rm reconstruction of cached reports, just force inference on "errorne… #38
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
Codecov Report
@@ Coverage Diff @@
## avi/constantpropagation #38 +/- ##
==========================================================
Coverage ? 85.77%
==========================================================
Files ? 9
Lines ? 689
Branches ? 0
==========================================================
Hits ? 591
Misses ? 98
Partials ? 0
Flags with carried forward coverage won't be shown. Click here to find out more. Continue to review full report at Codecov.
|
aviatesk
force-pushed
the
avi/nocache
branch
from
e491503
to
bad03a3
Compare
|
ah, it turns out the cache invalidation can really slow down profiling on really-world code, e.g.: using TypeProfiler, StatsPlots
@time @profile_call rand(100)
@time @profile_call density(rand(100))will take 350 secs from (previously) 15 secs |
…ous" frames - previous "cache clean up" logic potentially throws away true positive errors, while we need that to surpress false positives that can be revealed by constant propagation - this commit removes the whole reconstructions of cached reports, and just force inference on those "erronesout" `MethodInstance`s by returning `default` fallback value to `CC.get(tpc::TPCache, mi::MethodInstance, default)` - as a result, those "erronesous" frames will always re-run inference and so constant propagation will reveal false/true positives whatever it's cached (just because it ignores the cache) - this may slow down abstract interpretation performance, but cache ignorance is only done partially and so hopefully performance won't get so worse (I hope) - at the cost of that, this will simplify the whole code **a lot** and also resolves the previous problems of cache clean up of true positive reports
aviatesk
force-pushed
the
avi/nocache
branch
from
bad03a3
to
daf9e8e
Compare
|
Now above example finishes within 15 secs, which isn't worse than previous implementation at least. the test failure seems unrelated to this PR, going to merge |
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.
…ous" frames
errors, while we need that to surpress false positives that can be
revealed by constant propagation
just force inference on those "erronesout"
MethodInstances byreturning
defaultfallback value toCC.get(tpc::TPCache, mi::MethodInstance, default)so constant propagation will reveal false/true positives whatever it's cached
(just because it ignores the cache)
is only done partially and so hopefully performance won't get so worse (I hope)
the previous problems of cache clean up of true positive reports