Investigate why certain .ir files are bigger in the new format

[JaroslavTulach]

Investigate why certain .ir files like Standard/Database/Data/Table.ir are bigger in the new persistance format introduced by #8207 - the difference is 6616791 vs. 4245813 - e.g. 35%.

Possibly UUID instances are stored where they weren't previously. Possibly there is some duplication in the tree/graph caused by some objects being serialized multiple times.

[JaroslavTulach]

e0b27a2 contains a histogram of types inside of .ir files (the IR is traversed with .preorder). There is for example:

Loaded Standard.Base.System.File with 3850 IR elements
   4903 scala.None$
   3909 scala.collection.immutable.Nil$
   3836 org.enso.compiler.core.ir.IrLazyMap
   3836 org.enso.compiler.core.ir.DiagnosticStorage
   3836 org.enso.compiler.core.ir.MetadataStorage
   3487 scala.Some
   1813 org.enso.compiler.core.ir.IdentifiedLocation
   1813 org.enso.compiler.core.ir.Location
   1345 org.enso.compiler.core.ir.Name$Literal
   726 scala.collection.immutable.$colon$colon
   715 org.enso.compiler.core.ir.CallArgument$Specified
   390 org.enso.compiler.core.ir.expression.Application$Prefix
   232 org.enso.compiler.core.ir.Name$Self
   229 org.enso.compiler.core.ir.IrLazySeq
   224 org.enso.compiler.core.ir.DefinitionArgument$Specified
   133 org.enso.compiler.core.ir.Name$Qualified
   119 org.enso.compiler.core.ir.Function$Lambda
   107 org.enso.compiler.core.ir.Expression$Block
   98 java.util.UUID
   95 org.enso.compiler.core.ir.Literal$Text
   94 org.enso.compiler.core.ir.module.scope.definition.Method$Explicit
   94 org.enso.compiler.core.ir.Name$MethodReference
   76 org.enso.compiler.core.ir.Expression$Binding
   40 org.enso.compiler.core.ir.expression.Application$Force
   38 org.enso.compiler.core.ir.module.scope.Import$Module
   33 org.enso.compiler.core.ir.expression.Case$Branch
   22 org.enso.compiler.core.ir.expression.Application$Sequence
   20 org.enso.compiler.core.ir.Pattern$Constructor
   16 org.enso.compiler.core.ir.expression.Case$Expr
   11 org.enso.compiler.core.ir.Name$Blank

and here are the most frequent classes in the biggest file Table.ir:

 Loaded Standard.Table.Data.Table with 14428 IR elements
   18422 scala.None$
   14729 scala.collection.immutable.Nil$
   14504 org.enso.compiler.core.ir.DiagnosticStorage
   14504 org.enso.compiler.core.ir.MetadataStorage
   14504 org.enso.compiler.core.ir.IrLazyMap
   14325 scala.Some
   7290 org.enso.compiler.core.ir.Location
   7290 org.enso.compiler.core.ir.IdentifiedLocation
   5814 org.enso.compiler.core.ir.Name$Literal
   3274 org.enso.compiler.core.ir.CallArgument$Specified
   2531 scala.collection.immutable.$colon$colon
   1679 org.enso.compiler.core.ir.expression.Application$Prefix
   715 org.enso.compiler.core.ir.DefinitionArgument$Specified
   494 org.enso.compiler.core.ir.Name$Self
   475 org.enso.compiler.core.ir.IrLazySeq
   412 org.enso.compiler.core.ir.Expression$Binding
   378 org.enso.compiler.core.ir.Expression$Block
   278 org.enso.compiler.core.ir.Function$Lambda
   229 org.enso.compiler.core.ir.Name$Qualified
   213 java.util.UUID
   149 org.enso.compiler.core.ir.Name$MethodReference
   142 org.enso.compiler.core.ir.module.scope.definition.Method$Explicit
   135 org.enso.compiler.core.ir.expression.Case$Branch
   133 org.enso.compiler.core.ir.expression.Application$Force
   108 org.enso.compiler.core.ir.Literal$Text
   78 org.enso.compiler.core.ir.Literal$Number
   76 org.enso.compiler.core.ir.Name$Blank
   75 org.enso.compiler.core.ir.expression.Application$Sequence
   64 org.enso.compiler.core.ir.module.scope.Import$Module
   59 org.enso.compiler.core.ir.Pattern$Constructor

[JaroslavTulach]

With 634e9fb and when requesting serialization of Table library via runEngineDistribution --log-level debug --compile built-distribution/enso-engine-0.0.0-dev-linux-amd64/enso-0.0.0-dev/lib/Standard/Table/0.0.0-dev/ one can see that a lot of bytes (the first number) is occupied by AliasAnalysis pass, surprisingly there is just 801 instances (second number):

  456636 801 org.enso.compiler.pass.analyse.AliasAnalysis$Info$Occurrence
  308128 523 org.enso.compiler.pass.analyse.AliasAnalysis$Info$Scope$Child
  73980 905 org.enso.compiler.core.ir.Name$Literal
  68560 3428 org.enso.compiler.pass.analyse.DataflowAnalysis$DependencyInfo$Type$Static
  54136 2 scala.collection.mutable.HashMap
  38640 3220 scala.collection.immutable.Set$Set1
  31676 434 org.enso.compiler.core.ir.CallArgument$Specified
  27490 1742 org.enso.compiler.data.BindingsMap$Cons
  25776 942 org.enso.compiler.pass.analyse.AliasAnalysis$Graph$Scope
  22368 1848 org.enso.compiler.core.ir.IdentifiedLocation
  20360 232 org.enso.compiler.core.ir.expression.Application$Prefix
  19277 557 org.enso.compiler.pass.analyse.AliasAnalysis$Graph$Occurrence$Use
  14939 82 org.enso.compiler.data.BindingsMap$ResolvedConstructor
  10490 249 org.enso.compiler.data.BindingsMap$ModuleReference$Abstract
  7325 208 org.enso.compiler.pass.analyse.AliasAnalysis$Graph$Occurrence$Def

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2023-12-27):

Progress: - analyzing IR files: #8323 (comment)

• Down to 4MB: Store Graph as a reference #8635
• debugging vitest
• asking on Graal.js channel: https://graalvm.slack.com/archives/CN9L59P61/p1703648239128539 It should be finished by 2024-01-02.

Next Day: Continue making .ir file smaller

Slack

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2023-12-28):

Progress: - Addressing review comments

• Analyzing benchmarks: Fix build of engine benchmarks #8620 (comment)
• down to 3,5MB: 99284c1
• investigating better storage for Name.Literal
• investigating better storeage for Location
• analyzing second benchmarks: Good
• only one ModuleReference.Abstract: 2b178e4
• Fresh PerMap: a9d793c
• analyzing SerdeCompilerTest failure
• bugfix: b856ecf
• Issue for singe file cache: Use single cache file per library #8639
• rewritting to logging
• Updating espresso doc: Updating Espresso support & docs to match GraalVM for JDK21 #8641 It should be finished by 2024-01-02.

Next Day: Continue making .ir file smaller

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2023-12-29):

Progress: - merged: #8635

• track logging issue: Cannot enable org.enso.persist logger via a property #8644
• one Scala class less: 8b89ec8
• autoscoping of constructors: Autoscoping of constructors #8645 (comment) It should be finished by 2024-01-02.

Next Day: Whatever new year brings