Vm-dev June 2016

vm-dev@lists.squeakfoundation.org

50 participants
129 discussions

Functional callbacks on all Cog platforms
by Eliot Miranda 09 Jul '16

09 Jul '16

Hi All, just wanted to report that as of Alien-eem.35 and https://github.com/OpenSmalltalk/vm/commit/da3fafdec9444754af104e0ed9f613f6… we now have functional callbacks on all x86 platforms, ARM32 platforms and x86_64 platforms. There's a simple example Alien class>>exampleCqsort that shows how to use the system and tests the implementation using the C library's sort quick sort implementation that takes a function pointer as an argument to use for comparing pairs of elements. This is mapped onto a Smalltalk block by Alien's callback implementation. The example sorts 100 random double-precision floats using qsort and a callback into Smalltalk. _,,,^..^,,,_ best, Eliot

2 1

Loading VMMaker
by Yoshiki Ohshima 06 Jul '16

06 Jul '16

Hi, Just to refresh my skill on compiling the Squeak VM (I'd like to play with Kinect; I found some efforts like: http://tecnodacta.com.ar/gira/ but want to make a plugin that is suitable for my need.) Let us say I'd like to compile a VM and my plugins on Mac, what are the right steps? (I had some experiences with VMMaker but that was before I need to specify Interpreter class name in the VMMakerTool, for example. I added VMMaker repository in the Monticello browser, and opened it. As trying to just load VMMaker shows a warning of missing FFI constants, I turned to 'update' and load update-dtl.17.mcm. Is this a right thing to do? But then, trying to load update-dtl.17.mcm onto the VM and image in the 5.0 all in one package, the VM crashes. Do you see this? I'd like to get the files and image set up so that I can compile an external plugin, preferably for Cog/Spur. Some up to date guidance would be greatly appreciated. Thanks! -- -- Yoshiki

5 9

build products seem to go to the wrong branch
by Ben Coman 02 Jul '16

02 Jul '16

I just did my first vm build from git... $ git checkout -b firstbuild Cog $ cd image $ ./buildspurtrunkvmmakerimage.sh $ cd build.linux32x86/squeak.cog.spur/build $ ./mvm Which completes okay, but reports... Libraries have been installed in: /home/ben/Repos/OpenSmalltalk/vm/products/cogspurlinuxht/lib/squeak/5.0-201606221719-btc/condense-gitignore but condense-gitignore is not the current branch, but a previous branch I worked on where I didn't even do a build (I'm reasonably sure, but maybe I forget). $ git branch [snip] * firstbuild btc/condense-gitignore I poked around and found the message comes from ./platforms/unix/config/ltmain.sh but I don't see any git commands there to determine the branch, so it seems the branch must be passed to ltmain from somewhere else?? cheers -ben

3 3

[OT] git gui (was: Why OpenSmalltalk/vm not googlable)
by Bert Freudenberg 30 Jun '16

30 Jun '16

On Wed, Jun 29, 2016 at 10:24 PM, tim Rowledge <tim(a)rowledge.org> wrote: > the recommended ‘SourceTree’ > How do you like that app? I'm using Github's own app (https://desktop.github.com/) but I'm not a newbie, doing git on the command line often too. - Bert -

5 4

VM Maker: VMMaker.oscog-cb.1894.mcz
by commits＠source.squeak.org 30 Jun '16

30 Jun '16

ClementBera uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-cb.1894.mcz ==================== Summary ==================== Name: VMMaker.oscog-cb.1894 Author: cb Time: 30 June 2016, 2:02:37.049787 pm UUID: b3c32cc0-576b-4a14-9572-b5323dfa774f Ancestors: VMMaker.oscog-eem.1893 Accesses on remote instance variables can't be done on forwarders: to happen, the method with the inst var access needs to be inlined hence there's either a guard ensuring the object is not a forwarder or the type of the object is known (immutable literal, inlined object allocation such as Array, FullBlockClosure, immediate rcvr, etc.) =============== Diff against VMMaker.oscog-eem.1893 =============== Item was added: + ----- Method: StackToRegisterMappingCogit>>genLoadTemp:in: (in category 'bytecode generator support') ----- + genLoadTemp: objectIndex in: destReg + destReg = ReceiverResultReg ifTrue: [ optStatus isReceiverResultRegLive: false ]. + self ssAllocateRequiredReg: destReg. + self MoveMw: (self frameOffsetOfTemporary: objectIndex) r: FPReg R: destReg.! Item was removed: - ----- Method: StackToRegisterMappingCogit>>genLoadUnFwdTemp:in: (in category 'bytecode generator support') ----- - genLoadUnFwdTemp: objectIndex in: destReg - destReg = ReceiverResultReg ifTrue: [ optStatus isReceiverResultRegLive: false ]. - self ssAllocateRequiredReg: destReg. - self MoveMw: (self frameOffsetOfTemporary: objectIndex) r: FPReg R: destReg. - objectRepresentation - genEnsureOopInRegNotForwarded: destReg - scratchReg: TempReg.! Item was changed: ----- Method: StackToRegisterMappingCogit>>genPushMaybeContextRemoteInstVar:inObjectAt: (in category 'bytecode generator support') ----- genPushMaybeContextRemoteInstVar: slotIndex inObjectAt: index <inline: false> self ssAllocateCallReg: ReceiverResultReg and: SendNumArgsReg. + self genLoadTemp: index in: ReceiverResultReg. - self genLoadUnFwdTemp: index in: ReceiverResultReg. ^ self genPushMaybeContextSlotIndex: slotIndex! Item was changed: ----- Method: StackToRegisterMappingCogit>>genPushRemoteInstVar:inObjectAt: (in category 'bytecode generator support') ----- genPushRemoteInstVar: index inObjectAt: objectIndex <inline: false> | objectReg resultReg | self assert: needsFrame. objectReg := self allocateRegNotConflictingWith: 0. + self genLoadTemp: objectIndex in: objectReg. - self genLoadUnFwdTemp: objectIndex in: objectReg. resultReg := self availableRegOrNoneNotConflictingWith: (self registerMaskFor: objectReg). resultReg = NoReg ifTrue: [resultReg := objectReg]. objectRepresentation genLoadSlot: byte1 sourceReg: objectReg destReg: resultReg. ^self ssPushRegister: resultReg! Item was changed: ----- Method: StackToRegisterMappingCogit>>genStorePop:MaybeContextRemoteInstVar:ofObjectAt:needsStoreCheck: (in category 'bytecode generator support') ----- genStorePop: popBoolean MaybeContextRemoteInstVar: slotIndex ofObjectAt: objectIndex needsStoreCheck: needsStoreCheck <inline: false> "The reason we need a frame here is that assigning to an inst var of a context may involve wholesale reorganization of stack pages, and the only way to preserve the execution state of an activation in that case is if it has a frame." self assert: needsFrame. + self genLoadTemp: objectIndex in: ReceiverResultReg. - self genLoadUnFwdTemp: objectIndex in: ReceiverResultReg. self genStorePop: popBoolean MaybeContextSlotIndex: slotIndex needsStoreCheck: needsStoreCheck. ^ 0! Item was changed: ----- Method: StackToRegisterMappingCogit>>genStorePop:RemoteInstVar:ofObjectAt:needsStoreCheck: (in category 'bytecode generator support') ----- genStorePop: popBoolean RemoteInstVar: slotIndex ofObjectAt: objectIndex needsStoreCheck: needsStoreCheck <inline: false> self assert: needsFrame. + self genLoadTemp: objectIndex in: ReceiverResultReg. - self genLoadUnFwdTemp: objectIndex in: ReceiverResultReg. self genStorePop: popBoolean slotIndex: slotIndex destReg: ReceiverResultReg needsStoreCheck: needsStoreCheck needsRestoreRcvr: false. ^ 0!

1 0

Re: [Pharo-dev] [Vm-dev] Re: [ANN] Ephemeron Support is Ready
by Yuriy Tymchuk 30 Jun '16

30 Jun '16

Maybe I will find some time to hack the tree-matching rules to highlight the matched piece of code. It’s so frustrating that they don’t do it… Uko > On 30 Jun 2016, at 10:17, Marcus Denker <marcus.denker(a)inria.fr> wrote: > > > > > finalizationProcess > "The finalization process arranges to send mourn to each element of the VM's finalization queue, > which is accessed via primitiveFetchMourner. The VM signals FinalizationSemaphore whenever > the queue is non-empty. This process loops, waiting on the semaphore, fetches the first element > of the queue and then spawns a process at a higher priority to acually send the mourn messages. > If an error occurs in the higher priority mourn loop process then this process will simply spawn > another process, hence ensuring that errors in finalization methods don't break finalization. > > In addition this process also runs the old finalization scheme, supporting clients of the older, > WeakRegistry based scheme. Hopefully this will go away when all cleints have moved over." > | throttle firstMourner | > throttle := Semaphore new. > [FinalizationSemaphore wait; initSignals. > "Support the old scheme until things have changed over..." > self doOldFinalization. > [firstMourner := self primitiveFetchMourner. > firstMourner notNil] whileTrue: > [[throttle signal. > self mournLoopWith: firstMourner] forkAt: Processor activePriority + 1. > throttle wait]] > > Doh! So I wrote a naked block, not a loop !! So it did nothing. Thank you!! I think the compiler should warn about naked blocks. > > > I added a new Code-Critic rule "RBDeadBlockRule", with > > initialize > super initialize. > self matcher > matches: '`{:node | node isBlock and: [node parent isSequence > and: [ node isLastStatementInBlock not ]]}' > do: [ :node :answer | node ] > > This now shows a warning in the tools (the explanation is shown when clicking on the ? button): > It does not yet highlight the block and has no automatic transformation to fix the problem, but as > a first step it should be nice to have. > > > <Dead.jpg> > > This is in 60 update 126. > > (I did not read the mail again before doing it, so now it is called "dead", not "naked", we can change > that... dead somehow makes it look to be related a "dead context"...). > > Marcus > > >

1 0

Re: [Vm-dev] VM Maker: VMMaker.oscog-eem.1893.mcz
by Ben Coman 30 Jun '16

30 Jun '16

On Thu, Jun 30, 2016 at 8:36 AM, <commits(a)source.squeak.org> wrote: > > Eliot Miranda uploaded a new version of VMMaker to project VM Maker: > http://source.squeak.org/VMMaker/VMMaker.oscog-eem.1893.mcz > > ==================== Summary ==================== > > Name: VMMaker.oscog-eem.1893 > Author: eem > Time: 29 June 2016, 5:35:30.7422 pm > UUID: ed498bfa-0631-4cf2-8bfb-c49e20505d40 > Ancestors: VMMaker.oscog-eem.1892 > > Fix a couple of cases that should use mositiveMachineIntegerValue I didn't realise you had such close ties to this US Military project... http://www.snopes.com/photos/technology/insectdrone.asp Of course with git, you might do a "commit --amend" if you noticed this before pushing to the server ;) except we'd then be denied an occasional chuckle. cheers -ben

1 0

Malloc Benchmarking
by Ben Coman 30 Jun '16

30 Jun '16

I haven't worked out quite how to read this, but as a low-level benchmark thought it might be of random interest... http://highlandsun.com/hyc/malloc/ cheers -ben

1 0

VM Maker: VMMaker.oscog-eem.1893.mcz
by commits＠source.squeak.org 30 Jun '16

30 Jun '16

Eliot Miranda uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-eem.1893.mcz ==================== Summary ==================== Name: VMMaker.oscog-eem.1893 Author: eem Time: 29 June 2016, 5:35:30.7422 pm UUID: ed498bfa-0631-4cf2-8bfb-c49e20505d40 Ancestors: VMMaker.oscog-eem.1892 Fix a couple of cases that should use mositiveMachineIntegerValue: and nuke an unnecessary stackObjectValue:. In 1893: Riots of Mons during the Belgian general strike of 1893, The day after, Belgian parliament approved Universal suffrage. The Independent Labour Party of the United Kingdom has its first meeting. The United States overthrows the Kingdom of Hawaii: Lorrin A. Thurston and the Citizen's Committee of Public Safety in Hawaii with the intervention of the United States Marine Corps overthrow the government of Queen Liliuokalani. =============== Diff against VMMaker.oscog-eem.1892 =============== Item was changed: ----- Method: IA32ABIPlugin>>primBoxedFree (in category 'primitives-memory management') ----- primBoxedFree "Free the memory referenced by the receiver, an Alien." "proxy <Alien> primFree ^<Alien> <primitive: 'primBoxedFree' error: errorCode module: 'IA32ABI'>" | addr rcvr ptr sizeField | <export: true> <var: #ptr type: 'sqInt *'> <var: #sizeField type: 'long'> + rcvr := interpreterProxy stackValue: 0. - rcvr := interpreterProxy stackObjectValue: 0. (interpreterProxy byteSizeOf: rcvr) >= (2 * interpreterProxy bytesPerOop) ifFalse: [^interpreterProxy primitiveFailFor: PrimErrBadReceiver]. ptr := interpreterProxy firstIndexableField: rcvr. sizeField := ptr at: 0. addr := ptr at: 1. "Don't you dare to free Squeak's memory!!" (sizeField >= 0 or: [addr = 0 or: [interpreterProxy isInMemory: addr]]) ifTrue: [^interpreterProxy primitiveFailFor: PrimErrInappropriate]. self cCode: 'free((void *)addr)' inSmalltalk: [self Cfree: addr]. ptr at: 0 put: 0; at: 1 put: 0 "cleanup"! Item was changed: ----- Method: IA32ABIPlugin>>primFree (in category 'primitives-memory management') ----- primFree "Free the memory referenced by the argument, an integer." "<Anywhere> primFree: address <Integer> <primitive: 'primFree' error: errorCode module: 'IA32ABI'>" | addr | <export: true> + addr := interpreterProxy stackPositiveMachineIntegerValue: 0. - addr := interpreterProxy positive32BitValueOf: (interpreterProxy stackValue: 0). interpreterProxy failed ifTrue: [^interpreterProxy primitiveFailFor: PrimErrBadArgument]. "Don't you dare to free Squeak's memory!!" (addr = 0 or: [interpreterProxy isInMemory: addr]) ifTrue: [^interpreterProxy primitiveFailFor: PrimErrInappropriate]. self cCode: 'free((void *)addr)' inSmalltalk: [self Cfree: addr]. interpreterProxy pop: 1! Item was changed: ----- Method: IA32ABIPlugin>>primReturnFromContextThrough (in category 'primitives-callbacks') ----- primReturnFromContextThrough "Return a result from a callback to the callback's callee. The primitive has a signature of either of the forms: result <FFICallbackResult> primReturnFromContext: callbackContext <MethodContext> through: jmpBuf <Integer> result <FFICallbackResult> primSignal: aSemaphore <Semaphore> andReturnFromContext: callbackContext <MethodContext> through: jmpBuf <Integer> <primitive: 'primReturnFromContextThrough' error: errorCode module: 'IA32ABI'>. If of the second form answer true if this is not the most recent callback, and signal aSemaphore if it is, so as to implement LIFO ordering of callbacks." <export: true> <legacy> | mac vmCallbackContext vmCallbackReturnValue isMostRecent | <var: #vmCallbackContext type: #'VMCallbackContext *'> <var: #vmCallbackReturnValue type: #'VMCallbackReturnValue *'> + vmCallbackContext := self cCoerceSimple: (interpreterProxy positiveMachineIntegerValueOf: (interpreterProxy stackValue: 0)) - vmCallbackContext := self cCoerceSimple: (interpreterProxy positive32BitValueOf: (interpreterProxy stackValue: 0)) to: #'VMCallbackContext *'. (interpreterProxy failed or: [vmCallbackContext = 0]) ifTrue: [^interpreterProxy primitiveFailFor: PrimErrBadArgument]. (mac := interpreterProxy methodArgumentCount) = 3 ifTrue: [isMostRecent := vmCallbackContext = self getMostRecentCallbackContext. isMostRecent ifFalse: [interpreterProxy methodReturnValue: interpreterProxy trueObject. ^nil]. (interpreterProxy fetchClassOf: (interpreterProxy stackValue: 2)) = interpreterProxy classSemaphore ifFalse: [^interpreterProxy primitiveFailFor: PrimErrBadArgument]. [interpreterProxy signalNoResume: (interpreterProxy stackValue: 2)] whileFalse]. vmCallbackReturnValue := self cCoerceSimple: (self startOfData: (interpreterProxy stackValue: mac)) to: #'VMCallbackReturnValue *'. self cCode: "C needs a typedef for structs to be assigned, but that implies a struct class for just one assignment." [self mem: (self addressOf: vmCallbackContext rvs) cp: (self addressOf: vmCallbackReturnValue crvrvs) y: (self sizeof: vmCallbackContext rvs)] inSmalltalk: [vmCallbackContext rvs: vmCallbackReturnValue crvrvs]. (interpreterProxy returnAs: (interpreterProxy integerObjectOf: vmCallbackReturnValue type + 1) ThroughCallback: vmCallbackContext Context: (interpreterProxy stackValue: 1)) ifFalse: [^interpreterProxy primitiveFailFor: PrimErrBadArgument]. "NOTREACHED"!

1 0

VM Maker: VMMaker.oscog-eem.1892.mcz
by commits＠source.squeak.org 30 Jun '16

30 Jun '16

Eliot Miranda uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-eem.1892.mcz ==================== Summary ==================== Name: VMMaker.oscog-eem.1892 Author: eem Time: 29 June 2016, 4:59:05.109418 pm UUID: 51671a87-df9b-49c2-8972-b9fe8eb0cf5f Ancestors: VMMaker.oscog-eem.1891 Fix assert failure in primitiveVoidVMStateForMethod. The instructrionPointer must be pushed after the frame is married. In 1892 the Liverpool Football Club is founded by John Houlding, the owner of Anfield. Houlding decides to form his own team after Everton leaves Anfield in an argument over rent. =============== Diff against VMMaker.oscog-eem.1891 =============== Item was changed: ----- Method: CoInterpreterPrimitives>>primitiveVoidVMStateForMethod (in category 'system control primitives') ----- primitiveVoidVMStateForMethod "The receiver is a compiledMethod. Clear all VM state associated with the method, including any machine code, or machine code pcs in context objects." | activeContext methodObj hasCogMethod theFrame thePage | <var: #theFrame type: #'char *'> <var: #thePage type: #'StackPage *'> super primitiveFlushCacheByMethod. "One might think (as this author did) that the heap scan is unnecessary if the method does not have a cog method. But it could be the case that the code zone has recently been reclaimed and so not having a cog method is no indication that it didn't have a cog method some time in the recent past, and that there are indeed still contexts with machine code pcs out there. The only steps that can be avoided is divorcing frames in the stack zone, and scanning to unlink and free if there isn't a cog method." methodObj := self stackTop. + activeContext := self ensureFrameIsMarried: framePointer SP: stackPointer. + self ensurePushedInstructionPointer. - self push: instructionPointer. self externalWriteBackHeadFramePointers. - activeContext := self ensureFrameIsMarried: framePointer SP: stackPointer - objectMemory wordSize. (hasCogMethod := self methodHasCogMethod: methodObj) ifTrue: [self divorceMachineCodeFramesWithMethod: methodObj]. self ensureAllContextsWithMethodHaveBytecodePCs: methodObj. hasCogMethod ifTrue: [cogit unlinkSendsTo: methodObj andFreeIf: true]. (self isStillMarriedContext: activeContext) ifTrue: [theFrame := self frameOfMarriedContext: activeContext. thePage := stackPages stackPageFor: theFrame. self assert: thePage headFP = theFrame. self setStackPageAndLimit: thePage. stackPointer := thePage headSP. framePointer := thePage headFP. instructionPointer := self popStack. self assert: methodObj = self stackTop] ifFalse: [self zeroStackPage. "to avoid assert in marryContextInNewStackPageAndInitializeInterpreterRegisters:" self marryContextInNewStackPageAndInitializeInterpreterRegisters: activeContext. self popStack. "pop bogus machine-code instructionPointer" self assert: methodObj = self stackTop. self siglong: reenterInterpreter jmp: ReturnToInterpreter]!

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

Vm-dev June 2016