Speedup EnsoRootNode.buildFrameDescriptor

[Akirathan]

Follow-up of #10837 and #10729, part of #10833.

enso/engine/runtime/src/main/java/org/enso/interpreter/node/EnsoRootNode.java

Lines 65 to 74 in 396d70d

	private static FrameDescriptor buildFrameDescriptor(LocalScope localScope) {
	var descriptorBuilder = FrameDescriptor.newBuilder();
	descriptorBuilder.addSlot(FrameSlotKind.Object, LocalScope.monadicStateSlotName(), null);
	for (var definition : ScalaConversions.asJava(localScope.scope().allDefinitions())) {
	descriptorBuilder.addSlot(FrameSlotKind.Illegal, definition.symbol(), null);
	}
	descriptorBuilder.defaultValue(DataflowError.UNINITIALIZED);
	var frameDescriptor = descriptorBuilder.build();
	return frameDescriptor;
	}

still uses LocalScope.scope().allDefinitions() that forces deserialization of alias analysis metadata for every EnsoRootNode created, that is created for at least every Enso method. Let's remove the dependency on alias metadata from EnsoRootNode. Optimally, we should use only FramePointer metadata for that.

Further notes in #10837 (comment)

[JaroslavTulach]

Looks like FramePointer will not be enough. I need to associate result of allDefinitions() with every IR that contains AliasMetadata.RootScope I think now.

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2024-08-30):

Progress: .

• another fix to the CI output: Fold output of each benchmark #10914 (comment)
• trying to order messages properly: Fold output of each benchmark #10914 (comment)
• FramePointer updates while traversion: abb7e02
• rewriting logging in IrToTruffle:
• standup
• write test for FramePointer variable names: 1b28c4a
• folding fixes: Fold output of each benchmark #10914 (comment) It should be finished by 2024-09-05.

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2024-09-01):

Progress: .

• debugging IR: a41aab6
• testFnAsADefaultValue passes: fb37b39 It should be finished by 2024-09-05.

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2024-09-02):

Progress: .

• test/Base_Tests is starting without loading AliasAnalysis: Provide names of local variables via FramePointerAnalysis #10906
• Graph.freeze proposal: https://discord.com/channels/@me/959434339182067712/1280188201809612873
• standups
• meeting Dmitry (eliminate scala.Option in heap)
• meeting about BindingsMap with Hubert and Pavel
• demo meeting with demo: https:/orgs/enso-org/discussions/10943#discussioncomment-10520821 It should be finished by 2024-09-05.

Discord

Discord - Group Chat That’s All Fun & Games

Discord is great for playing games and chilling with friends, or even building a worldwide community. Customize your own space to talk, play, and hang out.

GitHub

[2024-09-02] Library & Engine Team Update · enso-org · Discussion #10943

James: Tuning S3_File doc comments. #10832 James: Wip/jd/fast excel read #10877 James: Update what needs to be re-signed. #10922 James: Small tweaks from QA #10941 Radosław: Fix Count_Distinct in S...

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2024-09-03):

Progress: .

• Encapsulating Graph with freeze: c975066
• another backport request for OracleLabs: Random CI failure on Oracle GraalVM: jdk.internal.vm.TranslatedException: Error decoding exception #9457 (comment)
• who attaches FramePointerMeta twice?
• Case.Expr.duplicate: Provide names of local variables via FramePointerAnalysis #10906 (comment)
• reported auto-TCO issue: Automatically detect suitable @Tail_Call optimization #10956
• planning meeting
• ready for review: Provide names of local variables via FramePointerAnalysis #10906 It should be finished by 2024-09-05.

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2024-09-04):

Progress: .

• startup benchmarks are OK: Provide names of local variables via FramePointerAnalysis #10906 (comment)
• reacting to review comments: 90a5e46
• removing Python and JavaScript from native image launcher: Get ready for python and js being optional when running test/Base_Tests #10975
• eliminate scala.Option: Eliminate scala.Some 10% overhead in Enso IR by inlining objects together #10973 (comment) It should be finished by 2024-09-05.

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2024-09-05):

Progress: .

• PR merged: Provide names of local variables via FramePointerAnalysis #10906
• mini phases issue: Avoid IR traversal by converting (some of) our Passes to mini passes #10981
• PR merged: Get ready for python and js being optional when running test/Base_Tests #10975
• libs standup
• unsuccessfully trying to remove Python and JS from NI
• checking deserialization problems - imports being resolved too early It should be finished by 2024-09-05.