IanButterworth
removed
the
merge me
aviatesk
pushed a commit
that referenced
this pull request
Apr 23, 2024
This is an alternative to #53642 The `dom_edges()` for an exit block in the CFG are empty when computing the PostDomTree so the loop below this may not actually run. In that case, the right semidominator is the ancestor from the DFSTree, which is the "virtual" -1 block. This resolves half of the issue in #53613: ```julia julia> let code = Any[ # block 1 GotoIfNot(Argument(2), 3), # block 2 ReturnNode(Argument(3)), # block 3 (we should visit this block) Expr(:call, throw, "potential throw"), ReturnNode(), # unreachable ] ir = make_ircode(code; slottypes=Any[Any,Bool,Bool]) visited = BitSet() @test !Core.Compiler.visit_conditional_successors(CC.LazyPostDomtree(ir), ir, #=bb=#1) do succ::Int push!(visited, succ) return false end @test 2 ∈ visited @test 3 ∈ visited end Test Passed ``` This needs some tests (esp. since I don't think we have any DomTree tests at all right now), but otherwise should be good to go.
topolarity
added
the
backport 1.10
topolarity
added a commit
that referenced
this pull request
Oct 30, 2024
This is an alternative to #53642 The `dom_edges()` for an exit block in the CFG are empty when computing the PostDomTree so the loop below this may not actually run. In that case, the right semidominator is the ancestor from the DFSTree, which is the "virtual" -1 block. This resolves half of the issue in #53613: ```julia julia> let code = Any[ # block 1 GotoIfNot(Argument(2), 3), # block 2 ReturnNode(Argument(3)), # block 3 (we should visit this block) Expr(:call, throw, "potential throw"), ReturnNode(), # unreachable ] ir = make_ircode(code; slottypes=Any[Any,Bool,Bool]) visited = BitSet() @test !Core.Compiler.visit_conditional_successors(CC.LazyPostDomtree(ir), ir, #=bb=#1) do succ::Int push!(visited, succ) return false end @test 2 ∈ visited @test 3 ∈ visited end Test Passed ``` This needs some tests (esp. since I don't think we have any DomTree tests at all right now), but otherwise should be good to go.
topolarity
removed
the
backport 1.10
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This is an alternative to #53642
The
dom_edges()for an exit block in the CFG are empty when computing the PostDomTree so the loop below this may not actually run. In that case, the right semidominator is the ancestor from the DFSTree, which is the "virtual" -1 block.This resolves half of the issue in #53613:
This needs some tests (esp. since I don't think we have any DomTree tests at all right now), but otherwise should be good to go.