Vm-dev July 2023

vm-dev@lists.squeakfoundation.org

4 participants
20 discussions

[OpenSmalltalk/opensmalltalk-vm] BitBlt rgbMul is broken for depth 1 and not correctly rounded above depth 4 (Issue #651)
by Nicolas Cellier 24 Nov '23

24 Nov '23

In depth 1, the resulting bits will always be 0. It's not a big problem because rgbMul is just a bitAnd operation at this depth. So a quick workaround would be to detect the case in BitBltSimulation destDepth = 1 ifTrue: [^self bitAnd: sourceWord with: destinationWord]. That would also accelerate the Bit BLock Transfer operation, so it's a good hack. But there is more. What we want is multiply ratios in interval [0,1]. dstRatio * srcRatio Our implementation is scaled ratio (scaled by `1 << nBits - 1`): src := (srcRatio * scale) rounded. dst := (dstRatio * scale) rounded. So what we want is: ((dst/scale) * (src/scale) * scale) rounded that is: (dst*src / (1<<nBits-1)) rounded Unfortunately, that's the other grief with the current implementation used for rounding: (dst+1)*(src+1) - 1 >> nBits It only equals correctly rounded operation for depths 2 and 4. For rounding we might use: (((dst/scale) * (src/scale) + 0.5) * scale) truncated. that is expressed with truncated division: dst*src + (scale+1//2) // scale So here is a nicer formulation for doing the job at any depth (including 5bits rgb channels for 16 bits depth) with correctly rounded division: aux := src * dst + (1 << (nBits - 1)). "add mid-scale for rounding" result := aux << (nBits - 1) + aux << (nBits -1). "divide by scale" This is because instead of dividing by scale, we can multiply by shifted inverse (sort of double precision), then shift right. (2 to: 32) allSatisfy: [:nBits | (1 << (nBits * 2) / (1 << nBits - 1)) rounded = (1 << nBits + 1)]. Multiplying by this inverse is easy and cheap: x * (1 << nBits + 1) = (x << nBits + x). And then applying the right shift `>> (2 * nBits)` is equivalent to: x >> nBits + x >> nBits. We must first add 0.5 (scaled), that is `src * dst + (1 << (nBits -1))` - our formulation of aux, and we're done. We verify: { (0 to: 1<<20-1) allSatisfy: [:i | (1<<9+i)>>10+ (1<<9+i)>>10 = (i/1023) rounded]. (0 to: 1<<18-1) allSatisfy: [:i | (1<<8+i)>>9+ (1<<8+i)>>9 = (i/511) rounded]. (0 to: 1<<16-1) allSatisfy: [:i | (1<<7+i)>>8+ (1<<7+i)>>8 = (i/255) rounded]. (0 to: 1<<14-1) allSatisfy: [:i | (1<<6+i)>>7+ (1<<6+i)>>7 = (i/127) rounded]. (0 to: 1<<12-1) allSatisfy: [:i | (1<<5+i)>>6+ (1<<5+i)>>6 = (i/63) rounded]. (0 to: 1<<10-1) allSatisfy: [:i | (1<<4+i)>>5+ (1<<4+i)>>5 = (i/31) rounded]. (0 to: 1<<8-1) allSatisfy: [:i | (1<<3+i)>>4+ (1<<3+i)>>4 = (i/15) rounded]. (0 to: 1<<6-1) allSatisfy: [:i | (1<<2+i)>>3+ (1<<2+i)>>3 = (i/7) rounded]. (0 to: 1<<4-1) allSatisfy: [:i | (1<<1+i)>>2+ (1<<1+i)>>2 = (i/3) rounded]. } allSatisfy: #yourself. The nice thing is that above down-scaling operation can be multiplexed. Suppose that we have p groups of 2*nBits `M` holding square-scale multiplication of each channel concatenated in a double-Word-Mul. doubleWordMul = Mp .... M5 M3 M1 Note we arrange to have odd channels in low word, and even channels in high word. We first form a `groupMask` on a word with (p+1)/2 groups of nBits alternating all one `i` and all zero `o`, `oioi...ioi`. channelMask := 1 << nBits - 1. groupMask := 0. 0 to: wordBits // (2 * nBits) do: [:i | groupMask = groupMask << (2 * nBits) + channelMask]. Where wordBits is the number of bits in a word (usually we want to operate on 32 bits words in BitBlt). We form the `doubleGroupMask` on a double-word with p groups of 2*nBits `oi`: doubleGroupMask := groupMask >> nBits. doubleGroupMask := doubleGroupMask << wordBits + groupMask. And we perform the division by scale: doubleWordMul := (doubleWordMul >> nBits bitAnd: doubleGroupMask) + doubleWord >> nBits bitAnd: doubleGroupMask. At this stage we obtain a double word containing scaled multiplicands interleaved with groups of nBits zeros: o mp ... o m3 o m1 Now the final result can be obtained by shifting back: doubleWordMul >> (wordBits - nBits) + (doubleWordMul bitAnd: groupMask) The only problem remaining is how to obtain the squared-scale multiplicands. It would be easy to form the alternate even-odd channels for each src and dst operands: doubleWordSrc := src >> nBits bitAnd: groupMask. doubleWordSrc := doubleWordSrc << wordBits + (src bitAnd: groupMask). doubleWordDst := dst >> nBits bitAnd: groupMask. doubleWordDst := doubleWordDst << wordBits + (dst bitAnd: groupMask). we now get `o sp ... o s3 o s1` and `0 dp ... o d3 o d1`, but we would now need a SIMD integer multiplication operating on groups of 2*nBits in parallel... We don't have that, at least in portable C code. So we still have to emulate it with a loop. half := 1 << (nBits - 1). shift := 0. doubleWordMul := 0 0 to: nChannels - 1 do: [:i | doubleWordMul := doubleWordMul + (((doubleWordSrc >> shift bitAnd: channelMask) * (doubleWordSrc >> shift bitAnd: channelMask) + half) << shift). shift := shift + nBits + nBits]. We know that each operation cannot overflow on upper neighbour group of 2*nBits, because the maximum value is: (1<<nBits-1) squared + (1 << (nBits-1)) = 1 << (2*nBits) - (2*(1<<nBits)) + (1 << (nBits-1)) - 1 < (1 << (2*nBits) - 1) It remains the odd case of 16 bits depth, which has 3 groups of 5 bits and a leading zero. I believe that above algorithm works without splitting in two half-words... To be tested. We have gathered the pieces for a correctly rounded almost-multiplexed rgbMul. Somehow have our cake and eat it too. -- Reply to this email directly or view it on GitHub: https://github.com/OpenSmalltalk/opensmalltalk-vm/issues/651 You are receiving this because you are subscribed to this thread. Message ID: <OpenSmalltalk/opensmalltalk-vm/issues/651(a)github.com>

2 4

Re: [Newbies] FreeBSD VM
by Tobias Pape 30 Sep '23

30 Sep '23

Hi David > On 30. Mar 2023, at 16:09, David Gray <dgray(a)iesl.forth.gr> wrote: > > I’ve been trying to get a working VM for FreeBSD. However even though I manage to get the system to start and update itself it gives a segmentation error when I open a workspace and type any character at all. I used the latest linux64x64 cog.spur. > Both the 6.0 and 5.3 images behaved the same. To get it to compile I had the comment out the Joystick plugin and edit the baud rates inside the serial plugin. I’m a complete beginner with squeak so I don’t know how to proceed from here. Based on your comments, I went and tried to verify that the VM still builds. I have made similar adjustments and was able to build a VM with the recents commits to the Cog branch of the VM. Can you please build a debug vm and try to recreate the error? I am very interested in that. Best regards -Tobias

4 5

In reference to: Minifying Woes
by Tom Braun 28 Aug '23

28 Aug '23

Hi all, I am new to the list, so I couldn’t answer directly to Minifying Woes (or at least didn’t know how to). I identified 75 VM intern objects in Tom’s (tobe) image, by loading it in the simulator and setting a halt after loading the image. With the following code: | collection | collection := OrderedCollection new. self allOldSpaceEntitiesDo: [:obj | (((self classIndexOf: obj) < self lastClassIndexPun) and: (self isImmediate: obj) not) ifTrue: [ collection add: obj] ]. (collection sorted: [:a :b | (self bytesInBody: a) > (self bytesInBody: b)]) collect: [:ea | ea hex -> (self bytesInBody: ea)] I get the following 75 objects: 1. A free chunk after all other objects. Can be ignored for the sake of minimising image size 2. Remembered set -> 1048592 byte 3. hiddenRootsObj -> 32848 byte 4 - 61. Pages of the mark and weakling stack -> respectively 32752 byte 62 - 74. arrays of the class table -> respectively 8208 byte 75. specialObjectsOop -> 520 byte As far as I can judge the stack pages and remembered set could be removed from the image to minimize it further. If I understand correctly the StackPages could be removed by using the SpurImagePreener. The remembered set Could be removed in the SpurImagePreener too (when the VM initialises the memory it initialises a new remembered set too, if it is nil). After a quick read of the SpurImagePreener I didn’t see that the remembered set gets removed. When I tried using the preener it resulted in an unusable image (both simulator and compiled VM couldn’t load it). I tried both: SpurImagePreener new preenImage: '/Users/tombraun/Desktop/Squeak6.0-22104-64bit copy.image' SpurImagePreener new writeDefaultHeader: true; savedWindowSize: 1@1; preenImage: '/Users/tombraun/Desktop/Squeak6.0-22104-64bit copy.image' Could be a me problem, as I did some changes to the memory management in my VMMaker image, although this shouldn’t influence the preener…. On this note @Eliot: why the decision to make the object stacks and the remembered set VM managed objects instead of allocating them separately? 1. We don’t need to keep them in a snapshot. All object stacks are empty after GC and as we flushed the new space pre snapshot The remembered set shouldn't need to be persisted too. 2. During GC we simply mark all stack pages and keep them alive. When we at least freed empty pages (after a limit, to prevent the running VM from having to allocate too many pages every GC?) I would see the value. What did I overlook or does it simply have historical reasons? Best, Tom (WoC)

5 5

VMMakerJS for JsSqueak
by Florin Mateoc 26 Jul '23

26 Jul '23

Hi all, I have now pushed to the JsSqueak github the last missing pieces, the VMMakerJS changeset for generating the JsSqueak plugins, together with some minimal instructions in the README. It is not standalone, in that it uses method nodes that are pre-processed the same way that regular (image-side) methods are before emitting the non-plugin JavaScript code. I have not tried it lately, but without pre-processing, at the vary least GOTOs would remain in the code, and JavaScript does not have them. Of course, while both image-side code and plugin code is translated to JavaScript, the emitted code is quite different. The image side emits code directly from the MethodNode tree, and uses yield* invocations, the plugin code uses JSMethod (TMethod equivalents) and it has to transform stack-passed arguments into regular arguments Florin

1 0

VM Maker: VMMaker.oscog.seperateMarking-eem.3341.mcz
by commits＠source.squeak.org 17 Jul '23

17 Jul '23

Eliot Miranda uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog.seperateMarking-eem.3341.mcz ==================== Summary ==================== Name: VMMaker.oscog.seperateMarking-eem.3341 Author: eem Time: 16 July 2023, 7:32:53.705085 pm UUID: 05d9b0a1-2923-4cb8-a433-6461710b54e1 Ancestors: VMMaker.oscog.seperateMarking-eem.3340 Integrate VMMaker.oscog-eem.3332 Cogit: When a link register is in use, fix the failure to save & restore the link register around the unforwarding call in a frameless block activation. =============== Diff against VMMaker.oscog.seperateMarking-eem.3340 =============== Item was changed: ----- Method: CogObjectRepresentationForSpur>>genEnsureOopInRegNotForwarded:scratchReg: (in category 'compile abstract instructions') ----- genEnsureOopInRegNotForwarded: reg scratchReg: scratch + <inline: true> ^ self genEnsureOopInRegNotForwarded: reg scratchReg: scratch jumpBackTo: cogit Label "label is just before the forwarder check"! Item was changed: ----- Method: CogObjectRepresentationForSpur>>genEnsureOopInRegNotForwarded:scratchReg:updatingSlot:in: (in category 'compile abstract instructions') ----- genEnsureOopInRegNotForwarded: reg scratchReg: scratch updatingSlot: index in: objReg "Make sure that the oop in reg is not forwarded, updating the slot in objReg with the value." | loop imm ok | <var: #ok type: #'AbstractInstruction *'> <var: #imm type: #'AbstractInstruction *'> <var: #loop type: #'AbstractInstruction *'> "Open-code self genEnsureOopInRegNotForwarded: reg scratchReg: scratch updatingMw: index * objectMemory wordSize + objectMemory baseHeaderSize r: objReg. to avoid calling the store check unless the receiver is forwarded." self assert: (reg ~= scratch and: [objReg ~= scratch]). loop := cogit Label. imm := self genJumpImmediate: reg. "notionally self genGetClassIndexOfNonImm: reg into: scratch. cogit CmpCq: objectMemory isForwardedObjectClassIndexPun R: TempReg. but the following is an instruction shorter:" cogit MoveMw: 0 r: reg R: scratch. cogit AndCq: objectMemory classIndexMask - objectMemory isForwardedObjectClassIndexPun R: scratch. ok := cogit JumpNonZero: 0. self genLoadSlot: 0 sourceReg: reg destReg: reg. cogit MoveR: reg Mw: index * objectMemory wordSize + objectMemory baseHeaderSize r: objReg. "Check that we're meeting the contract of ceStoreCheckContextReceiverTrampoline." self assert: (reg = Arg0Reg and: [scratch = TempReg and: [objReg = ReceiverResultReg]]). + "If a frame has been built then (if a link register is in use) the link register has been saved. + If not it must be saved and restored around the call since it will be used to return from the method." + cogit needsFrame + ifTrue: + [cogit CallRT: ceStoreCheckContextReceiverTrampoline] + ifFalse: + [cogit backEnd saveAndRestoreLinkRegAround: + [cogit CallRT: ceStoreCheckContextReceiverTrampoline]]. - cogit CallRT: ceStoreCheckContextReceiverTrampoline. cogit Jump: loop. ok jmpTarget: (imm jmpTarget: cogit Label). ^0! Item was added: + ----- Method: Cogit class>>emptyBlockMethod (in category 'tests') ----- + emptyBlockMethod + "Tom Braun found a bug in the code generator in July '23 with generation of returns for frameless full blocks. + Answer a suiitable empty block for in-image compilation tests." + ^[]! Item was changed: + ----- Method: Cogit class>>structureOfACogMethod (in category 'documentation') ----- - ----- Method: Cogit class>>structureOfACogMethod (in category 'tests') ----- structureOfACogMethod "A CogMethod is the machine code for executable code in the Cog VM, and in the simulator these are instances of CogMethod. In actuality they are structures in memory in the CogMethodZone.. There are four real kinds, defined by the cmType field, free space: CMFree, methods: CMMethod, closed PICs: CMClosedPIC (finite polymorphic inline caches with up to 6 entries), and open PICs: CMOpenPIC (infinite megamorphicinline caches that probe the first-level method lookup cache). There is a fifth + kind of method, which is merely a header, for old-style embedded (not full) blocks: CMBlock, one which + exists only within CMMethods, and exist only to allow block activations to refer to something that looks + like a CogMethod. - kind of method, which is merely a header, for blocks: CMBlock, one which exists only within CMMethods, - and exist only to allow block activations to refer to something that looks like a CogMethod. + The blockSize field in a CogMethod is the size in bytes of the entire method, including the header. Methods + are aligned to at least an 8 byte boundary in the CogMethodZone. See CogAbstractInstruction>> + #roundUpToMethodAlignment:. The size is used to iterate over the methods in the zone. - The blockSize field in a CogMethod is the size in bytes of the entire method, including the header. - Methods are aligned to an 8 byte boundary in the CogMethodZone. The size is used to iterate over - the methods in the zone. + Follwing the header is the abort and entry code. Starting immediately after the header is the call to the + abort routine called when either a send fails or a stack limit check fails. Following that is the checked + entry point that checks the receiver is of the right class, and this code ends in the unchecked entry point. + Following this is either primitive code, followed by frame building code, or frame-building code, or, for + frameless methods, the code for the frameless method. Following that is the code for the method. If the + method contains embedded blocks then following the method code will be a CMBlock header, followed by + code for the block, for each block, and following this will be the block dispatch code, which is indexed by + the blockEntryOffset field in the CogMethod. - Follwing the header is the abort and entry code. Starting immediately after the header is the call to - the abort routine called when either a send fails or a stack limit check fails. Following that is the - checked entry point that checks the receiver is of the right class, and this code ends in the unchecked - entry point. Following this is either primitive code, followed by frame building code, or frame-building - code, or, for frameless methods, the code for the frameless method. Following that is the code for the - method. If the method contains blocks then followng the method code will be a CMBlock header, - followed by code for the block, for each block, and following this will be the block dispatch code, - which is indexed by the blockEntryOffset field in the CogMethod. Following either the return in the method, or the block dispatch, is the method map, the meta data which identifies interesting points in the machine code. The map starts at the end of the structure and is read + backwards towards the last instruction of the method, and is terminated by a null byte. So the blockSize - backwards towards the last instruftion of the method, and is terminated by a null byte. So the blockSize is used to find the start of the map. The map reveals where object references, sends and pc-mapping points exist in the machine code. The map is parsed when garbage collecting to find and update object + references, when unlinking sends for method cache flushing, and to convert between bytecode and - references, and when unlinking sends for method cache flushing, and to convert between bytecode and machine code pcs by scanning both bytecode and machine code looking for matching points in the map."! Item was added: + Object subclass: #EmptyFullBlockBug + instanceVariableNames: '' + classVariableNames: '' + poolDictionaries: '' + category: 'VMMaker-Tests'! + + !EmptyFullBlockBug commentStamp: 'eem 7/16/2023 19:11' prior: 0! + An EmptyFullBlockBug is an example of a code generator bug Tom Braun found in July of '23. A frameless full block unforwarding its receiver on entry failed to save and restore the link register (if in use) around the call to follow the receiver. + + Bug doit: + + | receiver block | + receiver := EmptyFullBlockBug new. + block := receiver emptyFullBlock. + 100 timesRepeat: + [receiver becomeForward: Morph new. + block value]! Item was added: + ----- Method: EmptyFullBlockBug>>emptyFullBlock (in category 'accessing') ----- + emptyFullBlock + ^[]! Item was removed: - ----- Method: RegisterAllocatingCogit>>needsFrame (in category 'accessing') ----- - needsFrame - "for asserts" - <cmacro: '() needsFrame'> - ^needsFrame! Item was added: + ----- Method: SimpleStackBasedCogit>>needsFrame (in category 'accessing') ----- + needsFrame + "for asserts" + <cmacro: '() needsFrame'> + ^needsFrame!

1 0

[OpenSmalltalk/opensmalltalk-vm] 9f5bd0: Tidy up. Remove obsolete gnuify in win32 makefiles...
by Eliot Miranda 17 Jul '23

17 Jul '23

Branch: refs/heads/virtend Home: https://github.com/OpenSmalltalk/opensmalltalk-vm Commit: 9f5bd008813420514a67f532e5e8b25f4bcc1de6 https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/9f5bd008813420514a… Author: Eliot Miranda <eliot.miranda(a)gmail.com> Date: 2023-07-16 (Sun, 16 Jul 2023) Changed paths: R platforms/Plan9/vm/sqPlan9prof.c M platforms/RiscOS/vm/sqRPCMain.c M platforms/iOS/vm/OSX/SqSqueakOSXView.m.inc M platforms/iOS/vm/OSX/macintoshosxextra.c M platforms/iOS/vm/OSX/sqSqueakOSXApplication+clipboard.m M platforms/iOS/vm/OSX/sqSqueakOSXApplication+imageReadWrite.m M platforms/iOS/vm/OSX/sqSqueakOSXCGView.m M platforms/iOS/vm/OSX/sqSqueakOSXHeadlessView.m M platforms/iOS/vm/OSX/sqSqueakOSXMetalView.m M platforms/iOS/vm/OSX/sqSqueakOSXOpenGLView.m M platforms/iOS/vm/iPhone/macintoshextra.c M platforms/win32/Makefile.mingw32 M platforms/win32/misc/Makefile.mingw32 R platforms/win32/misc/gnuify Log Message: ----------- Tidy up. Remove obsolete gnuify in win32 makefiles (it's done in VMMaker now). Avoid reiterating a string alloc in the Mac keyboard event chain. Nuke some obsolete profiling stubs. Commit: 45ec5219b660a7407f96b80cd656a2716861ad96 https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/45ec5219b660a7407f… Author: Eliot Miranda <eliot.miranda(a)gmail.com> Date: 2023-07-16 (Sun, 16 Jul 2023) Changed paths: M src/spur32.cog.lowcode/cogit.h M src/spur32.cog.lowcode/cogitARMv5.c M src/spur32.cog.lowcode/cogitIA32.c M src/spur32.cog.lowcode/cointerp.c M src/spur32.cog.lowcode/gcc3x-cointerp.c M src/spur32.cog/cogit.h M src/spur32.cog/cogitARMv5.c M src/spur32.cog/cogitIA32.c M src/spur32.cog/cointerp.c M src/spur32.cog/cointerpmt.c M src/spur32.cog/cointerpmt.h M src/spur32.cog/gcc3x-cointerp.c M src/spur32.cog/gcc3x-cointerpmt.c M src/spur32.sista/cogit.h M src/spur32.sista/cogitARMv5.c M src/spur32.sista/cogitIA32.c M src/spur32.sista/cointerp.c M src/spur32.sista/gcc3x-cointerp.c M src/spur32.stack.lowcode/gcc3x-interp.c M src/spur32.stack.lowcode/interp.c M src/spur32.stack/gcc3x-interp.c M src/spur32.stack/interp.c M src/spur32.stack/validImage.c M src/spur64.cog.lowcode/cogit.h M src/spur64.cog.lowcode/cogitARMv8.c M src/spur64.cog.lowcode/cogitX64SysV.c M src/spur64.cog.lowcode/cogitX64WIN64.c M src/spur64.cog.lowcode/cointerp.c M src/spur64.cog.lowcode/gcc3x-cointerp.c M src/spur64.cog/cogit.h M src/spur64.cog/cogitARMv8.c M src/spur64.cog/cogitX64SysV.c M src/spur64.cog/cogitX64WIN64.c M src/spur64.cog/cointerp.c M src/spur64.cog/cointerpmt.c M src/spur64.cog/cointerpmt.h M src/spur64.cog/gcc3x-cointerp.c M src/spur64.cog/gcc3x-cointerpmt.c M src/spur64.sista/cogit.h M src/spur64.sista/cogitARMv8.c M src/spur64.sista/cogitX64SysV.c M src/spur64.sista/cogitX64WIN64.c M src/spur64.sista/cointerp.c M src/spur64.sista/gcc3x-cointerp.c M src/spur64.stack.lowcode/gcc3x-interp.c M src/spur64.stack.lowcode/interp.c M src/spur64.stack/gcc3x-interp.c M src/spur64.stack/interp.c M src/spur64.stack/validImage.c M src/v3.cog/cogit.h M src/v3.cog/cogitARMv5.c M src/v3.cog/cogitIA32.c M src/v3.cog/cointerp.c M src/v3.cog/gcc3x-cointerp.c M src/v3.stack/gcc3x-interp.c M src/v3.stack/interp.c Log Message: ----------- CogVM source as per VMMaker.oscog-eem.3332 Cogit: When a link register is in use, fix the failure to save & restore the link register around the unforwarding call in a frameless block activation. Thanks to Tom Braun for finding this bug. Compare: https://github.com/OpenSmalltalk/opensmalltalk-vm/compare/24053d2f52b5...45…

1 0

[OpenSmalltalk/opensmalltalk-vm] 57eda1: CogVM source as per VMMaker.oscog-eem.3332
by Eliot Miranda 17 Jul '23

17 Jul '23

Branch: refs/heads/Cog Home: https://github.com/OpenSmalltalk/opensmalltalk-vm Commit: 57eda1c587de3a56353ea64f192aa1b837a86794 https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/57eda1c587de3a5635… Author: Eliot Miranda <eliot.miranda(a)gmail.com> Date: 2023-07-16 (Sun, 16 Jul 2023) Changed paths: M src/spur32.cog.lowcode/cogit.h M src/spur32.cog.lowcode/cogitARMv5.c M src/spur32.cog.lowcode/cogitIA32.c M src/spur32.cog.lowcode/cointerp.c M src/spur32.cog.lowcode/gcc3x-cointerp.c M src/spur32.cog/cogit.h M src/spur32.cog/cogitARMv5.c M src/spur32.cog/cogitIA32.c M src/spur32.cog/cointerp.c M src/spur32.cog/cointerpmt.c M src/spur32.cog/cointerpmt.h M src/spur32.cog/gcc3x-cointerp.c M src/spur32.cog/gcc3x-cointerpmt.c M src/spur32.sista/cogit.h M src/spur32.sista/cogitARMv5.c M src/spur32.sista/cogitIA32.c M src/spur32.sista/cointerp.c M src/spur32.sista/gcc3x-cointerp.c M src/spur32.stack.lowcode/gcc3x-interp.c M src/spur32.stack.lowcode/interp.c M src/spur32.stack/gcc3x-interp.c M src/spur32.stack/interp.c M src/spur32.stack/validImage.c M src/spur64.cog.lowcode/cogit.h M src/spur64.cog.lowcode/cogitARMv8.c M src/spur64.cog.lowcode/cogitX64SysV.c M src/spur64.cog.lowcode/cogitX64WIN64.c M src/spur64.cog.lowcode/cointerp.c M src/spur64.cog.lowcode/gcc3x-cointerp.c M src/spur64.cog/cogit.h M src/spur64.cog/cogitARMv8.c M src/spur64.cog/cogitX64SysV.c M src/spur64.cog/cogitX64WIN64.c M src/spur64.cog/cointerp.c M src/spur64.cog/cointerpmt.c M src/spur64.cog/cointerpmt.h M src/spur64.cog/gcc3x-cointerp.c M src/spur64.cog/gcc3x-cointerpmt.c M src/spur64.sista/cogit.h M src/spur64.sista/cogitARMv8.c M src/spur64.sista/cogitX64SysV.c M src/spur64.sista/cogitX64WIN64.c M src/spur64.sista/cointerp.c M src/spur64.sista/gcc3x-cointerp.c M src/spur64.stack.lowcode/gcc3x-interp.c M src/spur64.stack.lowcode/interp.c M src/spur64.stack/gcc3x-interp.c M src/spur64.stack/interp.c M src/spur64.stack/validImage.c M src/v3.cog/cogit.h M src/v3.cog/cogitARMv5.c M src/v3.cog/cogitIA32.c M src/v3.cog/cointerp.c M src/v3.cog/gcc3x-cointerp.c M src/v3.stack/gcc3x-interp.c M src/v3.stack/interp.c Log Message: ----------- CogVM source as per VMMaker.oscog-eem.3332 Cogit: When a link register is in use, fix the failure to save & restore the link register around the unforwarding call in a frameless block activation. Thanks to Tom Braun for finding this bug.

1 0

VM Maker: VMMaker.oscog-eem.3332.mcz
by commits＠source.squeak.org 17 Jul '23

17 Jul '23

Eliot Miranda uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-eem.3332.mcz ==================== Summary ==================== Name: VMMaker.oscog-eem.3332 Author: eem Time: 16 July 2023, 7:15:08.57287 pm UUID: c611edf1-6978-4222-99d0-0370a8f40bec Ancestors: VMMaker.oscog-eem.3331 Cogit: When a link register is in use, fix the failure to save & restore the link register around the unforwarding call in a frameless block activation. Thanks to Tom Braun for finding this bug. =============== Diff against VMMaker.oscog-eem.3331 =============== Item was changed: ----- Method: CogObjectRepresentationForSpur>>genEnsureOopInRegNotForwarded:scratchReg: (in category 'compile abstract instructions') ----- genEnsureOopInRegNotForwarded: reg scratchReg: scratch + <inline: true> ^ self genEnsureOopInRegNotForwarded: reg scratchReg: scratch jumpBackTo: cogit Label "label is just before the forwarder check"! Item was changed: ----- Method: CogObjectRepresentationForSpur>>genEnsureOopInRegNotForwarded:scratchReg:updatingSlot:in: (in category 'compile abstract instructions') ----- genEnsureOopInRegNotForwarded: reg scratchReg: scratch updatingSlot: index in: objReg "Make sure that the oop in reg is not forwarded, updating the slot in objReg with the value." | loop imm ok | <var: #ok type: #'AbstractInstruction *'> <var: #imm type: #'AbstractInstruction *'> <var: #loop type: #'AbstractInstruction *'> "Open-code self genEnsureOopInRegNotForwarded: reg scratchReg: scratch updatingMw: index * objectMemory wordSize + objectMemory baseHeaderSize r: objReg. to avoid calling the store check unless the receiver is forwarded." self assert: (reg ~= scratch and: [objReg ~= scratch]). loop := cogit Label. imm := self genJumpImmediate: reg. "notionally self genGetClassIndexOfNonImm: reg into: scratch. cogit CmpCq: objectMemory isForwardedObjectClassIndexPun R: TempReg. but the following is an instruction shorter:" cogit MoveMw: 0 r: reg R: scratch. cogit AndCq: objectMemory classIndexMask - objectMemory isForwardedObjectClassIndexPun R: scratch. ok := cogit JumpNonZero: 0. self genLoadSlot: 0 sourceReg: reg destReg: reg. cogit MoveR: reg Mw: index * objectMemory wordSize + objectMemory baseHeaderSize r: objReg. "Check that we're meeting the contract of ceStoreCheckContextReceiverTrampoline." self assert: (reg = Arg0Reg and: [scratch = TempReg and: [objReg = ReceiverResultReg]]). + "If a frame has been built then (if a link register is in use) the link register has been saved. + If not it must be saved and restored around the call since it will be used to return from the method." + cogit needsFrame + ifTrue: + [cogit CallRT: ceStoreCheckContextReceiverTrampoline] + ifFalse: + [cogit backEnd saveAndRestoreLinkRegAround: + [cogit CallRT: ceStoreCheckContextReceiverTrampoline]]. - cogit CallRT: ceStoreCheckContextReceiverTrampoline. cogit Jump: loop. ok jmpTarget: (imm jmpTarget: cogit Label). ^0! Item was added: + ----- Method: Cogit class>>emptyBlockMethod (in category 'tests') ----- + emptyBlockMethod + "Tom Braun found a bug in the code generator in July '23 with generation of returns for frameless full blocks. + Answer a suiitable empty block for in-image compilation tests." + ^[]! Item was changed: + ----- Method: Cogit class>>structureOfACogMethod (in category 'documentation') ----- - ----- Method: Cogit class>>structureOfACogMethod (in category 'tests') ----- structureOfACogMethod "A CogMethod is the machine code for executable code in the Cog VM, and in the simulator these are instances of CogMethod. In actuality they are structures in memory in the CogMethodZone.. There are four real kinds, defined by the cmType field, free space: CMFree, methods: CMMethod, closed PICs: CMClosedPIC (finite polymorphic inline caches with up to 6 entries), and open PICs: CMOpenPIC (infinite megamorphicinline caches that probe the first-level method lookup cache). There is a fifth + kind of method, which is merely a header, for old-style embedded (not full) blocks: CMBlock, one which + exists only within CMMethods, and exist only to allow block activations to refer to something that looks + like a CogMethod. - kind of method, which is merely a header, for blocks: CMBlock, one which exists only within CMMethods, - and exist only to allow block activations to refer to something that looks like a CogMethod. + The blockSize field in a CogMethod is the size in bytes of the entire method, including the header. Methods + are aligned to at least an 8 byte boundary in the CogMethodZone. See CogAbstractInstruction>> + #roundUpToMethodAlignment:. The size is used to iterate over the methods in the zone. - The blockSize field in a CogMethod is the size in bytes of the entire method, including the header. - Methods are aligned to an 8 byte boundary in the CogMethodZone. The size is used to iterate over - the methods in the zone. + Follwing the header is the abort and entry code. Starting immediately after the header is the call to the + abort routine called when either a send fails or a stack limit check fails. Following that is the checked + entry point that checks the receiver is of the right class, and this code ends in the unchecked entry point. + Following this is either primitive code, followed by frame building code, or frame-building code, or, for + frameless methods, the code for the frameless method. Following that is the code for the method. If the + method contains embedded blocks then following the method code will be a CMBlock header, followed by + code for the block, for each block, and following this will be the block dispatch code, which is indexed by + the blockEntryOffset field in the CogMethod. - Follwing the header is the abort and entry code. Starting immediately after the header is the call to - the abort routine called when either a send fails or a stack limit check fails. Following that is the - checked entry point that checks the receiver is of the right class, and this code ends in the unchecked - entry point. Following this is either primitive code, followed by frame building code, or frame-building - code, or, for frameless methods, the code for the frameless method. Following that is the code for the - method. If the method contains blocks then followng the method code will be a CMBlock header, - followed by code for the block, for each block, and following this will be the block dispatch code, - which is indexed by the blockEntryOffset field in the CogMethod. Following either the return in the method, or the block dispatch, is the method map, the meta data which identifies interesting points in the machine code. The map starts at the end of the structure and is read + backwards towards the last instruction of the method, and is terminated by a null byte. So the blockSize - backwards towards the last instruftion of the method, and is terminated by a null byte. So the blockSize is used to find the start of the map. The map reveals where object references, sends and pc-mapping points exist in the machine code. The map is parsed when garbage collecting to find and update object + references, when unlinking sends for method cache flushing, and to convert between bytecode and - references, and when unlinking sends for method cache flushing, and to convert between bytecode and machine code pcs by scanning both bytecode and machine code looking for matching points in the map."! Item was added: + Object subclass: #EmptyFullBlockBug + instanceVariableNames: '' + classVariableNames: '' + poolDictionaries: '' + category: 'VMMaker-Tests'! + + !EmptyFullBlockBug commentStamp: 'eem 7/16/2023 19:11' prior: 0! + An EmptyFullBlockBug is an example of a code generator bug Tom Braun found in July of '23. A frameless full block unforwarding its receiver on entry failed to save and restore the link register (if in use) around the call to follow the receiver. + + Bug doit: + + | receiver block | + receiver := EmptyFullBlockBug new. + block := receiver emptyFullBlock. + 100 timesRepeat: + [receiver becomeForward: Morph new. + block value]! Item was added: + ----- Method: EmptyFullBlockBug>>emptyFullBlock (in category 'accessing') ----- + emptyFullBlock + ^[]! Item was removed: - ----- Method: RegisterAllocatingCogit>>needsFrame (in category 'accessing') ----- - needsFrame - "for asserts" - <cmacro: '() needsFrame'> - ^needsFrame! Item was added: + ----- Method: SimpleStackBasedCogit>>needsFrame (in category 'accessing') ----- + needsFrame + "for asserts" + <cmacro: '() needsFrame'> + ^needsFrame!

1 0

VM Maker: VMMaker.oscog.seperateMarking-eem.3340.mcz
by commits＠source.squeak.org 15 Jul '23

15 Jul '23

Eliot Miranda uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog.seperateMarking-eem.3340.mcz ==================== Summary ==================== Name: VMMaker.oscog.seperateMarking-eem.3340 Author: eem Time: 15 July 2023, 3:29:33.607877 pm UUID: b5db574e-13b4-4d71-bc6e-eff73b40f34e Ancestors: VMMaker.oscog.seperateMarking-WoC.3339 Declare lilliputianChunksToFree in SpurIncrementalSweeper Merge VMMaker.oscog-eem.3331 The StackInterpreter needs a version of followForwardingPointersOfReceiverAndTemporariesInStackZone, not just CoInterpreter. =============== Diff against VMMaker.oscog.seperateMarking-WoC.3339 =============== Item was changed: SpurCompactor subclass: #SpurIncrementalSweeper + instanceVariableNames: 'currentSweepingEntity isCurrentlySweeping currentSegmentUsed currentSegmentUnused currentSegmentsIndex currentsCycleSeenObjectCount currentSegmentsBridge sStartTime lilliputianList lilliputianListSize lilliputianListEnd lilliputianChunksToFree sweepingStart sweptEntities' - instanceVariableNames: 'currentSweepingEntity isCurrentlySweeping currentSegmentUsed currentSegmentUnused currentSegmentsIndex currentsCycleSeenObjectCount currentSegmentsBridge sStartTime lilliputianList lilliputianListSize lilliputianListEnd sweepingStart sweptEntities' classVariableNames: 'MaxObjectsToFree' poolDictionaries: '' category: 'VMMaker-SpurGarbageCollector'! Item was added: + ----- Method: StackInterpreter>>followForwardingPointersOfReceiverAndTemporariesInStackZone (in category 'object memory support') ----- + followForwardingPointersOfReceiverAndTemporariesInStackZone + "A more thorough version of followForwardingPointersInStackZone that also follows all temporaries (but not stack contents after the temps). + This allows removal of the TempVectReadBarrier in the IGC" + <inline: false> + + stackPage ifNil: "the system must be snapshotting; nothing to do..." + [self assert: (stackPages mostRecentlyUsedPage isNil or: [stackPages mostRecentlyUsedPage isFree]). + self cCode: [] inSmalltalk: [self assert: stackPages allPagesFree]. + ^self]. + + self externalWriteBackHeadFramePointers. + + 0 to: numStackPages - 1 do: + [:i| | thePage theFP theSP theIPPtr callerFP offset oop frameRcvrOffset methodHeader | + thePage := stackPages stackPageAt: i. + thePage isFree ifFalse: + [self assert: (self ifCurrentStackPageHasValidHeadPointers: thePage). + theFP := thePage headFP. + "Skip the instruction pointer on top of stack of inactive pages." + theIPPtr := thePage = stackPage ifFalse: [thePage headSP asUnsignedInteger]. + [self assert: (thePage addressIsInPage: theFP). + self assert: (theIPPtr isNil or: [thePage addressIsInPage: theIPPtr asVoidPointer]). + oop := stackPages longAt: theFP + FoxReceiver. + (objectMemory isOopForwarded: oop) ifTrue: + [stackPages + longAt: theFP + FoxReceiver + put: (objectMemory followForwarded: oop)]. + ((self frameHasContext: theFP) + and: [(objectMemory isForwarded: (self frameContext: theFP))]) ifTrue: + [stackPages + longAt: theFP + FoxThisContext + put: (objectMemory followForwarded: (self frameContext: theFP))]. + oop := self frameMethod: theFP. + (objectMemory isForwarded: oop) ifTrue: + [| newOop delta | + newOop := objectMemory followForwarded: oop. + theIPPtr ifNotNil: + [self assert: (stackPages longAt: theIPPtr) > (self frameMethod: theFP). + delta := newOop - oop. + stackPages + longAt: theIPPtr + put: (stackPages longAt: theIPPtr) + delta]. + stackPages + longAt: theFP + FoxMethod + put: (oop := newOop)]. + methodHeader := objectMemory methodHeaderOf: oop. + offset := self frameStackedReceiverOffsetNumArgs: (self argumentCountOfMethodHeader: methodHeader). + oop := stackPages longAt: theFP + offset. + (objectMemory isOopForwarded: oop) ifTrue: + [stackPages + longAt: theFP + offset + put: (objectMemory followForwarded: oop)]. + frameRcvrOffset := theFP + FoxIFReceiver. + theSP := frameRcvrOffset - ((self temporaryCountOfMethodHeader: methodHeader) * objectMemory wordSize). + [theSP <= frameRcvrOffset] whileTrue: + [oop := stackPages longAt: theSP. + (objectMemory isOopForwarded: oop) ifTrue: + [oop := objectMemory followForwarded: oop. + stackPages longAt: theSP put: oop]. + theSP := theSP + objectMemory wordSize]. + (callerFP := self frameCallerFP: theFP) ~= 0] whileTrue: + [theIPPtr := (theFP + FoxCallerSavedIP) asUnsignedInteger. + theFP := callerFP]. + "And finally follow the caller context." + self assert: theFP = thePage baseFP. + oop := self frameCallerContext: theFP. + (objectMemory isForwarded: oop) ifTrue: + [self frameCallerContext: theFP put: (objectMemory followForwarded: oop)]]]!

1 0

VM Maker: VMMaker.oscog-eem.3331.mcz
by commits＠source.squeak.org 15 Jul '23

15 Jul '23

Eliot Miranda uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-eem.3331.mcz ==================== Summary ==================== Name: VMMaker.oscog-eem.3331 Author: eem Time: 15 July 2023, 3:09:20.265325 pm UUID: a3f6d0ef-e74c-414c-a9f1-79fb2c7be0f4 Ancestors: VMMaker.oscog-eem.3330 The StackInterpreter needs a version of followForwardingPointersOfReceiverAndTemporariesInStackZone, not just CoInterpreter. =============== Diff against VMMaker.oscog-eem.3330 =============== Item was added: + ----- Method: StackInterpreter>>followForwardingPointersOfReceiverAndTemporariesInStackZone (in category 'object memory support') ----- + followForwardingPointersOfReceiverAndTemporariesInStackZone + "A more thorough version of followForwardingPointersInStackZone that also follows all temporaries (but not stack contents after the temps). + This allows removal of the TempVectReadBarrier in the IGC" + <inline: false> + + stackPage ifNil: "the system must be snapshotting; nothing to do..." + [self assert: (stackPages mostRecentlyUsedPage isNil or: [stackPages mostRecentlyUsedPage isFree]). + self cCode: [] inSmalltalk: [self assert: stackPages allPagesFree]. + ^self]. + + self externalWriteBackHeadFramePointers. + + 0 to: numStackPages - 1 do: + [:i| | thePage theFP theSP theIPPtr callerFP offset oop frameRcvrOffset methodHeader | + thePage := stackPages stackPageAt: i. + thePage isFree ifFalse: + [self assert: (self ifCurrentStackPageHasValidHeadPointers: thePage). + theFP := thePage headFP. + "Skip the instruction pointer on top of stack of inactive pages." + theIPPtr := thePage = stackPage ifFalse: [thePage headSP asUnsignedInteger]. + [self assert: (thePage addressIsInPage: theFP). + self assert: (theIPPtr isNil or: [thePage addressIsInPage: theIPPtr asVoidPointer]). + oop := stackPages longAt: theFP + FoxReceiver. + (objectMemory isOopForwarded: oop) ifTrue: + [stackPages + longAt: theFP + FoxReceiver + put: (objectMemory followForwarded: oop)]. + ((self frameHasContext: theFP) + and: [(objectMemory isForwarded: (self frameContext: theFP))]) ifTrue: + [stackPages + longAt: theFP + FoxThisContext + put: (objectMemory followForwarded: (self frameContext: theFP))]. + oop := self frameMethod: theFP. + (objectMemory isForwarded: oop) ifTrue: + [| newOop delta | + newOop := objectMemory followForwarded: oop. + theIPPtr ifNotNil: + [self assert: (stackPages longAt: theIPPtr) > (self frameMethod: theFP). + delta := newOop - oop. + stackPages + longAt: theIPPtr + put: (stackPages longAt: theIPPtr) + delta]. + stackPages + longAt: theFP + FoxMethod + put: (oop := newOop)]. + methodHeader := objectMemory methodHeaderOf: oop. + offset := self frameStackedReceiverOffsetNumArgs: (self argumentCountOfMethodHeader: methodHeader). + oop := stackPages longAt: theFP + offset. + (objectMemory isOopForwarded: oop) ifTrue: + [stackPages + longAt: theFP + offset + put: (objectMemory followForwarded: oop)]. + frameRcvrOffset := theFP + FoxIFReceiver. + theSP := frameRcvrOffset - ((self temporaryCountOfMethodHeader: methodHeader) * objectMemory wordSize). + [theSP <= frameRcvrOffset] whileTrue: + [oop := stackPages longAt: theSP. + (objectMemory isOopForwarded: oop) ifTrue: + [oop := objectMemory followForwarded: oop. + stackPages longAt: theSP put: oop]. + theSP := theSP + objectMemory wordSize]. + (callerFP := self frameCallerFP: theFP) ~= 0] whileTrue: + [theIPPtr := (theFP + FoxCallerSavedIP) asUnsignedInteger. + theFP := callerFP]. + "And finally follow the caller context." + self assert: theFP = thePage baseFP. + oop := self frameCallerContext: theFP. + (objectMemory isForwarded: oop) ifTrue: + [self frameCallerContext: theFP put: (objectMemory followForwarded: oop)]]]!

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

Vm-dev July 2023