Rework :inbounds effects tainting #48246
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
This works to fix #48243, by only tanting effects if an `@inbounds` statement is actually reached. Further, it refines the `noinbounds` effect to be IPO-cached and used to track whether a particular method read the inbounds state. A `:boundscheck` expression now does not immediately taint consistencty, but instead, taints `noinbounds` only. Then, if a method that has `:noinbounds` tainted is called within an `@inbounds` region, consistency is tainted. Similarly, a tainted `:noinbounds` disables constant propagation at `@inbounds` statements or if the method propagates inbounds.
aviatesk
reviewed
Jan 13, 2023
| @goto delay_effects_analysis | ||
| end | ||
| end | ||
| flag = sv.src.ssaflags[sv.currpc] |
There was a problem hiding this comment.
We can use get_curr_ssaflag(sv) for this.
| noinbounds = inbounds === :on || (inbounds === :default && all(flag::UInt8->iszero(flag&IR_FLAG_INBOUNDS), src.ssaflags)) | ||
| consistent = noinbounds ? ALWAYS_TRUE : ALWAYS_FALSE | ||
| ipo_effects = Effects(EFFECTS_TOTAL; consistent, noinbounds) | ||
| ipo_effects = Effects(EFFECTS_TOTAL) |
aviatesk
added a commit
that referenced
this pull request
Jan 13, 2023
aviatesk
added a commit
to aviatesk/JET.jl
that referenced
this pull request
Jan 13, 2023
aviatesk
added a commit
that referenced
this pull request
Jan 13, 2023
aviatesk
added a commit
that referenced
this pull request
Jan 13, 2023
41 tasks
Keno
added a commit
that referenced
this pull request
Jan 16, 2023
LLVM can prove this inbounds and the annotation weakens the inferable effects for tuple iteration, which has a surprisingly large inference performance and precision impact. Unfortunately, my previous changes to :inbounds tainting weren't quite strong enough yet, because `getfield` was still tainting consistency on unknown boundscheck arguments. To fix that, we pass through the fargs into the fetfield effects to check if we're getting a literal `:boundscheck`, in which case the `:noinbounds` consistency-tainting logic I added in #48246 is sufficient to not require additional consistency tainting. Also add a test for both effects and codegen to make sure this doens't regress.
Keno
added a commit
that referenced
this pull request
Jan 16, 2023
LLVM can prove this inbounds and the annotation weakens the inferable effects for tuple iteration, which has a surprisingly large inference performance and precision impact. Unfortunately, my previous changes to :inbounds tainting weren't quite strong enough yet, because `getfield` was still tainting consistency on unknown boundscheck arguments. To fix that, we pass through the fargs into the fetfield effects to check if we're getting a literal `:boundscheck`, in which case the `:noinbounds` consistency-tainting logic I added in #48246 is sufficient to not require additional consistency tainting. Also add a test for both effects and codegen to make sure this doens't regress.
Keno
added a commit
that referenced
this pull request
Jan 16, 2023
LLVM can prove this inbounds and the annotation weakens the inferable effects for tuple iteration, which has a surprisingly large inference performance and precision impact. Unfortunately, my previous changes to :inbounds tainting weren't quite strong enough yet, because `getfield` was still tainting consistency on unknown boundscheck arguments. To fix that, we pass through the fargs into the fetfield effects to check if we're getting a literal `:boundscheck`, in which case the `:noinbounds` consistency-tainting logic I added in #48246 is sufficient to not require additional consistency tainting. Also add a test for both effects and codegen to make sure this doens't regress.
KristofferC
pushed a commit
that referenced
this pull request
Jan 16, 2023
This works to fix #48243, by only tanting effects if an `@inbounds` statement is actually reached. Further, it refines the `noinbounds` effect to be IPO-cached and used to track whether a particular method read the inbounds state. A `:boundscheck` expression now does not immediately taint consistencty, but instead, taints `noinbounds` only. Then, if a method that has `:noinbounds` tainted is called within an `@inbounds` region, consistency is tainted. Similarly, a tainted `:noinbounds` disables constant propagation at `@inbounds` statements or if the method propagates inbounds. (cherry picked from commit d544e78)
Keno
added a commit
that referenced
this pull request
Jan 17, 2023
* Remove @inbounds in tuple iteration LLVM can prove this inbounds and the annotation weakens the inferable effects for tuple iteration, which has a surprisingly large inference performance and precision impact. Unfortunately, my previous changes to :inbounds tainting weren't quite strong enough yet, because `getfield` was still tainting consistency on unknown boundscheck arguments. To fix that, we pass through the fargs into the fetfield effects to check if we're getting a literal `:boundscheck`, in which case the `:noinbounds` consistency-tainting logic I added in #48246 is sufficient to not require additional consistency tainting. Also add a test for both effects and codegen to make sure this doens't regress. * Int64 -> Int * fixup typo
aviatesk
added a commit
that referenced
this pull request
Jan 18, 2023
Keno
added a commit
that referenced
this pull request
Feb 14, 2023
We used to have this hack before #48246, but I removed it because I had hoped we don't need. Unfortunately, we weren't able to infer consistency of ``` @inbounds (1,2)[1] ``` With this hack, constprop is able to independently prove inbounded-ness, overriding the usual consistency taintaing that happens for inbounds.
aviatesk
pushed a commit
that referenced
this pull request
Feb 15, 2023
We used to have this hack before #48246, but I removed it because I had hoped we don't need. Unfortunately, we weren't able to infer consistency of ``` @inbounds (1,2)[1] ``` With this hack, constprop is able to independently prove inbounded-ness, overriding the usual consistency taintaing that happens for inbounds.
KristofferC
pushed a commit
that referenced
this pull request
Feb 20, 2023
We used to have this hack before #48246, but I removed it because I had hoped we don't need. Unfortunately, we weren't able to infer consistency of ``` @inbounds (1,2)[1] ``` With this hack, constprop is able to independently prove inbounded-ness, overriding the usual consistency taintaing that happens for inbounds. (cherry picked from commit 113c2f3)
KristofferC
pushed a commit
that referenced
this pull request
Feb 20, 2023
We used to have this hack before #48246, but I removed it because I had hoped we don't need. Unfortunately, we weren't able to infer consistency of ``` @inbounds (1,2)[1] ``` With this hack, constprop is able to independently prove inbounded-ness, overriding the usual consistency taintaing that happens for inbounds. (cherry picked from commit 113c2f3)
KristofferC
pushed a commit
that referenced
this pull request
Feb 21, 2023
We used to have this hack before #48246, but I removed it because I had hoped we don't need. Unfortunately, we weren't able to infer consistency of ``` @inbounds (1,2)[1] ``` With this hack, constprop is able to independently prove inbounded-ness, overriding the usual consistency taintaing that happens for inbounds. (cherry picked from commit 113c2f3)
KristofferC
pushed a commit
that referenced
this pull request
Feb 21, 2023
We used to have this hack before #48246, but I removed it because I had hoped we don't need. Unfortunately, we weren't able to infer consistency of ``` @inbounds (1,2)[1] ``` With this hack, constprop is able to independently prove inbounded-ness, overriding the usual consistency taintaing that happens for inbounds. (cherry picked from commit 113c2f3)
KristofferC
pushed a commit
that referenced
this pull request
Feb 21, 2023
We used to have this hack before #48246, but I removed it because I had hoped we don't need. Unfortunately, we weren't able to infer consistency of ``` @inbounds (1,2)[1] ``` With this hack, constprop is able to independently prove inbounded-ness, overriding the usual consistency taintaing that happens for inbounds. (cherry picked from commit 113c2f3)
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 works to fix #48243, by only tanting effects if an
@inboundsstatement is actually reached. Further, it refines thenoinboundseffect to be IPO-cached and used to track whether a particular method read the inbounds state. A:boundscheckexpression now does not immediately taint consistencty, but instead, taintsnoinboundsonly. Then, if a method that has:noinboundstainted is called within an@inboundsregion, consistency is tainted. Similarly, a tainted:noinboundsdisables constant propagation at@inboundsstatements or if the method propagates inbounds.