Iterate type interfaces without materialising a list during super type matching #1578
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
richardstartin
force-pushed
the
richardstartin/reduce-startup-arraylist-allocation
branch
from
4ae4e0a
to
a550113
Compare
jbachorik
reviewed
Jun 12, 2020
| if (!checkedInterfaces.contains(type)) { | ||
| checkedInterfaces.add(type); | ||
| if (checkedInterfaces.add(type)) { |
...ing/src/main/java/datadog/trace/agent/tooling/bytebuddy/matcher/SafeHasSuperTypeMatcher.java
Show resolved
Hide resolved
| } | ||
|
|
||
| @Override | ||
| public void remove() {} |
There was a problem hiding this comment.
What about throwing UnsupportedOperationException to make it totally clear that this method is not supposed to be called at all.
...ent-tooling/src/main/java/datadog/trace/agent/tooling/bytebuddy/matcher/TypeComparators.java
Outdated
Show resolved
Hide resolved
...ent-tooling/src/main/java/datadog/trace/agent/tooling/bytebuddy/matcher/TypeComparators.java
Outdated
Show resolved
Hide resolved
...src/test/groovy/datadog/trace/agent/tooling/bytebuddy/matcher/HasInterfaceMatcherTest.groovy
Outdated
Show resolved
Hide resolved
richardstartin
force-pushed
the
richardstartin/reduce-startup-arraylist-allocation
branch
from
a550113
to
6801e71
Compare
|
Regarding percentage, I think percentage of total is the number to focus on because we happen to allocate four |
jbachorik
approved these changes
Jun 12, 2020
richardstartin
force-pushed
the
richardstartin/reduce-startup-arraylist-allocation
branch
from
6801e71
to
b2bd5e1
Compare
|
This provokes an interesting failure detected by the netty tests: Why are we trying to resolve a type description for |
bantonsson
reviewed
Jun 12, 2020
| * | ||
| * <p>The caller MUST call hasNext() before calling next(). | ||
| * | ||
| * <p>This wrapper exists to allow getting interfaces even if the lookup on one fails. |
There was a problem hiding this comment.
I know that this comment and behavior is the same as the old one, but maybe clarify that it stops at the first exception.
bantonsson
approved these changes
Jun 12, 2020
richardstartin
force-pushed
the
richardstartin/reduce-startup-arraylist-allocation
branch
from
b2bd5e1
to
b5ca36b
Compare
The handling of this case does turn out to be important here. ByteBuddy first constructs a lazy TypeDescriptor without loading the underlying class file. If later, it needs to load the class file, but cannot -- then an exception is raised. I'd previously done an experiment with creating a custom Iterator that catches the exception to avoid the extra ArrayList-s while handling the exception. At that the time, we were mostly focused on start-up, so I don't end up putting the change in. |
bantonsson
reviewed
Jun 12, 2020
|
|
||
| @Override | ||
| public int compare(TypeDescription o1, TypeDescription o2) { | ||
| return o1.getSimpleName().compareTo(o2.getSimpleName()); |
There was a problem hiding this comment.
I switched to getName and the :dd-java-agent:instrumentation:netty-4.1:testJava11Generated succeeded locally, while getSimpleName fails.
public int compare(TypeDescription o1, TypeDescription o2) {
if (o1 == o2) {
return 0;
}
return o1.getName().compareTo(o2.getName());
}
There was a problem hiding this comment.
Is name sufficient here? We have to be a bit carefully about duplicates being loaded across different ClassLoaders, but in a limited context, name should work.
There was a problem hiding this comment.
I think I'll just change this back to HashSet to avoid edge cases.
richardstartin
force-pushed
the
richardstartin/reduce-startup-arraylist-allocation
branch
from
bf43a07
to
e8302ef
Compare
richardstartin
added
tag: low hanging fruit
richardstartin
changed the title
richardstartin
force-pushed
the
richardstartin/reduce-startup-arraylist-allocation
branch
from
e8302ef
to
9527c77
Compare
richardstartin
deleted the
richardstartin/reduce-startup-arraylist-allocation
branch
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.
reduces allocation during type matching at startup.
Evaluation:
run spring pet-clinic with
-XX:+UnlockExperimentalVMOptions -XX:+UseEpsilonGC -mx1G -XX:+HeapDumpOnOutOfMemoryErrorto capture every allocated object in the first GB allocated. Results in reduction inArrayListallocations by 1% of total (4th most commonly allocated object during startup)Some tests weren't actually testing handling of an exception thrown during iteration, but simulated this by throwing on the
getInterfaces()call - I fixed those.