Eliot Miranda uploaded a new version of Compiler to project The Trunk:
http://source.squeak.org/trunk/Compiler-eem.248.mcz ==================== Summary ==================== Name: Compiler-eem.248 Author: eem Time: 8 February 2013, 11:33:58.616 am UUID: c4da1b1b-b487-44bb-b8ef-392432ca421d Ancestors: Compiler-cwp.247 Slightly less circuitous generation of methods with temp names. Since schematic temp names string is computed during generation, best to generate a method and copy it with the computed temp names than generate two separate methods. =============== Diff against Compiler-cwp.247 =============== Item was changed: ----- Method: MethodNode>>generateWithTempNames (in category 'code generation') ----- generateWithTempNames + "Answer a CompiledMethod with temps names encoded in its trailer." + | methodSansTempNames | + "The encoder computes the schematic temp names during generation, so + generate a method without temp names first. If the method is quick there are + no temps and hence no temp names." + methodSansTempNames := self + generate: CompiledMethodTrailer empty + using: CompiledMethod. + ^methodSansTempNames + copyWithTempNames: (methodSansTempNames isQuick + ifTrue: [''] + ifFalse: [encoder schematicTempNamesString])! - "Answer a CompiledMethod with temps names encoded in trailer" - ^self - generate: (CompiledMethodTrailer new tempNames: self schematicTempNamesString) - using: CompiledMethod! |
On 8 February 2013 19:34, <[hidden email]> wrote:
> Eliot Miranda uploaded a new version of Compiler to project The Trunk: > http://source.squeak.org/trunk/Compiler-eem.248.mcz > > ==================== Summary ==================== > > Name: Compiler-eem.248 > Author: eem > Time: 8 February 2013, 11:33:58.616 am > UUID: c4da1b1b-b487-44bb-b8ef-392432ca421d > Ancestors: Compiler-cwp.247 > > Slightly less circuitous generation of methods with temp > names. Since schematic temp names string is computed > during generation, best to generate a method and copy > it with the computed temp names than generate two > separate methods. > > =============== Diff against Compiler-cwp.247 =============== It'd be great to automate reporting of test regressions. This (*) causes two regressions: Tests.Compiler.ClosureCompilerTest.testBlockDoitDecompilation [1] Tests.Compiler.ClosureCompilerTest.testDecompiledDoitMethodTempNames [2] [1] http://build.squeak.org/job/SqueakTrunk/160/testReport/junit/Tests.Compiler/ClosureCompilerTest/testBlockDoitDecompilation/ [2] http://build.squeak.org/job/SqueakTrunk/160/testReport/junit/Tests.Compiler/ClosureCompilerTest/testDecompiledDoitMethodTempNames/ frank (*) "This" really just being "this with high probability" because there's currently no indication in Jenkins as to how build N+1 differs source-wise from build N. |
Thanks, Frank. Fixed.
On Fri, Feb 8, 2013 at 12:08 PM, Frank Shearar <[hidden email]> wrote:
best, Eliot
|
Free forum by Nabble | Edit this page |