Hi, 
Guest
Pic
Sign In

pharo-dev@lists.pharo.org

Pharo Development List

View all threads

This week (46/2022) on the Pharo Issue Tracker

MD
Marcus Denker
Fri, Nov 18, 2022 7:15 AM

We merged 18 PRs

Lots of improvements related to  BlockClosures this week.

First some helper methods on the AST:

  • #hasNonLocalReturn: check if there is a non-local return
  • #isConstant: Blocks of the kind [#justSomeLiteral]
  • #constantValue: return that literal (and nil for empty [] )
  • #numArgs for RBBlockNode, as we have already for BlockClosure  "[:a | ] sourceNode numArgs"

Some improvements for isClean:

  • unify between AST and BlockClosure
  • more tests
  • inlined blocks can be clean, important as it is recursive: [true ifTrue: []] should be clean
  • cleanup code in CleanBlock, receiver is known to be nil

Compiler support for ConstantBlockClosure:

  • We can specialize clean blocks of the kind [#justSomeLiteral] to execute faster
  • add ConstantBlockClosure, add compiler support, active when compiling with clean blocks enabled

Compiler support for full block without outer context:

  • we only need the outerContext for non-local returns.
  • add support to IRBuilder for full blocks without outer context
  • implement compiler support (option #optionBlockClosureOptionalOuter, false for now)

Quick return for Blocks:

  • Compile quickreturn for CompiledBlock, just like CompiledMethod (e.g. for [^1]).
  • speeds up execution a little (but of course not creation)

Block Closures

Fixes

Cleanup

Comments

We merged 18 PRs Lots of improvements related to BlockClosures this week. First some helper methods on the AST: - #hasNonLocalReturn: check if there is a non-local return - #isConstant: Blocks of the kind [#justSomeLiteral] - #constantValue: return that literal (and nil for empty [] ) - #numArgs for RBBlockNode, as we have already for BlockClosure "[:a | ] sourceNode numArgs" Some improvements for isClean: - unify between AST and BlockClosure - more tests - inlined blocks can be clean, important as it is recursive: [true ifTrue: []] should be clean - cleanup code in CleanBlock, receiver is known to be nil Compiler support for ConstantBlockClosure: - We can specialize clean blocks of the kind [#justSomeLiteral] to execute faster - add ConstantBlockClosure, add compiler support, active when compiling with clean blocks enabled Compiler support for full block without outer context: - we only need the outerContext for non-local returns. - add support to IRBuilder for full blocks without outer context - implement compiler support (option #optionBlockClosureOptionalOuter, false for now) Quick return for Blocks: - Compile quickreturn for CompiledBlock, just like CompiledMethod (e.g. for [^1]). - speeds up execution a little (but of course not creation) Block Closures ============== - AST: add a recursive #hasNonLocalReturn and #isConstant #11882 https://github.com/pharo-project/pharo/pull/11882 - add #numArgs to RBBlockNode #11902 https://github.com/pharo-project/pharo/pull/11902 - CleanBlock cleanup code: there should be no receiver ivar #11884 https://github.com/pharo-project/pharo/pull/11884 - 11881-RBBlockNodeisClean-takes-every-optimized-block-to-be-not-clean #11890 https://github.com/pharo-project/pharo/pull/11890 - 1883-CleanBlockChecker-does-not-recurse-into-the-blocks #11900 https://github.com/pharo-project/pharo/pull/11900 - ConstantBlockClosure: active when compiling with clean blocks enabled #11894 https://github.com/pharo-project/pharo/pull/11894 - CompiledBlocks: compile quick returns #11907 https://github.com/pharo-project/pharo/pull/11907 - IRBuilder: add support for outerContext flag for Closure creation #11905 https://github.com/pharo-project/pharo/pull/11905 - 11911-Compiler-add-compiler-option-for-outerContextNeeded #11917 https://github.com/pharo-project/pharo/pull/11917 Fixes ===== - Register-as-basicInspector #431 https://github.com/pharo-spec/NewTools/pull/431 - 11803 return value of inspect should be the receiver and not a sp window presenter #11886 https://github.com/pharo-project/pharo/pull/11886 - Fixes #429, VersionBrowser: left pane empty #430 https://github.com/pharo-spec/NewTools/pull/430 - Turn off ReLongMethodsRule for test methods #11892 https://github.com/pharo-project/pharo/pull/11892 - ClassParser: Implement #asSlot to return a correct Slot instances for complex slots #11898 https://github.com/pharo-project/pharo/pull/11898 - Fixes #11802 - Adds full consistency checks to class Time #11891 https://github.com/pharo-project/pharo/pull/11891 - Fixing to correctly get the VM for running the tests #11897 https://github.com/pharo-project/pharo/pull/11897 - Fixes #429, VersionBrowser: left pane empty #430 https://github.com/pharo-spec/NewTools/pull/430 Cleanup ======= - 11810-FT2Handle-classstartUp-compiled-code-on-image-startup #11811 https://github.com/pharo-project/pharo/pull/11811 Comments ========= - fixing comments in package Regex #11888 https://github.com/pharo-project/pharo/pull/11888
TM
Tim Mackinnon
Fri, Nov 18, 2022 9:06 AM

The attention to detail on these cleanups to make everything symmetrical is very nice - cool work.

Tim

On Fri, 18 Nov 2022, at 7:15 AM, Marcus Denker wrote:

We merged 18 PRs

Lots of improvements related to  BlockClosures this week.

First some helper methods on the AST:

  • #hasNonLocalReturn: check if there is a non-local return
  • #isConstant: Blocks of the kind [#justSomeLiteral]
  • #constantValue: return that literal (and nil for empty [] )
  • #numArgs for RBBlockNode, as we have already for BlockClosure  "[:a |
    ] sourceNode numArgs"

Some improvements for isClean:

  • unify between AST and BlockClosure
  • more tests
  • inlined blocks can be clean, important as it is recursive: [true
    ifTrue: []] should be clean
  • cleanup code in CleanBlock, receiver is known to be nil

Compiler support for ConstantBlockClosure:

  • We can specialize clean blocks of the kind [#justSomeLiteral] to
    execute faster
  • add ConstantBlockClosure, add compiler support, active when compiling
    with clean blocks enabled

Compiler support for full block without outer context:

  • we only need the outerContext for non-local returns.
  • add support to IRBuilder for full blocks without outer context
  • implement compiler support (option #optionBlockClosureOptionalOuter,
    false for now)

Quick return for Blocks:

  • Compile quickreturn for CompiledBlock, just like CompiledMethod (e.g.
    for [^1]).
  • speeds up execution a little (but of course not creation)

Block Closures

Fixes

Cleanup

Comments

The attention to detail on these cleanups to make everything symmetrical is very nice - cool work. Tim On Fri, 18 Nov 2022, at 7:15 AM, Marcus Denker wrote: > We merged 18 PRs > > Lots of improvements related to BlockClosures this week. > > First some helper methods on the AST: > - #hasNonLocalReturn: check if there is a non-local return > - #isConstant: Blocks of the kind [#justSomeLiteral] > - #constantValue: return that literal (and nil for empty [] ) > - #numArgs for RBBlockNode, as we have already for BlockClosure "[:a | > ] sourceNode numArgs" > > Some improvements for isClean: > - unify between AST and BlockClosure > - more tests > - inlined blocks can be clean, important as it is recursive: [true > ifTrue: []] should be clean > - cleanup code in CleanBlock, receiver is known to be nil > > Compiler support for ConstantBlockClosure: > - We can specialize clean blocks of the kind [#justSomeLiteral] to > execute faster > - add ConstantBlockClosure, add compiler support, active when compiling > with clean blocks enabled > > Compiler support for full block without outer context: > - we only need the outerContext for non-local returns. > - add support to IRBuilder for full blocks without outer context > - implement compiler support (option #optionBlockClosureOptionalOuter, > false for now) > > Quick return for Blocks: > - Compile quickreturn for CompiledBlock, just like CompiledMethod (e.g. > for [^1]). > - speeds up execution a little (but of course not creation) > > Block Closures > ============== > > - AST: add a recursive #hasNonLocalReturn and #isConstant #11882 > https://github.com/pharo-project/pharo/pull/11882 > > - add #numArgs to RBBlockNode #11902 > https://github.com/pharo-project/pharo/pull/11902 > > - CleanBlock cleanup code: there should be no receiver ivar #11884 > https://github.com/pharo-project/pharo/pull/11884 > > - 11881-RBBlockNodeisClean-takes-every-optimized-block-to-be-not-clean #11890 > https://github.com/pharo-project/pharo/pull/11890 > > - 1883-CleanBlockChecker-does-not-recurse-into-the-blocks #11900 > https://github.com/pharo-project/pharo/pull/11900 > > - ConstantBlockClosure: active when compiling with clean blocks enabled #11894 > https://github.com/pharo-project/pharo/pull/11894 > > - CompiledBlocks: compile quick returns #11907 > https://github.com/pharo-project/pharo/pull/11907 > > - IRBuilder: add support for outerContext flag for Closure creation #11905 > https://github.com/pharo-project/pharo/pull/11905 > > - 11911-Compiler-add-compiler-option-for-outerContextNeeded #11917 > https://github.com/pharo-project/pharo/pull/11917 > > > > Fixes > ===== > > - Register-as-basicInspector #431 > https://github.com/pharo-spec/NewTools/pull/431 > > - 11803 return value of inspect should be the receiver and not a sp > window presenter #11886 > https://github.com/pharo-project/pharo/pull/11886 > > - Fixes #429, VersionBrowser: left pane empty #430 > https://github.com/pharo-spec/NewTools/pull/430 > > - Turn off ReLongMethodsRule for test methods #11892 > https://github.com/pharo-project/pharo/pull/11892 > > - ClassParser: Implement #asSlot to return a correct Slot instances for > complex slots #11898 > https://github.com/pharo-project/pharo/pull/11898 > > - Fixes #11802 - Adds full consistency checks to class Time #11891 > https://github.com/pharo-project/pharo/pull/11891 > > - Fixing to correctly get the VM for running the tests #11897 > https://github.com/pharo-project/pharo/pull/11897 > > - Fixes #429, VersionBrowser: left pane empty #430 > https://github.com/pharo-spec/NewTools/pull/430 > > Cleanup > ======= > - 11810-FT2Handle-classstartUp-compiled-code-on-image-startup #11811 > https://github.com/pharo-project/pharo/pull/11811 > > > > Comments > ========= > > - fixing comments in package Regex #11888 > https://github.com/pharo-project/pharo/pull/11888
MD
Marcus Denker
Fri, Nov 18, 2022 10:19 AM

On 18 Nov 2022, at 08:15, Marcus Denker marcus.denker@inria.fr wrote:

We merged 18 PRs

Lots of improvements related to  BlockClosures this week.

First some helper methods on the AST:

  • #hasNonLocalReturn: check if there is a non-local return
  • #isConstant: Blocks of the kind [#justSomeLiteral]
  • #constantValue: return that literal (and nil for empty [] )
  • #numArgs for RBBlockNode, as we have already for BlockClosure  "[:a | ] sourceNode numArgs”

Some numbers:

"there are lots of blocks"
allBlocks := Smalltalk globals allMethods flatCollect: [:meth | meth ast blockNodes ].
allBlocks size. "100140"

"but many are compiled inline (eg argunments of #ifTrue:)"
currentFullBlocks := allBlocks select: [:blockNode | blockNode isInlined not].
currentFullBlocks size. "45193"

"What we could compile as CleanBlockClosure"
cleanBlocks := currentFullBlocks select: [:blockNode | blockNode isClean].
cleanBlocks size. "11090"

"many clean blocks are actually constant"
constantBlocks := cleanBlocks select: [:blockNode | blockNode isConstant].
constantBlocks size. "3200"

"FullBlocks that need the outerContext to return"
fullBocksWithReturn := currentFullBlocks select: [ :each  | each hasNonLocalReturn ].
fullBocksWithReturn size “3816”

We can inspect the collections and then use the inspector to browse the code (with the block
that you look at hightlighted), to get a feel where these are used, e.g. for constant:

> On 18 Nov 2022, at 08:15, Marcus Denker <marcus.denker@inria.fr> wrote: > > We merged 18 PRs > > Lots of improvements related to BlockClosures this week. > > First some helper methods on the AST: > - #hasNonLocalReturn: check if there is a non-local return > - #isConstant: Blocks of the kind [#justSomeLiteral] > - #constantValue: return that literal (and nil for empty [] ) > - #numArgs for RBBlockNode, as we have already for BlockClosure "[:a | ] sourceNode numArgs” > Some numbers: "there are lots of blocks" allBlocks := Smalltalk globals allMethods flatCollect: [:meth | meth ast blockNodes ]. allBlocks size. "100140" "but many are compiled inline (eg argunments of #ifTrue:)" currentFullBlocks := allBlocks select: [:blockNode | blockNode isInlined not]. currentFullBlocks size. "45193" "What we could compile as CleanBlockClosure" cleanBlocks := currentFullBlocks select: [:blockNode | blockNode isClean]. cleanBlocks size. "11090" "many clean blocks are actually constant" constantBlocks := cleanBlocks select: [:blockNode | blockNode isConstant]. constantBlocks size. "3200" "FullBlocks that need the outerContext to return" fullBocksWithReturn := currentFullBlocks select: [ :each | each hasNonLocalReturn ]. fullBocksWithReturn size “3816” We can inspect the collections and then use the inspector to browse the code (with the block that you look at hightlighted), to get a feel where these are used, e.g. for constant:
Attached image
TM
Tim Mackinnon
Fri, Nov 18, 2022 11:09 AM

Wow - thats even neater... all done with our normal tools - on the fly. So cool. Not sure many environments/languages can reason about things in this way.

Worthy of a blog post for the wider world I would say - as this kind of stuff should be on more people's radars.

Thanks again for all this kind of work. It would be so easy to just continue to think the job was done 20 years ago, where in fact we've only scratched the surface as our understanding of effective software development has improved.

Tim

On Fri, 18 Nov 2022, at 10:19 AM, Marcus Denker wrote:

On 18 Nov 2022, at 08:15, Marcus Denker marcus.denker@inria.fr wrote:

We merged 18 PRs

Lots of improvements related to  BlockClosures this week.

First some helper methods on the AST:

  • #hasNonLocalReturn: check if there is a non-local return
  • #isConstant: Blocks of the kind [#justSomeLiteral]
  • #constantValue: return that literal (and nil for empty [] )
  • #numArgs for RBBlockNode, as we have already for BlockClosure  "[:a | ] sourceNode numArgs”

Some numbers:

"there are lots of blocks"
allBlocks := Smalltalk globals allMethods flatCollect: [:meth | meth ast blockNodes ].
allBlocks size. "100140"

"but many are compiled inline (eg argunments of #ifTrue:)"
currentFullBlocks := allBlocks select: [:blockNode | blockNode isInlined not].
currentFullBlocks size. "45193"

"What we could compile as CleanBlockClosure"
cleanBlocks := currentFullBlocks select: [:blockNode | blockNode isClean].
cleanBlocks size. "11090"

"many clean blocks are actually constant"
constantBlocks := cleanBlocks select: [:blockNode | blockNode isConstant].
constantBlocks size. "3200"

"FullBlocks that need the outerContext to return"
fullBocksWithReturn := currentFullBlocks select: [ :each  | each hasNonLocalReturn ].
fullBocksWithReturn size “3816”

We can inspect the collections and then use the inspector to browse the code (with the block
that you look at hightlighted), to get a feel where these are used, e.g. for constant:

Wow - thats even neater... all done with our normal tools - on the fly. So cool. Not sure many environments/languages can reason about things in this way. Worthy of a blog post for the wider world I would say - as this kind of stuff should be on more people's radars. Thanks again for all this kind of work. It would be so easy to just continue to think the job was done 20 years ago, where in fact we've only scratched the surface as our understanding of effective software development has improved. Tim On Fri, 18 Nov 2022, at 10:19 AM, Marcus Denker wrote: > > > > On 18 Nov 2022, at 08:15, Marcus Denker <marcus.denker@inria.fr> wrote: > > > > We merged 18 PRs > > > > Lots of improvements related to BlockClosures this week. > > > > First some helper methods on the AST: > > - #hasNonLocalReturn: check if there is a non-local return > > - #isConstant: Blocks of the kind [#justSomeLiteral] > > - #constantValue: return that literal (and nil for empty [] ) > > - #numArgs for RBBlockNode, as we have already for BlockClosure "[:a | ] sourceNode numArgs” > > > > Some numbers: > > "there are lots of blocks" > allBlocks := Smalltalk globals allMethods flatCollect: [:meth | meth ast blockNodes ]. > allBlocks size. "100140" > > "but many are compiled inline (eg argunments of #ifTrue:)" > currentFullBlocks := allBlocks select: [:blockNode | blockNode isInlined not]. > currentFullBlocks size. "45193" > > "What we could compile as CleanBlockClosure" > cleanBlocks := currentFullBlocks select: [:blockNode | blockNode isClean]. > cleanBlocks size. "11090" > > "many clean blocks are actually constant" > constantBlocks := cleanBlocks select: [:blockNode | blockNode isConstant]. > constantBlocks size. "3200" > > "FullBlocks that need the outerContext to return" > fullBocksWithReturn := currentFullBlocks select: [ :each | each hasNonLocalReturn ]. > fullBocksWithReturn size “3816” > > We can inspect the collections and then use the inspector to browse the code (with the block > that you look at hightlighted), to get a feel where these are used, e.g. for constant: > > > >
Attached image
MD
Marcus Denker
Fri, Nov 25, 2022 12:40 PM

Hi

T

On 18 Nov 2022, at 12:09, Tim Mackinnon tim@testit.works wrote:

Wow - thats even neater... all done with our normal tools - on the fly. So cool. Not sure many environments/languages can reason about things in this way.

Worthy of a blog post for the wider world I would say - as this kind of stuff should be on more people's radars.

Yes, I will do that.

Thanks again for all this kind of work. It would be so easy to just continue to think the job was done 20 years ago, where in fact we've only scratched the surface as our understanding of effective software development has improved.

Thanks. Sometimes I think my problem was that when I discovered Smalltalk I did not see perfection in it  but an interesting starting point…

Marcus
Hi T > On 18 Nov 2022, at 12:09, Tim Mackinnon <tim@testit.works> wrote: > > Wow - thats even neater... all done with our normal tools - on the fly. So cool. Not sure many environments/languages can reason about things in this way. > > Worthy of a blog post for the wider world I would say - as this kind of stuff should be on more people's radars. Yes, I will do that. > > Thanks again for all this kind of work. It would be so easy to just continue to think the job was done 20 years ago, where in fact we've only scratched the surface as our understanding of effective software development has improved. > Thanks. Sometimes I think my problem was that when I discovered Smalltalk I did *not* see perfection in it but an interesting starting point… Marcus
Empathy v1.0   2024 ©Harmonylists.com