Vm-dev October 2020

vm-dev@lists.squeakfoundation.org

26 participants
351 discussions

[OpenSmalltalk/opensmalltalk-vm] 17625f: CogVM source as per VMMaker.oscog-eem.2859
by Eliot Miranda 29 Oct '20

29 Oct '20

Branch: refs/heads/Cog Home: https://github.com/OpenSmalltalk/opensmalltalk-vm Commit: 17625fc20fa1f55ed11a728329f002e9feddb459 https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/17625fc20fa1f55ed1… Author: Eliot Miranda <eliot.miranda(a)gmail.com> Date: 2020-10-28 (Wed, 28 Oct 2020) Changed paths: M nsspur64src/vm/cogit.c M spur64src/vm/cogit.c M spurlowcode64src/vm/cogit.c M spursista64src/vm/cogit.c Log Message: ----------- CogVM source as per VMMaker.oscog-eem.2859 Allow processor predefined macros to include expressions (i.e. __ARM_ARCH >= 8). [ci skip]

1 0

VM Maker: Cog-eem.420.mcz
by commits＠source.squeak.org 29 Oct '20

29 Oct '20

Eliot Miranda uploaded a new version of Cog to project VM Maker: http://source.squeak.org/VMMaker/Cog-eem.420.mcz ==================== Summary ==================== Name: Cog-eem.420 Author: eem Time: 28 October 2020, 7:15:44.003408 pm UUID: b03ad887-622e-4bf4-9fe4-7470af59aa6c Ancestors: Cog-eem.419 Fix the stack stitching code. Fix MultiProcessor's desNotUnderstand: handler to return the non-virtual processor to the second trap continuation. ANd no breaks for the wicked... =============== Diff against Cog-eem.419 =============== Item was changed: ----- Method: MultiProcessor>>doesNotUnderstand: (in category 'message forwarding') ----- doesNotUnderstand: aMessage "Forward a message to the actual processor, managing a thread-switch if necessary. Catch ProcessorSimulationTraps and raise them outside of the critical section to avoid deadlock when reentering the VM from a trap and switching threads in the run-time." | selector result trap contextInCritical | selector := aMessage selector. (guardedProcessorProtocol includes: selector) ifFalse: [^(unguardedProcessorProtocol includes: selector) ifTrue: [processor perform: selector withArguments: aMessage arguments] ifFalse: [super doesNotUnderstand: aMessage]]. result := [mutex critical: [contextInCritical := thisContext. owner ~~ mutex owningProcess ifTrue: [owner ifNotNil: [registerState at: owner put: processor registerState]. (registerState at: (owner := mutex owningProcess) ifAbsent: nil) ifNil: [coInterpreter initializeProcessorForThreadIndex: (threadIndex := threadIndex + 1)] ifNotNil: [:newState| processor setRegisterState: newState]]. processor perform: selector withArguments: aMessage arguments]] on: ProcessorSimulationTrap, Error, AssertionFailure do: [:ex| ex class == ProcessorSimulationTrap ifFalse: [ex pass]. "Alas things are not so simple with ARMv8 ldp/stp... The comments in these two methods explain..." (self isAboutToReturn: ex signalerContext) ifFalse: [self saveStackFor: ex signalerContext above: contextInCritical homeContext: thisContext home]. trap := ex]. + ^trap + ifNil: [result] + ifNotNil: [trap signal ifNotNil: + [:resultOfSignal| + resultOfSignal == self + ifTrue: [processor] + ifFalse: [resultOfSignal]]]! - ^trap ifNil: [result] ifNotNil: [trap signal]! Item was changed: ----- Method: MultiProcessor>>saveStackFor:above:homeContext: (in category 'private-deep hackery') ----- saveStackFor: signalerContext above: contextInCritical homeContext: doesNotUnderstandContext "Some processors, notably ARMv8, support load/store pair instructions which may raise two ProcessorSimulationTraps. To handle this we must preserve the continuation following the + raising of the first simulation trap. This cuts up, and restitches the stack so that when the - raising of the first simulaiton trap. This cuts up, and restitches the stack so that when the first ProcessorSimulationTrap is raised outside the critical: block in our doesNotUnderstand: method, control continues to the second trap." + | contextAboveContextInCritical contextAboveSignalerContext | + "This is the bottom context in the saved continuation, which must be stitched in beneath the doesNotUnderstand: handler." + contextAboveContextInCritical := signalerContext findContextSuchThat: [:ctxt| ctxt sender == contextInCritical]. - | contextAboveSignalerContext | "This is our continuation, which should be changed to continue back into the critical section." contextAboveSignalerContext := thisContext findContextSuchThat: [:ctxt| ctxt sender == signalerContext]. self deny: contextAboveSignalerContext isNil. "signalerContext through the context above contextInCritical is the stack that is the continuation for the trap. This must be inserted between the doesNotUnderstandContext and its sender. N.B. Context>>cut: cuts above the argument!!!!" signalerContext cut: contextInCritical. + contextAboveContextInCritical privSender: doesNotUnderstandContext sender. - contextInCritical privSender: doesNotUnderstandContext sender. doesNotUnderstandContext privSender: signalerContext. "Now arrange that the handler returns to contextInCritical; and we're done." contextAboveSignalerContext privSender: contextInCritical! Item was changed: ----- Method: ProcessorSimulatorPlugin>>singleStepCPU:In:Size:MinAddressRead:Write: (in category 'simulation') ----- singleStepCPU: cpu In: memoryCArray Size: memorySize MinAddressRead: minAddress Write: minWriteMaxExecAddress <doNotGenerate> "*now* we need derived pointers. Ho hum... But all we need is one more level of indirection..." | result | - self break. result := mySimulatorAlien rawPrimitiveSingleStepInMemory: interpreterProxy memory offsetBy: memoryCArray ptrAddress minimumAddress: minAddress readOnlyBelow: minWriteMaxExecAddress. ^result == mySimulatorAlien ifTrue: [0] ifFalse: [result isPrimitiveError ifTrue: [result errorCode] ifFalse: [result]]!

1 0

Failed: OpenSmalltalk/opensmalltalk-vm#2288 (Cog - 6477267)
by Travis CI 29 Oct '20

29 Oct '20

Build Update for OpenSmalltalk/opensmalltalk-vm ------------------------------------- Build: #2288 Status: Failed Duration: 2 hrs, 35 mins, and 43 secs Commit: 6477267 (Cog) Author: Eliot Miranda Message: ProcessorSimulator plugins as per Cog-eem.418 IA32ABI plugin as per VMMaker.oscog-eem.2857 Add the offset version of the run/step primitives which allow simulation of the simulator. IA32ABI: make most of the simple acessors simulate. This makes it possible to simulate the simulator ;-) Slang: generate neater code for pointer indirection. Instead of indexing with zero, dereference with *. View the changeset: https://github.com/OpenSmalltalk/opensmalltalk-vm/compare/b223cd542dc6...64… View the full build log and details: https://travis-ci.org/github/OpenSmalltalk/opensmalltalk-vm/builds/73969391… -- You can unsubscribe from build emails from the OpenSmalltalk/opensmalltalk-vm repository going to https://travis-ci.org/account/preferences/unsubscribe?repository=8795279&ut…. Or unsubscribe from *all* email updating your settings at https://travis-ci.org/account/preferences/unsubscribe?utm_medium=notificati…. Or configure specific recipients for build notifications in your .travis.yml file. See https://docs.travis-ci.com/user/notifications.

1 0

Still Failing: OpenSmalltalk/opensmalltalk-vm#2284 (Cog - 75c8e99)
by Travis CI 29 Oct '20

29 Oct '20

Build Update for OpenSmalltalk/opensmalltalk-vm ------------------------------------- Build: #2284 Status: Still Failing Duration: 9 hrs, 17 mins, and 0 secs Commit: 75c8e99 (Cog) Author: Eliot Miranda Message: Merge pull request #530 from smalltalking/Cog Simplify sqUnixMain.c's imgInit() View the changeset: https://github.com/OpenSmalltalk/opensmalltalk-vm/compare/c3f4c20a157b...75… View the full build log and details: https://travis-ci.org/github/OpenSmalltalk/opensmalltalk-vm/builds/73965762… -- You can unsubscribe from build emails from the OpenSmalltalk/opensmalltalk-vm repository going to https://travis-ci.org/account/preferences/unsubscribe?repository=8795279&ut…. Or unsubscribe from *all* email updating your settings at https://travis-ci.org/account/preferences/unsubscribe?utm_medium=notificati…. Or configure specific recipients for build notifications in your .travis.yml file. See https://docs.travis-ci.com/user/notifications.

1 0

[OpenSmalltalk/opensmalltalk-vm] 2fd7f4: Balloon-Engine-Pools no longer needed for a VMMake...
by Eliot Miranda 29 Oct '20

29 Oct '20

Branch: refs/heads/Cog Home: https://github.com/OpenSmalltalk/opensmalltalk-vm Commit: 2fd7f4385ed58a907202954d7ad72b26dba679d8 https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/2fd7f4385ed58a9072… Author: Eliot Miranda <eliot.miranda(a)gmail.com> Date: 2020-10-28 (Wed, 28 Oct 2020) Changed paths: M .gitignore M image/BuildSqueakSpurTrunkVMMakerImage.st Log Message: ----------- Balloon-Engine-Pools no longer needed for a VMMaker image. Balloon is in trunk. ignore workspaces in image. [ci skip]

1 0

VM Maker: VMMaker.oscog-eem.2859.mcz
by commits＠source.squeak.org 29 Oct '20

29 Oct '20

Eliot Miranda uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-eem.2859.mcz ==================== Summary ==================== Name: VMMaker.oscog-eem.2859 Author: eem Time: 28 October 2020, 5:16:02.046216 pm UUID: cf8dc817-56ad-44b6-a593-7ec5a163df5f Ancestors: VMMaker.oscog-eem.2858 Allow processor predefined macros to include expressions (i.e. __ARM_ARCH >= 8). =============== Diff against VMMaker.oscog-eem.2858 =============== Item was changed: ----- Method: Cogit class>>generateCodeStringForCogitDotC (in category 'translation') ----- generateCodeStringForCogitDotC "Generate a skeletal cogit.c that includes the relevant cogitFOO.c for the appropriate subclasses of CogAbstractInstruction. self generateCodeStringForCogitDotC" | string insertPosition abis defaultDef | abis := OrderedCollection new. string := String streamContents: [:s| s nextPutAll: '/* Automatically generated by\ ' withCRs. s nextPutAll: (CCodeGenerator monticelloDescriptionFor: self). s cr; nextPutAll: ' */'. s cr; cr; nextPut: $#. insertPosition := s position. self translateableInstructionSubclassesAndInstalledOptionsDo: [:class | | abi | s nextPutAll: 'if '. (abi := InitializationOptions at: #ABI ifAbsent: []) ifNotNil: [s nextPutAll: (abis addLast: abi); nextPutAll: ' && (']. class identifyingPredefinedMacros + do: [:predefinedMacro| + (predefinedMacro includes: Character space) "not a macr name but an expression" + ifTrue: [s nextPut: $(; nextPutAll: predefinedMacro] + ifFalse: [s nextPutAll: 'defined('; nextPutAll: predefinedMacro]. + s nextPut: $)] - do: [:predefinedMacro| s nextPutAll: 'defined('; nextPutAll: predefinedMacro; nextPut: $)] separatedBy: [s nextPutAll: ' || ']. abi ifNotNil: [s nextPut: $)]. s cr; cr; nextPutAll: '# include "'; nextPutAll: class moduleName; nextPutAll: '.c"'. s cr; cr; nextPutAll: '#el']. s nextPutAll: 'se'. #( 'As yet no Cogit implementation appears to exist for your platform.' 'Consider implementing it, starting by adding a subclass of CogAbstractInstruction.') do: [:msg| s cr; nextPutAll: '# error '; nextPutAll: msg]. s cr; nextPutAll: '#endif'; cr]. abis isEmpty ifTrue: [^string]. defaultDef := String streamContents: [:s| s nextPutAll: '#if !!'. abis do: [:abi| s nextPutAll: abi] separatedBy: [s nextPutAll: ' && !!']. s cr; nextPutAll: '# define '; nextPutAll: abis first; nextPutAll: ' 1'; cr. s nextPutAll: '#endif'; cr; cr]. ^string copyReplaceFrom: insertPosition to: insertPosition - 1 with: defaultDef!

1 0

Build completed: opensmalltalk-vm 1.0.2289
by AppVeyor 29 Oct '20

29 Oct '20

1 0

VM Maker: VMMaker.oscog-eem.2858.mcz
by commits＠source.squeak.org 29 Oct '20

29 Oct '20

Eliot Miranda uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-eem.2858.mcz ==================== Summary ==================== Name: VMMaker.oscog-eem.2858 Author: eem Time: 28 October 2020, 5:04:18.475277 pm UUID: a82bc143-8937-4842-8fe8-58e4d9d3ef85 Ancestors: VMMaker.oscog-eem.2857 We need a rump setInterruptCheckChain: to be able to simulate the processor plugins. =============== Diff against VMMaker.oscog-eem.2857 =============== Item was added: + ----- Method: SpurMemoryManager>>setInterruptCheckChain: (in category 'simulation only') ----- + setInterruptCheckChain: ignored + <doNotGenerate> + ^0!

1 0

VM Maker: Cog-eem.419.mcz
by commits＠source.squeak.org 29 Oct '20

29 Oct '20

Eliot Miranda uploaded a new version of Cog to project VM Maker: http://source.squeak.org/VMMaker/Cog-eem.419.mcz ==================== Summary ==================== Name: Cog-eem.419 Author: eem Time: 28 October 2020, 5:00:42.614467 pm UUID: 82e5725f-bce2-42a6-a9e0-2c048b5553cb Ancestors: Cog-eem.418 Add the simulating-the-simulator primitives for the processor aliens. Use these to simulate runCPU:In:Size:MinAddressRead:Write: & singleStepCPU:In:Size:MinAddressRead:Write:. Fix my misunderstanding of the use of Context>>cut: in MultiProcessor>>saveStackFor:above:homeContext: and consequently the COGMTVM happily simulates more than 1k methods into simulating-the-simulator before an assertt-fail. COGMTVM is officially undead ;-) =============== Diff against Cog-eem.418 =============== Item was added: + ----- Method: BochsIA32Alien>>rawPrimitiveRunInMemory:offsetBy:minimumAddress:readOnlyBelow: (in category 'primitives-simulation') ----- + rawPrimitiveRunInMemory: memoryArray offsetBy: offset minimumAddress: minimumAddress readOnlyBelow: minimumWritableAddress + "A version of primitiveRunInMemory:offsetBy:minimumAddress:readOnlyBelow: for simulation" + <primitive: 'primitiveRunInMemoryOffsetMinimumAddressReadWrite' module: 'BochsIA32Plugin' error: ec> + ^ec! Item was added: + ----- Method: BochsIA32Alien>>rawPrimitiveSingleStepInMemory:offsetBy:minimumAddress:readOnlyBelow: (in category 'primitives-simulation') ----- + rawPrimitiveSingleStepInMemory: memoryArray offsetBy: offset minimumAddress: minimumAddress readOnlyBelow: minimumWritableAddress + "A version of primitiveSingleStepInMemory:offsetBy:minimumAddress:readOnlyBelow: for simulation" + <primitive: 'primitiveSingleStepInMemoryOffsetMinimumAddressReadWrite' module: 'BochsIA32Plugin' error: ec> + ^ec! Item was added: + ----- Method: BochsX64Alien>>rawPrimitiveRunInMemory:offsetBy:minimumAddress:readOnlyBelow: (in category 'primitives-simulation') ----- + rawPrimitiveRunInMemory: memoryArray offsetBy: offset minimumAddress: minimumAddress readOnlyBelow: minimumWritableAddress + "A version of primitiveRunInMemory:offsetBy:minimumAddress:readOnlyBelow: for simulation" + <primitive: 'primitiveRunInMemoryOffsetMinimumAddressReadWrite' module: 'BochsX64Plugin' error: ec> + ^ec! Item was added: + ----- Method: BochsX64Alien>>rawPrimitiveSingleStepInMemory:offsetBy:minimumAddress:readOnlyBelow: (in category 'primitives-simulation') ----- + rawPrimitiveSingleStepInMemory: memoryArray offsetBy: offset minimumAddress: minimumAddress readOnlyBelow: minimumWritableAddress + "A version of primitiveSingleStepInMemory:offsetBy:minimumAddress:readOnlyBelow: for simulation" + <primitive: 'primitiveSingleStepInMemoryOffsetMinimumAddressReadWrite' module: 'BochsX64Plugin' error: ec> + ^ec! Item was added: + ----- Method: GdbARMAlien>>rawPrimitiveRunInMemory:offsetBy:minimumAddress:readOnlyBelow: (in category 'primitives-simulation') ----- + rawPrimitiveRunInMemory: memoryArray offsetBy: offset minimumAddress: minimumAddress readOnlyBelow: minimumWritableAddress + "A version of primitiveRunInMemory:offsetBy:minimumAddress:readOnlyBelow: for simulation" + <primitive: 'primitiveRunInMemoryOffsetMinimumAddressReadWrite' module: 'GdbARMPlugin' error: ec> + ^ec! Item was added: + ----- Method: GdbARMAlien>>rawPrimitiveSingleStepInMemory:offsetBy:minimumAddress:readOnlyBelow: (in category 'primitives-simulation') ----- + rawPrimitiveSingleStepInMemory: memoryArray offsetBy: offset minimumAddress: minimumAddress readOnlyBelow: minimumWritableAddress + "A version of primitiveSingleStepInMemory:offsetBy:minimumAddress:readOnlyBelow: for simulation" + <primitive: 'primitiveSingleStepInMemoryOffsetMinimumAddressReadWrite' module: 'GdbARMPlugin' error: ec> + ^ec! Item was changed: ----- Method: MultiProcessor>>saveStackFor:above:homeContext: (in category 'private-deep hackery') ----- saveStackFor: signalerContext above: contextInCritical homeContext: doesNotUnderstandContext "Some processors, notably ARMv8, support load/store pair instructions which may raise two ProcessorSimulationTraps. To handle this we must preserve the continuation following the raising of the first simulaiton trap. This cuts up, and restitches the stack so that when the first ProcessorSimulationTrap is raised outside the critical: block in our doesNotUnderstand: method, control continues to the second trap." + | contextAboveSignalerContext | - | contextAboveContextInCritical contextAboveSignalerContext | - contextAboveContextInCritical := signalerContext findContextSuchThat: [:ctxt| ctxt sender == contextInCritical]. "This is our continuation, which should be changed to continue back into the critical section." contextAboveSignalerContext := thisContext findContextSuchThat: [:ctxt| ctxt sender == signalerContext]. - self deny: contextAboveContextInCritical isNil. self deny: contextAboveSignalerContext isNil. + "signalerContext through the context above contextInCritical is the stack that is the continuation + for the trap. This must be inserted between the doesNotUnderstandContext and its sender. + N.B. Context>>cut: cuts above the argument!!!!" + signalerContext cut: contextInCritical. + contextInCritical privSender: doesNotUnderstandContext sender. - "signalerContext through contextAboveContextInCritical is the stack that is the continuation - for the trap. This must be inserted between the doesNotUnderstandContext and its sender." - signalerContext cut: contextAboveContextInCritical. - contextAboveContextInCritical privSender: doesNotUnderstandContext sender. doesNotUnderstandContext privSender: signalerContext. "Now arrange that the handler returns to contextInCritical; and we're done." + contextAboveSignalerContext privSender: contextInCritical! - contextAboveSignalerContext privSender: contextInCritical. - self deny: (thisContext findContextSuchThat: [:ctxt| ctxt == signalerContext]) isNil. - self deny: (thisContext findContextSuchThat: [:ctxt| ctxt == contextInCritical]) isNil. - self deny: (thisContext findContextSuchThat: [:ctxt| ctxt == doesNotUnderstandContext]) isNil! Item was changed: ----- Method: ProcessorSimulatorPlugin>>runCPU:In:Size:MinAddressRead:Write: (in category 'simulation') ----- runCPU: cpu In: memoryCArray Size: memorySize MinAddressRead: minAddress Write: minWriteMaxExecAddress <doNotGenerate> "*now* we need derived pointers. Ho hum... But all we need is one more level of indirection..." + | result | + result := mySimulatorAlien + rawPrimitiveRunInMemory: interpreterProxy memory + offsetBy: memoryCArray ptrAddress + minimumAddress: minAddress + readOnlyBelow: minWriteMaxExecAddress. + ^result == mySimulatorAlien + ifTrue: [0] + ifFalse: + [result isPrimitiveError + ifTrue: [result errorCode] + ifFalse: [result]]! - ^mySimulatorAlien - primitiveRunInMemory: interpreterProxy memory - offsetBy: memoryCArray ptrAddress - minimumAddress: minAddress - readOnlyBelow: minWriteMaxExecAddress! Item was changed: ----- Method: ProcessorSimulatorPlugin>>singleStepCPU:In:Size:MinAddressRead:Write: (in category 'simulation') ----- singleStepCPU: cpu In: memoryCArray Size: memorySize MinAddressRead: minAddress Write: minWriteMaxExecAddress <doNotGenerate> "*now* we need derived pointers. Ho hum... But all we need is one more level of indirection..." + | result | + self break. + result := mySimulatorAlien + rawPrimitiveSingleStepInMemory: interpreterProxy memory + offsetBy: memoryCArray ptrAddress + minimumAddress: minAddress + readOnlyBelow: minWriteMaxExecAddress. + ^result == mySimulatorAlien + ifTrue: [0] + ifFalse: + [result isPrimitiveError + ifTrue: [result errorCode] + ifFalse: [result]]! - ^mySimulatorAlien - primitiveSingleStepInMemory: interpreterProxy memory - offsetBy: memoryCArray ptrAddress - minimumAddress: minAddress - readOnlyBelow: minWriteMaxExecAddress!

1 0

Build failed: opensmalltalk-vm 1.0.2288
by AppVeyor 28 Oct '20

28 Oct '20

1 0

← Newer
1
2
3
4
5
6
7
8
...
36
Older →

Jump to page:

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

Vm-dev October 2020