Vm-dev December 2016

vm-dev@lists.squeakfoundation.org

34 participants
102 discussions

[OpenSmalltalk/opensmalltalk-vm] 1c6147: CogVM source as per VMMaker.oscog-eem.2040
by GitHub 15 Dec '16

15 Dec '16

Branch: refs/heads/Cog Home: https://github.com/OpenSmalltalk/opensmalltalk-vm Commit: 1c6147597cef6cba31456cedc9692d16e6b9fecd https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/1c6147597cef6cba31… Author: Eliot Miranda <eliot.miranda(a)gmail.com> Date: 2016-12-15 (Thu, 15 Dec 2016) Changed paths: M nsspur64src/vm/cogit.h M nsspur64src/vm/cogitX64.c M nsspur64src/vm/cointerp.c M nsspur64src/vm/cointerp.h M nsspur64src/vm/gcc3x-cointerp.c M nsspursrc/vm/cogit.h M nsspursrc/vm/cogitARMv5.c M nsspursrc/vm/cogitIA32.c M nsspursrc/vm/cogitMIPSEL.c M nsspursrc/vm/cointerp.c M nsspursrc/vm/cointerp.h M nsspursrc/vm/gcc3x-cointerp.c M nsspurstack64src/vm/gcc3x-interp.c M nsspurstack64src/vm/interp.c M nsspurstacksrc/vm/gcc3x-interp.c M nsspurstacksrc/vm/interp.c M platforms/Cross/vm/sqSCCSVersion.h M spur64src/vm/cogit.h M spur64src/vm/cogitX64.c M spur64src/vm/cointerp.c M spur64src/vm/cointerp.h M spur64src/vm/gcc3x-cointerp.c M spursista64src/vm/cogit.h M spursista64src/vm/cogitX64.c M spursista64src/vm/cointerp.c M spursista64src/vm/cointerp.h M spursista64src/vm/gcc3x-cointerp.c M spursistasrc/vm/cogit.h M spursistasrc/vm/cogitARMv5.c M spursistasrc/vm/cogitIA32.c M spursistasrc/vm/cogitMIPSEL.c M spursistasrc/vm/cointerp.c M spursistasrc/vm/cointerp.h M spursistasrc/vm/gcc3x-cointerp.c M spursrc/vm/cogit.h M spursrc/vm/cogitARMv5.c M spursrc/vm/cogitIA32.c M spursrc/vm/cogitMIPSEL.c M spursrc/vm/cointerp.c M spursrc/vm/cointerp.h M spursrc/vm/gcc3x-cointerp.c M spurstack64src/vm/gcc3x-interp.c M spurstack64src/vm/interp.c M spurstacksrc/vm/gcc3x-interp.c M spurstacksrc/vm/interp.c M src/plugins/B2DPlugin/B2DPlugin.c M src/plugins/BitBltPlugin/BitBltPlugin.c M src/plugins/DSAPrims/DSAPrims.c M src/plugins/IA32ABI/IA32ABI.c M src/plugins/LargeIntegers/LargeIntegers.c M src/plugins/MacMenubarPlugin/MacMenubarPlugin.c M src/plugins/SqueakFFIPrims/ARM32FFIPlugin.c M src/plugins/SqueakFFIPrims/IA32FFIPlugin.c M src/plugins/SqueakFFIPrims/X64SysVFFIPlugin.c M src/plugins/SqueakFFIPrims/X64Win64FFIPlugin.c M src/plugins/UnixOSProcessPlugin/UnixOSProcessPlugin.c M src/plugins/VMProfileMacSupportPlugin/VMProfileMacSupportPlugin.c M src/plugins/Win32OSProcessPlugin/Win32OSProcessPlugin.c M src/plugins/ZipPlugin/ZipPlugin.c M src/vm/cogit.h M src/vm/cogitARMv5.c M src/vm/cogitIA32.c M src/vm/cogitMIPSEL.c M src/vm/cointerp.c M src/vm/cointerp.h M src/vm/cointerpmt.c M src/vm/cointerpmt.h M src/vm/gcc3x-cointerp.c M src/vm/gcc3x-cointerpmt.c M stacksrc/vm/gcc3x-interp.c M stacksrc/vm/interp.c Log Message: ----------- CogVM source as per VMMaker.oscog-eem.2040 ThreadedX64SysVFFIPlugin Fix regression due to faulty merge. Slang: Make functional methods that start with an assert inlineable and hence make isSmallFloatZero: inlineable, given that it is marked <inline: #always>. To this end: Refactor tryToInlineMethodsIn: into tryToInlineMethodsIn:, tryToInlineMethodExpressionsIn: & tryToInlineMethodStatementsIn:statementListsInto:. Choose to apply tryToInlineMethodStatementsIn:statementListsInto: first (reversing the order of the previous tryToInlineMethodExpressionsIn:) because doing so creates less methods with long comma-chained expressions, which IME can be a source of C compiler bugs. Add a check for failure to inline <inline: #always> methods. Hence regenerate all sources (unchanged plugins have been reverted). StackInterpreter: Remove nsMethodCache in non NewspeakVMs.

1 0

VM Maker: VMMaker.oscog-eem.2040.mcz
by commits＠source.squeak.org 15 Dec '16

15 Dec '16

Eliot Miranda uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-eem.2040.mcz ==================== Summary ==================== Name: VMMaker.oscog-eem.2040 Author: eem Time: 15 December 2016, 1:24:23.427888 pm UUID: 5c1c9278-c66a-4508-8b7c-fb1b0c8b3e46 Ancestors: VMMaker.oscog-eem.2039 ThreadedX64SysVFFIPlugin Fix regression due to faulty merge. Slang: Make functional methods that start with an assert inlineable and hence4 make isSmallFloatZero: inlineable, given that it is marked <inline: #always>. To this end: Refactor tryToInlineMethodsIn: into tryToInlineMethodsIn:, tryToInlineMethodExpressionsIn: & tryToInlineMethodStatementsIn:statementListsInto:. Choose to apply tryToInlineMethodStatementsIn:statementListsInto: first (reversing the order of the previous tryToInlineMethodExpressionsIn:) because doing so creates less methods with long comma-chained expressions, which IME can be a source of C compiler bugs. Add a check for failure to inline <inline: #always> methods. Fix TAssignmentNode>>emitCCodeAsExpressionOn:level:generator: StackInterpreter: Remove nsMethodCache in non NewspeakVMs. =============== Diff against VMMaker.oscog-eem.2039 =============== Item was changed: ----- Method: StackInterpreter class>>declareCVarsIn: (in category 'translation') ----- declareCVarsIn: aCCodeGenerator | vmClass | self class == thisContext methodClass ifFalse: [^self]. "Don't duplicate decls in subclasses" vmClass := aCCodeGenerator vmClass. "Generate primitiveTable etc based on vmClass, not just StackInterpreter" aCCodeGenerator addHeaderFile:'<stddef.h> /* for e.g. alloca */'; addHeaderFile:'<setjmp.h>'; addHeaderFile:'<wchar.h> /* for wint_t */'; addHeaderFile:'"vmCallback.h"'; addHeaderFile:'"sqMemoryFence.h"'; addHeaderFile:'"dispdbg.h"'. LowcodeVM ifTrue: [ aCCodeGenerator addHeaderFile:'"sqLowcodeFFI.h"']. vmClass declareInterpreterVersionIn: aCCodeGenerator defaultName: 'Stack'. aCCodeGenerator var: #interpreterProxy type: #'struct VirtualMachine*'. aCCodeGenerator declareVar: #sendTrace type: 'volatile int'; declareVar: #byteCount type: #usqInt. "These need to be pointers or unsigned." self declareC: #(instructionPointer method newMethod) as: #usqInt in: aCCodeGenerator. "These are all pointers; char * because Slang has no support for C pointer arithmetic." self declareC: #(localIP localSP localFP nativeSP stackPointer framePointer stackLimit breakSelector nativeStackPointer nativeFramePointer shadowCallStack) as: #'char *' in: aCCodeGenerator. aCCodeGenerator var: #breakSelectorLength declareC: 'sqInt breakSelectorLength = MinSmallInteger'. self declareC: #(stackPage overflowedPage) as: #'StackPage *' in: aCCodeGenerator. aCCodeGenerator removeVariable: 'stackPages'. "this is an implicit receiver in the translated code." - NewspeakVM ifFalse: - [aCCodeGenerator - removeVariable: 'localAbsentReceiver'; - removeVariable: 'localAbsentReceiverOrZero'; - removeVariable: 'nsMethodCache']. "This defines bytecodeSetSelector as 0 if MULTIPLEBYTECODESETS is not defined, for the benefit of the interpreter on slow machines." aCCodeGenerator addConstantForBinding: (self bindingOf: #MULTIPLEBYTECODESETS). MULTIPLEBYTECODESETS == false ifTrue: [aCCodeGenerator removeVariable: 'bytecodeSetSelector']. BytecodeSetHasExtensions == false ifTrue: [aCCodeGenerator removeVariable: 'extA'; removeVariable: 'extB']. aCCodeGenerator var: #methodCache declareC: 'sqIntptr_t methodCache[MethodCacheSize + 1 /* ', (MethodCacheSize + 1) printString, ' */]'. + NewspeakVM + ifTrue: + [aCCodeGenerator + var: #nsMethodCache + declareC: 'sqIntptr_t nsMethodCache[NSMethodCacheSize + 1 /* ', (NSMethodCacheSize + 1) printString, ' */]'] + ifFalse: + [aCCodeGenerator + removeVariable: 'localAbsentReceiver'; + removeVariable: 'localAbsentReceiverOrZero']. - aCCodeGenerator - var: #nsMethodCache - declareC: 'sqIntptr_t nsMethodCache[NSMethodCacheSize + 1 /* ', (NSMethodCacheSize + 1) printString, ' */]'. AtCacheTotalSize isInteger ifTrue: [aCCodeGenerator var: #atCache declareC: 'sqInt atCache[AtCacheTotalSize + 1 /* ', (AtCacheTotalSize + 1) printString, ' */]']. aCCodeGenerator var: #primitiveTable declareC: 'void (*primitiveTable[MaxPrimitiveIndex + 2 /* ', (MaxPrimitiveIndex + 2) printString, ' */])(void) = ', vmClass primitiveTableString. vmClass primitiveTable do: [:symbolOrNot| (symbolOrNot isSymbol and: [symbolOrNot ~~ #primitiveFail]) ifTrue: [(aCCodeGenerator methodNamed: symbolOrNot) ifNotNil: [:tMethod| tMethod returnType: #void]]]. vmClass objectMemoryClass hasSpurMemoryManagerAPI ifTrue: [aCCodeGenerator var: #primitiveAccessorDepthTable type: 'signed char' sizeString: 'MaxPrimitiveIndex + 2 /* ', (MaxPrimitiveIndex + 2) printString, ' */' array: vmClass primitiveAccessorDepthTable] ifFalse: [aCCodeGenerator removeVariable: #primitiveAccessorDepthTable]. aCCodeGenerator var: #primitiveFunctionPointer declareC: 'void (*primitiveFunctionPointer)()'. aCCodeGenerator var: #externalPrimitiveTable declareC: 'void (*externalPrimitiveTable[MaxExternalPrimitiveTableSize + 1 /* ', (MaxExternalPrimitiveTableSize + 1) printString, ' */])(void)'. aCCodeGenerator var: #showSurfaceFn type: #'void *'. aCCodeGenerator var: #jmpBuf declareC: 'jmp_buf jmpBuf[MaxJumpBuf + 1 /* ', (MaxJumpBuf + 1) printString, ' */]'. aCCodeGenerator var: #suspendedCallbacks declareC: 'usqInt suspendedCallbacks[MaxJumpBuf + 1 /* ', (MaxJumpBuf + 1) printString, ' */]'. aCCodeGenerator var: #suspendedMethods declareC: 'usqInt suspendedMethods[MaxJumpBuf + 1 /* ', (MaxJumpBuf + 1) printString, ' */]'. aCCodeGenerator var: #interruptCheckChain declareC: 'void (*interruptCheckChain)(void) = 0'. self declareCAsUSqLong: #(nextPollUsecs nextWakeupUsecs longRunningPrimitiveGCUsecs longRunningPrimitiveStartUsecs longRunningPrimitiveStopUsecs "these are high-frequency enough that they're overflowing quite quickly on modern hardware" statProcessSwitch statIOProcessEvents statForceInterruptCheck statCheckForEvents statStackOverflow statStackPageDivorce) in: aCCodeGenerator. aCCodeGenerator var: #nextProfileTick type: #sqLong. LowcodeVM ifTrue: [ aCCodeGenerator var: #shadowCallStackPointer type: #'char*'. aCCodeGenerator var: #lowcodeCalloutState type: #'sqLowcodeCalloutState*' ].! Item was changed: ----- Method: TAssignmentNode>>emitCCodeAsArgumentOn:level:generator: (in category 'C code generation') ----- emitCCodeAsArgumentOn: aStream level: level generator: aCodeGen + ^self emitCCodeAsExpressionOn: aStream level: level generator: aCodeGen! - aStream nextPut: $(. - self emitCCodeOn: aStream level: level generator: aCodeGen. - aStream nextPut: $)! Item was changed: ----- Method: TAssignmentNode>>emitCCodeAsExpressionOn:level:generator: (in category 'C code generation') ----- emitCCodeAsExpressionOn: aStream level: level generator: aCodeGen + (expression isStmtList and: [expression statements size > 1]) ifTrue: + [^self emitStatementListExpansionAsExpression: expression on: aStream level: level generator: aCodeGen]. aStream nextPut: $(. self emitCCodeOn: aStream level: level generator: aCodeGen. aStream nextPut: $)! Item was added: + ----- Method: TAssignmentNode>>emitStatementListExpansionAsExpression:on:level:generator: (in category 'C code generation') ----- + emitStatementListExpansionAsExpression: stmtList on: aStream level: level generator: aCodeGen + stmtList statements last = variable ifTrue: + [^expression emitCCodeAsExpressionOn: aStream level: level generator: aCodeGen]. + stmtList copy + assignLastExpressionTo: variable; + emitCCodeAsExpressionOn: aStream level: level generator: aCodeGen! Item was added: + ----- Method: TMethod>>checkForRequiredInlinability (in category 'testing') ----- + checkForRequiredInlinability + "This is used in methods answering inlinability. + Always answer false. But if the receiver is marked as something that must be inlined (inline == #always) raise an error." + (inline == #always and: [complete]) ifTrue: + [self error: 'cannot inline method ', selector, ' marked as <inline: #always>']. + ^false! Item was changed: ----- Method: TMethod>>inlineBuiltin:in: (in category 'inlining') ----- inlineBuiltin: aSendNode in: aCodeGen | sel meth inlinedReplacement | (aSendNode selector beginsWith: 'perform:') ifTrue: [^self inlineFunctionCall: aSendNode asTransformedConstantPerform in: aCodeGen]. sel := aSendNode receiver selector. meth := aCodeGen methodNamed: sel. (meth notNil and: [meth inline == true]) ifFalse: [^nil]. + (meth isFunctionalIn: aCodeGen) ifTrue: - meth isFunctional ifTrue: [inlinedReplacement := (aCodeGen methodNamed: aSendNode receiver selector) copy inlineFunctionCall: aSendNode receiver in: aCodeGen. ^TSendNode new setSelector: aSendNode selector receiver: inlinedReplacement arguments: aSendNode args copy]. (self isInlineableConditional: aSendNode in: aCodeGen) ifTrue: [^self inlineConditional: aSendNode in: aCodeGen]. ^nil! Item was changed: ----- Method: TMethod>>inlineFunctionCall:in: (in category 'inlining') ----- inlineFunctionCall: aSendNode in: aCodeGen "Answer the body of the called function, substituting the actual parameters for the formal argument variables in the method body. Assume caller has established that: 1. the method arguments are all substitutable nodes, and 2. the method to be inlined contains no additional embedded returns." | sel meth doNotRename argsForInlining substitutionDict | sel := aSendNode selector. meth := (aCodeGen methodNamed: sel) copy. meth ifNil: [^self inlineBuiltin: aSendNode in: aCodeGen]. doNotRename := Set withAll: args. argsForInlining := aSendNode argumentsForInliningCodeGenerator: aCodeGen. meth args with: argsForInlining do: [ :argName :exprNode | exprNode isLeaf ifTrue: [doNotRename add: argName]]. (meth statements size = 2 and: [meth statements first isSend and: [meth statements first selector == #flag:]]) ifTrue: [meth statements removeFirst]. meth renameVarsForInliningInto: self except: doNotRename in: aCodeGen. meth renameLabelsForInliningInto: self. self addVarsDeclarationsAndLabelsOf: meth except: doNotRename. substitutionDict := Dictionary new: meth args size * 2. meth args with: argsForInlining do: [ :argName :exprNode | substitutionDict at: argName put: exprNode. (doNotRename includes: argName) ifFalse: [locals remove: argName]]. meth parseTree bindVariablesIn: substitutionDict. + ^meth parseTree endsWithReturn + ifTrue: [meth parseTree copyWithoutReturn] - ^meth statements first isReturn - ifTrue: [meth statements first expression] ifFalse: [meth parseTree]! Item was changed: ----- Method: TMethod>>inlineSend:directReturn:exitVar:in: (in category 'inlining') ----- inlineSend: aSendNode directReturn: directReturn exitVar: exitVar in: aCodeGen "Answer a collection of statements to replace the given send. directReturn indicates that the send is the expression in a return statement, so returns can be left in the body of the inlined method. If exitVar is nil, the value returned by the send is not used; thus, returns need not assign to the output variable. Types are propagated to as-yet-untyped variables when inlining a send that is assigned, otherwise the assignee variable type must match the return type of the inlinee. Return types are not propagated." | sel meth methArgs exitLabel inlineStmts label exitType | sel := aSendNode selector. meth := aCodeGen methodNamed: sel. methArgs := meth args. "convenient for debugging..." aCodeGen maybeBreakForInlineOf: aSendNode in: self. (methArgs notEmpty and: [methArgs first beginsWith: 'self_in_']) ifTrue: [methArgs := methArgs allButFirst]. methArgs size = aSendNode args size ifFalse: [^nil]. meth := meth copy. + (meth statements size > 1 + and: [meth statements first isSend + and: [meth statements first selector == #flag:]]) ifTrue: + [meth statements removeFirst]. + "Propagate the return type of an inlined method" (directReturn or: [exitVar notNil]) ifTrue: [exitType := directReturn ifTrue: [returnType] ifFalse: [(self typeFor: exitVar in: aCodeGen) ifNil: [#sqInt]]. (exitType = #void or: [exitType = meth returnType]) ifFalse: [meth propagateReturnIn: aCodeGen]]. "Propagate any unusual argument types to untyped argument variables" methArgs with: aSendNode args do: [:formal :actual| (meth declarationAt: formal ifAbsent: nil) ifNil: [(self typeFor: actual in: aCodeGen) ifNotNil: [:type| type ~= #sqInt ifTrue: [meth declarationAt: formal put: (type last = $* ifTrue: [type, formal] ifFalse: [type, ' ', formal])]]]]. meth renameVarsForInliningInto: self except: #() in: aCodeGen. meth renameLabelsForInliningInto: self. self addVarsDeclarationsAndLabelsOf: meth except: #(). meth hasReturn ifTrue: [directReturn ifFalse: [exitLabel := self unusedLabelForInliningInto: self. (meth exitVar: exitVar label: exitLabel) "is label used?" ifTrue: [ labels add: exitLabel ] ifFalse: [ exitLabel := nil ]]]. (inlineStmts := OrderedCollection new: meth statements size + meth args size + 2) add: (label := TLabeledCommentNode new setComment: 'begin ', sel); addAll: (self argAssignmentsFor: meth send: aSendNode in: aCodeGen); addAll: meth statements. "method body" directReturn ifTrue: [meth endsWithReturn ifTrue: [exitVar ifNotNil: "don't remove the returns if being invoked in the context of a return" [inlineStmts at: inlineStmts size put: inlineStmts last copyWithoutReturn]] ifFalse: [inlineStmts add: (TReturnNode new setExpression: (TVariableNode new setName: 'nil'))]]. exitLabel ifNotNil: [inlineStmts add: (TLabeledCommentNode new setLabel: exitLabel comment: 'end ', meth selector)]. inlineStmts size = 1 ifTrue: "Nuke empty methods; e.g. override of flushAtCache" [self assert: inlineStmts first isComment. inlineStmts removeFirst]. ^inlineStmts! Item was changed: ----- Method: TMethod>>inlineableFunctionCall:in: (in category 'inlining') ----- inlineableFunctionCall: aNode in: aCodeGen + "Answer if the given send node is a call to a 'functional' method--a method whose body is a single return statement of some expression and whose actual parameters can all be directly substituted." - "Answer true if the given send node is a call to a 'functional' method--a method whose body is a single return statement of some expression and whose actual parameters can all be directly substituted." aCodeGen maybeBreakForTestToInline: aNode in: self. aNode isSend ifFalse: [^false]. ^(aCodeGen methodNamed: aNode selector) ifNil: [aNode asTransformedConstantPerform ifNil: [self isInlineableConditional: aNode in: aCodeGen] ifNotNil: [:n| self inlineableFunctionCall: n in: aCodeGen]] ifNotNil: [:m| + (m ~~ self + and: [(m isFunctionalIn: aCodeGen) + and: [(aCodeGen mayInline: m selector) + and: [aNode args allSatisfy: [:a| self isSubstitutableNode: a intoMethod: m in: aCodeGen]]]]) + or: [m checkForRequiredInlinability]]! - m ~~ self - and: [m isFunctional - and: [(aCodeGen mayInline: m selector) - and: [aNode args allSatisfy: [ :a | self isSubstitutableNode: a intoMethod: m in: aCodeGen]]]]]! Item was changed: ----- Method: TMethod>>inlineableSend:in: (in category 'inlining') ----- inlineableSend: aNode in: aCodeGen + "Answer if the given send node is a call to a method that can be inlined." - "Answer true if the given send node is a call to a method that can be inlined." | m | aCodeGen maybeBreakForTestToInline: aNode in: self. + aNode isSend ifFalse: [^false]. - aNode isSend ifFalse: [ ^false ]. m := aCodeGen methodNamed: aNode selector. "nil if builtin or external function" + ^m ~= nil + and: [m ~~ self + and: [(m isComplete and: [aCodeGen mayInline: m selector]) + or: [m checkForRequiredInlinability]]]! - ^m ~= nil and: [m ~~ self and: [m isComplete and: [aCodeGen mayInline: m selector]]]! Item was removed: - ----- Method: TMethod>>isFunctional (in category 'inlining') ----- - isFunctional - "Answer true if the receiver is a functional method. That is, if it - consists of a single return statement of an expression that contains - no other returns. - - Answer false for methods with return types other than the simple - integer types to work around bugs in the inliner." - - parseTree statements isEmpty ifTrue: - [^false]. - parseTree statements last isReturn ifFalse: - [^false]. - parseTree statements size = 1 ifFalse: - [(parseTree statements size = 2 - and: [parseTree statements first isSend - and: [parseTree statements first selector == #flag:]]) ifFalse: - [^false]]. - parseTree statements last expression nodesDo: - [ :n | n isReturn ifTrue: [^false]]. - ^#(int #'unsigned int' #long #'unsigned long' #'long long' #'unsigned long long' - sqInt usqInt #'sqIntptr_t' #'usqIntptr_t' sqLong usqLong - #'int *' #'unsigned int *' #'sqInt *' #'usqInt *' #'sqLong *' #'usqLong *' #'CogMethod *' #'char *') includes: returnType! Item was added: + ----- Method: TMethod>>isFunctionalIn: (in category 'inlining') ----- + isFunctionalIn: aCodeGen + "Answer if the receiver is a functional method. That is, if it + consists of a single return statement of an expression that + contains no other returns, or an assert or flag followed by + such a statement. + + Answer false for methods with return types other than the simple + integer types to work around bugs in the inliner." + + parseTree statements size = 1 ifFalse: + [(parseTree statements size = 2 + and: [parseTree statements first isSend + and: [parseTree statements first selector == #flag: + or: [(aCodeGen isAssertSelector: parseTree statements first selector) + and: [parseTree statements first selector ~~ #asserta:]]]]) ifFalse: + [^false]]. + parseTree statements last isReturn ifFalse: + [^false]. + parseTree statements last expression nodesDo: + [ :n | n isReturn ifTrue: [^false]]. + ^#(int #'unsigned int' #long #'unsigned long' #'long long' #'unsigned long long' + sqInt usqInt #'sqIntptr_t' #'usqIntptr_t' sqLong usqLong + #'int *' #'unsigned int *' #'sqInt *' #'usqInt *' #'sqLong *' #'usqLong *' #'CogMethod *' #'char *') includes: returnType! Item was added: + ----- Method: TMethod>>tryToInlineMethodExpressionsIn: (in category 'inlining') ----- + tryToInlineMethodExpressionsIn: aCodeGen + "Expand any (complete) inline methods sent by this method as receivers or parameters. + Answer if anything was inlined." + + | sendsToInline | + sendsToInline := Dictionary new: 100. + parseTree + nodesDo: + [:node| + (self transformConditionalAssignment: node in: aCodeGen) ifNotNil: + [:replacement| + sendsToInline at: node put: replacement]. + (self inlineableFunctionCall: node in: aCodeGen) ifTrue: + [(self inlineFunctionCall: node in: aCodeGen) ifNotNil: + [:replacement| + sendsToInline at: node put: replacement]]] + unless: "Don't inline the arguments to asserts to keep the asserts readable" + [:node| + node isSend + and: [node selector == #cCode:inSmalltalk: + or: [aCodeGen isAssertSelector: node selector]]]. + + sendsToInline isEmpty ifTrue: + [^false]. + parseTree := parseTree replaceNodesIn: sendsToInline. + ^true! Item was added: + ----- Method: TMethod>>tryToInlineMethodStatementsIn:statementListsInto: (in category 'inlining') ----- + tryToInlineMethodStatementsIn: aCodeGen statementListsInto: aBlock + "Expand any (complete) inline methods sent by this method as top-level statements. + Answer if anything was inlined." + + | stmtLists didSomething newStatements returningNodes | + didSomething := false. + returningNodes := Set new. + parseTree nodesDo: + [:node| + node isReturn ifTrue: + [returningNodes add: node expression. + node expression isConditionalSend ifTrue: + [returningNodes addAll: (node expression args collect: [:stmtList| stmtList statements last])]]]. + stmtLists := self statementsListsForInliningIn: aCodeGen. + stmtLists do: + [:stmtList| + newStatements := OrderedCollection new: stmtList statements size. + stmtList statements do: + [:stmt| + (self inlineCodeOrNilForStatement: stmt returningNodes: returningNodes in: aCodeGen) + ifNil: [newStatements addLast: stmt] + ifNotNil: [:inlinedStmts| + didSomething := true. + newStatements addAllLast: inlinedStmts]]. + stmtList setStatements: newStatements asArray]. + + "This is a hack; forgive me. The inlining above tends to keep return statements in statement lists. + In the case of returning ifs we don't want the returns in case the returning if is generated as an expression." + returningNodes do: + [:returningNode| + (returningNode isConditionalSend + and: [returningNode args anySatisfy: [:alternativeNode| alternativeNode endsWithReturn]]) ifTrue: + [returningNode args withIndexDo: + [:alternativeNode :index| + alternativeNode endsWithReturn ifTrue: + [returningNode args at: index put: alternativeNode copyWithoutReturn]]]]. + + aBlock value: stmtLists. + + ^didSomething! Item was changed: ----- Method: TMethod>>tryToInlineMethodsIn: (in category 'inlining') ----- tryToInlineMethodsIn: aCodeGen + "Expand any (complete) inline methods sent by this method. + Set the complete flag when all inlining has been done. + Answer if something was inlined." - "Expand any (complete) inline methods called by this method. Set the complete bit when all inlining has been done. Return true if something was inlined." + | didSomething statementLists | - | stmtLists didSomething newStatements sendsToInline returningNodes | self definedAsMacro ifTrue: [complete := true. ^false]. + didSomething := self tryToInlineMethodStatementsIn: aCodeGen statementListsInto: [:stmtLists| statementLists := stmtLists]. + didSomething := (self tryToInlineMethodExpressionsIn: aCodeGen) or: [didSomething]. - didSomething := false. - sendsToInline := Dictionary new: 100. - parseTree - nodesDo: - [:node| - (self transformConditionalAssignment: node in: aCodeGen) ifNotNil: - [:replacement| - sendsToInline at: node put: replacement]. - (self inlineableFunctionCall: node in: aCodeGen) ifTrue: - [(self inlineFunctionCall: node in: aCodeGen) ifNotNil: - [:replacement| - sendsToInline at: node put: replacement]]] - unless: "Don't inline the arguments to asserts to keep the asserts readable" - [:node| - node isSend - and: [node selector == #cCode:inSmalltalk: - or: [aCodeGen isAssertSelector: node selector]]]. - sendsToInline isEmpty ifFalse: - [didSomething := true. - parseTree := parseTree replaceNodesIn: sendsToInline]. - didSomething ifTrue: [writtenToGlobalVarsCache := nil. ^didSomething]. - returningNodes := Set new. - parseTree nodesDo: - [:node| - node isReturn ifTrue: - [returningNodes add: node expression. - node expression isConditionalSend ifTrue: - [returningNodes addAll: (node expression args collect: [:stmtList| stmtList statements last])]]]. - stmtLists := self statementsListsForInliningIn: aCodeGen. - stmtLists do: - [:stmtList| - newStatements := OrderedCollection new: stmtList statements size. - stmtList statements do: - [:stmt| - (self inlineCodeOrNilForStatement: stmt returningNodes: returningNodes in: aCodeGen) - ifNil: [newStatements addLast: stmt] - ifNotNil: [:inlinedStmts| - didSomething := true. - newStatements addAllLast: inlinedStmts]]. - stmtList setStatements: newStatements asArray]. - - "This is a hack; forgive me. The inlining abiove tends to keep return statements in statement lists. - In the case of returning ifs we don't want the returns in case the returning if is generated as an expression." - returningNodes do: - [:returningNode| - (returningNode isConditionalSend - and: [returningNode args anySatisfy: [:alternativeNode| alternativeNode endsWithReturn]]) ifTrue: - [returningNode args withIndexDo: - [:alternativeNode :index| - alternativeNode endsWithReturn ifTrue: - [returningNode args at: index put: alternativeNode copyWithoutReturn]]]]. - - didSomething ifTrue: - [writtenToGlobalVarsCache := nil. - ^didSomething]. - complete ifFalse: + [self checkForCompleteness: statementLists in: aCodeGen. + complete ifTrue: [didSomething := true]]. "marking a method complete is progress" - [self checkForCompleteness: stmtLists in: aCodeGen. - complete ifTrue: [ didSomething := true ]]. "marking a method complete is progress" ^didSomething! Item was changed: ----- Method: TStmtListNode>>copyWithoutReturn (in category 'transformations') ----- copyWithoutReturn self assert: self endsWithReturn. + statements size = 1 ifTrue: + [^statements last expression]. ^self class new setArguments: arguments statements: statements allButLast, {statements last copyWithoutReturn}; yourself! Item was changed: ----- Method: ThreadedX64SysVFFIPlugin>>ffiCalloutTo:SpecOnStack:in: (in category 'callout support') ----- ffiCalloutTo: procAddr SpecOnStack: specOnStack in: calloutState <var: #procAddr type: #'void *'> <var: #calloutState type: #'CalloutState *'> <var: #loadFloatRegs declareC: 'extern void loadFloatRegs(double, double, double, double, double, double, double, double)'> "Go out, call this guy and create the return value. This *must* be inlined because of the alloca of the outgoing stack frame in ffiCall:WithFlags:NumArgs:Args:AndTypes:" | myThreadIndex atomicType floatRet intRet loadFloatRegs | <var: #floatRet type: #double> + <var: #intRet type: #SixteenByteReturn> - <var: #intRet type: 'SixteenByteReturn'> <inline: true> self cCode: '' inSmalltalk: [loadFloatRegs := #used. loadFloatRegs class]. self maybeDisownVM: calloutState threadIndexInto: [:threadIndex| myThreadIndex := threadIndex]. calloutState floatRegisterIndex > 0 ifTrue: [self load: (calloutState floatRegisters at: 0) Flo: (calloutState floatRegisters at: 1) a: (calloutState floatRegisters at: 2) t: (calloutState floatRegisters at: 3) R: (calloutState floatRegisters at: 4) e: (calloutState floatRegisters at: 5) g: (calloutState floatRegisters at: 6) s: (calloutState floatRegisters at: 7)]. (self allocaLiesSoSetSpBeforeCall or: [self mustAlignStack]) ifTrue: [self setsp: calloutState argVector]. atomicType := self atomicTypeOf: calloutState ffiRetHeader. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [atomicType = FFITypeSingleFloat ifTrue: [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'float (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3) with: (calloutState integerRegisters at: 4) with: (calloutState integerRegisters at: 5)] ifFalse: "atomicType = FFITypeDoubleFloat" [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3) with: (calloutState integerRegisters at: 4) with: (calloutState integerRegisters at: 5)]. self maybeOwnVM: calloutState threadIndex: myThreadIndex. ^interpreterProxy floatObjectOf: floatRet]. intRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'SixteenByteReturn (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3) with: (calloutState integerRegisters at: 4) with: (calloutState integerRegisters at: 5). self maybeOwnVM: calloutState threadIndex: myThreadIndex. (calloutState ffiRetHeader anyMask: FFIFlagPointer+FFIFlagStructure) ifTrue: ["Note: Order is important here since FFIFlagPointer + FFIFlagStructure is used to represent 'typedef void* VoidPointer' and VoidPointer must be returned as pointer *not* as struct." (calloutState ffiRetHeader anyMask: FFIFlagPointer) ifTrue: + [^self ffiReturnPointer: intRet a ofType: (self ffiReturnType: specOnStack) in: calloutState]. - [^self ffiReturnPointer: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. ^self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. + ^self ffiCreateIntegralResultOop: intRet a ofAtomicType: atomicType in: calloutState! - ^self ffiCreateIntegralResultOop: intRet ofAtomicType: atomicType in: calloutState! Item was changed: ----- Method: ThreadedX64SysVFFIPlugin>>ffiReturnStruct:ofType:in: (in category 'callout support') ----- ffiReturnStruct: sixteenByteRet ofType: ffiRetType in: calloutState + <var: #sixteenByteRet type: #SixteenByteReturn> - <var: #sixteenByteRet type: 'SixteenByteReturn'> <var: #calloutState type: #'CalloutState *'> "Create a structure return value from an external function call. The value has been stored in alloca'ed space pointed to by the calloutState or in the return value." | retOop retClass oop | <inline: true> retClass := interpreterProxy fetchPointer: 1 ofObject: ffiRetType. retOop := interpreterProxy instantiateClass: retClass indexableSize: 0. self remapOop: retOop in: [oop := interpreterProxy instantiateClass: interpreterProxy classByteArray indexableSize: calloutState structReturnSize]. self mem: (interpreterProxy firstIndexableField: oop) cp: ((self returnStructInRegisters: calloutState structReturnSize) ifTrue: [(self addressOf: sixteenByteRet) asVoidPointer] ifFalse: [calloutState limit]) y: calloutState structReturnSize. interpreterProxy storePointer: 0 ofObject: retOop withValue: oop. ^retOop!

1 0

[OpenSmalltalk/opensmalltalk-vm] ba2b5a: Undo the wreckage from my clumsy merge :-(
by GitHub 15 Dec '16

15 Dec '16

Branch: refs/heads/Cog Home: https://github.com/OpenSmalltalk/opensmalltalk-vm Commit: ba2b5a71a0e276c5c27d9f5bd15bd507a69a5e6f https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/ba2b5a71a0e276c5c2… Author: Eliot Miranda <eliot.miranda(a)gmail.com> Date: 2016-12-14 (Wed, 14 Dec 2016) Changed paths: M src/plugins/SqueakFFIPrims/ARM32FFIPlugin.c M src/plugins/SqueakFFIPrims/IA32FFIPlugin.c M src/plugins/SqueakFFIPrims/X64SysVFFIPlugin.c M src/plugins/SqueakFFIPrims/X64Win64FFIPlugin.c Log Message: ----------- Undo the wreckage from my clumsy merge :-(

1 0

[OpenSmalltalk/opensmalltalk-vm] 6653d1: CogVM source as per VMMaker.oscog-eem.2039
by GitHub 15 Dec '16

15 Dec '16

Branch: refs/heads/Cog Home: https://github.com/OpenSmalltalk/opensmalltalk-vm Commit: 6653d19a0ebd0d05e8cfc5d7bb0fcc80f5d81f98 https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/6653d19a0ebd0d05e8… Author: Eliot Miranda <eliot.miranda(a)gmail.com> Date: 2016-12-14 (Wed, 14 Dec 2016) Changed paths: M nsspur64src/vm/cogitX64.c M nsspur64src/vm/cointerp.c M nsspur64src/vm/cointerp.h M nsspur64src/vm/gcc3x-cointerp.c M nsspursrc/vm/cogitARMv5.c M nsspursrc/vm/cogitIA32.c M nsspursrc/vm/cogitMIPSEL.c M nsspursrc/vm/cointerp.c M nsspursrc/vm/cointerp.h M nsspursrc/vm/gcc3x-cointerp.c M nsspurstack64src/vm/gcc3x-interp.c M nsspurstack64src/vm/interp.c M nsspurstacksrc/vm/gcc3x-interp.c M nsspurstacksrc/vm/interp.c M spur64src/vm/cogitX64.c M spur64src/vm/cointerp.c M spur64src/vm/cointerp.h M spur64src/vm/gcc3x-cointerp.c M spursista64src/vm/cogitX64.c M spursista64src/vm/cointerp.c M spursista64src/vm/cointerp.h M spursista64src/vm/gcc3x-cointerp.c M spursistasrc/vm/cogitARMv5.c M spursistasrc/vm/cogitIA32.c M spursistasrc/vm/cogitMIPSEL.c M spursistasrc/vm/cointerp.c M spursistasrc/vm/cointerp.h M spursistasrc/vm/gcc3x-cointerp.c M spursrc/vm/cogitARMv5.c M spursrc/vm/cogitIA32.c M spursrc/vm/cogitMIPSEL.c M spursrc/vm/cointerp.c M spursrc/vm/cointerp.h M spursrc/vm/gcc3x-cointerp.c M spurstack64src/vm/gcc3x-interp.c M spurstack64src/vm/interp.c M spurstacksrc/vm/gcc3x-interp.c M spurstacksrc/vm/interp.c M src/plugins/SqueakFFIPrims/ARM32FFIPlugin.c M src/plugins/SqueakFFIPrims/IA32FFIPlugin.c M src/plugins/SqueakFFIPrims/X64SysVFFIPlugin.c M src/plugins/SqueakFFIPrims/X64Win64FFIPlugin.c M src/vm/cogitARMv5.c M src/vm/cogitIA32.c M src/vm/cogitMIPSEL.c M src/vm/cointerp.c M src/vm/cointerp.h M src/vm/cointerpmt.c M src/vm/cointerpmt.h M src/vm/gcc3x-cointerp.c M src/vm/gcc3x-cointerpmt.c M stacksrc/vm/gcc3x-interp.c M stacksrc/vm/interp.c Log Message: ----------- CogVM source as per VMMaker.oscog-eem.2039 ThreadedFFIPlugin (the FFI plugin) Fix callout failures for callouts that invoke callbacks. The VM checks that interpreter primitive calls (including external calls) have correctly-balanced stacks on return. The ThreadedFFIPlugin (and indeed the old FFIPlugin) would use argumentCount to compute the stack on return, but a callback can alter argumentCount to arbitrary values, and callback return can't easily determine the value of argumentCount to restore (especially in a threaded context). So the plugin must remember the argumentCount before it calls out and should use this cached value in restoring the stack. To this end, remove all uses of methodReturnValue: in the support code for th ecallouts and use pop:thenPush: in ffiCall:ArgArrayOrNil:NumArgs: to update the stack correctly. Streamline the callouts in the plugin to eliminate an extra test on the return type being float. Neaten the code for (dis)owning the VM. Revise ALLOCA_LIES_SO_USE_GET_SP The extra space allocated below the value returned by alloca() is used by mingw32 gcc (it is used for marshalling args of sub-functions - those who marshall ffi call args) So we cannot use it for marshalling FFI arguments. Thus we must change the strategy: rather than using getsp() for changing the value returned by alloca() before marshalling args (which will lead to SEGV) we must rather use setsp() for changing the stack pointer before calling the target function. Thus we rename the macro into ALLOCA_LIES_SO_SET_SP_BEFORE_CALL Remind that __clang__'s alloca does not lie. (clang compiler does define __GNUC__) --- Revise FFI stack alignment: 1) use STACK_BYTES_ALIGNMENT and getsp() as provided by include file sqCogStackAlignment.h 2) use a new macro MUST_ALIGN_STACK that check STACK_BYTES_ALIGNMENT against sizeof(void *). Indeed, some nice STACK_BYTES_ALIGNMENT might be > 0 now, which does not necessarily mean that we must realign the stack. This macro is used both for aligning alloca'ed space before marshalling and for aligning SP before call. It is still questionable whether we must do some further alignment over alloca(), because alloca() should be aware of ABI requirements, but if it ain't broken... --- nuke STACK_OFFSET_BYTES from the preamble since it is unused. --- get rid of registerArgsSlop It was intended for reserving stack space for shadowing register args as mandated by Win64 X64 ABI (and maybe PPC). But this is completely unecessary as this stack space will be reserved by the callout itself (at least on WIN64). For PPC, don't do anything, it's currently unsupported (unmaintained/untested/unnecessary) --- Let cogit generate remove float args from stack only if more than 0 float args (by symmetry with int args). This was forgotten in a previous commit. Note that this is only used by lowcode experiments and still waits to be ported to architectures other than IA32. StackInterpreter (helping resolve callback issue above): Refactor handleStackOverflow into checkForStackOverflow and handleStackOverflow to get better inlining/better readability. Fix a bug in checkStackPointerIndexForFrame: for the current frame that caused many false assert-fails during callbacks. SistaV1: Added support for BranchIfInstanceOf in addition to BranchIfNotInstanceOf by using the sign bit of extB as annotation. Commit: 78489f4909b34c9497af948b656a0ef34626ce74 https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/78489f4909b34c9497… Author: Eliot Miranda <eliot.miranda(a)gmail.com> Date: 2016-12-14 (Wed, 14 Dec 2016) Changed paths: M platforms/Cross/vm/sqCogStackAlignment.h M platforms/unix/vm/sqUnixMain.c M platforms/win32/vm/sqWin32Main.c M platforms/win32/vm/sqWin32Window.c M src/plugins/SqueakFFIPrims/ARM32FFIPlugin.c M src/plugins/SqueakFFIPrims/IA32FFIPlugin.c M src/plugins/SqueakFFIPrims/SqueakFFIPrims.c M src/plugins/SqueakFFIPrims/X64SysVFFIPlugin.c M src/plugins/SqueakFFIPrims/X64Win64FFIPlugin.c M third-party/libpng.spec.win Log Message: ----------- Merge branch 'Cog' of http://github.com/OpenSmalltalk/vm into Cog Clumsy merge with 3a6a4e05ed7fe09f6f6702bc8207059f88b3c7a5 Commit: ce197ec08e8c54887b8a5010bbe0097b0ff29242 https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/ce197ec08e8c54887b… Author: Eliot Miranda <eliot.miranda(a)gmail.com> Date: 2016-12-14 (Wed, 14 Dec 2016) Changed paths: M src/plugins/SqueakFFIPrims/ARM32FFIPlugin.c M src/plugins/SqueakFFIPrims/IA32FFIPlugin.c M src/plugins/SqueakFFIPrims/X64SysVFFIPlugin.c M src/plugins/SqueakFFIPrims/X64Win64FFIPlugin.c Log Message: ----------- Merge 3a6a4e05ed7fe09f6f6702bc8207059f88b3c7a5 & 6653d19a0ebd0d05e8cfc5d7bb0fcc80f5d81f98 Compare: https://github.com/OpenSmalltalk/opensmalltalk-vm/compare/3a6a4e05ed7f...ce…

1 0

VM Maker: VMMaker.oscog-eem.2039.mcz
by commits＠source.squeak.org 15 Dec '16

15 Dec '16

Eliot Miranda uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-eem.2039.mcz ==================== Summary ==================== Name: VMMaker.oscog-eem.2039 Author: eem Time: 14 December 2016, 3:19:42.124426 pm UUID: 78dde0e7-e826-47ad-b2ef-4a8aff88b963 Ancestors: VMMaker.oscog-nice.2038 ThreadedFFIPlugin (the FFI plugin) Fix callout failures for callouts that invoke callbacks. The VM checks that interpreter primitive calls (including external calls) have correctly-balanced stacks on return. The ThreadedFFIPlugin (and indeed the old FFIPlugin) would use argumentCount to compute the stack on return, but a callback can alter argumentCount to arbitrary values, and callback return can't easily determine the value of argumentCount to restore (especially in a threaded context). So the plugin must remember the argumentCount before it calls out and should use this cached value in restoring the stack. To this end, remove all uses of methodReturnValue: in the support code for th ecallouts and use pop:thenPush: in ffiCall:ArgArrayOrNil:NumArgs: to update the stack correctly. Streamline the callouts in the plugin to eliminate an extra test on the return type being float. Neaten the code for (dis)owning the VM. StackInterpreter (helping resolve callback issue above): Refactor handleStackOverflow into checkForStackOverflow and handleStackOverflow to get better inlining/better readability. Fix a bug in checkStackPointerIndexForFrame: for the current frame that caused many false assert-fails during callbacks. Nuke the recently-added callback debugging machinery now that the problem is fixed. Slang: Don't output methods marked as <inline: #always> which are used to allow limited use of literal blocks in Slang. These never want to be generated as C functions. SistaRegisterAllocatingCogit: Eliminate ssFlushTo:'s in several methods. SistaCogit: Use genCallMustBeBooleanFor: in the countertrip-only jump generator. Cogit: Remobe a couple of unused var types and variables. =============== Diff against VMMaker.oscog-nice.2038 =============== Item was changed: ----- Method: CCodeGenerator>>shouldGenerateMethod: (in category 'utilities') ----- shouldGenerateMethod: aTMethod ^(self isBuiltinSelector: aTMethod selector) ifTrue: [requiredSelectors includes: aTMethod selector] + ifFalse: [aTMethod inline ~~ #always]! - ifFalse: [true]! Item was added: + ----- Method: CogAbstractInstruction>>usesTempRegForAbsoluteLoads (in category 'testing') ----- + usesTempRegForAbsoluteLoads + "Answer if TempReg is used in absolute memory loads (as it is on x64). By default answer false, allowing subclasses to override." + <inline: true> + ^false! Item was changed: ----- Method: CogObjectRepresentationForSpur>>mixed:branchIf:instanceOfBehaviors:target: (in category 'sista support') ----- mixed: numNonImmediates branchIf: reg instanceOfBehaviors: arrayObj target: targetFixUp | jmpImmediate numCases classObj index jmp | - <var: #label type: #'AbstractInstruction *'> <var: #jmpImmediate type: #'AbstractInstruction *'> - <var: #jmp type: #'AbstractInstruction *'> <var: #targetFixUp type: #'AbstractInstruction *'> + <var: #jmp type: #'AbstractInstruction *'> numCases := objectMemory numSlotsOf: arrayObj. jmpImmediate := self genJumpImmediate: reg. "Rcvr is non immediate" self genGetClassIndexOfNonImm: reg into: TempReg. index := 0. 0 to: numCases - 1 do: [:i| classObj := objectMemory fetchPointer: i ofObject: arrayObj. (objectMemory isImmediateClass: classObj) ifFalse: [ self genCmpClassIndex: (objectMemory classTagForClass: classObj) R: TempReg. cogit JumpZero: targetFixUp. index := index + 1 ] ]. jmp := cogit Jump: 0. "Rcvr is immediate" jmpImmediate jmpTarget: cogit Label. numCases - numNonImmediates "num Immediates allowed" caseOf: {[ 1 ] -> [ "1 immediate needs to jump. Find it and jump." 0 to: numCases - 1 do: [ :j | classObj := objectMemory fetchPointer: j ofObject: arrayObj. (objectMemory isImmediateClass: classObj) ifTrue: [ self branchIf: reg hasImmediateTag: (objectMemory classTagForClass: classObj) target: targetFixUp ] ] ] . [ 2 ] -> [ "2 immediates needs to jump" self branch2CasesIf: reg instanceOfBehaviors: arrayObj target: targetFixUp ] . [ 3 ] -> [ "all 3 needs to jump" self Jump: targetFixUp ] }. jmp jmpTarget: self Label. ^ 0 ! Item was changed: ----- Method: CogObjectRepresentationForSpur>>noneImmediateBranchIf:instanceOfBehaviors:target: (in category 'sista support') ----- noneImmediateBranchIf: reg instanceOfBehaviors: arrayObj target: targetFixUp "All classes in arrayObj are not immediate" | classObj jmp | <var: #targetFixUp type: #'AbstractInstruction *'> - <var: #label type: #'AbstractInstruction *'> <var: #jmp type: #'AbstractInstruction *'> jmp := self genJumpImmediate: reg. self genGetClassIndexOfNonImm: reg into: TempReg. 0 to: (objectMemory numSlotsOf: arrayObj) - 1 do: [:i| classObj := objectMemory fetchPointer: i ofObject: arrayObj. self genCmpClassIndex: (objectMemory classTagForClass: classObj) R: TempReg. cogit JumpZero: targetFixUp ]. jmp jmpTarget: self Label. ^0! Item was added: + ----- Method: CogX64Compiler>>usesTempRegForAbsoluteLoads (in category 'testing') ----- + usesTempRegForAbsoluteLoads + "Answer if TempReg is used in absolute memory loads (as it is on x64). By default answer false, allowing subclasses to override." + <inline: true> + ^true! Item was changed: ----- Method: RegisterAllocatingCogit>>genSpecialSelectorComparison (in category 'bytecode generators') ----- genSpecialSelectorComparison | nextPC postBranchPC targetBytecodePC primDescriptor branchDescriptor + rcvrIsInt argIsInt argInt jumpNotSmallInts inlineCAB index rcvrReg argReg | - rcvrIsInt argIsInt argInt jumpNotSmallInts inlineCAB index rcvrReg argReg regMask | <var: #primDescriptor type: #'BytecodeDescriptor *'> <var: #branchDescriptor type: #'BytecodeDescriptor *'> <var: #jumpNotSmallInts type: #'AbstractInstruction *'> primDescriptor := self generatorAt: byte0. argIsInt := self ssTop type = SSConstant and: [objectMemory isIntegerObject: (argInt := self ssTop constant)]. rcvrIsInt := (self ssValue: 1) type = SSConstant and: [objectMemory isIntegerObject: (self ssValue: 1) constant]. (argIsInt and: [rcvrIsInt]) ifTrue: [^ self genStaticallyResolvedSpecialSelectorComparison]. self extractMaybeBranchDescriptorInto: [ :descr :next :postBranch :target | branchDescriptor := descr. nextPC := next. postBranchPC := postBranch. targetBytecodePC := target ]. "Only interested in inlining if followed by a conditional branch." inlineCAB := branchDescriptor isBranchTrue or: [branchDescriptor isBranchFalse]. "Further, only interested in inlining = and ~= if there's a SmallInteger constant involved. The relational operators successfully statically predict SmallIntegers; the equality operators do not." (inlineCAB and: [primDescriptor opcode = JumpZero or: [primDescriptor opcode = JumpNonZero]]) ifTrue: [inlineCAB := argIsInt or: [rcvrIsInt]]. inlineCAB ifFalse: [^self genSpecialSelectorSend]. "In-line the comparison and the jump, but if the types are not SmallInteger then we will need to do a send and fall through to the following conditional branch. Since we're allocating values in registers we would like to keep those registers live on the inlined path and reload registers along the non-inlined send path. The merge logic at the branch destinations handles this." argIsInt ifTrue: [rcvrReg := self allocateRegForStackEntryAt: 1. (self ssValue: 1) popToReg: rcvrReg. + self MoveR: rcvrReg R: TempReg] - self MoveR: rcvrReg R: TempReg. - regMask := self registerMaskFor: rcvrReg] ifFalse: [self allocateRegForStackTopTwoEntriesInto: [:rTop :rNext| argReg := rTop. rcvrReg := rNext]. rcvrReg = Arg0Reg ifTrue: [rcvrReg := argReg. argReg := Arg0Reg]. self ssTop popToReg: argReg. (self ssValue: 1) popToReg: rcvrReg. + self MoveR: argReg R: TempReg]. - self MoveR: argReg R: TempReg. - regMask := self registerMaskFor: rcvrReg and: argReg]. self ssPop: 2. jumpNotSmallInts := (argIsInt or: [rcvrIsInt]) ifTrue: [objectRepresentation genJumpNotSmallIntegerInScratchReg: TempReg] ifFalse: [objectRepresentation genJumpNotSmallIntegersIn: rcvrReg andScratch: TempReg scratch: ClassReg]. argIsInt ifTrue: [self CmpCq: argInt R: rcvrReg] ifFalse: [self CmpR: argReg R: rcvrReg]. "Cmp is weird/backwards so invert the comparison. Further since there is a following conditional jump bytecode define non-merge fixups and leave the cond bytecode to set the mergeness." self genConditionalBranch: (branchDescriptor isBranchTrue ifTrue: [primDescriptor opcode] ifFalse: [self inverseBranchFor: primDescriptor opcode]) operand: (self ensureFixupAt: targetBytecodePC - initialPC) asUnsignedInteger. self Jump: (self ensureFixupAt: postBranchPC - initialPC). jumpNotSmallInts jmpTarget: self Label. self ssFlushTo: simStackPtr. self deny: rcvrReg = Arg0Reg. argIsInt ifTrue: [self MoveCq: argInt R: Arg0Reg] ifFalse: [argReg ~= Arg0Reg ifTrue: [self MoveR: argReg R: Arg0Reg]]. rcvrReg ~= ReceiverResultReg ifTrue: [self MoveR: rcvrReg R: ReceiverResultReg]. index := byte0 - self firstSpecialSelectorBytecodeOffset. + ^self genMarshalledSend: index negated - 1 numArgs: 1 sendTable: ordinarySendTrampolines! - self genMarshalledSend: index negated - 1 numArgs: 1 sendTable: ordinarySendTrampolines. - ^0! Item was changed: ----- Method: SistaCogit>>genCounterTripOnlyJumpIf:to: (in category 'bytecode generator support') ----- genCounterTripOnlyJumpIf: boolean to: targetBytecodePC "Specific version if the branch is only reached while falling through if the counter trips after an inlined #== branch. We do not regenerate the counter logic in this case to avoid 24 bytes instructions." <var: #ok type: #'AbstractInstruction *'> <var: #mustBeBooleanTrampoline type: #'AbstractInstruction *'> | ok mustBeBooleanTrampoline | extA := 0. self ssFlushTo: simStackPtr - 1. self ssTop popToReg: TempReg. self ssPop: 1. counterIndex := counterIndex + 1. "counters are increased / decreased in the inlined branch" "We need SendNumArgsReg because of the mustBeBooleanTrampoline" self ssAllocateRequiredReg: SendNumArgsReg. self MoveCq: 1 R: SendNumArgsReg. "The first time this is reached, it calls necessarily the counter trip for the trampoline because SendNumArgsReg is non zero" + mustBeBooleanTrampoline := self genCallMustBeBooleanFor: boolean. - mustBeBooleanTrampoline := self CallRT: (boolean == objectMemory falseObject - ifTrue: [ceSendMustBeBooleanAddFalseTrampoline] - ifFalse: [ceSendMustBeBooleanAddTrueTrampoline]). - self annotateBytecode: self Label. "Cunning trick by LPD. If true and false are contiguous subtract the smaller. Correct result is either 0 or the distance between them. If result is not 0 or their distance send mustBeBoolean." self assert: (objectMemory objectAfter: objectMemory falseObject) = objectMemory trueObject. self genSubConstant: boolean R: TempReg. self JumpZero: (self ensureFixupAt: targetBytecodePC - initialPC). + self CmpCq: (boolean = objectMemory falseObject - self CmpCq: (boolean == objectMemory falseObject ifTrue: [objectMemory trueObject - objectMemory falseObject] ifFalse: [objectMemory falseObject - objectMemory trueObject]) R: TempReg. ok := self JumpZero: 0. self MoveCq: 0 R: SendNumArgsReg. "if counterReg is 0 this is a mustBeBoolean, not a counter trip." self Jump: mustBeBooleanTrampoline. ok jmpTarget: self Label. ^0! Item was changed: ----- Method: SistaCogit>>genJumpIf:to: (in category 'bytecode generator support') ----- genJumpIf: boolean to: targetBytecodePC "The heart of performance counting in Sista. Conditional branches are 6 times less frequent than sends and can provide basic block frequencies (send counters can't). Each conditional has a 32-bit counter split into an upper 16 bits counting executions and a lower half counting untaken executions of the branch. Executing the branch decrements the upper half, tripping if the count goes negative. Not taking the branch decrements the lower half. N.B. We *do not* eliminate dead branches (true ifTrue:/true ifFalse:) so that scanning for send and branch data is simplified and that branch data is correct." <inline: false> | ok counterAddress countTripped retry nextPC nextDescriptor desc | <var: #ok type: #'AbstractInstruction *'> <var: #desc type: #'CogSimStackEntry *'> <var: #retry type: #'AbstractInstruction *'> <var: #countTripped type: #'AbstractInstruction *'> <var: #nextDescriptor type: #'BytecodeDescriptor *'> "In optimized code we don't generate counters to improve performance" (coInterpreter isOptimizedMethod: methodObj) ifTrue: [ ^ super genJumpIf: boolean to: targetBytecodePC ]. "If the branch is reached only for the counter trip trampoline (typically, var1 == var2 ifTrue: falls through to the branch only for the trampoline) we generate a specific path to drastically reduce the number of machine instructions" branchReachedOnlyForCounterTrip ifTrue: [ branchReachedOnlyForCounterTrip := false. ^ self genCounterTripOnlyJumpIf: boolean to: targetBytecodePC ]. "We detect and: / or:, if found, we don't generate the counters to avoid pathological counter slow down" + boolean = objectMemory falseObject ifTrue: - boolean == objectMemory falseObject ifTrue: [ nextPC := bytecodePC + (self generatorAt: byte0) numBytes. nextDescriptor := self generatorAt: (objectMemory fetchByte: nextPC ofObject: methodObj) + bytecodeSetOffset. nextDescriptor generator == #genPushConstantTrueBytecode ifTrue: [ ^ super genJumpIf: boolean to: targetBytecodePC ]. nextDescriptor := self generatorAt: (objectMemory fetchByte: targetBytecodePC ofObject: methodObj) + bytecodeSetOffset. + nextDescriptor generator == #genPushConstantFalseBytecode ifTrue: [ ^ super genJumpIf: boolean to: targetBytecodePC ]. ]. - nextDescriptor generator == #genPushConstantFalseBytecode ifTrue: [ ^ super genJumpIf: boolean to: targetBytecodePC ]. ]. extA := 0. "We ignore the noMustBeBoolean flag. It should not be present in methods with counters, and if it is we don't care." "We don't generate counters on branches on true/false, the basicblock usage can be inferred" desc := self ssTop. (desc type == SSConstant and: [desc constant = objectMemory trueObject or: [desc constant = objectMemory falseObject]]) ifTrue: [ ^ super genJumpIf: boolean to: targetBytecodePC ]. self ssFlushTo: simStackPtr - 1. desc popToReg: TempReg. self ssPop: 1. "We need SendNumArgsReg because of the mustBeBooleanTrampoline" self ssAllocateRequiredReg: SendNumArgsReg. retry := self Label. self genExecutionCountLogicInto: [ :cAddress :countTripBranch | counterAddress := cAddress. countTripped := countTripBranch ] counterReg: SendNumArgsReg. counterIndex := counterIndex + 1. "Cunning trick by LPD. If true and false are contiguous subtract the smaller. Correct result is either 0 or the distance between them. If result is not 0 or their distance send mustBeBoolean." self assert: (objectMemory objectAfter: objectMemory falseObject) = objectMemory trueObject. self genSubConstant: boolean R: TempReg. self JumpZero: (self ensureFixupAt: targetBytecodePC - initialPC). self genFallsThroughCountLogicCounterReg: SendNumArgsReg counterAddress: counterAddress. self CmpCq: (boolean = objectMemory falseObject ifTrue: [objectMemory trueObject - objectMemory falseObject] ifFalse: [objectMemory falseObject - objectMemory trueObject]) R: TempReg. ok := self JumpZero: 0. self MoveCq: 0 R: SendNumArgsReg. "if counterReg is 0 this is a mustBeBoolean, not a counter trip." countTripped jmpTarget: (self genCallMustBeBooleanFor: boolean). "If we're in an image which hasn't got the Sista code loaded then the ceCounterTripped: trampoline will return directly to machine code, returning the boolean. So the code should jump back to the retry point. The trampoline makes sure that TempReg has been reloaded." + + "Clément: For some reason if I write self annotateBytecode: (self Jump: retry) the annotation is not at the correct place." + "Eliot: Annotations apply the the address following an instruction, and the annotation must be for the return address + of the call (since this is the address the run-time sees), so it must be on a label before the jump, not after the jump." + self annotateBytecode: self Label. - self annotateBytecode: self Label. "For some reason if I write self annotateBytecode: (self Jump: retry) the annotation is not at the correct place." self Jump: retry. ok jmpTarget: self Label. ^0! Item was removed: - ----- Method: SistaCogitClone>>genCounterTripOnlyJumpIf:to: (in category 'bytecode generator support') ----- - genCounterTripOnlyJumpIf: boolean to: targetBytecodePC - "Specific version if the branch is only reached while falling through if the counter trips after an inlined #== branch. We do not regenerate the counter logic in this case to avoid 24 bytes instructions." - - <var: #ok type: #'AbstractInstruction *'> - <var: #mustBeBooleanTrampoline type: #'AbstractInstruction *'> - - | ok mustBeBooleanTrampoline | - - extA := 0. - - self ssFlushTo: simStackPtr - 1. - - self ssTop popToReg: TempReg. - - self ssPop: 1. - - counterIndex := counterIndex + 1. "counters are increased / decreased in the inlined branch" - - "We need SendNumArgsReg because of the mustBeBooleanTrampoline" - self ssAllocateRequiredReg: SendNumArgsReg. - self MoveCq: 1 R: SendNumArgsReg. - - "The first time this is reached, it calls necessarily the counter trip for the trampoline because SendNumArgsReg is non zero" - mustBeBooleanTrampoline := self CallRT: (boolean == objectMemory falseObject - ifTrue: [ceSendMustBeBooleanAddFalseTrampoline] - ifFalse: [ceSendMustBeBooleanAddTrueTrampoline]). - - self annotateBytecode: self Label. - - "Cunning trick by LPD. If true and false are contiguous subtract the smaller. - Correct result is either 0 or the distance between them. If result is not 0 or - their distance send mustBeBoolean." - self assert: (objectMemory objectAfter: objectMemory falseObject) = objectMemory trueObject. - self genSubConstant: boolean R: TempReg. - self JumpZero: (self ensureFixupAt: targetBytecodePC - initialPC). - - self CmpCq: (boolean == objectMemory falseObject - ifTrue: [objectMemory trueObject - objectMemory falseObject] - ifFalse: [objectMemory falseObject - objectMemory trueObject]) - R: TempReg. - - ok := self JumpZero: 0. - self MoveCq: 0 R: SendNumArgsReg. "if counterReg is 0 this is a mustBeBoolean, not a counter trip." - - self Jump: mustBeBooleanTrampoline. - - ok jmpTarget: self Label. - ^0! Item was removed: - ----- Method: SistaCogitClone>>genJumpIf:to: (in category 'bytecode generator support') ----- - genJumpIf: boolean to: targetBytecodePC - "The heart of performance counting in Sista. Conditional branches are 6 times less - frequent than sends and can provide basic block frequencies (send counters can't). - Each conditional has a 32-bit counter split into an upper 16 bits counting executions - and a lower half counting untaken executions of the branch. Executing the branch - decrements the upper half, tripping if the count goes negative. Not taking the branch - decrements the lower half. N.B. We *do not* eliminate dead branches (true ifTrue:/true ifFalse:) - so that scanning for send and branch data is simplified and that branch data is correct." - <inline: false> - | ok counterAddress countTripped retry nextPC nextDescriptor desc | - <var: #ok type: #'AbstractInstruction *'> - <var: #desc type: #'CogSimStackEntry *'> - <var: #retry type: #'AbstractInstruction *'> - <var: #countTripped type: #'AbstractInstruction *'> - <var: #nextDescriptor type: #'BytecodeDescriptor *'> - - "In optimized code we don't generate counters to improve performance" - (coInterpreter isOptimizedMethod: methodObj) ifTrue: [ ^ super genJumpIf: boolean to: targetBytecodePC ]. - - "If the branch is reached only for the counter trip trampoline - (typically, var1 == var2 ifTrue: falls through to the branch only for the trampoline) - we generate a specific path to drastically reduce the number of machine instructions" - branchReachedOnlyForCounterTrip ifTrue: - [ branchReachedOnlyForCounterTrip := false. - ^ self genCounterTripOnlyJumpIf: boolean to: targetBytecodePC ]. - - "We detect and: / or:, if found, we don't generate the counters to avoid pathological counter slow down" - boolean == objectMemory falseObject ifTrue: - [ nextPC := bytecodePC + (self generatorAt: byte0) numBytes. - nextDescriptor := self generatorAt: (objectMemory fetchByte: nextPC ofObject: methodObj) + bytecodeSetOffset. - nextDescriptor generator == #genPushConstantTrueBytecode ifTrue: [ ^ super genJumpIf: boolean to: targetBytecodePC ]. - nextDescriptor := self generatorAt: (objectMemory fetchByte: targetBytecodePC ofObject: methodObj) + bytecodeSetOffset. - nextDescriptor generator == #genPushConstantFalseBytecode ifTrue: [ ^ super genJumpIf: boolean to: targetBytecodePC ]. ]. - - extA := 0. "We ignore the noMustBeBoolean flag. It should not be present in methods with counters, and if it is we don't care." - - "We don't generate counters on branches on true/false, the basicblock usage can be inferred" - desc := self ssTop. - (desc type == SSConstant - and: [desc constant = objectMemory trueObject or: [desc constant = objectMemory falseObject]]) ifTrue: - [ ^ super genJumpIf: boolean to: targetBytecodePC ]. - - self ssFlushTo: simStackPtr - 1. - desc popToReg: TempReg. - self ssPop: 1. - - "We need SendNumArgsReg because of the mustBeBooleanTrampoline" - self ssAllocateRequiredReg: SendNumArgsReg. - - retry := self Label. - self - genExecutionCountLogicInto: [ :cAddress :countTripBranch | - counterAddress := cAddress. - countTripped := countTripBranch ] - counterReg: SendNumArgsReg. - counterIndex := counterIndex + 1. - - "Cunning trick by LPD. If true and false are contiguous subtract the smaller. - Correct result is either 0 or the distance between them. If result is not 0 or - their distance send mustBeBoolean." - self assert: (objectMemory objectAfter: objectMemory falseObject) = objectMemory trueObject. - self genSubConstant: boolean R: TempReg. - self JumpZero: (self ensureFixupAt: targetBytecodePC - initialPC). - - self genFallsThroughCountLogicCounterReg: SendNumArgsReg counterAddress: counterAddress. - - self CmpCq: (boolean = objectMemory falseObject - ifTrue: [objectMemory trueObject - objectMemory falseObject] - ifFalse: [objectMemory falseObject - objectMemory trueObject]) - R: TempReg. - ok := self JumpZero: 0. - self MoveCq: 0 R: SendNumArgsReg. "if counterReg is 0 this is a mustBeBoolean, not a counter trip." - - countTripped jmpTarget: (self genCallMustBeBooleanFor: boolean). - - "If we're in an image which hasn't got the Sista code loaded then the ceCounterTripped: - trampoline will return directly to machine code, returning the boolean. So the code should - jump back to the retry point. The trampoline makes sure that TempReg has been reloaded." - self annotateBytecode: self Label. "For some reason if I write self annotateBytecode: (self Jump: retry) the annotation is not at the correct place." - self Jump: retry. - - ok jmpTarget: self Label. - ^0! Item was removed: - ----- Method: SistaCogitClone>>genSpecialSelectorComparison (in category 'bytecode generators') ----- - genSpecialSelectorComparison - "Override to count inlined branches if followed by a conditional branch. - We borrow the following conditional branch's counter and when about to - inline the comparison we decrement the counter (without writing it back) - and if it trips simply abort the inlining, falling back to the normal send which - will then continue to the conditional branch which will trip and enter the abort." - | nextPC postBranchPC targetBytecodePC primDescriptor branchDescriptor - rcvrIsInt argIsInt argInt jumpNotSmallInts inlineCAB - counterAddress countTripped counterReg index | - <var: #countTripped type: #'AbstractInstruction *'> - <var: #primDescriptor type: #'BytecodeDescriptor *'> - <var: #jumpNotSmallInts type: #'AbstractInstruction *'> - <var: #branchDescriptor type: #'BytecodeDescriptor *'> - - (coInterpreter isOptimizedMethod: methodObj) ifTrue: [ ^ self genSpecialSelectorComparisonWithoutCounters ]. - - self ssFlushTo: simStackPtr - 2. - primDescriptor := self generatorAt: byte0. - argIsInt := self ssTop type = SSConstant - and: [objectMemory isIntegerObject: (argInt := self ssTop constant)]. - rcvrIsInt := (self ssValue: 1) type = SSConstant - and: [objectMemory isIntegerObject: (self ssValue: 1) constant]. - - "short-cut the jump if operands are SmallInteger constants." - (argIsInt and: [rcvrIsInt]) ifTrue: - [^ self genStaticallyResolvedSpecialSelectorComparison]. - - self extractMaybeBranchDescriptorInto: [ :descr :next :postBranch :target | - branchDescriptor := descr. nextPC := next. postBranchPC := postBranch. targetBytecodePC := target ]. - - "Only interested in inlining if followed by a conditional branch." - inlineCAB := branchDescriptor isBranchTrue or: [branchDescriptor isBranchFalse]. - "Further, only interested in inlining = and ~= if there's a SmallInteger constant involved. - The relational operators successfully statically predict SmallIntegers; the equality operators do not." - (inlineCAB and: [primDescriptor opcode = JumpZero or: [primDescriptor opcode = JumpNonZero]]) ifTrue: - [inlineCAB := argIsInt or: [rcvrIsInt]]. - inlineCAB ifFalse: - [^self genSpecialSelectorSend]. - - argIsInt - ifTrue: - [(self ssValue: 1) popToReg: ReceiverResultReg. - self ssPop: 2. - self MoveR: ReceiverResultReg R: TempReg] - ifFalse: - [self marshallSendArguments: 1. - self MoveR: Arg0Reg R: TempReg]. - jumpNotSmallInts := (argIsInt or: [rcvrIsInt]) - ifTrue: [objectRepresentation genJumpNotSmallIntegerInScratchReg: TempReg] - ifFalse: [objectRepresentation genJumpNotSmallIntegersIn: ReceiverResultReg andScratch: TempReg scratch: ClassReg]. - - counterReg := self allocateRegNotConflictingWith: (self registerMaskFor: ReceiverResultReg and: Arg0Reg). - self - genExecutionCountLogicInto: [ :cAddress :countTripBranch | - counterAddress := cAddress. - countTripped := countTripBranch ] - counterReg: counterReg. - - argIsInt - ifTrue: [self CmpCq: argInt R: ReceiverResultReg] - ifFalse: [self CmpR: Arg0Reg R: ReceiverResultReg]. - "Cmp is weird/backwards so invert the comparison. Further since there is a following conditional - jump bytecode define non-merge fixups and leave the cond bytecode to set the mergeness." - self genConditionalBranch: (branchDescriptor isBranchTrue - ifTrue: [primDescriptor opcode] - ifFalse: [self inverseBranchFor: primDescriptor opcode]) - operand: (self ensureNonMergeFixupAt: targetBytecodePC - initialPC) asUnsignedInteger. - - self genFallsThroughCountLogicCounterReg: counterReg counterAddress: counterAddress. - - self Jump: (self ensureNonMergeFixupAt: postBranchPC - initialPC). - countTripped jmpTarget: (jumpNotSmallInts jmpTarget: self Label). - - argIsInt ifTrue: - [self MoveCq: argInt R: Arg0Reg]. - index := byte0 - self firstSpecialSelectorBytecodeOffset. - ^self genMarshalledSend: index negated - 1 numArgs: 1 sendTable: ordinarySendTrampolines! Item was added: + ----- Method: SistaRegisterAllocatingCogit class>>implementation:isNewerThan: (in category 'clone maintennance') ----- + implementation: aMethodReference isNewerThan: bMethodReference + ^(self dateAndTimeFrom: aMethodReference timeStamp) > (self dateAndTimeFrom: bMethodReference timeStamp)! Item was added: + ----- Method: SistaRegisterAllocatingCogit>>genCounterTripOnlyJumpIf:to: (in category 'bytecode generator support') ----- + genCounterTripOnlyJumpIf: boolean to: targetBytecodePC + "Specific version if the branch is only reached while falling through if the counter trips after an inlined #== branch. We do not regenerate the counter logic in this case to avoid 24 bytes instructions." + + <var: #ok type: #'AbstractInstruction *'> + <var: #mustBeBooleanTrampoline type: #'AbstractInstruction *'> + + | ok mustBeBooleanTrampoline | + + extA := 0. + + self ssTop popToReg: TempReg. + self ssPop: 1. + + counterIndex := counterIndex + 1. "counters are increased / decreased in the inlined branch" + + "We need SendNumArgsReg because of the mustBeBooleanTrampoline" + self ssAllocateRequiredReg: SendNumArgsReg. + self MoveCq: 1 R: SendNumArgsReg. + + "The first time this is reached, it calls necessarily the counter trip for the trampoline because SendNumArgsReg is non zero" + mustBeBooleanTrampoline := self genCallMustBeBooleanFor: boolean. + + self annotateBytecode: self Label. + + "Cunning trick by LPD. If true and false are contiguous subtract the smaller. + Correct result is either 0 or the distance between them. If result is not 0 or + their distance send mustBeBoolean." + self assert: (objectMemory objectAfter: objectMemory falseObject) = objectMemory trueObject. + self genSubConstant: boolean R: TempReg. + self JumpZero: (self ensureFixupAt: targetBytecodePC - initialPC). + + self CmpCq: (boolean = objectMemory falseObject + ifTrue: [objectMemory trueObject - objectMemory falseObject] + ifFalse: [objectMemory falseObject - objectMemory trueObject]) + R: TempReg. + + ok := self JumpZero: 0. + self MoveCq: 0 R: SendNumArgsReg. "if counterReg is 0 this is a mustBeBoolean, not a counter trip." + + self Jump: mustBeBooleanTrampoline. + + ok jmpTarget: self Label. + ^0! Item was added: + ----- Method: SistaRegisterAllocatingCogit>>genForwardersInlinedIdenticalOrNotIf: (in category 'bytecode generators') ----- + genForwardersInlinedIdenticalOrNotIf: orNot + "Override to count inlined branches if followed by a conditional branch. + We borrow the following conditional branch's counter and when about to + inline the comparison we decrement the counter (without writing it back) + and if it trips simply abort the inlining, falling back to the normal send which + will then continue to the conditional branch which will trip and enter the abort." + | nextPC postBranchPC targetBytecodePC branchDescriptor counterReg fixup jumpEqual jumpNotEqual + counterAddress countTripped unforwardArg unforwardRcvr argReg rcvrReg regMask | + <var: #fixup type: #'BytecodeFixup *'> + <var: #countTripped type: #'AbstractInstruction *'> + <var: #label type: #'AbstractInstruction *'> + <var: #branchDescriptor type: #'BytecodeDescriptor *'> + <var: #jumpEqual type: #'AbstractInstruction *'> + <var: #jumpNotEqual type: #'AbstractInstruction *'> + + ((coInterpreter isOptimizedMethod: methodObj) or: [needsFrame not]) ifTrue: + [^super genForwardersInlinedIdenticalOrNotIf: orNot]. + + regMask := 0. + + self extractMaybeBranchDescriptorInto: [ :descr :next :postBranch :target | + branchDescriptor := descr. nextPC := next. postBranchPC := postBranch. targetBytecodePC := target ]. + + unforwardRcvr := (objectRepresentation isUnannotatableConstant: (self ssValue: 1)) not. + unforwardArg := (objectRepresentation isUnannotatableConstant: self ssTop) not. + + "If an operand is an annotable constant, it may be forwarded, so we need to store it into a + register so the forwarder check can jump back to the comparison after unforwarding the constant. + However, if one of the operand is an unnanotable constant, does not allocate a register for it + (machine code will use operations on constants)." + rcvrReg:= argReg := NoReg. + self + allocateEqualsEqualsRegistersArgNeedsReg: unforwardArg + rcvrNeedsReg: unforwardRcvr + into: [ :rcvr :arg | rcvrReg:= rcvr. argReg := arg ]. + + argReg ~= NoReg ifTrue: [ regMask := self registerMaskFor: argReg ]. + rcvrReg ~= NoReg ifTrue: [ regMask := regMask bitOr: (self registerMaskFor: rcvrReg) ]. + + "Only interested in inlining if followed by a conditional branch." + (branchDescriptor isBranchTrue or: [branchDescriptor isBranchFalse]) ifFalse: + [^ self + genIdenticalNoBranchArgIsConstant: unforwardArg not + rcvrIsConstant: unforwardRcvr not + argReg: argReg + rcvrReg: rcvrReg + orNotIf: orNot]. + + unforwardArg ifTrue: [ objectRepresentation genEnsureOopInRegNotForwarded: argReg scratchReg: TempReg ]. + unforwardRcvr ifTrue: [ objectRepresentation genEnsureOopInRegNotForwarded: rcvrReg scratchReg: TempReg ]. + + counterReg := self allocateRegNotConflictingWith: regMask. + self + genExecutionCountLogicInto: [ :cAddress :countTripBranch | + counterAddress := cAddress. + countTripped := countTripBranch ] + counterReg: counterReg. + + self assert: (unforwardArg or: [ unforwardRcvr ]). + self genCmpArgIsConstant: unforwardArg not rcvrIsConstant: unforwardRcvr not argReg: argReg rcvrReg: rcvrReg. + self ssPop: 2. + + "We could use (branchDescriptor isBranchTrue xor: orNot) to simplify this." + orNot + ifFalse: [branchDescriptor isBranchTrue + ifTrue: + [ fixup := (self ensureNonMergeFixupAt: postBranchPC - initialPC) asUnsignedInteger. + self JumpZero: (self ensureNonMergeFixupAt: targetBytecodePC - initialPC) asUnsignedInteger ] + ifFalse: "branchDescriptor is branchFalse" + [ fixup := (self ensureNonMergeFixupAt: targetBytecodePC - initialPC) asUnsignedInteger. + self JumpZero: (self ensureNonMergeFixupAt: postBranchPC - initialPC) asUnsignedInteger ]] + ifTrue: [branchDescriptor isBranchTrue + ifFalse: "branchDescriptor is branchFalse" + [ fixup := (self ensureNonMergeFixupAt: postBranchPC - initialPC) asUnsignedInteger. + self JumpZero: (self ensureNonMergeFixupAt: targetBytecodePC - initialPC) asUnsignedInteger ] + ifTrue: + [ fixup := (self ensureNonMergeFixupAt: targetBytecodePC - initialPC) asUnsignedInteger. + self JumpZero: (self ensureNonMergeFixupAt: postBranchPC - initialPC) asUnsignedInteger ]]. + + self genFallsThroughCountLogicCounterReg: counterReg counterAddress: counterAddress. + self Jump: fixup. + + countTripped jmpTarget: self Label. + + "inlined version of #== ignoring the branchDescriptor if the counter trips to have normal state for the optimizer" + self ssPop: -2. + self genCmpArgIsConstant: unforwardArg not rcvrIsConstant: unforwardRcvr not argReg: argReg rcvrReg: rcvrReg. + self ssPop: 2. + + "This code necessarily directly falls through the jumpIf: code which pops the top of the stack into TempReg. + We therefore directly assign the result to TempReg to save one move instruction" + jumpEqual := orNot ifFalse: [self JumpZero: 0] ifTrue: [self JumpNonZero: 0]. + self genMoveFalseR: TempReg. + jumpNotEqual := self Jump: 0. + jumpEqual jmpTarget: (self genMoveTrueR: TempReg). + jumpNotEqual jmpTarget: self Label. + self ssPushRegister: TempReg. + + (self fixupAt: nextPC - initialPC) notAFixup ifTrue: [ branchReachedOnlyForCounterTrip := true ]. + + ^ 0! Item was added: + ----- Method: SistaRegisterAllocatingCogit>>genJumpIf:to: (in category 'bytecode generator support') ----- + genJumpIf: boolean to: targetBytecodePC + "The heart of performance counting in Sista. Conditional branches are 6 times less + frequent than sends and can provide basic block frequencies (send counters can't). + Each conditional has a 32-bit counter split into an upper 16 bits counting executions + and a lower half counting untaken executions of the branch. Executing the branch + decrements the upper half, tripping if the count goes negative. Not taking the branch + decrements the lower half. N.B. We *do not* eliminate dead branches (true ifTrue:/true ifFalse:) + so that scanning for send and branch data is simplified and that branch data is correct." + <inline: false> + | ok counterAddress countTripped retry nextPC nextDescriptor desc reg | + <var: #ok type: #'AbstractInstruction *'> + <var: #desc type: #'CogSimStackEntry *'> + <var: #retry type: #'AbstractInstruction *'> + <var: #countTripped type: #'AbstractInstruction *'> + <var: #nextDescriptor type: #'BytecodeDescriptor *'> + + "In optimized code we don't generate counters to improve performance" + (coInterpreter isOptimizedMethod: methodObj) ifTrue: + [^super genJumpIf: boolean to: targetBytecodePC]. + + "If the branch is reached only for the counter trip trampoline + (typically, var1 == var2 ifTrue: falls through to the branch only for the trampoline) + we generate a specific path to drastically reduce the number of machine instructions" + branchReachedOnlyForCounterTrip ifTrue: + [branchReachedOnlyForCounterTrip := false. + ^self genCounterTripOnlyJumpIf: boolean to: targetBytecodePC]. + + "We detect and: / or:, if found, we don't generate the counters to avoid pathological counter slow down" + boolean = objectMemory falseObject ifTrue: + [ nextPC := bytecodePC + (self generatorAt: byte0) numBytes. + nextDescriptor := self generatorAt: (objectMemory fetchByte: nextPC ofObject: methodObj) + bytecodeSetOffset. + nextDescriptor generator == #genPushConstantTrueBytecode ifTrue: [ ^ super genJumpIf: boolean to: targetBytecodePC ]. + nextDescriptor := self generatorAt: (objectMemory fetchByte: targetBytecodePC ofObject: methodObj) + bytecodeSetOffset. + nextDescriptor generator == #genPushConstantFalseBytecode ifTrue: [ ^ super genJumpIf: boolean to: targetBytecodePC ]. ]. + + extA := 0. "We ignore the noMustBeBoolean flag. It should not be present in methods with counters, and if it is we don't care." + + "We don't generate counters on branches on true/false, the basicblock usage can be inferred" + desc := self ssTop. + (desc type == SSConstant + and: [desc constant = objectMemory trueObject or: [desc constant = objectMemory falseObject]]) ifTrue: + [ ^ super genJumpIf: boolean to: targetBytecodePC ]. + + self flag: 'Because of the restriction on x64 that absolute loads must target %rax, it would perhaps be a better choice to use TempReg (%rax) for the counter reg and SendNumArgsReg for the boolean.'. + "try and use the top entry's register if ant, but only if it can be destroyed." + reg := (desc type ~= SSRegister + or: [(self anyReferencesToRegister: desc register inAllButTopNItems: 0) + or: [(desc register = ReceiverResultReg and: [optStatus isReceiverResultRegLive])]]) + ifTrue: [TempReg] + ifFalse: [desc register]. + desc popToReg: reg. + self ssPop: 1. + + "We need SendNumArgsReg because of the mustBeBooleanTrampoline" + self ssAllocateRequiredReg: SendNumArgsReg. + + retry := self Label. + self + genExecutionCountLogicInto: [ :cAddress :countTripBranch | + counterAddress := cAddress. + countTripped := countTripBranch ] + counterReg: SendNumArgsReg. + counterIndex := counterIndex + 1. + + "Cunning trick by LPD. If true and false are contiguous subtract the smaller. + Correct result is either 0 or the distance between them. If result is not 0 or + their distance send mustBeBoolean." + self assert: (objectMemory objectAfter: objectMemory falseObject) = objectMemory trueObject. + self genSubConstant: boolean R: reg. + self JumpZero: (self ensureFixupAt: targetBytecodePC - initialPC). + + self genFallsThroughCountLogicCounterReg: SendNumArgsReg counterAddress: counterAddress. + + self CmpCq: (boolean = objectMemory falseObject + ifTrue: [objectMemory trueObject - objectMemory falseObject] + ifFalse: [objectMemory falseObject - objectMemory trueObject]) + R: reg. + ok := self JumpZero: 0. + self MoveCq: 0 R: SendNumArgsReg. "if counterReg is 0 this is a mustBeBoolean, not a counter trip." + reg ~= TempReg ifTrue: + [self MoveR: reg R: TempReg]. + countTripped jmpTarget: self Label. + self copySimStackToScratch: simSpillBase. + self ssFlushTo: simStackPtr. + self genCallMustBeBooleanFor: boolean. + + "If we're in an image which hasn't got the Sista code loaded then the ceCounterTripped: trampoline + will return directly to machine code, returning the boolean. So the code should jump back to the + retry point. The trampoline preserves register state when taking the ceCounterTripped: path." + "Clément: For some reason if I write self annotateBytecode: (self Jump: retry) the annotation is not at the correct place." + "Eliot: Annotations apply the the address following an instruction, and the annotation must be for the return address + of the call (since this is the address the run-time sees), so it must be on a label before the jump, not after the jump." + self annotateBytecode: self Label. + simSpillBase ~= scratchSpillBase ifTrue: + [self assert: simSpillBase > scratchSpillBase. + self AddCq: simSpillBase - scratchSpillBase * objectMemory wordSize R: SPReg]. + self Jump: retry. + + ok jmpTarget: self Label. + self restoreSimStackFromScratch. + ^0! Item was added: + ----- Method: SistaRegisterAllocatingCogit>>genSpecialSelectorComparison (in category 'bytecode generators') ----- + genSpecialSelectorComparison + "Override to count inlined branches if followed by a conditional branch. + We borrow the following conditional branch's counter and when about to + inline the comparison we decrement the counter (without writing it back) + and if it trips simply abort the inlining, falling back to the normal send which + will then continue to the conditional branch which will trip and enter the abort." + | nextPC postBranchPC targetBytecodePC primDescriptor branchDescriptor + rcvrIsInt argIsInt argInt jumpNotSmallInts inlineCAB + counterAddress countTripped counterReg index rcvrReg argReg | + <var: #countTripped type: #'AbstractInstruction *'> + <var: #primDescriptor type: #'BytecodeDescriptor *'> + <var: #jumpNotSmallInts type: #'AbstractInstruction *'> + <var: #branchDescriptor type: #'BytecodeDescriptor *'> + + (coInterpreter isOptimizedMethod: methodObj) ifTrue: + [^self genSpecialSelectorComparisonWithoutCounters]. + + primDescriptor := self generatorAt: byte0. + argIsInt := self ssTop type = SSConstant + and: [objectMemory isIntegerObject: (argInt := self ssTop constant)]. + rcvrIsInt := (self ssValue: 1) type = SSConstant + and: [objectMemory isIntegerObject: (self ssValue: 1) constant]. + + "short-cut the jump if operands are SmallInteger constants." + (argIsInt and: [rcvrIsInt]) ifTrue: + [^ self genStaticallyResolvedSpecialSelectorComparison]. + + self extractMaybeBranchDescriptorInto: [ :descr :next :postBranch :target | + branchDescriptor := descr. nextPC := next. postBranchPC := postBranch. targetBytecodePC := target ]. + + "Only interested in inlining if followed by a conditional branch." + inlineCAB := branchDescriptor isBranchTrue or: [branchDescriptor isBranchFalse]. + "Further, only interested in inlining = and ~= if there's a SmallInteger constant involved. + The relational operators successfully statically predict SmallIntegers; the equality operators do not." + (inlineCAB and: [primDescriptor opcode = JumpZero or: [primDescriptor opcode = JumpNonZero]]) ifTrue: + [inlineCAB := argIsInt or: [rcvrIsInt]]. + inlineCAB ifFalse: + [^self genSpecialSelectorSend]. + + "In-line the comparison and the jump, but if the types are not SmallInteger then we will need + to do a send and fall through to the following conditional branch. Since we're allocating values + in registers we would like to keep those registers live on the inlined path and reload registers + along the non-inlined send path. The merge logic at the branch destinations handles this." + argIsInt + ifTrue: + [rcvrReg := self allocateRegForStackEntryAt: 1. + (self ssValue: 1) popToReg: rcvrReg. + self MoveR: rcvrReg R: TempReg. + counterReg := self allocateRegNotConflictingWith: (self registerMaskFor: rcvrReg)] + ifFalse: + [self allocateRegForStackTopTwoEntriesInto: [:rTop :rNext| argReg := rTop. rcvrReg := rNext]. + rcvrReg = Arg0Reg ifTrue: + [rcvrReg := argReg. argReg := Arg0Reg]. + self ssTop popToReg: argReg. + (self ssValue: 1) popToReg: rcvrReg. + self MoveR: argReg R: TempReg. + counterReg := self allocateRegNotConflictingWith: (self registerMaskFor: rcvrReg and: argReg)]. + jumpNotSmallInts := (argIsInt or: [rcvrIsInt]) + ifTrue: [objectRepresentation genJumpNotSmallIntegerInScratchReg: TempReg] + ifFalse: [objectRepresentation genJumpNotSmallIntegersIn: rcvrReg andScratch: TempReg scratch: ClassReg]. + + self + genExecutionCountLogicInto: [ :cAddress :countTripBranch | + counterAddress := cAddress. + countTripped := countTripBranch ] + counterReg: counterReg. + + argIsInt + ifTrue: [self CmpCq: argInt R: rcvrReg] + ifFalse: [self CmpR: argReg R: rcvrReg]. + "Cmp is weird/backwards so invert the comparison. Further since there is a following conditional + jump bytecode define non-merge fixups and leave the cond bytecode to set the mergeness." + self genConditionalBranch: (branchDescriptor isBranchTrue + ifTrue: [primDescriptor opcode] + ifFalse: [self inverseBranchFor: primDescriptor opcode]) + operand: (self ensureFixupAt: targetBytecodePC - initialPC) asUnsignedInteger. + + self genFallsThroughCountLogicCounterReg: counterReg counterAddress: counterAddress. + + self Jump: (self ensureFixupAt: postBranchPC - initialPC). + countTripped jmpTarget: (jumpNotSmallInts jmpTarget: self Label). + + self ssFlushTo: simStackPtr. + self deny: rcvrReg = Arg0Reg. + argIsInt + ifTrue: [self MoveCq: argInt R: Arg0Reg] + ifFalse: [argReg ~= Arg0Reg ifTrue: [self MoveR: argReg R: Arg0Reg]]. + rcvrReg ~= ReceiverResultReg ifTrue: [self MoveR: rcvrReg R: ReceiverResultReg]. + index := byte0 - self firstSpecialSelectorBytecodeOffset. + ^self genMarshalledSend: index negated - 1 numArgs: 1 sendTable: ordinarySendTrampolines! Item was changed: InterpreterPrimitives subclass: #StackInterpreter + instanceVariableNames: 'currentBytecode bytecodeSetSelector localFP localIP localSP stackLimit stackPage stackPages method instructionPointer stackPointer framePointer localReturnValue localAbsentReceiver localAbsentReceiverOrZero extA extB primitiveFunctionPointer methodCache nsMethodCache atCache lkupClassTag lkupClass methodDictLinearSearchLimit highestRunnableProcessPriority nextWakeupUsecs nextPollUsecs inIOProcessEvents interruptKeycode interruptPending savedWindowSize imageHeaderFlags fullScreenFlag deferDisplayUpdates pendingFinalizationSignals extraVMMemory interpreterProxy showSurfaceFn primitiveTable primitiveAccessorDepthTable externalPrimitiveTable externalPrimitiveTableFirstFreeIndex overflowedPage extraFramesToMoveOnOverflow globalSessionID jmpBuf jmpDepth suspendedCallbacks suspendedMethods numStackPages desiredNumStackPages desiredEdenBytes classNameIndex thisClassIndex metaclassNumSlots interruptCheckChain suppressHeartbeatFlag breakSelector breakSelectorLength l ongRunningPrimitiveCheckMethod longRunningPrimitiveCheckSemaphore longRunningPrimitiveStartUsecs longRunningPrimitiveStopUsecs longRunningPrimitiveGCUsecs longRunningPrimitiveCheckSequenceNumber longRunningPrimitiveSignalUndelivered checkAllocFiller tempOop tempOop2 tempOop3 theUnknownShort the2ndUnknownShort imageFloatsBigEndian maxExtSemTabSizeSet lastMethodCacheProbeWrite gcSemaphoreIndex classByteArrayCompactIndex checkedPluginName statForceInterruptCheck statStackOverflow statStackPageDivorce statCheckForEvents statProcessSwitch statIOProcessEvents statPendingFinalizationSignals nativeSP nativeStackPointer lowcodeCalloutState shadowCallStackPointer' - instanceVariableNames: 'currentBytecode bytecodeSetSelector localFP localIP localSP stackLimit stackPage stackPages method instructionPointer stackPointer framePointer localReturnValue localAbsentReceiver localAbsentReceiverOrZero extA extB primitiveFunctionPointer methodCache nsMethodCache atCache lkupClassTag lkupClass methodDictLinearSearchLimit highestRunnableProcessPriority nextWakeupUsecs nextPollUsecs inIOProcessEvents interruptKeycode interruptPending savedWindowSize imageHeaderFlags fullScreenFlag deferDisplayUpdates pendingFinalizationSignals extraVMMemory interpreterProxy showSurfaceFn primitiveTable primitiveAccessorDepthTable externalPrimitiveTable externalPrimitiveTableFirstFreeIndex overflowedPage extraFramesToMoveOnOverflow globalSessionID jmpBuf jmpDepth suspendedCallbacks suspendedMethods numStackPages desiredNumStackPages desiredEdenBytes classNameIndex thisClassIndex metaclassNumSlots interruptCheckChain suppressHeartbeatFlag breakSelector breakSelectorLength l ongRunningPrimitiveCheckMethod longRunningPrimitiveCheckSemaphore longRunningPrimitiveStartUsecs longRunningPrimitiveStopUsecs longRunningPrimitiveGCUsecs longRunningPrimitiveCheckSequenceNumber longRunningPrimitiveSignalUndelivered checkAllocFiller tempOop tempOop2 tempOop3 theUnknownShort the2ndUnknownShort imageFloatsBigEndian maxExtSemTabSizeSet lastMethodCacheProbeWrite gcSemaphoreIndex classByteArrayCompactIndex checkedPluginName statForceInterruptCheck statStackOverflow statStackPageDivorce statCheckForEvents statProcessSwitch statIOProcessEvents statPendingFinalizationSignals nativeSP nativeStackPointer lowcodeCalloutState shadowCallStackPointer debugCallbackPath callbackReturnSP' classVariableNames: 'AccessModifierPrivate AccessModifierProtected AccessModifierPublic AltBytecodeEncoderClassName AltLongStoreBytecode AlternateHeaderHasPrimFlag AlternateHeaderIsOptimizedFlag AlternateHeaderNumLiteralsMask AtCacheFixedFields AtCacheFmt AtCacheMask AtCacheOop AtCacheSize AtCacheTotalSize AtPutBase BytecodeEncoderClassName BytecodeTable CacheProbeMax DirBadPath DirEntryFound DirNoMoreEntries DumpStackOnLowSpace EnclosingMixinIndex EnclosingObjectIndex EnforceAccessControl FailImbalancedPrimitives LongStoreBytecode MaxExternalPrimitiveTableSize MaxJumpBuf MaxPrimitiveIndex MaxQuickPrimitiveIndex MethodHeaderArgCountShift MethodHeaderFlagBitPosition MethodHeaderTempCountShift MixinIndex PrimNumberDoExternalCall PrimNumberExternalCall PrimNumberFFICall PrimitiveExternalCallIndex PrimitiveTable StackPageReachedButUntraced StackPageTraceInvalid StackPageTraced StackPageUnreached V3PrimitiveBitsMask' poolDictionaries: 'VMBasicConstants VMBytecodeConstants VMMethodCacheConstants VMObjectIndices VMSpurObjectRepresentationConstants VMSqueakClassIndices VMStackFrameOffsets' category: 'VMMaker-Interpreter'! !StackInterpreter commentStamp: 'eem 12/5/2014 11:32' prior: 0! This class is a complete implementation of the Smalltalk-80 virtual machine, derived originally from the Blue Book specification but quite different in some areas. This VM supports Closures but *not* old-style BlockContexts. It has been modernized with 32-bit pointers, better management of Contexts (see next item), and attention to variable use that allows the CCodeGenerator (qv) to keep, eg, the instruction pointer and stack pointer in registers as well as keeping most simple variables in a global array that seems to improve performance for most platforms. The VM does not use Contexts directly. Instead Contexts serve as proxies for a more conventional stack format that is invisible to the image. There is considerable explanation at http://www.mirandabanda.org/cogblog/2009/01/14/under-cover-contexts-and-the…. The VM maintains a fixed-size stack zone divided into pages, each page being capable of holding several method/block activations. A send establishes a new frame in the current stack page, a return returns to the previous frame. This eliminates allocation/deallocation of contexts and the moving of receiver and arguments from caller to callee on each send/return. Contexts are created lazily when an activation needs a context (creating a block, explicit use of thisContext, access to sender when sender is a frame, or linking of stack pages together). Contexts are either conventional and heap-resident ("single") or "married" and serve as proxies for their corresponding frame or "widowed", meaning that their spouse f rame has been returned from (died). A married context is specially marked (more details in the code) and refers to its frame. Likewise a married frame is specially marked and refers to its context. In addition to SmallInteger arithmetic and Floats, the VM supports logic on 32-bit PositiveLargeIntegers, thus allowing it to simulate itself much more effectively than would otherwise be the case. StackInterpreter and subclasses support multiple memory managers. Currently there are two. NewMemoryManager is a slightly refined version of ObjectMemory, and is the memory manager and garbage collector for the original Squeak object representation as described in "Back to the Future The Story of Squeak, A Practical Smalltalk Written in Itself", see http://ftp.squeak.org/docs/OOPSLA.Squeak.html. Spur is a faster, more regular object representation that is designed for more performance and functionality, and to have a common header format for both 32-bit and 64-bit versions. You can read about it in SpurMemoryManager's class comment. There is also a video of a presentation at ESUG 2014 (https://www.youtube.com/watch?v=k0nBNS1aHZ4), along with slides (http://www.slideshare.net/esug/spur-a-new-object-representation-for-cog?rel…).! Item was changed: ----- Method: StackInterpreter class>>mustBeGlobal: (in category 'translation') ----- mustBeGlobal: var "Answer if a variable must be global and exported. Used for inst vars that are accessed from VM support code." ^(super mustBeGlobal: var) or: [(self objectMemoryClass mustBeGlobal: var) or: [(#('interpreterProxy' 'interpreterVersion' 'inIOProcessEvents' 'deferDisplayUpdates' 'extraVMMemory' 'showSurfaceFn' 'desiredNumStackPages' 'desiredEdenBytes' 'breakSelector' 'breakSelectorLength' 'sendTrace' 'checkAllocFiller' 'checkedPluginName' + 'suppressHeartbeatFlag') includes: var) - 'suppressHeartbeatFlag' 'debugCallbackPath' 'callbackReturnSP') includes: var) or: [ "This allows slow machines to define bytecodeSetSelector as 0 to avoid the interpretation overhead." MULTIPLEBYTECODESETS not and: [var = 'bytecodeSetSelector']]]]! Item was added: + ----- Method: StackInterpreter>>checkForStackOverflow (in category 'message sending') ----- + checkForStackOverflow + "Check for stack overflow, moving frames to another stack if so." + <inline: true> + "After checkForInterrupts another event check may have been forced, setting both + stackLimit and stackPage stackLimit to all ones. So here we must check against + the real stackLimit, not the effective stackLimit." + stackPointer < stackPage realStackLimit ifTrue: + [self handleStackOverflow]! Item was changed: ----- Method: StackInterpreter>>checkStackPointerIndexForFrame: (in category 'frame access') ----- checkStackPointerIndexForFrame: theFP "Version of stackPointerIndexForFrame: that does not depend on writing back head frame pointers. Used for assertion checking. Safe only in external primitives (framePointer valid). Answer the 0-based index rel to the given frame. (This is what stackPointer used to be before conversion to pointer)" "In the StackInterpreter stacks grow down." | thePage theSP | <inline: false> <var: #theFP type: #'char *'> <var: #thePage type: #'StackPage *'> <var: #theSP type: #'char *'> + theFP = framePointer ifTrue: + [^self stackPointerIndexForFrame: theFP WithSP: stackPointer]. thePage := stackPages stackPageFor: theFP. theSP := self findSPOrNilOf: theFP on: thePage startingFrom: (thePage = stackPage ifTrue: [framePointer] ifFalse: [thePage headFP]). ^self stackPointerIndexForFrame: theFP WithSP: theSP! Item was changed: ----- Method: StackInterpreter>>handleStackOverflow (in category 'message sending') ----- handleStackOverflow + "Check for stack overflow, moving frames to another stack if so. + This should *only* be sent from checkForStackOverflow." + <inline: #never> - "Check for stack overflow, moving frames to another stack if so." | newPage theFP callerFP overflowLimitAddress overflowCount | <var: #newPage type: #'StackPage *'> <var: #theFP type: #'char *'> <var: #callerFP type: #'char *'> <var: #overflowLimitAddress type: #'char *'> + self assert: stackPointer < stackPage realStackLimit. - "After checkForInterrupts another event check may have been forced, setting both - stackLimit and stackPage stackLimit to all ones. So here we must check against - the real stackLimit, not the effective stackLimit." - stackPointer < stackPage realStackLimit ifFalse: - [^self]. self maybeTraceStackOverflow. statStackOverflow := statStackOverflow + 1. "The stack has overflowed this page. If the system is executing some recursive algorithm, e.g. fibonacci, then the system could thrash overflowing the stack if the call soon returns back to the current page. To avoid thrashing, since overflow is quite slow, we can move more than one frame. The idea is to record which page has overflowed, and the first time it overflows move one frame, the second time two frames, and so on. We move no more frames than would leave the page half occupied." theFP := framePointer. stackPage = overflowedPage ifTrue: [overflowLimitAddress := stackPage baseAddress - stackPages overflowLimit. overflowCount := extraFramesToMoveOnOverflow := extraFramesToMoveOnOverflow + 1. [(overflowCount := overflowCount - 1) >= 0 and: [(callerFP := self frameCallerFP: theFP) < overflowLimitAddress and: [(self isBaseFrame: callerFP) not]]] whileTrue: [theFP := callerFP]] ifFalse: [overflowedPage := stackPage. extraFramesToMoveOnOverflow := 0]. self ensureCallerContext: theFP. newPage := stackPages newStackPage. self moveFramesIn: stackPage through: theFP toPage: newPage. self setStackPageAndLimit: newPage. framePointer := stackPage headFP. stackPointer := stackPage headSP. self isCog ifFalse: "To overflow the stack this must be a new frame, but in Cog base frames are married." [self assert: (self frameHasContext: framePointer) not. self assert: (self validInstructionPointer: instructionPointer + 1 inMethod: method framePointer: framePointer)] ifTrue: [self assert: (self validInstructionPointer: instructionPointer + 1 inFrame: framePointer). self assert: ((self frameHasContext: framePointer) not or: [objectMemory isContext: (self frameContext: framePointer)])]! Item was changed: ----- Method: StackInterpreter>>handleStackOverflowOrEventAllowContextSwitch: (in category 'message sending') ----- handleStackOverflowOrEventAllowContextSwitch: mayContextSwitch "The stackPointer is below the stackLimit. This is either because of a stack overflow or the setting of stackLimit to indicate a possible interrupt. Check for interrupts and stackOverflow and deal with each appropriately. Answer if a context switch occurred." | switched | <inline: false> "If the stackLimit differs from the realStackLimit then the stackLimit has been set to indicate an event or interrupt that needs servicing." stackLimit = stackPage realStackLimit ifTrue: [self externalWriteBackHeadFramePointers. switched := false] ifFalse: [switched := self checkForEventsMayContextSwitch: mayContextSwitch]. + self checkForStackOverflow. - self handleStackOverflow. ^switched! Item was changed: ----- Method: StackInterpreter>>returnAs:ThroughCallback:Context: (in category 'callback support') ----- returnAs: returnTypeOop ThroughCallback: vmCallbackContext Context: callbackMethodContext "callbackMethodContext is an activation of invokeCallback:[stack:registers:jmpbuf:]. Its sender is the VM's state prior to the callback. Reestablish that state (via longjmp), and mark callbackMethodContext as dead." <export: true> <var: #vmCallbackContext type: #'VMCallbackContext *'> | calloutMethodContext theFP thePage | <var: #theFP type: #'char *'> <var: #thePage type: #'StackPage *'> - debugCallbackPath := 0. ((self isIntegerObject: returnTypeOop) and: [self isLiveContext: callbackMethodContext]) ifFalse: + [^false]. - [debugCallbackPath := 1. - ^false]. calloutMethodContext := self externalInstVar: SenderIndex ofContext: callbackMethodContext. (self isLiveContext: calloutMethodContext) ifFalse: + [^false]. - [debugCallbackPath := 2. - ^false]. - debugCallbackPath := 4. "We're about to leave this stack page; must save the current frame's instructionPointer." self push: instructionPointer. self externalWriteBackHeadFramePointers. "Mark callbackMethodContext as dead; the common case is that it is the current frame. We go the extra mile for the debugger." (self isSingleContext: callbackMethodContext) + ifTrue: [self markContextAsDead: callbackMethodContext] - ifTrue: [self markContextAsDead: callbackMethodContext. debugCallbackPath := debugCallbackPath bitOr: 8] ifFalse: + [theFP := self frameOfMarriedContext: callbackMethodContext. - [debugCallbackPath := debugCallbackPath bitOr: 16. - theFP := self frameOfMarriedContext: callbackMethodContext. framePointer = theFP "common case" ifTrue: + [(self isBaseFrame: theFP) ifFalse: "calloutMethodContext is immediately below on the same page. Make it current." + [instructionPointer := (self frameCallerSavedIP: framePointer) asUnsignedInteger. + stackPointer := framePointer + (self frameStackedReceiverOffset: framePointer) + objectMemory wordSize. + framePointer := self frameCallerFP: framePointer. + self setMethod: (self frameMethodObject: framePointer). + self restoreCStackStateForCallbackContext: vmCallbackContext. + "N.B. siglongjmp is defines as _longjmp on non-win32 platforms. + This matches the use of _setjmp in ia32abicc.c." + self siglong: vmCallbackContext trampoline jmp: (self integerValueOf: returnTypeOop). + ^true]. + stackPages freeStackPage: stackPage] - [debugCallbackPath := debugCallbackPath bitOr: 32. - (self isBaseFrame: theFP) - ifTrue: [stackPages freeStackPage: stackPage. debugCallbackPath := debugCallbackPath bitOr: 64] - ifFalse: "calloutMethodContext is immediately below on the same page. Make it current." - [debugCallbackPath := debugCallbackPath bitOr: 128. - instructionPointer := (self frameCallerSavedIP: framePointer) asUnsignedInteger. - stackPointer := framePointer + (self frameStackedReceiverOffset: framePointer) + objectMemory wordSize. - framePointer := self frameCallerFP: framePointer. - self setMethod: (self frameMethodObject: framePointer). - self restoreCStackStateForCallbackContext: vmCallbackContext. - callbackReturnSP := stackPointer. - "N.B. siglongjmp is defines as _longjmp on non-win32 platforms. - This matches the use of _setjmp in ia32abicc.c." - self siglong: vmCallbackContext trampoline jmp: (self integerValueOf: returnTypeOop). - ^true]] ifFalse: + [self externalDivorceFrame: theFP andContext: callbackMethodContext. - [debugCallbackPath := debugCallbackPath bitOr: 256. - self externalDivorceFrame: theFP andContext: callbackMethodContext. self markContextAsDead: callbackMethodContext]]. "Make the calloutMethodContext the active frame. The case where calloutMethodContext is immediately below callbackMethodContext on the same page is handled above." (self isStillMarriedContext: calloutMethodContext) ifTrue: + [theFP := self frameOfMarriedContext: calloutMethodContext. - [debugCallbackPath := debugCallbackPath bitOr: 512. - theFP := self frameOfMarriedContext: calloutMethodContext. thePage := stackPages stackPageFor: theFP. "findSPOf:on: points to the word beneath the instructionPointer, but there is no instructionPointer on the top frame of the current page." self assert: thePage ~= stackPage. + stackPointer := thePage headFP = theFP + ifTrue: [thePage headSP] + ifFalse: [(self findSPOf: theFP on: thePage) - objectMemory wordSize]. - stackPointer := (self findSPOf: theFP on: thePage) - objectMemory wordSize. framePointer := theFP] ifFalse: + [thePage := self makeBaseFrameFor: calloutMethodContext. - [debugCallbackPath := debugCallbackPath bitOr: 1024. - thePage := self makeBaseFrameFor: calloutMethodContext. framePointer := thePage headFP. stackPointer := thePage headSP]. instructionPointer := self popStack. self setMethod: (objectMemory fetchPointer: MethodIndex ofObject: calloutMethodContext). self setStackPageAndLimit: thePage. self restoreCStackStateForCallbackContext: vmCallbackContext. + "N.B. siglongjmp is defines as _longjmp on non-win32 platforms. - debugCallbackPath := debugCallbackPath bitOr: 2048. - callbackReturnSP := stackPointer. - "N.B. siglongjmp is defines as _longjmp on non-win32 platforms. This matches the use of _setjmp in ia32abicc.c." self siglong: vmCallbackContext trampoline jmp: (self integerValueOf: returnTypeOop). "NOTREACHED" ^true! Item was changed: ----- Method: StackInterpreter>>sendInvokeCallback:Stack:Registers:Jmpbuf: (in category 'callback support') ----- sendInvokeCallback: thunkPtr Stack: stackPtr Registers: regsPtr Jmpbuf: jmpBufPtr "Send the 4 argument callback message invokeCallback:stack:registers:jmpbuf: to Alien class with the supplied args. The arguments are raw C addresses and are converted to integer objects on the way." <export: true> | classTag | classTag := self fetchClassTagOfNonImm: (self splObj: ClassAlien). messageSelector := self splObj: SelectorInvokeCallback. argumentCount := 4. (self lookupInMethodCacheSel: messageSelector classTag: classTag) ifFalse: [(self lookupOrdinaryNoMNUEtcInClass: (objectMemory classForClassTag: classTag)) ~= 0 ifTrue: [^false]]. ((self argumentCountOf: newMethod) = 4 and: [primitiveFunctionPointer = 0]) ifFalse: [^false]. self push: (self splObj: ClassAlien). "receiver" self push: (self positiveMachineIntegerFor: thunkPtr). self push: (self positiveMachineIntegerFor: stackPtr). self push: (self positiveMachineIntegerFor: regsPtr). self push: (self positiveMachineIntegerFor: jmpBufPtr). self ifAppropriateCompileToNativeCode: newMethod selector: messageSelector. self justActivateNewMethod. (self isMachineCodeFrame: framePointer) ifFalse: [self maybeFlagMethodAsInterpreted: newMethod]. self externalWriteBackHeadFramePointers. + self checkForStackOverflow. - self handleStackOverflow. self enterSmalltalkExecutiveFromCallback. "not reached" ^true! Item was changed: ----- Method: StackInterpreter>>sendInvokeCallbackContext: (in category 'callback support') ----- sendInvokeCallbackContext: vmCallbackContext "Send the calllback message to Alien class with the supplied arg(s). Use either the 1 arg invokeCallbackContext: or the 4 arg invokeCallback:stack:registers:jmpbuf: message, depending on what selector is installed in the specialObjectsArray. Note that if invoking the legacy invokeCallback:stack:registers:jmpbuf: we pass the vmCallbackContext as the jmpbuf argument (see reestablishContextPriorToCallback:). The arguments are raw C addresses and are converted to integer objects on the way." <export: true> <var: #vmCallbackContext type: #'VMCallbackContext *'> | classTag | classTag := self fetchClassTagOfNonImm: (self splObj: ClassAlien). messageSelector := self splObj: SelectorInvokeCallback. (self lookupInMethodCacheSel: messageSelector classTag: classTag) ifFalse: [(self lookupOrdinaryNoMNUEtcInClass: (objectMemory classForClassTag: classTag)) ~= 0 ifTrue: [^false]]. primitiveFunctionPointer ~= 0 ifTrue: [^false]. self saveCStackStateForCallbackContext: vmCallbackContext. self push: (self splObj: ClassAlien). "receiver" (self argumentCountOf: newMethod) = 4 ifTrue: [self push: (self positiveMachineIntegerFor: vmCallbackContext thunkp asUnsignedInteger). self push: (self positiveMachineIntegerFor: vmCallbackContext stackp asUnsignedInteger). self push: (self positiveMachineIntegerFor: vmCallbackContext intregargsp asUnsignedInteger)]. self push: (self positiveMachineIntegerFor: vmCallbackContext asUnsignedInteger). self ifAppropriateCompileToNativeCode: newMethod selector: messageSelector. self justActivateNewMethod. (self isMachineCodeFrame: framePointer) ifFalse: [self maybeFlagMethodAsInterpreted: newMethod]. self externalWriteBackHeadFramePointers. + self checkForStackOverflow. - self handleStackOverflow. self enterSmalltalkExecutiveFromCallback. "not reached" ^true! Item was changed: ----- Method: ThreadedARMFFIPlugin>>ffiCalloutTo:SpecOnStack:in: (in category 'callout support') ----- ffiCalloutTo: procAddr SpecOnStack: specOnStack in: calloutState <var: #procAddr type: #'void *'> <var: #calloutState type: #'CalloutState *'> <var: #loadFloatRegs declareC: 'extern void loadFloatRegs(double, double, double, double, double, double, double, double)'> "Go out, call this guy and create the return value. This *must* be inlined because of the alloca of the outgoing stack frame in ffiCall:WithFlags:NumArgs:Args:AndTypes:" + | myThreadIndex atomicType floatRet intRet loadFloatRegs | - | myThreadIndex atomicType floatRet intRet loadFloatRegs oop | <var: #floatRet type: #double> <var: #intRet type: #usqLong> <inline: true> self cCode: '' inSmalltalk: [loadFloatRegs := #used. loadFloatRegs class]. + self maybeDisownVM: calloutState threadIndexInto: [:threadIndex| myThreadIndex := threadIndex]. - self cppIf: COGMTVM ifTrue: - [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: - [myThreadIndex := interpreterProxy disownVM: 0]]. calloutState floatRegisterIndex > 0 ifTrue: [self load: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 0)) to: 'double *') at: 0) Flo: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 2)) to: 'double *') at: 0) a: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 4)) to: 'double *') at: 0) t: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 6)) to: 'double *') at: 0) R: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 8)) to: 'double *') at: 0) e: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 10)) to: 'double *') at: 0) g: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 12)) to: 'double *') at: 0) s: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 14)) to: 'double *') at: 0)]. (self allocaLiesSoSetSpBeforeCall or: [self mustAlignStack]) ifTrue: [self setsp: calloutState argVector]. atomicType := self atomicTypeOf: calloutState ffiRetHeader. + (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: + [atomicType = FFITypeSingleFloat + ifTrue: + [floatRet := self + dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'float (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') + with: (calloutState integerRegisters at: 0) + with: (calloutState integerRegisters at: 1) + with: (calloutState integerRegisters at: 2) + with: (calloutState integerRegisters at: 3)] + ifFalse: "atomicType = FFITypeDoubleFloat" + [floatRet := self + dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') + with: (calloutState integerRegisters at: 0) + with: (calloutState integerRegisters at: 1) + with: (calloutState integerRegisters at: 2) + with: (calloutState integerRegisters at: 3)]. + + "undo any callee argument pops because it may confuse stack management with the alloca." + (self isCalleePopsConvention: calloutState callFlags) ifTrue: + [self setsp: calloutState argVector]. + self maybeOwnVM: calloutState threadIndex: myThreadIndex. + + ^interpreterProxy floatObjectOf: floatRet]. + + intRet := self - (atomicType >> 1) = (FFITypeSingleFloat >> 1) - ifTrue: - [atomicType = FFITypeSingleFloat - ifTrue: - [floatRet := self - dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'float (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') - with: (calloutState integerRegisters at: 0) - with: (calloutState integerRegisters at: 1) - with: (calloutState integerRegisters at: 2) - with: (calloutState integerRegisters at: 3)] - ifFalse: "atomicType = FFITypeDoubleFloat" - [floatRet := self - dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') - with: (calloutState integerRegisters at: 0) - with: (calloutState integerRegisters at: 1) - with: (calloutState integerRegisters at: 2) - with: (calloutState integerRegisters at: 3)]] - ifFalse: - [intRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'usqIntptr_t (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) + with: (calloutState integerRegisters at: 3). + - with: (calloutState integerRegisters at: 3)]. "undo any callee argument pops because it may confuse stack management with the alloca." (self isCalleePopsConvention: calloutState callFlags) ifTrue: [self setsp: calloutState argVector]. + self maybeOwnVM: calloutState threadIndex: myThreadIndex. - self cppIf: COGMTVM ifTrue: - [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: - [interpreterProxy ownVM: myThreadIndex]]. - (calloutState ffiRetHeader anyMask: FFIFlagPointer+FFIFlagStructure) ifTrue: ["Note: Order is important here since FFIFlagPointer + FFIFlagStructure is used to represent 'typedef void* VoidPointer' and VoidPointer must be returned as pointer *not* as struct." + (calloutState ffiRetHeader anyMask: FFIFlagPointer) ifTrue: + [^self ffiReturnPointer: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. + ^self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. - (calloutState ffiRetHeader anyMask: FFIFlagPointer) - ifTrue: - [oop := self ffiReturnPointer: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState] - ifFalse: - [oop := self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. - ^oop]. + ^self ffiCreateIntegralResultOop: intRet ofAtomicType: atomicType in: calloutState! - (atomicType >> 1) = (FFITypeSingleFloat >> 1) - ifTrue: - [oop := interpreterProxy floatObjectOf: floatRet] - ifFalse: - [oop := self ffiCreateIntegralResultOop: intRet - ofAtomicType: atomicType - in: calloutState]. - ^interpreterProxy methodReturnValue: oop! Item was changed: ----- Method: ThreadedARMFFIPlugin>>ffiReturnStruct:ofType:in: (in category 'callout support') ----- ffiReturnStruct: longLongRet ofType: ffiRetType in: calloutState <var: #longLongRet type: #usqLong> <var: #calloutState type: #'CalloutState *'> "Create a structure return value from an external function call. The value has been stored in alloca'ed space pointed to by the calloutState or in the return value." | retOop retClass oop | <inline: true> retClass := interpreterProxy fetchPointer: 1 ofObject: ffiRetType. retOop := interpreterProxy instantiateClass: retClass indexableSize: 0. self remapOop: retOop in: [oop := interpreterProxy instantiateClass: interpreterProxy classByteArray indexableSize: calloutState structReturnSize]. self mem: (interpreterProxy firstIndexableField: oop) cp: ((self returnStructInRegisters: calloutState structReturnSize) ifTrue: [self addressOf: longLongRet] ifFalse: [calloutState limit]) y: calloutState structReturnSize. interpreterProxy storePointer: 0 ofObject: retOop withValue: oop. + ^retOop! - ^interpreterProxy methodReturnValue: retOop! Item was changed: ----- Method: ThreadedFFIPlugin>>ffiCall:ArgArrayOrNil:NumArgs: (in category 'callout support') ----- ffiCall: externalFunction ArgArrayOrNil: argArrayOrNil NumArgs: nArgs "Generic callout. Does the actual work. If argArrayOrNil is nil it takes args from the stack and the spec from the method. If argArrayOrNil is not nil takes args from argArrayOrNil and the spec from the receiver." + | flags argTypeArray address argType oop argSpec argClass err theCalloutState calloutState requiredStackSize stackSize allocation result primNumArgs | - | flags argTypeArray address argType oop argSpec argClass err theCalloutState calloutState requiredStackSize stackSize allocation result | <inline: true> <var: #theCalloutState type: #'CalloutState'> <var: #calloutState type: #'CalloutState *'> <var: #allocation type: #'char *'> + primNumArgs := interpreterProxy methodArgumentCount. (interpreterProxy is: externalFunction KindOfClass: interpreterProxy classExternalFunction) ifFalse: [^self ffiFail: FFIErrorNotFunction]. "Load and check the values in the externalFunction before we call out" flags := interpreterProxy fetchInteger: ExternalFunctionFlagsIndex ofObject: externalFunction. interpreterProxy failed ifTrue: [^self ffiFail: FFIErrorBadArgs]. "This must come early for compatibility with the old FFIPlugin. Image-level code may assume the function pointer is loaded eagerly. Thanks to Nicolas Cellier." address := self ffiLoadCalloutAddress: externalFunction. interpreterProxy failed ifTrue: [^0 "error code already set by ffiLoadCalloutAddress:"]. argTypeArray := interpreterProxy fetchPointer: ExternalFunctionArgTypesIndex ofObject: externalFunction. "must be array of arg types" ((interpreterProxy isArray: argTypeArray) and: [(interpreterProxy slotSizeOf: argTypeArray) = (nArgs + 1)]) ifFalse: [^self ffiFail: FFIErrorBadArgs]. "check if the calling convention is supported" self cppIf: COGMTVM ifTrue: [(self ffiSupportsCallingConvention: (flags bitAnd: FFICallTypesMask)) ifFalse: [^self ffiFail: FFIErrorCallType]] ifFalse: "not masking causes threaded calls to fail, which is as they should if the plugin is not threaded." [(self ffiSupportsCallingConvention: flags) ifFalse: [^self ffiFail: FFIErrorCallType]]. requiredStackSize := self externalFunctionHasStackSizeSlot ifTrue: [interpreterProxy fetchInteger: ExternalFunctionStackSizeIndex ofObject: externalFunction] ifFalse: [-1]. interpreterProxy failed ifTrue: [^interpreterProxy primitiveFailFor: (argArrayOrNil isNil ifTrue: [PrimErrBadMethod] ifFalse: [PrimErrBadReceiver])]. stackSize := requiredStackSize < 0 ifTrue: [DefaultMaxStackSize] ifFalse: [requiredStackSize]. self cCode: [] inSmalltalk: [theCalloutState := self class calloutStateClass new]. calloutState := self addressOf: theCalloutState. self cCode: [self me: calloutState ms: 0 et: (self sizeof: #CalloutState)]. calloutState callFlags: flags. "Fetch return type and args" argType := interpreterProxy fetchPointer: 0 ofObject: argTypeArray. argSpec := interpreterProxy fetchPointer: 0 ofObject: argType. argClass := interpreterProxy fetchPointer: 1 ofObject: argType. (err := self ffiCheckReturn: argSpec With: argClass in: calloutState) ~= 0 ifTrue: [^self ffiFail: err]. "cannot return" "alloca the outgoing stack frame, leaving room for marshalling args, and including space for the return struct, if any. Additional space reserved for saving register args like mandated by Win64 X64 or PPC ABI, will be managed by the call itself" allocation := self alloca: stackSize + calloutState structReturnSize + self cStackAlignment. self mustAlignStack ifTrue: + [allocation := self cCoerce: (allocation asUnsignedIntegerPtr bitClear: self cStackAlignment - 1) to: #'char *']. - [allocation := self cCoerce: (allocation asUnsignedIntegerPtr bitClear: self cStackAlignment - 1) - to: #'char *']. calloutState argVector: allocation; currentArg: allocation; limit: allocation + stackSize. (calloutState structReturnSize > 0 and: [self nonRegisterStructReturnIsViaImplicitFirstArgument and: [(self returnStructInRegisters: calloutState structReturnSize) not]]) ifTrue: [err := self ffiPushPointer: calloutState limit in: calloutState. err ~= 0 ifTrue: [self cleanupCalloutState: calloutState. self cppIf: COGMTVM ifTrue: [err = PrimErrObjectMayMove negated ifTrue: [^PrimErrObjectMayMove]]. "N.B. Do not fail if object may move because caller will GC and retry." ^self ffiFail: err]]. 1 to: nArgs do: [:i| argType := interpreterProxy fetchPointer: i ofObject: argTypeArray. argSpec := interpreterProxy fetchPointer: 0 ofObject: argType. argClass := interpreterProxy fetchPointer: 1 ofObject: argType. oop := argArrayOrNil isNil ifTrue: [interpreterProxy stackValue: nArgs - i] ifFalse: [interpreterProxy fetchPointer: i - 1 ofObject: argArrayOrNil]. err := self ffiArgument: oop Spec: argSpec Class: argClass in: calloutState. err ~= 0 ifTrue: [self cleanupCalloutState: calloutState. self cppIf: COGMTVM ifTrue: [err = PrimErrObjectMayMove negated ifTrue: [^PrimErrObjectMayMove]]. "N.B. Do not fail if object may move because caller will GC and retry." ^self ffiFail: err]]. "coercion failed or out of stack space" "Failures must be reported back from ffiArgument:Spec:Class:in:. Should not fail from here on in." self assert: interpreterProxy failed not. self ffiLogCallout: externalFunction. (requiredStackSize < 0 and: [self externalFunctionHasStackSizeSlot]) ifTrue: [stackSize := calloutState currentArg - calloutState argVector. interpreterProxy storeInteger: ExternalFunctionStackSizeIndex ofObject: externalFunction withValue: stackSize]. "Go out and call this guy" result := self ffiCalloutTo: address SpecOnStack: argArrayOrNil notNil in: calloutState. self cleanupCalloutState: calloutState. + "Can not safely use argumentCount (via e.g. methodReturnValue:) since it may have been changed by a callback." + interpreterProxy pop: primNumArgs + 1 thenPush: result. ^result! Item was changed: ----- Method: ThreadedFFIPlugin>>ffiReturnCStringFrom: (in category 'callout support') ----- ffiReturnCStringFrom: cPointer "Create a Smalltalk string from a zero terminated C string" | strLen strOop cString strPtr | <var: #cString type: #'char *'> <var: #strPtr type: #'char *'> + cPointer ifNil: [^interpreterProxy nilObject]. "nil always returns as nil" + cString := self cCoerce: cPointer to: #'char *'. - cPointer = nil ifTrue:[ - ^interpreterProxy methodReturnValue: interpreterProxy nilObject]. "nil always returns as nil" - cString := self cCoerce: cPointer to:'char *'. strLen := 0. [(cString at: strLen) = 0] whileFalse:[strLen := strLen+1]. strOop := interpreterProxy instantiateClass: interpreterProxy classString indexableSize: strLen. strPtr := interpreterProxy firstIndexableField: strOop. + 0 to: strLen-1 do: [:i| strPtr at: i put: (cString at: i)]. + ^strOop! - 0 to: strLen-1 do:[:i| strPtr at: i put: (cString at: i)]. - ^interpreterProxy methodReturnValue: strOop! Item was changed: ----- Method: ThreadedFFIPlugin>>ffiReturnPointer:ofType:in: (in category 'callout support') ----- ffiReturnPointer: retVal ofType: retType in: calloutState <var: #calloutState type: #'CalloutState *'> <var: #retVal type: #usqLong> "Generic callout support. Create a pointer return value from an external function call" | retClass atomicType retOop oop ptr classOop | <var: #ptr type: #'sqInt *'> retClass := interpreterProxy fetchPointer: 1 ofObject: retType. retClass = interpreterProxy nilObject ifTrue: ["Create ExternalData upon return" atomicType := self atomicTypeOf: calloutState ffiRetHeader. (atomicType >> 1) = (FFITypeSignedChar >> 1) ifTrue: "String return" + [retOop := self ffiReturnCStringFrom: (self cCoerceSimple: retVal to: #usqInt). + ^retOop]. - [^self ffiReturnCStringFrom: (self cCoerceSimple: retVal to: #usqInt)]. "generate external data" self remapOop: retType in: [oop := interpreterProxy instantiateClass: interpreterProxy classExternalAddress indexableSize: BytesPerWord. ptr := interpreterProxy firstIndexableField: oop. ptr at: 0 put: (self cCoerceSimple: retVal to: #sqInt). self remapOop: oop in: [retOop := interpreterProxy instantiateClass: interpreterProxy classExternalData indexableSize: 0]. interpreterProxy storePointer: 0 ofObject: retOop withValue: oop]. interpreterProxy storePointer: 1 ofObject: retOop withValue: retType. + ^retOop]. - ^interpreterProxy methodReturnValue: retOop]. "non-atomic pointer return" classOop := (calloutState ffiRetHeader anyMask: FFIFlagStructure) ifTrue:[interpreterProxy classByteArray] ifFalse:[interpreterProxy classExternalAddress]. self remapOop: retClass in: [oop := interpreterProxy instantiateClass: classOop indexableSize: BytesPerWord]. ptr := interpreterProxy firstIndexableField: oop. ptr at: 0 put: (self cCoerceSimple: retVal to: #sqInt). self remapOop: oop in: [retOop := interpreterProxy instantiateClass: retClass indexableSize: 0]. interpreterProxy storePointer: 0 ofObject: retOop withValue: oop. + ^retOop! - ^interpreterProxy methodReturnValue: retOop! Item was added: + ----- Method: ThreadedFFIPlugin>>maybeDisownVM:threadIndexInto: (in category 'primitive support') ----- + maybeDisownVM: calloutState threadIndexInto: aBlock + <var: #calloutState type: #'CalloutState *'> + <inline: #always> + self cppIf: COGMTVM + ifTrue: + [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: + [aBlock value: (interpreterProxy disownVM: 0)]]! Item was added: + ----- Method: ThreadedFFIPlugin>>maybeOwnVM:threadIndex: (in category 'primitive support') ----- + maybeOwnVM: calloutState threadIndex: myThreadIndex + <var: #calloutState type: #'CalloutState *'> + <inline: #always> + self cppIf: COGMTVM ifTrue: + [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: + [interpreterProxy ownVM: myThreadIndex]]! Item was changed: ----- Method: ThreadedFFIPlugin>>mustAlignStack (in category 'marshalling') ----- mustAlignStack "Many ABIs mandate a particular stack alignment greater than the natural word size. If so, this macro will answer true. See class-side preambleCCode." <cmacro: '() MUST_ALIGN_STACK'> + ^false! - ^0! Item was changed: ----- Method: ThreadedFFIPlugin>>primitiveLoadSymbolFromModule (in category 'primitives') ----- primitiveLoadSymbolFromModule "Attempt to find the address of a symbol in a loaded library. loadSymbol: aSymbol fromModule: moduleName <primitive: 'primitiveLoadSymbolFromModule' error: errorCode module: 'SqueakFFIPrims'> " <export: true> | symbol module moduleHandle address oop ptr | <var: #address type: #'void *'> <var: #ptr type: #'void **'> interpreterProxy methodArgumentCount = 2 ifFalse: [^interpreterProxy primitiveFailFor: PrimErrBadNumArgs]. module := interpreterProxy stackValue: 0. symbol := interpreterProxy stackValue: 1. + moduleHandle := module ~= interpreterProxy nilObject ifTrue: + [self ffiLoadCalloutModule: module]. - module ~= interpreterProxy nilObject - ifTrue: [ moduleHandle := self ffiLoadCalloutModule: module ] - ifFalse: [ moduleHandle := nil ]. interpreterProxy failed ifTrue: [^interpreterProxy primitiveFailFor: PrimErrNotFound]. address := interpreterProxy ioLoadSymbol: (self cCoerce: (interpreterProxy firstIndexableField: symbol) to: #sqInt) OfLength: (interpreterProxy byteSizeOf: symbol) FromModule: moduleHandle. (interpreterProxy failed or: [address = 0]) ifTrue: [^interpreterProxy primitiveFailFor: PrimErrNotFound]. oop := interpreterProxy instantiateClass: interpreterProxy classExternalAddress indexableSize: (self sizeof: #'void *'). ptr := interpreterProxy firstIndexableField: oop. ptr at: 0 put: address. ^interpreterProxy methodReturnValue: oop! Item was changed: ----- Method: ThreadedIA32FFIPlugin>>ffiCalloutTo:SpecOnStack:in: (in category 'callout support') ----- ffiCalloutTo: procAddr SpecOnStack: specOnStack in: calloutState <var: #procAddr type: #'void *'> <var: #calloutState type: #'CalloutState *'> "Go out, call this guy and create the return value. This *must* be inlined because of the alloca of the outgoing stack frame in ffiCall:WithFlags:NumArgs:Args:AndTypes:" + | myThreadIndex atomicType floatRet intRet | - | myThreadIndex atomicType floatRet intRet oop | <var: #floatRet type: #double> <var: #intRet type: #usqLong> <inline: true> + self maybeDisownVM: calloutState threadIndexInto: [:threadIndex| myThreadIndex := threadIndex]. - self cppIf: COGMTVM ifTrue: - [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: - [myThreadIndex := interpreterProxy disownVM: 0]]. (self allocaLiesSoSetSpBeforeCall or: [self mustAlignStack]) ifTrue: [self setsp: calloutState argVector]. atomicType := self atomicTypeOf: calloutState ffiRetHeader. + (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: + [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)()'). + + "undo any callee argument pops because it may confuse stack management with the alloca." + (self isCalleePopsConvention: calloutState callFlags) ifTrue: + [self setsp: calloutState argVector]. + self maybeOwnVM: calloutState threadIndex: myThreadIndex. + + ^interpreterProxy floatObjectOf: floatRet]. + + intRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'usqLong (*)()'). + - (atomicType >> 1) = (FFITypeSingleFloat >> 1) - ifTrue: - [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)()')] - ifFalse: - [intRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'usqLong (*)()')]. "undo any callee argument pops because it may confuse stack management with the alloca." (self isCalleePopsConvention: calloutState callFlags) ifTrue: [self setsp: calloutState argVector]. + self maybeOwnVM: calloutState threadIndex: myThreadIndex. - self cppIf: COGMTVM ifTrue: - [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: - [interpreterProxy ownVM: myThreadIndex]]. - (calloutState ffiRetHeader anyMask: FFIFlagPointer+FFIFlagStructure) ifTrue: ["Note: Order is important here since FFIFlagPointer + FFIFlagStructure is used to represent 'typedef void* VoidPointer' and VoidPointer must be returned as pointer *not* as struct." + (calloutState ffiRetHeader anyMask: FFIFlagPointer) ifTrue: + [^self ffiReturnPointer: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. + ^self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. - (calloutState ffiRetHeader anyMask: FFIFlagPointer) - ifTrue: - [oop := self ffiReturnPointer: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState] - ifFalse: - [oop := self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. - ^oop]. + ^self ffiCreateIntegralResultOop: intRet ofAtomicType: atomicType in: calloutState! - (atomicType >> 1) = (FFITypeSingleFloat >> 1) - ifTrue: - [oop := interpreterProxy floatObjectOf: floatRet] - ifFalse: - [oop := self ffiCreateIntegralResultOop: intRet - ofAtomicType: atomicType - in: calloutState]. - ^interpreterProxy methodReturnValue: oop! Item was changed: ----- Method: ThreadedIA32FFIPlugin>>ffiReturnStruct:ofType:in: (in category 'callout support') ----- ffiReturnStruct: longLongRet ofType: ffiRetType in: calloutState <var: #longLongRet type: #usqLong> <var: #calloutState type: #'CalloutState *'> "Create a structure return value from an external function call. The value as been stored in alloca'ed space pointed to by the calloutState." | retOop retClass oop | <inline: true> retClass := interpreterProxy fetchPointer: 1 ofObject: ffiRetType. retOop := interpreterProxy instantiateClass: retClass indexableSize: 0. self remapOop: retOop in: [oop := interpreterProxy instantiateClass: interpreterProxy classByteArray indexableSize: calloutState structReturnSize]. + self mem: (interpreterProxy firstIndexableField: oop) + cp: ((self returnStructInRegisters: calloutState structReturnSize) + ifTrue: [(self addressOf: longLongRet) asVoidPointer] + ifFalse: [calloutState limit]) + y: calloutState structReturnSize. - (self returnStructInRegisters: calloutState structReturnSize) - ifTrue: - [self mem: (interpreterProxy firstIndexableField: oop) cp: (self addressOf: longLongRet) y: calloutState structReturnSize] - ifFalse: - [self mem: (interpreterProxy firstIndexableField: oop) cp: calloutState limit y: calloutState structReturnSize]. interpreterProxy storePointer: 0 ofObject: retOop withValue: oop. + ^retOop! - ^interpreterProxy methodReturnValue: retOop! Item was changed: ----- Method: ThreadedX64SysVFFIPlugin>>ffiCalloutTo:SpecOnStack:in: (in category 'callout support') ----- ffiCalloutTo: procAddr SpecOnStack: specOnStack in: calloutState <var: #procAddr type: #'void *'> <var: #calloutState type: #'CalloutState *'> <var: #loadFloatRegs declareC: 'extern void loadFloatRegs(double, double, double, double, double, double, double, double)'> "Go out, call this guy and create the return value. This *must* be inlined because of the alloca of the outgoing stack frame in ffiCall:WithFlags:NumArgs:Args:AndTypes:" + | myThreadIndex atomicType floatRet intRet loadFloatRegs | - | myThreadIndex atomicType floatRet intRet loadFloatRegs oop | <var: #floatRet type: #double> <var: #intRet type: 'SixteenByteReturn'> <inline: true> self cCode: '' inSmalltalk: [loadFloatRegs := #used. loadFloatRegs class]. + self maybeDisownVM: calloutState threadIndexInto: [:threadIndex| myThreadIndex := threadIndex]. - self cppIf: COGMTVM ifTrue: - [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: - [myThreadIndex := interpreterProxy disownVM: 0]]. calloutState floatRegisterIndex > 0 ifTrue: [self load: (calloutState floatRegisters at: 0) Flo: (calloutState floatRegisters at: 1) a: (calloutState floatRegisters at: 2) t: (calloutState floatRegisters at: 3) R: (calloutState floatRegisters at: 4) e: (calloutState floatRegisters at: 5) g: (calloutState floatRegisters at: 6) s: (calloutState floatRegisters at: 7)]. (self allocaLiesSoSetSpBeforeCall or: [self mustAlignStack]) ifTrue: [self setsp: calloutState argVector]. atomicType := self atomicTypeOf: calloutState ffiRetHeader. + (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: + [atomicType = FFITypeSingleFloat + ifTrue: + [floatRet := self + dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'float (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') + with: (calloutState integerRegisters at: 0) + with: (calloutState integerRegisters at: 1) + with: (calloutState integerRegisters at: 2) + with: (calloutState integerRegisters at: 3) + with: (calloutState integerRegisters at: 4) + with: (calloutState integerRegisters at: 5)] + ifFalse: "atomicType = FFITypeDoubleFloat" + [floatRet := self + dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') + with: (calloutState integerRegisters at: 0) + with: (calloutState integerRegisters at: 1) + with: (calloutState integerRegisters at: 2) + with: (calloutState integerRegisters at: 3) + with: (calloutState integerRegisters at: 4) + with: (calloutState integerRegisters at: 5)]. + + self maybeOwnVM: calloutState threadIndex: myThreadIndex. + + ^interpreterProxy floatObjectOf: floatRet]. + + intRet := self - (atomicType >> 1) = (FFITypeSingleFloat >> 1) - ifTrue: - [atomicType = FFITypeSingleFloat - ifTrue: - [floatRet := self - dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'float (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') - with: (calloutState integerRegisters at: 0) - with: (calloutState integerRegisters at: 1) - with: (calloutState integerRegisters at: 2) - with: (calloutState integerRegisters at: 3) - with: (calloutState integerRegisters at: 4) - with: (calloutState integerRegisters at: 5)] - ifFalse: "atomicType = FFITypeDoubleFloat" - [floatRet := self - dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') - with: (calloutState integerRegisters at: 0) - with: (calloutState integerRegisters at: 1) - with: (calloutState integerRegisters at: 2) - with: (calloutState integerRegisters at: 3) - with: (calloutState integerRegisters at: 4) - with: (calloutState integerRegisters at: 5)]] - ifFalse: - [intRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'SixteenByteReturn (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3) with: (calloutState integerRegisters at: 4) + with: (calloutState integerRegisters at: 5). - with: (calloutState integerRegisters at: 5)]. - "undo any callee argument pops because it may confuse stack management with the alloca." - (self isCalleePopsConvention: calloutState callFlags) ifTrue: - [self setsp: calloutState argVector]. + self maybeOwnVM: calloutState threadIndex: myThreadIndex. - self cppIf: COGMTVM ifTrue: - [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: - [interpreterProxy ownVM: myThreadIndex]]. (calloutState ffiRetHeader anyMask: FFIFlagPointer+FFIFlagStructure) ifTrue: ["Note: Order is important here since FFIFlagPointer + FFIFlagStructure is used to represent 'typedef void* VoidPointer' and VoidPointer must be returned as pointer *not* as struct." + (calloutState ffiRetHeader anyMask: FFIFlagPointer) ifTrue: + [^self ffiReturnPointer: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. + ^self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. - (calloutState ffiRetHeader anyMask: FFIFlagPointer) - ifTrue: - [oop := self ffiReturnPointer: intRet a ofType: (self ffiReturnType: specOnStack) in: calloutState] - ifFalse: - [oop := self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. - ^oop]. + ^self ffiCreateIntegralResultOop: intRet ofAtomicType: atomicType in: calloutState! - (atomicType >> 1) = (FFITypeSingleFloat >> 1) - ifTrue: - [oop := interpreterProxy floatObjectOf: floatRet] - ifFalse: - [oop := self ffiCreateIntegralResultOop: intRet a - ofAtomicType: atomicType - in: calloutState]. - ^interpreterProxy methodReturnValue: oop! Item was changed: ----- Method: ThreadedX64SysVFFIPlugin>>ffiReturnStruct:ofType:in: (in category 'callout support') ----- ffiReturnStruct: sixteenByteRet ofType: ffiRetType in: calloutState <var: #sixteenByteRet type: 'SixteenByteReturn'> <var: #calloutState type: #'CalloutState *'> "Create a structure return value from an external function call. The value has been stored in alloca'ed space pointed to by the calloutState or in the return value." | retOop retClass oop | <inline: true> retClass := interpreterProxy fetchPointer: 1 ofObject: ffiRetType. retOop := interpreterProxy instantiateClass: retClass indexableSize: 0. self remapOop: retOop in: [oop := interpreterProxy instantiateClass: interpreterProxy classByteArray indexableSize: calloutState structReturnSize]. self mem: (interpreterProxy firstIndexableField: oop) cp: ((self returnStructInRegisters: calloutState structReturnSize) ifTrue: [(self addressOf: sixteenByteRet) asVoidPointer] ifFalse: [calloutState limit]) y: calloutState structReturnSize. interpreterProxy storePointer: 0 ofObject: retOop withValue: oop. + ^retOop! - ^interpreterProxy methodReturnValue: retOop! Item was changed: ----- Method: ThreadedX64Win64FFIPlugin>>ffiCalloutTo:SpecOnStack:in: (in category 'callout support') ----- ffiCalloutTo: procAddr SpecOnStack: specOnStack in: calloutState <var: #procAddr type: #'void *'> <var: #calloutState type: #'CalloutState *'> <var: #loadFloatRegs declareC: 'extern void loadFloatRegs(double, double, double, double)'> "Go out, call this guy and create the return value. This *must* be inlined because of the alloca of the outgoing stack frame in ffiCall:WithFlags:NumArgs:Args:AndTypes:" + | myThreadIndex atomicType floatRet intRet loadFloatRegs | - | myThreadIndex atomicType floatRet intRet loadFloatRegs oop | <var: #floatRet type: #double> <var: #intRet type: #usqLong> <inline: true> self cCode: '' inSmalltalk: [loadFloatRegs := #used. loadFloatRegs class]. + self maybeDisownVM: calloutState threadIndexInto: [:threadIndex| myThreadIndex := threadIndex]. - self cppIf: COGMTVM ifTrue: - [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: - [myThreadIndex := interpreterProxy disownVM: 0]]. calloutState floatRegisterSignature > 0 ifTrue: [self load: (calloutState floatRegisters at: 0) Flo: (calloutState floatRegisters at: 1) at: (calloutState floatRegisters at: 2) Re: (calloutState floatRegisters at: 3) gs: (calloutState floatRegisters at: 4)]. (self allocaLiesSoSetSpBeforeCall or: [self mustAlignStack]) ifTrue: [self setsp: calloutState argVector]. + - atomicType := self atomicTypeOf: calloutState ffiRetHeader. + (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: + [atomicType = FFITypeSingleFloat + ifTrue: + [floatRet := self + dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'float (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') + with: (calloutState integerRegisters at: 0) + with: (calloutState integerRegisters at: 1) + with: (calloutState integerRegisters at: 2) + with: (calloutState integerRegisters at: 3)] + ifFalse: "atomicType = FFITypeDoubleFloat" + [floatRet := self + dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') + with: (calloutState integerRegisters at: 0) + with: (calloutState integerRegisters at: 1) + with: (calloutState integerRegisters at: 2) + with: (calloutState integerRegisters at: 3)]. + + "undo any callee argument pops because it may confuse stack management with the alloca." + (self isCalleePopsConvention: calloutState callFlags) ifTrue: + [self setsp: calloutState argVector]. + self maybeOwnVM: calloutState threadIndex: myThreadIndex. + + ^interpreterProxy floatObjectOf: floatRet]. + + intRet := self - (atomicType >> 1) = (FFITypeSingleFloat >> 1) - ifTrue: - [atomicType = FFITypeSingleFloat - ifTrue: - [floatRet := self - dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'float (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') - with: (calloutState integerRegisters at: 0) - with: (calloutState integerRegisters at: 1) - with: (calloutState integerRegisters at: 2) - with: (calloutState integerRegisters at: 3)] - ifFalse: "atomicType = FFITypeDoubleFloat" - [floatRet := self - dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') - with: (calloutState integerRegisters at: 0) - with: (calloutState integerRegisters at: 1) - with: (calloutState integerRegisters at: 2) - with: (calloutState integerRegisters at: 3)]] - ifFalse: - [intRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'usqIntptr_t (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) + with: (calloutState integerRegisters at: 3). + - with: (calloutState integerRegisters at: 3)]. "undo any callee argument pops because it may confuse stack management with the alloca." (self isCalleePopsConvention: calloutState callFlags) ifTrue: [self setsp: calloutState argVector]. + self maybeOwnVM: calloutState threadIndex: myThreadIndex. - self cppIf: COGMTVM ifTrue: - [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: - [interpreterProxy ownVM: myThreadIndex]]. - (calloutState ffiRetHeader anyMask: FFIFlagPointer+FFIFlagStructure) ifTrue: ["Note: Order is important here since FFIFlagPointer + FFIFlagStructure is used to represent 'typedef void* VoidPointer' and VoidPointer must be returned as pointer *not* as struct." + (calloutState ffiRetHeader anyMask: FFIFlagPointer) ifTrue: + [^self ffiReturnPointer: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. + ^self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. - (calloutState ffiRetHeader anyMask: FFIFlagPointer) - ifTrue: - [oop := self ffiReturnPointer: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState] - ifFalse: - [oop := self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. - ^oop]. + ^self ffiCreateIntegralResultOop: intRet ofAtomicType: atomicType in: calloutState! - (atomicType >> 1) = (FFITypeSingleFloat >> 1) - ifTrue: - [oop := interpreterProxy floatObjectOf: floatRet] - ifFalse: - [oop := self ffiCreateIntegralResultOop: intRet - ofAtomicType: atomicType - in: calloutState]. - ^interpreterProxy methodReturnValue: oop! Item was changed: ----- Method: ThreadedX64Win64FFIPlugin>>ffiReturnStruct:ofType:in: (in category 'callout support') ----- ffiReturnStruct: intRet ofType: ffiRetType in: calloutState <var: #intRet type: #usqLong> <var: #calloutState type: #'CalloutState *'> "Create a structure return value from an external function call. The value has been stored in alloca'ed space pointed to by the calloutState or in the return value." | retOop retClass oop | <inline: true> retClass := interpreterProxy fetchPointer: 1 ofObject: ffiRetType. retOop := interpreterProxy instantiateClass: retClass indexableSize: 0. self remapOop: retOop in: [oop := interpreterProxy instantiateClass: interpreterProxy classByteArray indexableSize: calloutState structReturnSize]. self mem: (interpreterProxy firstIndexableField: oop) cp: ((self returnStructInRegisters: calloutState structReturnSize) ifTrue: [self addressOf: intRet] ifFalse: [calloutState limit]) y: calloutState structReturnSize. interpreterProxy storePointer: 0 ofObject: retOop withValue: oop. + ^retOop! - ^interpreterProxy methodReturnValue: retOop!

1 0

[OpenSmalltalk/opensmalltalk-vm] 3a6a4e: Fix an incorrect return for compiling linux VM wit...
by GitHub 14 Dec '16

14 Dec '16

Branch: refs/heads/Cog Home: https://github.com/OpenSmalltalk/opensmalltalk-vm Commit: 3a6a4e05ed7fe09f6f6702bc8207059f88b3c7a5 https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/3a6a4e05ed7fe09f6f… Author: Nicolas Cellier <nicolas.cellier.aka.nice(a)gmail.com> Date: 2016-12-14 (Wed, 14 Dec 2016) Changed paths: M platforms/unix/vm/sqUnixMain.c Log Message: ----------- Fix an incorrect return for compiling linux VM with clang Indeed, a non void function should return a value. Arbitrarily choose 0 since we don't seem to care.

1 0

Unsubscrible
by MR.WANGYOUCHU 14 Dec '16

14 Dec '16

2016-12-14 MR.WANGYOUCHU 发件人：vm-dev-request(a)lists.squeakfoundation.org 发送时间：2016-12-14 06:13 主题：Vm-dev Digest, Vol 126, Issue 29 收件人："vm-dev"<vm-dev(a)lists.squeakfoundation.org> 抄送： Send Vm-dev mailing list submissions to vm-dev(a)lists.squeakfoundation.org To subscribe or unsubscribe via the World Wide Web, visit http://lists.squeakfoundation.org/mailman/listinfo/vm-dev or, via email, send a message with subject or body 'help' to vm-dev-request(a)lists.squeakfoundation.org You can reach the person managing the list at vm-dev-owner(a)lists.squeakfoundation.org When replying, please edit your Subject line so it is more specific than "Re: Contents of Vm-dev digest..." Today's Topics: 1. [OpenSmalltalk/opensmalltalk-vm] 1acd02: Merge commit '5a949459043392bbc093d78574c2d7c07b10... (GitHub) 2. VM Maker: VMMaker.oscog-nice.2038.mcz (commits(a)source.squeak.org) ---------------------------------------------------------------------- Message: 1 Date: Tue, 13 Dec 2016 10:47:15 -0800 From: GitHub <noreply(a)github.com> To: vm-dev(a)lists.squeakfoundation.org Subject: [Vm-dev] [OpenSmalltalk/opensmalltalk-vm] 1acd02: Merge commit '5a949459043392bbc093d78574c2d7c07b10... Message-ID: <585042338c44f_42c83fa21117713458182(a)hookshot-fe5-cp1-prd.iad.github.net.mail> Content-Type: text/plain; charset="utf-8" Branch: refs/heads/Cog Home: https://github.com/OpenSmalltalk/opensmalltalk-vm Commit: 1acd02683a9e25936615e0573c712b0ecfd8d06e https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/1acd02683a9e259366… Author: Esteban Lorenzano <estebanlm(a)gmail.com> Date: 2016-12-09 (Fri, 09 Dec 2016) Changed paths: M .gitignore M build.macos32x86/pharo.stack.spur/plugins.ext M build.macos32x86/pharo.stack.spur/plugins.int M nsspur64src/vm/cogit.h M nsspur64src/vm/cogitX64.c M nsspur64src/vm/cointerp.c M nsspur64src/vm/cointerp.h M nsspur64src/vm/gcc3x-cointerp.c M nsspursrc/vm/cogit.h M nsspursrc/vm/cogitARMv5.c M nsspursrc/vm/cogitIA32.c M nsspursrc/vm/cogitMIPSEL.c M nsspursrc/vm/cointerp.c M nsspursrc/vm/cointerp.h M nsspursrc/vm/gcc3x-cointerp.c M nsspurstack64src/vm/gcc3x-interp.c M nsspurstack64src/vm/interp.c M nsspurstacksrc/vm/gcc3x-interp.c M nsspurstacksrc/vm/interp.c M platforms/iOS/vm/Common/Classes/sqSqueakMainApp.m M platforms/unix/vm/sqUnixMain.c M platforms/win32/vm/sqWin32Backtrace.c M spur64src/vm/cogit.h M spur64src/vm/cogitX64.c M spur64src/vm/cointerp.c M spur64src/vm/cointerp.h M spur64src/vm/gcc3x-cointerp.c M spursista64src/vm/cogit.h M spursista64src/vm/cogitX64.c M spursista64src/vm/cointerp.c M spursista64src/vm/cointerp.h M spursista64src/vm/gcc3x-cointerp.c M spursistasrc/vm/cogit.h M spursistasrc/vm/cogitARMv5.c M spursistasrc/vm/cogitIA32.c M spursistasrc/vm/cogitMIPSEL.c M spursistasrc/vm/cointerp.c M spursistasrc/vm/cointerp.h M spursistasrc/vm/gcc3x-cointerp.c M spursrc/vm/cogit.h M spursrc/vm/cogitARMv5.c M spursrc/vm/cogitIA32.c M spursrc/vm/cogitMIPSEL.c M spursrc/vm/cointerp.c M spursrc/vm/cointerp.h M spursrc/vm/gcc3x-cointerp.c M spurstack64src/vm/gcc3x-interp.c M spurstack64src/vm/interp.c M spurstacksrc/vm/gcc3x-interp.c M spurstacksrc/vm/interp.c M src/plugins/AsynchFilePlugin/AsynchFilePlugin.c M src/plugins/B2DPlugin/B2DPlugin.c M src/plugins/BitBltPlugin/BitBltPlugin.c M src/plugins/FilePlugin/FilePlugin.c M src/plugins/MiscPrimitivePlugin/MiscPrimitivePlugin.c M src/plugins/SqueakFFIPrims/ARM32FFIPlugin.c M src/plugins/SqueakFFIPrims/IA32FFIPlugin.c M src/plugins/SqueakFFIPrims/X64SysVFFIPlugin.c M src/plugins/SqueakFFIPrims/X64Win64FFIPlugin.c M src/vm/cogit.h M src/vm/cogitARMv5.c M src/vm/cogitIA32.c M src/vm/cogitMIPSEL.c M src/vm/cointerp.c M src/vm/cointerp.h M src/vm/cointerpmt.c M src/vm/cointerpmt.h M src/vm/gcc3x-cointerp.c M src/vm/gcc3x-cointerpmt.c M stacksrc/vm/gcc3x-interp.c M stacksrc/vm/interp.c Log Message: ----------- Merge commit '5a949459043392bbc093d78574c2d7c07b10fcb2' Commit: 7a7283139b72e56572f53caa167cb735a53ba949 https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/7a7283139b72e56572… Author: Esteban Lorenzano <estebanlm(a)gmail.com> Date: 2016-12-09 (Fri, 09 Dec 2016) Changed paths: M third-party/libpng.spec.win Log Message: ----------- fix appveyor build script to fail when… it fails :) fix png download url. Commit: c8d42164e557e5107f4452f0d1e6267f6e65feaa https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/c8d42164e557e5107f… Author: Esteban Lorenzano <estebanlm(a)gmail.com> Date: 2016-12-13 (Tue, 13 Dec 2016) Changed paths: M nsspur64src/vm/cogit.h M nsspur64src/vm/cogitX64.c M nsspur64src/vm/cointerp.c M nsspur64src/vm/cointerp.h M nsspur64src/vm/gcc3x-cointerp.c M nsspursrc/vm/cogit.h M nsspursrc/vm/cogitARMv5.c M nsspursrc/vm/cogitIA32.c M nsspursrc/vm/cogitMIPSEL.c M nsspursrc/vm/cointerp.c M nsspursrc/vm/cointerp.h M nsspursrc/vm/gcc3x-cointerp.c M nsspurstack64src/vm/gcc3x-interp.c M nsspurstack64src/vm/interp.c M nsspurstacksrc/vm/gcc3x-interp.c M nsspurstacksrc/vm/interp.c M spur64src/vm/cogit.h M spur64src/vm/cogitX64.c M spur64src/vm/cointerp.c M spur64src/vm/cointerp.h M spur64src/vm/gcc3x-cointerp.c M spursista64src/vm/cogit.h M spursista64src/vm/cogitX64.c M spursista64src/vm/cointerp.c M spursista64src/vm/cointerp.h M spursista64src/vm/gcc3x-cointerp.c M spursistasrc/vm/cogit.h M spursistasrc/vm/cogitARMv5.c M spursistasrc/vm/cogitIA32.c M spursistasrc/vm/cogitMIPSEL.c M spursistasrc/vm/cointerp.c M spursistasrc/vm/cointerp.h M spursistasrc/vm/gcc3x-cointerp.c M spursrc/vm/cogit.h M spursrc/vm/cogitARMv5.c M spursrc/vm/cogitIA32.c M spursrc/vm/cogitMIPSEL.c M spursrc/vm/cointerp.c M spursrc/vm/cointerp.h M spursrc/vm/gcc3x-cointerp.c M spurstack64src/vm/gcc3x-interp.c M spurstack64src/vm/interp.c M spurstacksrc/vm/gcc3x-interp.c M spurstacksrc/vm/interp.c M src/vm/cogit.h M src/vm/cogitARMv5.c M src/vm/cogitIA32.c M src/vm/cogitMIPSEL.c M src/vm/cointerp.c M src/vm/cointerp.h M src/vm/cointerpmt.c M src/vm/cointerpmt.h M src/vm/gcc3x-cointerp.c M src/vm/gcc3x-cointerpmt.c M stacksrc/vm/gcc3x-interp.c M stacksrc/vm/interp.c Log Message: ----------- Merge commit '8a77b2517d9ec3d467cb0f1ae427e94c4a5eeb10' Commit: 945fbc37d97f2db5ddcb1abfd14e9a1e7cb55c9c https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/945fbc37d97f2db5dd… Author: Esteban Lorenzano <estebanlm(a)gmail.com> Date: 2016-12-13 (Tue, 13 Dec 2016) Changed paths: M platforms/win32/vm/sqWin32Main.c M platforms/win32/vm/sqWin32Window.c Log Message: ----------- add double dash for Pharo on windows (the only remaining) Commit: 99fdc442dc59f3bbe3fc48b2a037c873b9529dcb https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/99fdc442dc59f3bbe3… Author: Esteban Lorenzano <estebanlm(a)gmail.com> Date: 2016-12-13 (Tue, 13 Dec 2016) Changed paths: M platforms/win32/vm/sqWin32Main.c M platforms/win32/vm/sqWin32Window.c M third-party/libpng.spec.win Log Message: ----------- Merge pull request #107 from estebanlm/Cog add double dash for Pharo on windows (the only remaining) Compare: https://github.com/OpenSmalltalk/opensmalltalk-vm/compare/8a77b2517d9e...99… ------------------------------ Message: 2 Date: Tue, 13 Dec 2016 22:13:18 0000 From: commits(a)source.squeak.org To: vm-dev(a)lists.squeakfoundation.org Subject: [Vm-dev] VM Maker: VMMaker.oscog-nice.2038.mcz Message-ID: <E1cGvKe-0003L2-V8@andreas> Nicolas Cellier uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-nice.2038.mcz ==================== Summary ==================== Name: VMMaker.oscog-nice.2038 Author: nice Time: 13 December 2016, 11:11:51.599692 pm UUID: f98d3e8c-e81d-fb42-b790-8e514777bfff Ancestors: VMMaker.oscog-eem.2037 Revise ALLOCA_LIES_SO_USE_GET_SP The extra space allocated below the value returned by alloca() is used by mingw32 gcc (it is used for marshalling args of sub-functions - those who marshall ffi call args) So we cannot use it for marshalling FFI arguments. Thus we must change the strategy: rather than using getsp() for changing the value returned by alloca() before marshalling args (which will lead to SEGV) we must rather use setsp() for changing the stack pointer before calling the target function. Thus we rename the macro into ALLOCA_LIES_SO_SET_SP_BEFORE_CALL Remind that __clang__'s alloca does not lie. (clang compiler does define __GNUC__) --- Revise FFI stack alignment: 1) use STACK_BYTES_ALIGNMENT and getsp() as provided by include file sqCogStackAlignment.h 2) use a new macro MUST_ALIGN_STACK that check STACK_BYTES_ALIGNMENT against sizeof(void *) indeed, some nice STACK_BYTES_ALIGNMENT might be > 0 now, which does not necessarily mean that we must realign the stack This macro is used both for aligning alloca'ed space before marshalling and for aligning SP before call. It is still questionable whether we must do some further alignment over alloca(), because alloca() should be aware of ABI requirements, but well, if it ain't broken... TO THINK ABOUT: maybe SP should be saved before call/reset after call. By now, this does not seem necessary. --- nuke STACK_OFFSET_BYTES from the preamble since it is unused. --- get rid of registerArgsSlop It was intended for reserving stack space for shadowing register args as mandated by Win64 X64 ABI (and maybe PPC). But this is completely unecessary as this stack space will be reserved by the callout itself (at least on WIN64). For PPC, don't do anything, it's currently unsupported (unmaintained/untested/unnecessary/you name it) --- Let cogit generate remove float args from stack only if more than 0 float args (by symmetry with int args) This was forgotten in a previous commit Note that this is only used by lowcode experiments and still wait to be ported to other architectures than IA32. =============== Diff against VMMaker.oscog-eem.2037 =============== Item was changed: ----- Method: CogIA32Compiler>>genRemoveNFloatArgsFromStack: (in category 'abi') ----- genRemoveNFloatArgsFromStack: n + n > 0 ifTrue: [cogit AddCq: n * 8 R: ESP]. - cogit AddCq: n * 8 R: ESP. ^0! Item was changed: ----- Method: ThreadedARMFFIPlugin>>ffiCalloutTo:SpecOnStack:in: (in category 'callout support') ----- ffiCalloutTo: procAddr SpecOnStack: specOnStack in: calloutState <var: #procAddr type: #'void *'> <var: #calloutState type: #'CalloutState *'> <var: #loadFloatRegs declareC: 'extern void loadFloatRegs(double, double, double, double, double, double, double, double)'> "Go out, call this guy and create the return value. This *must* be inlined because of the alloca of the outgoing stack frame in ffiCall:WithFlags:NumArgs:Args:AndTypes:" | myThreadIndex atomicType floatRet intRet loadFloatRegs oop | <var: #floatRet type: #double> <var: #intRet type: #usqLong> <inline: true> self cCode: '' inSmalltalk: [loadFloatRegs := #used. loadFloatRegs class]. self cppIf: COGMTVM ifTrue: [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: [myThreadIndex := interpreterProxy disownVM: 0]]. - self registerArgsSlop + self cStackAlignment > 0 ifTrue: - [self setsp: calloutState argVector]. - calloutState floatRegisterIndex > 0 ifTrue: [self load: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 0)) to: 'double *') at: 0) Flo: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 2)) to: 'double *') at: 0) a: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 4)) to: 'double *') at: 0) t: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 6)) to: 'double *') at: 0) R: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 8)) to: 'double *') at: 0) e: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 10)) to: 'double *') at: 0) g: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 12)) to: 'double *') at: 0) s: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 14)) to: 'double *') at: 0)]. + (self allocaLiesSoSetSpBeforeCall or: [self mustAlignStack]) ifTrue: + [self setsp: calloutState argVector]. + atomicType := self atomicTypeOf: calloutState ffiRetHeader. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [atomicType = FFITypeSingleFloat ifTrue: [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'float (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3)] ifFalse: "atomicType = FFITypeDoubleFloat" [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3)]] ifFalse: [intRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'usqIntptr_t (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3)]. "undo any callee argument pops because it may confuse stack management with the alloca." (self isCalleePopsConvention: calloutState callFlags) ifTrue: [self setsp: calloutState argVector]. self cppIf: COGMTVM ifTrue: [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: [interpreterProxy ownVM: myThreadIndex]]. (calloutState ffiRetHeader anyMask: FFIFlagPointer+FFIFlagStructure) ifTrue: ["Note: Order is important here since FFIFlagPointer + FFIFlagStructure is used to represent 'typedef void* VoidPointer' and VoidPointer must be returned as pointer *not* as struct." (calloutState ffiRetHeader anyMask: FFIFlagPointer) ifTrue: [oop := self ffiReturnPointer: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState] ifFalse: [oop := self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. ^oop]. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [oop := interpreterProxy floatObjectOf: floatRet] ifFalse: [oop := self ffiCreateIntegralResultOop: intRet ofAtomicType: atomicType in: calloutState]. ^interpreterProxy methodReturnValue: oop! Item was removed: - ----- Method: ThreadedARMFFIPlugin>>registerArgsSlop (in category 'marshalling') ----- - registerArgsSlop - "Answer any space needed to prevent the alloca'ed outgoing arguments marshalling area from - being overwritten by any register arguments during calls during marshalling. On ARM we - believe this is zero." - ^0! Item was changed: ----- Method: ThreadedFFIPlugin class>>preambleCCode (in category 'translation') ----- preambleCCode "For a source of builtin defines grep for builtin_define in a gcc release config directory." ^' #include "sqAssert.h" /* for assert */ #define ThreadedFFIPlugin 1 /* to filter-out unwanted declarations from sqFFI.h */ #include "sqFFI.h" /* for logging and surface functions */ + #include "sqCogStackAlignment.h" /* for STACK_ALIGN_BYTES and getsp() */ #ifdef _MSC_VER # define alloca _alloca #endif #if defined(__GNUC__) && (defined(_X86_) || defined(i386) || defined(__i386) || defined(__i386__)) # define setsp(sp) asm volatile ("movl %0,%%esp" : : "m"(sp)) - # define getsp() ({ void *esp; asm volatile ("movl %%esp,%0" : "=r"(esp) : ); esp;}) # elif defined(__GNUC__) && (defined(__amd64__) || defined(__x86_64__) || defined(__amd64) || defined(__x86_64)) # define setsp(sp) asm volatile ("movq %0,%%rsp" : : "m"(sp)) - # define getsp() ({ void *rsp; asm volatile ("movq %%rsp,%0" : "=r"(rsp) : ); rsp;}) # elif defined(__GNUC__) && (defined(__arm__)) # define setsp(sp) asm volatile ("ldr %%sp, %0" : : "m"(sp)) - # define getsp() ({ void *sp; asm volatile ("mov %0, %%sp" : "=r"(sp) : ); sp;}) #endif #if !!defined(getsp) # define getsp() 0 #endif #if !!defined(setsp) # define setsp(ignored) 0 #endif #if !!defined(STACK_ALIGN_BYTES) - # if __APPLE__ && __MACH__ && __i386__ - # define STACK_ALIGN_BYTES 16 - # elif __linux__ && __i386__ - # define STACK_ALIGN_BYTES 16 - # elif defined(__amd64__) || defined(__x86_64__) || defined(__amd64) || defined(__x86_64) - # define STACK_ALIGN_BYTES 16 - # elif defined(powerpc) || defined(__powerpc__) || defined(_POWER) || defined(__POWERPC__) || defined(__PPC__) - # define STACK_ALIGN_BYTES 16 - # elif defined(__sparc64__) || defined(__sparcv9__) || defined(__sparc_v9__) /* must precede 32-bit sparc defs */ - # define STACK_ALIGN_BYTES 16 - # elif defined(sparc) || defined(__sparc__) || defined(__sparclite__) - # define STACK_ALIGN_BYTES 8 - # elif defined(__arm__) - # define STACK_ALIGN_BYTES 8 - # else # define STACK_ALIGN_BYTES 0 - # endif #endif /* !!defined(STACK_ALIGN_BYTES) */ + /* For ABI that require stack alignment greater than natural word size */ + #define MUST_ALIGN_STACK (STACK_ALIGN_BYTES > sizeof(void*)) - #if !!defined(STACK_OFFSET_BYTES) - # define STACK_OFFSET_BYTES 0 - #endif #if defined(_X86_) || defined(i386) || defined(__i386) || defined(__i386__) /* Both Mac OS X x86 and Win32 x86 return structs of a power of two in size * less than or equal to eight bytes in length in registers. Linux never does so. */ # if __linux__ # define WIN32_X86_STRUCT_RETURN 0 # else # define WIN32_X86_STRUCT_RETURN 1 # endif # if WIN32 # define PLATFORM_API_USES_CALLEE_POPS_CONVENTION 1 # endif # elif defined(__amd64__) || defined(__x86_64__) || defined(__amd64) || defined(__x86_64) # if WIN32 | WIN64 # define PLATFORM_API_USES_CALLEE_POPS_CONVENTION 1 # endif #endif /* defined(_X86_) || defined(i386) || defined(__i386) || defined(__i386__) */ + #if !!defined(ALLOCA_LIES_SO_SETSP_BEFORE_CALL) + # if defined(__MINGW32__) && !!defined(__clang__) && (__GNUC__ >= 3) && (defined(_X86_) || defined(i386) || defined(__i386) || defined(__i386__)) - #if !!defined(ALLOCA_LIES_SO_USE_GETSP) - # if defined(__MINGW32__) && (__GNUC__ >= 3) && (defined(_X86_) || defined(i386) || defined(__i386) || defined(__i386__)) /* * cygwin -mno-cygwin (MinGW) gcc 3.4.x''s alloca is a library routine that answers + * %esp + xx, so the outgoing stack is offset by one or more word if uncorrected. - * %esp + 4, so the outgoing stack is offset by one word if uncorrected. * Grab the actual stack pointer to correct. */ + # define ALLOCA_LIES_SO_SETSP_BEFORE_CALL 1 - # define ALLOCA_LIES_SO_USE_GETSP 1 # else + # define ALLOCA_LIES_SO_SETSP_BEFORE_CALL 0 - # define ALLOCA_LIES_SO_USE_GETSP 0 # endif + #endif /* !!defined(ALLOCA_LIES_SO_SETSP_BEFORE_CALL) */ - #endif /* !!defined(ALLOCA_LIES_SO_USE_GETSP) */ #if !!defined(PLATFORM_API_USES_CALLEE_POPS_CONVENTION) # define PLATFORM_API_USES_CALLEE_POPS_CONVENTION 0 #endif /* The dispatchOn:in:with:with: generates an unwanted call on error. Just squash it. */ #define error(foo) 0 #ifndef SQUEAK_BUILTIN_PLUGIN /* but print assert failures. */ void warning(char *s) { /* Print an error message but don''t exit. */ printf("\n%s\n", s); } #endif /* sanitize */ #ifdef SQUEAK_BUILTIN_PLUGIN # define EXTERN #else # define EXTERN extern #endif '! Item was added: + ----- Method: ThreadedFFIPlugin>>allocaLiesSoSetSpBeforeCall (in category 'marshalling') ----- + allocaLiesSoSetSpBeforeCall + "At least one alloca implementation does not answer the actual top of stack. + If so we need to reset the actual stack pointer just before the call. + Answer whether this is necessary." + <cmacro: '() ALLOCA_LIES_SO_SETSP_BEFORE_CALL'> + ^false! Item was removed: - ----- Method: ThreadedFFIPlugin>>allocaLiesSoUseGetsp (in category 'marshalling') ----- - allocaLiesSoUseGetsp - "At least one alloca implementation does not answer the actual top of stack. - If so we need to get the actual stack pointer. Answer whether this is necessary." - <cmacro: '() ALLOCA_LIES_SO_USE_GETSP'> - ^false! Item was changed: ----- Method: ThreadedFFIPlugin>>ffiCall:ArgArrayOrNil:NumArgs: (in category 'callout support') ----- ffiCall: externalFunction ArgArrayOrNil: argArrayOrNil NumArgs: nArgs "Generic callout. Does the actual work. If argArrayOrNil is nil it takes args from the stack and the spec from the method. If argArrayOrNil is not nil takes args from argArrayOrNil and the spec from the receiver." | flags argTypeArray address argType oop argSpec argClass err theCalloutState calloutState requiredStackSize stackSize allocation result | <inline: true> <var: #theCalloutState type: #'CalloutState'> <var: #calloutState type: #'CalloutState *'> <var: #allocation type: #'char *'> (interpreterProxy is: externalFunction KindOfClass: interpreterProxy classExternalFunction) ifFalse: [^self ffiFail: FFIErrorNotFunction]. "Load and check the values in the externalFunction before we call out" flags := interpreterProxy fetchInteger: ExternalFunctionFlagsIndex ofObject: externalFunction. interpreterProxy failed ifTrue: [^self ffiFail: FFIErrorBadArgs]. "This must come early for compatibility with the old FFIPlugin. Image-level code may assume the function pointer is loaded eagerly. Thanks to Nicolas Cellier." address := self ffiLoadCalloutAddress: externalFunction. interpreterProxy failed ifTrue: [^0 "error code already set by ffiLoadCalloutAddress:"]. argTypeArray := interpreterProxy fetchPointer: ExternalFunctionArgTypesIndex ofObject: externalFunction. "must be array of arg types" ((interpreterProxy isArray: argTypeArray) and: [(interpreterProxy slotSizeOf: argTypeArray) = (nArgs + 1)]) ifFalse: [^self ffiFail: FFIErrorBadArgs]. "check if the calling convention is supported" self cppIf: COGMTVM ifTrue: [(self ffiSupportsCallingConvention: (flags bitAnd: FFICallTypesMask)) ifFalse: [^self ffiFail: FFIErrorCallType]] ifFalse: "not masking causes threaded calls to fail, which is as they should if the plugin is not threaded." [(self ffiSupportsCallingConvention: flags) ifFalse: [^self ffiFail: FFIErrorCallType]]. requiredStackSize := self externalFunctionHasStackSizeSlot ifTrue: [interpreterProxy fetchInteger: ExternalFunctionStackSizeIndex ofObject: externalFunction] ifFalse: [-1]. interpreterProxy failed ifTrue: [^interpreterProxy primitiveFailFor: (argArrayOrNil isNil ifTrue: [PrimErrBadMethod] ifFalse: [PrimErrBadReceiver])]. stackSize := requiredStackSize < 0 ifTrue: [DefaultMaxStackSize] ifFalse: [requiredStackSize]. self cCode: [] inSmalltalk: [theCalloutState := self class calloutStateClass new]. calloutState := self addressOf: theCalloutState. self cCode: [self me: calloutState ms: 0 et: (self sizeof: #CalloutState)]. calloutState callFlags: flags. "Fetch return type and args" argType := interpreterProxy fetchPointer: 0 ofObject: argTypeArray. argSpec := interpreterProxy fetchPointer: 0 ofObject: argType. argClass := interpreterProxy fetchPointer: 1 ofObject: argType. (err := self ffiCheckReturn: argSpec With: argClass in: calloutState) ~= 0 ifTrue: [^self ffiFail: err]. "cannot return" + "alloca the outgoing stack frame, leaving room for marshalling args, and including space for the return struct, if any. + Additional space reserved for saving register args like mandated by Win64 X64 or PPC ABI, will be managed by the call itself" + allocation := self alloca: stackSize + calloutState structReturnSize + self cStackAlignment. + self mustAlignStack ifTrue: + [allocation := self cCoerce: (allocation asUnsignedIntegerPtr bitClear: self cStackAlignment - 1) - "alloca the outgoing stack frame, leaving room for register args while marshalling, and including space for the return struct, if any." - allocation := self alloca: stackSize + calloutState structReturnSize + self registerArgsSlop + self cStackAlignment. - self allocaLiesSoUseGetsp ifTrue: - [allocation := self getsp]. - self cStackAlignment ~= 0 ifTrue: - [allocation := self cCoerce: (allocation asUnsignedInteger bitClear: self cStackAlignment - 1) to: #'char *']. calloutState argVector: allocation; + currentArg: allocation; + limit: allocation + stackSize. - currentArg: allocation + self registerArgsSlop; - limit: allocation + stackSize + self registerArgsSlop. (calloutState structReturnSize > 0 and: [self nonRegisterStructReturnIsViaImplicitFirstArgument and: [(self returnStructInRegisters: calloutState structReturnSize) not]]) ifTrue: [err := self ffiPushPointer: calloutState limit in: calloutState. err ~= 0 ifTrue: [self cleanupCalloutState: calloutState. self cppIf: COGMTVM ifTrue: [err = PrimErrObjectMayMove negated ifTrue: [^PrimErrObjectMayMove]]. "N.B. Do not fail if object may move because caller will GC and retry." ^self ffiFail: err]]. 1 to: nArgs do: [:i| argType := interpreterProxy fetchPointer: i ofObject: argTypeArray. argSpec := interpreterProxy fetchPointer: 0 ofObject: argType. argClass := interpreterProxy fetchPointer: 1 ofObject: argType. oop := argArrayOrNil isNil ifTrue: [interpreterProxy stackValue: nArgs - i] ifFalse: [interpreterProxy fetchPointer: i - 1 ofObject: argArrayOrNil]. err := self ffiArgument: oop Spec: argSpec Class: argClass in: calloutState. err ~= 0 ifTrue: [self cleanupCalloutState: calloutState. self cppIf: COGMTVM ifTrue: [err = PrimErrObjectMayMove negated ifTrue: [^PrimErrObjectMayMove]]. "N.B. Do not fail if object may move because caller will GC and retry." ^self ffiFail: err]]. "coercion failed or out of stack space" "Failures must be reported back from ffiArgument:Spec:Class:in:. Should not fail from here on in." self assert: interpreterProxy failed not. self ffiLogCallout: externalFunction. (requiredStackSize < 0 and: [self externalFunctionHasStackSizeSlot]) ifTrue: [stackSize := calloutState currentArg - calloutState argVector. interpreterProxy storeInteger: ExternalFunctionStackSizeIndex ofObject: externalFunction withValue: stackSize]. "Go out and call this guy" result := self ffiCalloutTo: address SpecOnStack: argArrayOrNil notNil in: calloutState. self cleanupCalloutState: calloutState. ^result! Item was added: + ----- Method: ThreadedFFIPlugin>>mustAlignStack (in category 'marshalling') ----- + mustAlignStack + "Many ABIs mandate a particular stack alignment greater than the natural word size. + If so, this macro will answer true. See class-side preambleCCode." + <cmacro: '() MUST_ALIGN_STACK'> + ^0! Item was removed: - ----- Method: ThreadedFFIPlugin>>registerArgsSlop (in category 'marshalling') ----- - registerArgsSlop - "Answer any space needed to prevent the alloca'ed outgoing arguments marshalling area from - being overwritten by any register arguments during calls during marshalling. For example, on - PowerPC, which has 8 register arguments in the calling convention, register arguments are also - written to the stack. So unless space is left for them, calls during marshalling prior to the actual - callout (e.g. to interpreterProxy object manipulation routines) can end up overwriting the - marshalling stack as register arguments are written to the stack during calls." - self subclassResponsibility! Item was changed: ----- Method: ThreadedIA32FFIPlugin>>ffiCalloutTo:SpecOnStack:in: (in category 'callout support') ----- ffiCalloutTo: procAddr SpecOnStack: specOnStack in: calloutState <var: #procAddr type: #'void *'> <var: #calloutState type: #'CalloutState *'> "Go out, call this guy and create the return value. This *must* be inlined because of the alloca of the outgoing stack frame in ffiCall:WithFlags:NumArgs:Args:AndTypes:" | myThreadIndex atomicType floatRet intRet oop | <var: #floatRet type: #double> <var: #intRet type: #usqLong> <inline: true> self cppIf: COGMTVM ifTrue: [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: [myThreadIndex := interpreterProxy disownVM: 0]]. + (self allocaLiesSoSetSpBeforeCall or: [self mustAlignStack]) ifTrue: - self registerArgsSlop + self cStackAlignment > 0 ifTrue: [self setsp: calloutState argVector]. atomicType := self atomicTypeOf: calloutState ffiRetHeader. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)()')] ifFalse: [intRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'usqLong (*)()')]. "undo any callee argument pops because it may confuse stack management with the alloca." (self isCalleePopsConvention: calloutState callFlags) ifTrue: [self setsp: calloutState argVector]. self cppIf: COGMTVM ifTrue: [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: [interpreterProxy ownVM: myThreadIndex]]. (calloutState ffiRetHeader anyMask: FFIFlagPointer+FFIFlagStructure) ifTrue: ["Note: Order is important here since FFIFlagPointer + FFIFlagStructure is used to represent 'typedef void* VoidPointer' and VoidPointer must be returned as pointer *not* as struct." (calloutState ffiRetHeader anyMask: FFIFlagPointer) ifTrue: [oop := self ffiReturnPointer: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState] ifFalse: [oop := self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. ^oop]. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [oop := interpreterProxy floatObjectOf: floatRet] ifFalse: [oop := self ffiCreateIntegralResultOop: intRet ofAtomicType: atomicType in: calloutState]. ^interpreterProxy methodReturnValue: oop! Item was removed: - ----- Method: ThreadedIA32FFIPlugin>>registerArgsSlop (in category 'marshalling') ----- - registerArgsSlop - "Answer any space needed to prevent the alloca'ed outgoing arguments marshalling area from - being overwritten by any register arguments during calls during marshalling. On x86 this is 0" - ^0! Item was removed: - ----- Method: ThreadedPPCBEFFIPlugin>>registerArgsSlop (in category 'marshalling') ----- - registerArgsSlop - "Answer any space needed to prevent the alloca'ed outgoing arguments marshalling area from - being overwritten by any register arguments during calls during marshalling. On PowerPC, which - has 8 register arguments in the calling convention, register arguments are also written to the stack. - So we must leave room for 8 * 4 bytes to avoid overwriting the marshalling stack as register - arguments are written to the stack during calls to interpreterProxy etc." - ^32! Item was removed: - ----- Method: ThreadedX64FFIPlugin>>registerArgsSlop (in category 'marshalling') ----- - registerArgsSlop - "Answer any space needed to prevent the alloca'ed outgoing arguments marshalling area from - being overwritten by any register arguments during calls during marshalling. On ARM we - believe this is zero." - ^0! Item was changed: ----- Method: ThreadedX64SysVFFIPlugin>>ffiCalloutTo:SpecOnStack:in: (in category 'callout support') ----- ffiCalloutTo: procAddr SpecOnStack: specOnStack in: calloutState <var: #procAddr type: #'void *'> <var: #calloutState type: #'CalloutState *'> <var: #loadFloatRegs declareC: 'extern void loadFloatRegs(double, double, double, double, double, double, double, double)'> "Go out, call this guy and create the return value. This *must* be inlined because of the alloca of the outgoing stack frame in ffiCall:WithFlags:NumArgs:Args:AndTypes:" | myThreadIndex atomicType floatRet intRet loadFloatRegs oop | <var: #floatRet type: #double> <var: #intRet type: 'SixteenByteReturn'> <inline: true> self cCode: '' inSmalltalk: [loadFloatRegs := #used. loadFloatRegs class]. self cppIf: COGMTVM ifTrue: [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: [myThreadIndex := interpreterProxy disownVM: 0]]. - self registerArgsSlop + self cStackAlignment > 0 ifTrue: - [self setsp: calloutState argVector]. - calloutState floatRegisterIndex > 0 ifTrue: [self load: (calloutState floatRegisters at: 0) Flo: (calloutState floatRegisters at: 1) a: (calloutState floatRegisters at: 2) t: (calloutState floatRegisters at: 3) R: (calloutState floatRegisters at: 4) e: (calloutState floatRegisters at: 5) g: (calloutState floatRegisters at: 6) s: (calloutState floatRegisters at: 7)]. + (self allocaLiesSoSetSpBeforeCall or: [self mustAlignStack]) ifTrue: + [self setsp: calloutState argVector]. + atomicType := self atomicTypeOf: calloutState ffiRetHeader. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [atomicType = FFITypeSingleFloat ifTrue: [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'float (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3) with: (calloutState integerRegisters at: 4) with: (calloutState integerRegisters at: 5)] ifFalse: "atomicType = FFITypeDoubleFloat" [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3) with: (calloutState integerRegisters at: 4) with: (calloutState integerRegisters at: 5)]] ifFalse: [intRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'SixteenByteReturn (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3) with: (calloutState integerRegisters at: 4) with: (calloutState integerRegisters at: 5)]. "undo any callee argument pops because it may confuse stack management with the alloca." (self isCalleePopsConvention: calloutState callFlags) ifTrue: [self setsp: calloutState argVector]. self cppIf: COGMTVM ifTrue: [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: [interpreterProxy ownVM: myThreadIndex]]. (calloutState ffiRetHeader anyMask: FFIFlagPointer+FFIFlagStructure) ifTrue: ["Note: Order is important here since FFIFlagPointer + FFIFlagStructure is used to represent 'typedef void* VoidPointer' and VoidPointer must be returned as pointer *not* as struct." (calloutState ffiRetHeader anyMask: FFIFlagPointer) ifTrue: [oop := self ffiReturnPointer: intRet a ofType: (self ffiReturnType: specOnStack) in: calloutState] ifFalse: [oop := self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. ^oop]. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [oop := interpreterProxy floatObjectOf: floatRet] ifFalse: [oop := self ffiCreateIntegralResultOop: intRet a ofAtomicType: atomicType in: calloutState]. ^interpreterProxy methodReturnValue: oop! Item was changed: ----- Method: ThreadedX64Win64FFIPlugin>>ffiCalloutTo:SpecOnStack:in: (in category 'callout support') ----- ffiCalloutTo: procAddr SpecOnStack: specOnStack in: calloutState <var: #procAddr type: #'void *'> <var: #calloutState type: #'CalloutState *'> <var: #loadFloatRegs declareC: 'extern void loadFloatRegs(double, double, double, double)'> "Go out, call this guy and create the return value. This *must* be inlined because of the alloca of the outgoing stack frame in ffiCall:WithFlags:NumArgs:Args:AndTypes:" | myThreadIndex atomicType floatRet intRet loadFloatRegs oop | <var: #floatRet type: #double> <var: #intRet type: #usqLong> <inline: true> self cCode: '' inSmalltalk: [loadFloatRegs := #used. loadFloatRegs class]. self cppIf: COGMTVM ifTrue: [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: [myThreadIndex := interpreterProxy disownVM: 0]]. - self registerArgsSlop + self cStackAlignment > 0 ifTrue: - [self setsp: calloutState argVector]. - calloutState floatRegisterSignature > 0 ifTrue: [self load: (calloutState floatRegisters at: 0) Flo: (calloutState floatRegisters at: 1) at: (calloutState floatRegisters at: 2) Re: (calloutState floatRegisters at: 3) gs: (calloutState floatRegisters at: 4)]. + (self allocaLiesSoSetSpBeforeCall or: [self mustAlignStack]) ifTrue: + [self setsp: calloutState argVector]. + atomicType := self atomicTypeOf: calloutState ffiRetHeader. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [atomicType = FFITypeSingleFloat ifTrue: [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'float (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3)] ifFalse: "atomicType = FFITypeDoubleFloat" [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3)]] ifFalse: [intRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'usqIntptr_t (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3)]. "undo any callee argument pops because it may confuse stack management with the alloca." (self isCalleePopsConvention: calloutState callFlags) ifTrue: [self setsp: calloutState argVector]. self cppIf: COGMTVM ifTrue: [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: [interpreterProxy ownVM: myThreadIndex]]. (calloutState ffiRetHeader anyMask: FFIFlagPointer+FFIFlagStructure) ifTrue: ["Note: Order is important here since FFIFlagPointer + FFIFlagStructure is used to represent 'typedef void* VoidPointer' and VoidPointer must be returned as pointer *not* as struct." (calloutState ffiRetHeader anyMask: FFIFlagPointer) ifTrue: [oop := self ffiReturnPointer: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState] ifFalse: [oop := self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. ^oop]. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [oop := interpreterProxy floatObjectOf: floatRet] ifFalse: [oop := self ffiCreateIntegralResultOop: intRet ofAtomicType: atomicType in: calloutState]. ^interpreterProxy methodReturnValue: oop! ------------------------------ Subject: Digest Footer _______________________________________________ Vm-dev mailing list Vm-dev(a)lists.squeakfoundation.org http://lists.squeakfoundation.org/mailman/listinfo/vm-dev ------------------------------ End of Vm-dev Digest, Vol 126, Issue 29 ***************************************

1 0

VM Maker: VMMaker.oscog-nice.2038.mcz
by commits＠source.squeak.org 14 Dec '16

14 Dec '16

Nicolas Cellier uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-nice.2038.mcz ==================== Summary ==================== Name: VMMaker.oscog-nice.2038 Author: nice Time: 13 December 2016, 11:11:51.599692 pm UUID: f98d3e8c-e81d-fb42-b790-8e514777bfff Ancestors: VMMaker.oscog-eem.2037 Revise ALLOCA_LIES_SO_USE_GET_SP The extra space allocated below the value returned by alloca() is used by mingw32 gcc (it is used for marshalling args of sub-functions - those who marshall ffi call args) So we cannot use it for marshalling FFI arguments. Thus we must change the strategy: rather than using getsp() for changing the value returned by alloca() before marshalling args (which will lead to SEGV) we must rather use setsp() for changing the stack pointer before calling the target function. Thus we rename the macro into ALLOCA_LIES_SO_SET_SP_BEFORE_CALL Remind that __clang__'s alloca does not lie. (clang compiler does define __GNUC__) --- Revise FFI stack alignment: 1) use STACK_BYTES_ALIGNMENT and getsp() as provided by include file sqCogStackAlignment.h 2) use a new macro MUST_ALIGN_STACK that check STACK_BYTES_ALIGNMENT against sizeof(void *) indeed, some nice STACK_BYTES_ALIGNMENT might be > 0 now, which does not necessarily mean that we must realign the stack This macro is used both for aligning alloca'ed space before marshalling and for aligning SP before call. It is still questionable whether we must do some further alignment over alloca(), because alloca() should be aware of ABI requirements, but well, if it ain't broken... TO THINK ABOUT: maybe SP should be saved before call/reset after call. By now, this does not seem necessary. --- nuke STACK_OFFSET_BYTES from the preamble since it is unused. --- get rid of registerArgsSlop It was intended for reserving stack space for shadowing register args as mandated by Win64 X64 ABI (and maybe PPC). But this is completely unecessary as this stack space will be reserved by the callout itself (at least on WIN64). For PPC, don't do anything, it's currently unsupported (unmaintained/untested/unnecessary/you name it) --- Let cogit generate remove float args from stack only if more than 0 float args (by symmetry with int args) This was forgotten in a previous commit Note that this is only used by lowcode experiments and still wait to be ported to other architectures than IA32. =============== Diff against VMMaker.oscog-eem.2037 =============== Item was changed: ----- Method: CogIA32Compiler>>genRemoveNFloatArgsFromStack: (in category 'abi') ----- genRemoveNFloatArgsFromStack: n + n > 0 ifTrue: [cogit AddCq: n * 8 R: ESP]. - cogit AddCq: n * 8 R: ESP. ^0! Item was changed: ----- Method: ThreadedARMFFIPlugin>>ffiCalloutTo:SpecOnStack:in: (in category 'callout support') ----- ffiCalloutTo: procAddr SpecOnStack: specOnStack in: calloutState <var: #procAddr type: #'void *'> <var: #calloutState type: #'CalloutState *'> <var: #loadFloatRegs declareC: 'extern void loadFloatRegs(double, double, double, double, double, double, double, double)'> "Go out, call this guy and create the return value. This *must* be inlined because of the alloca of the outgoing stack frame in ffiCall:WithFlags:NumArgs:Args:AndTypes:" | myThreadIndex atomicType floatRet intRet loadFloatRegs oop | <var: #floatRet type: #double> <var: #intRet type: #usqLong> <inline: true> self cCode: '' inSmalltalk: [loadFloatRegs := #used. loadFloatRegs class]. self cppIf: COGMTVM ifTrue: [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: [myThreadIndex := interpreterProxy disownVM: 0]]. - self registerArgsSlop + self cStackAlignment > 0 ifTrue: - [self setsp: calloutState argVector]. - calloutState floatRegisterIndex > 0 ifTrue: [self load: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 0)) to: 'double *') at: 0) Flo: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 2)) to: 'double *') at: 0) a: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 4)) to: 'double *') at: 0) t: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 6)) to: 'double *') at: 0) R: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 8)) to: 'double *') at: 0) e: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 10)) to: 'double *') at: 0) g: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 12)) to: 'double *') at: 0) s: ((self cCoerceSimple: (self addressOf: (calloutState floatRegisters at: 14)) to: 'double *') at: 0)]. + (self allocaLiesSoSetSpBeforeCall or: [self mustAlignStack]) ifTrue: + [self setsp: calloutState argVector]. + atomicType := self atomicTypeOf: calloutState ffiRetHeader. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [atomicType = FFITypeSingleFloat ifTrue: [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'float (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3)] ifFalse: "atomicType = FFITypeDoubleFloat" [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3)]] ifFalse: [intRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'usqIntptr_t (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3)]. "undo any callee argument pops because it may confuse stack management with the alloca." (self isCalleePopsConvention: calloutState callFlags) ifTrue: [self setsp: calloutState argVector]. self cppIf: COGMTVM ifTrue: [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: [interpreterProxy ownVM: myThreadIndex]]. (calloutState ffiRetHeader anyMask: FFIFlagPointer+FFIFlagStructure) ifTrue: ["Note: Order is important here since FFIFlagPointer + FFIFlagStructure is used to represent 'typedef void* VoidPointer' and VoidPointer must be returned as pointer *not* as struct." (calloutState ffiRetHeader anyMask: FFIFlagPointer) ifTrue: [oop := self ffiReturnPointer: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState] ifFalse: [oop := self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. ^oop]. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [oop := interpreterProxy floatObjectOf: floatRet] ifFalse: [oop := self ffiCreateIntegralResultOop: intRet ofAtomicType: atomicType in: calloutState]. ^interpreterProxy methodReturnValue: oop! Item was removed: - ----- Method: ThreadedARMFFIPlugin>>registerArgsSlop (in category 'marshalling') ----- - registerArgsSlop - "Answer any space needed to prevent the alloca'ed outgoing arguments marshalling area from - being overwritten by any register arguments during calls during marshalling. On ARM we - believe this is zero." - ^0! Item was changed: ----- Method: ThreadedFFIPlugin class>>preambleCCode (in category 'translation') ----- preambleCCode "For a source of builtin defines grep for builtin_define in a gcc release config directory." ^' #include "sqAssert.h" /* for assert */ #define ThreadedFFIPlugin 1 /* to filter-out unwanted declarations from sqFFI.h */ #include "sqFFI.h" /* for logging and surface functions */ + #include "sqCogStackAlignment.h" /* for STACK_ALIGN_BYTES and getsp() */ #ifdef _MSC_VER # define alloca _alloca #endif #if defined(__GNUC__) && (defined(_X86_) || defined(i386) || defined(__i386) || defined(__i386__)) # define setsp(sp) asm volatile ("movl %0,%%esp" : : "m"(sp)) - # define getsp() ({ void *esp; asm volatile ("movl %%esp,%0" : "=r"(esp) : ); esp;}) # elif defined(__GNUC__) && (defined(__amd64__) || defined(__x86_64__) || defined(__amd64) || defined(__x86_64)) # define setsp(sp) asm volatile ("movq %0,%%rsp" : : "m"(sp)) - # define getsp() ({ void *rsp; asm volatile ("movq %%rsp,%0" : "=r"(rsp) : ); rsp;}) # elif defined(__GNUC__) && (defined(__arm__)) # define setsp(sp) asm volatile ("ldr %%sp, %0" : : "m"(sp)) - # define getsp() ({ void *sp; asm volatile ("mov %0, %%sp" : "=r"(sp) : ); sp;}) #endif #if !!defined(getsp) # define getsp() 0 #endif #if !!defined(setsp) # define setsp(ignored) 0 #endif #if !!defined(STACK_ALIGN_BYTES) - # if __APPLE__ && __MACH__ && __i386__ - # define STACK_ALIGN_BYTES 16 - # elif __linux__ && __i386__ - # define STACK_ALIGN_BYTES 16 - # elif defined(__amd64__) || defined(__x86_64__) || defined(__amd64) || defined(__x86_64) - # define STACK_ALIGN_BYTES 16 - # elif defined(powerpc) || defined(__powerpc__) || defined(_POWER) || defined(__POWERPC__) || defined(__PPC__) - # define STACK_ALIGN_BYTES 16 - # elif defined(__sparc64__) || defined(__sparcv9__) || defined(__sparc_v9__) /* must precede 32-bit sparc defs */ - # define STACK_ALIGN_BYTES 16 - # elif defined(sparc) || defined(__sparc__) || defined(__sparclite__) - # define STACK_ALIGN_BYTES 8 - # elif defined(__arm__) - # define STACK_ALIGN_BYTES 8 - # else # define STACK_ALIGN_BYTES 0 - # endif #endif /* !!defined(STACK_ALIGN_BYTES) */ + /* For ABI that require stack alignment greater than natural word size */ + #define MUST_ALIGN_STACK (STACK_ALIGN_BYTES > sizeof(void*)) - #if !!defined(STACK_OFFSET_BYTES) - # define STACK_OFFSET_BYTES 0 - #endif #if defined(_X86_) || defined(i386) || defined(__i386) || defined(__i386__) /* Both Mac OS X x86 and Win32 x86 return structs of a power of two in size * less than or equal to eight bytes in length in registers. Linux never does so. */ # if __linux__ # define WIN32_X86_STRUCT_RETURN 0 # else # define WIN32_X86_STRUCT_RETURN 1 # endif # if WIN32 # define PLATFORM_API_USES_CALLEE_POPS_CONVENTION 1 # endif # elif defined(__amd64__) || defined(__x86_64__) || defined(__amd64) || defined(__x86_64) # if WIN32 | WIN64 # define PLATFORM_API_USES_CALLEE_POPS_CONVENTION 1 # endif #endif /* defined(_X86_) || defined(i386) || defined(__i386) || defined(__i386__) */ + #if !!defined(ALLOCA_LIES_SO_SETSP_BEFORE_CALL) + # if defined(__MINGW32__) && !!defined(__clang__) && (__GNUC__ >= 3) && (defined(_X86_) || defined(i386) || defined(__i386) || defined(__i386__)) - #if !!defined(ALLOCA_LIES_SO_USE_GETSP) - # if defined(__MINGW32__) && (__GNUC__ >= 3) && (defined(_X86_) || defined(i386) || defined(__i386) || defined(__i386__)) /* * cygwin -mno-cygwin (MinGW) gcc 3.4.x''s alloca is a library routine that answers + * %esp + xx, so the outgoing stack is offset by one or more word if uncorrected. - * %esp + 4, so the outgoing stack is offset by one word if uncorrected. * Grab the actual stack pointer to correct. */ + # define ALLOCA_LIES_SO_SETSP_BEFORE_CALL 1 - # define ALLOCA_LIES_SO_USE_GETSP 1 # else + # define ALLOCA_LIES_SO_SETSP_BEFORE_CALL 0 - # define ALLOCA_LIES_SO_USE_GETSP 0 # endif + #endif /* !!defined(ALLOCA_LIES_SO_SETSP_BEFORE_CALL) */ - #endif /* !!defined(ALLOCA_LIES_SO_USE_GETSP) */ #if !!defined(PLATFORM_API_USES_CALLEE_POPS_CONVENTION) # define PLATFORM_API_USES_CALLEE_POPS_CONVENTION 0 #endif /* The dispatchOn:in:with:with: generates an unwanted call on error. Just squash it. */ #define error(foo) 0 #ifndef SQUEAK_BUILTIN_PLUGIN /* but print assert failures. */ void warning(char *s) { /* Print an error message but don''t exit. */ printf("\n%s\n", s); } #endif /* sanitize */ #ifdef SQUEAK_BUILTIN_PLUGIN # define EXTERN #else # define EXTERN extern #endif '! Item was added: + ----- Method: ThreadedFFIPlugin>>allocaLiesSoSetSpBeforeCall (in category 'marshalling') ----- + allocaLiesSoSetSpBeforeCall + "At least one alloca implementation does not answer the actual top of stack. + If so we need to reset the actual stack pointer just before the call. + Answer whether this is necessary." + <cmacro: '() ALLOCA_LIES_SO_SETSP_BEFORE_CALL'> + ^false! Item was removed: - ----- Method: ThreadedFFIPlugin>>allocaLiesSoUseGetsp (in category 'marshalling') ----- - allocaLiesSoUseGetsp - "At least one alloca implementation does not answer the actual top of stack. - If so we need to get the actual stack pointer. Answer whether this is necessary." - <cmacro: '() ALLOCA_LIES_SO_USE_GETSP'> - ^false! Item was changed: ----- Method: ThreadedFFIPlugin>>ffiCall:ArgArrayOrNil:NumArgs: (in category 'callout support') ----- ffiCall: externalFunction ArgArrayOrNil: argArrayOrNil NumArgs: nArgs "Generic callout. Does the actual work. If argArrayOrNil is nil it takes args from the stack and the spec from the method. If argArrayOrNil is not nil takes args from argArrayOrNil and the spec from the receiver." | flags argTypeArray address argType oop argSpec argClass err theCalloutState calloutState requiredStackSize stackSize allocation result | <inline: true> <var: #theCalloutState type: #'CalloutState'> <var: #calloutState type: #'CalloutState *'> <var: #allocation type: #'char *'> (interpreterProxy is: externalFunction KindOfClass: interpreterProxy classExternalFunction) ifFalse: [^self ffiFail: FFIErrorNotFunction]. "Load and check the values in the externalFunction before we call out" flags := interpreterProxy fetchInteger: ExternalFunctionFlagsIndex ofObject: externalFunction. interpreterProxy failed ifTrue: [^self ffiFail: FFIErrorBadArgs]. "This must come early for compatibility with the old FFIPlugin. Image-level code may assume the function pointer is loaded eagerly. Thanks to Nicolas Cellier." address := self ffiLoadCalloutAddress: externalFunction. interpreterProxy failed ifTrue: [^0 "error code already set by ffiLoadCalloutAddress:"]. argTypeArray := interpreterProxy fetchPointer: ExternalFunctionArgTypesIndex ofObject: externalFunction. "must be array of arg types" ((interpreterProxy isArray: argTypeArray) and: [(interpreterProxy slotSizeOf: argTypeArray) = (nArgs + 1)]) ifFalse: [^self ffiFail: FFIErrorBadArgs]. "check if the calling convention is supported" self cppIf: COGMTVM ifTrue: [(self ffiSupportsCallingConvention: (flags bitAnd: FFICallTypesMask)) ifFalse: [^self ffiFail: FFIErrorCallType]] ifFalse: "not masking causes threaded calls to fail, which is as they should if the plugin is not threaded." [(self ffiSupportsCallingConvention: flags) ifFalse: [^self ffiFail: FFIErrorCallType]]. requiredStackSize := self externalFunctionHasStackSizeSlot ifTrue: [interpreterProxy fetchInteger: ExternalFunctionStackSizeIndex ofObject: externalFunction] ifFalse: [-1]. interpreterProxy failed ifTrue: [^interpreterProxy primitiveFailFor: (argArrayOrNil isNil ifTrue: [PrimErrBadMethod] ifFalse: [PrimErrBadReceiver])]. stackSize := requiredStackSize < 0 ifTrue: [DefaultMaxStackSize] ifFalse: [requiredStackSize]. self cCode: [] inSmalltalk: [theCalloutState := self class calloutStateClass new]. calloutState := self addressOf: theCalloutState. self cCode: [self me: calloutState ms: 0 et: (self sizeof: #CalloutState)]. calloutState callFlags: flags. "Fetch return type and args" argType := interpreterProxy fetchPointer: 0 ofObject: argTypeArray. argSpec := interpreterProxy fetchPointer: 0 ofObject: argType. argClass := interpreterProxy fetchPointer: 1 ofObject: argType. (err := self ffiCheckReturn: argSpec With: argClass in: calloutState) ~= 0 ifTrue: [^self ffiFail: err]. "cannot return" + "alloca the outgoing stack frame, leaving room for marshalling args, and including space for the return struct, if any. + Additional space reserved for saving register args like mandated by Win64 X64 or PPC ABI, will be managed by the call itself" + allocation := self alloca: stackSize + calloutState structReturnSize + self cStackAlignment. + self mustAlignStack ifTrue: + [allocation := self cCoerce: (allocation asUnsignedIntegerPtr bitClear: self cStackAlignment - 1) - "alloca the outgoing stack frame, leaving room for register args while marshalling, and including space for the return struct, if any." - allocation := self alloca: stackSize + calloutState structReturnSize + self registerArgsSlop + self cStackAlignment. - self allocaLiesSoUseGetsp ifTrue: - [allocation := self getsp]. - self cStackAlignment ~= 0 ifTrue: - [allocation := self cCoerce: (allocation asUnsignedInteger bitClear: self cStackAlignment - 1) to: #'char *']. calloutState argVector: allocation; + currentArg: allocation; + limit: allocation + stackSize. - currentArg: allocation + self registerArgsSlop; - limit: allocation + stackSize + self registerArgsSlop. (calloutState structReturnSize > 0 and: [self nonRegisterStructReturnIsViaImplicitFirstArgument and: [(self returnStructInRegisters: calloutState structReturnSize) not]]) ifTrue: [err := self ffiPushPointer: calloutState limit in: calloutState. err ~= 0 ifTrue: [self cleanupCalloutState: calloutState. self cppIf: COGMTVM ifTrue: [err = PrimErrObjectMayMove negated ifTrue: [^PrimErrObjectMayMove]]. "N.B. Do not fail if object may move because caller will GC and retry." ^self ffiFail: err]]. 1 to: nArgs do: [:i| argType := interpreterProxy fetchPointer: i ofObject: argTypeArray. argSpec := interpreterProxy fetchPointer: 0 ofObject: argType. argClass := interpreterProxy fetchPointer: 1 ofObject: argType. oop := argArrayOrNil isNil ifTrue: [interpreterProxy stackValue: nArgs - i] ifFalse: [interpreterProxy fetchPointer: i - 1 ofObject: argArrayOrNil]. err := self ffiArgument: oop Spec: argSpec Class: argClass in: calloutState. err ~= 0 ifTrue: [self cleanupCalloutState: calloutState. self cppIf: COGMTVM ifTrue: [err = PrimErrObjectMayMove negated ifTrue: [^PrimErrObjectMayMove]]. "N.B. Do not fail if object may move because caller will GC and retry." ^self ffiFail: err]]. "coercion failed or out of stack space" "Failures must be reported back from ffiArgument:Spec:Class:in:. Should not fail from here on in." self assert: interpreterProxy failed not. self ffiLogCallout: externalFunction. (requiredStackSize < 0 and: [self externalFunctionHasStackSizeSlot]) ifTrue: [stackSize := calloutState currentArg - calloutState argVector. interpreterProxy storeInteger: ExternalFunctionStackSizeIndex ofObject: externalFunction withValue: stackSize]. "Go out and call this guy" result := self ffiCalloutTo: address SpecOnStack: argArrayOrNil notNil in: calloutState. self cleanupCalloutState: calloutState. ^result! Item was added: + ----- Method: ThreadedFFIPlugin>>mustAlignStack (in category 'marshalling') ----- + mustAlignStack + "Many ABIs mandate a particular stack alignment greater than the natural word size. + If so, this macro will answer true. See class-side preambleCCode." + <cmacro: '() MUST_ALIGN_STACK'> + ^0! Item was removed: - ----- Method: ThreadedFFIPlugin>>registerArgsSlop (in category 'marshalling') ----- - registerArgsSlop - "Answer any space needed to prevent the alloca'ed outgoing arguments marshalling area from - being overwritten by any register arguments during calls during marshalling. For example, on - PowerPC, which has 8 register arguments in the calling convention, register arguments are also - written to the stack. So unless space is left for them, calls during marshalling prior to the actual - callout (e.g. to interpreterProxy object manipulation routines) can end up overwriting the - marshalling stack as register arguments are written to the stack during calls." - self subclassResponsibility! Item was changed: ----- Method: ThreadedIA32FFIPlugin>>ffiCalloutTo:SpecOnStack:in: (in category 'callout support') ----- ffiCalloutTo: procAddr SpecOnStack: specOnStack in: calloutState <var: #procAddr type: #'void *'> <var: #calloutState type: #'CalloutState *'> "Go out, call this guy and create the return value. This *must* be inlined because of the alloca of the outgoing stack frame in ffiCall:WithFlags:NumArgs:Args:AndTypes:" | myThreadIndex atomicType floatRet intRet oop | <var: #floatRet type: #double> <var: #intRet type: #usqLong> <inline: true> self cppIf: COGMTVM ifTrue: [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: [myThreadIndex := interpreterProxy disownVM: 0]]. + (self allocaLiesSoSetSpBeforeCall or: [self mustAlignStack]) ifTrue: - self registerArgsSlop + self cStackAlignment > 0 ifTrue: [self setsp: calloutState argVector]. atomicType := self atomicTypeOf: calloutState ffiRetHeader. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)()')] ifFalse: [intRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'usqLong (*)()')]. "undo any callee argument pops because it may confuse stack management with the alloca." (self isCalleePopsConvention: calloutState callFlags) ifTrue: [self setsp: calloutState argVector]. self cppIf: COGMTVM ifTrue: [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: [interpreterProxy ownVM: myThreadIndex]]. (calloutState ffiRetHeader anyMask: FFIFlagPointer+FFIFlagStructure) ifTrue: ["Note: Order is important here since FFIFlagPointer + FFIFlagStructure is used to represent 'typedef void* VoidPointer' and VoidPointer must be returned as pointer *not* as struct." (calloutState ffiRetHeader anyMask: FFIFlagPointer) ifTrue: [oop := self ffiReturnPointer: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState] ifFalse: [oop := self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. ^oop]. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [oop := interpreterProxy floatObjectOf: floatRet] ifFalse: [oop := self ffiCreateIntegralResultOop: intRet ofAtomicType: atomicType in: calloutState]. ^interpreterProxy methodReturnValue: oop! Item was removed: - ----- Method: ThreadedIA32FFIPlugin>>registerArgsSlop (in category 'marshalling') ----- - registerArgsSlop - "Answer any space needed to prevent the alloca'ed outgoing arguments marshalling area from - being overwritten by any register arguments during calls during marshalling. On x86 this is 0" - ^0! Item was removed: - ----- Method: ThreadedPPCBEFFIPlugin>>registerArgsSlop (in category 'marshalling') ----- - registerArgsSlop - "Answer any space needed to prevent the alloca'ed outgoing arguments marshalling area from - being overwritten by any register arguments during calls during marshalling. On PowerPC, which - has 8 register arguments in the calling convention, register arguments are also written to the stack. - So we must leave room for 8 * 4 bytes to avoid overwriting the marshalling stack as register - arguments are written to the stack during calls to interpreterProxy etc." - ^32! Item was removed: - ----- Method: ThreadedX64FFIPlugin>>registerArgsSlop (in category 'marshalling') ----- - registerArgsSlop - "Answer any space needed to prevent the alloca'ed outgoing arguments marshalling area from - being overwritten by any register arguments during calls during marshalling. On ARM we - believe this is zero." - ^0! Item was changed: ----- Method: ThreadedX64SysVFFIPlugin>>ffiCalloutTo:SpecOnStack:in: (in category 'callout support') ----- ffiCalloutTo: procAddr SpecOnStack: specOnStack in: calloutState <var: #procAddr type: #'void *'> <var: #calloutState type: #'CalloutState *'> <var: #loadFloatRegs declareC: 'extern void loadFloatRegs(double, double, double, double, double, double, double, double)'> "Go out, call this guy and create the return value. This *must* be inlined because of the alloca of the outgoing stack frame in ffiCall:WithFlags:NumArgs:Args:AndTypes:" | myThreadIndex atomicType floatRet intRet loadFloatRegs oop | <var: #floatRet type: #double> <var: #intRet type: 'SixteenByteReturn'> <inline: true> self cCode: '' inSmalltalk: [loadFloatRegs := #used. loadFloatRegs class]. self cppIf: COGMTVM ifTrue: [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: [myThreadIndex := interpreterProxy disownVM: 0]]. - self registerArgsSlop + self cStackAlignment > 0 ifTrue: - [self setsp: calloutState argVector]. - calloutState floatRegisterIndex > 0 ifTrue: [self load: (calloutState floatRegisters at: 0) Flo: (calloutState floatRegisters at: 1) a: (calloutState floatRegisters at: 2) t: (calloutState floatRegisters at: 3) R: (calloutState floatRegisters at: 4) e: (calloutState floatRegisters at: 5) g: (calloutState floatRegisters at: 6) s: (calloutState floatRegisters at: 7)]. + (self allocaLiesSoSetSpBeforeCall or: [self mustAlignStack]) ifTrue: + [self setsp: calloutState argVector]. + atomicType := self atomicTypeOf: calloutState ffiRetHeader. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [atomicType = FFITypeSingleFloat ifTrue: [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'float (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3) with: (calloutState integerRegisters at: 4) with: (calloutState integerRegisters at: 5)] ifFalse: "atomicType = FFITypeDoubleFloat" [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3) with: (calloutState integerRegisters at: 4) with: (calloutState integerRegisters at: 5)]] ifFalse: [intRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'SixteenByteReturn (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3) with: (calloutState integerRegisters at: 4) with: (calloutState integerRegisters at: 5)]. "undo any callee argument pops because it may confuse stack management with the alloca." (self isCalleePopsConvention: calloutState callFlags) ifTrue: [self setsp: calloutState argVector]. self cppIf: COGMTVM ifTrue: [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: [interpreterProxy ownVM: myThreadIndex]]. (calloutState ffiRetHeader anyMask: FFIFlagPointer+FFIFlagStructure) ifTrue: ["Note: Order is important here since FFIFlagPointer + FFIFlagStructure is used to represent 'typedef void* VoidPointer' and VoidPointer must be returned as pointer *not* as struct." (calloutState ffiRetHeader anyMask: FFIFlagPointer) ifTrue: [oop := self ffiReturnPointer: intRet a ofType: (self ffiReturnType: specOnStack) in: calloutState] ifFalse: [oop := self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. ^oop]. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [oop := interpreterProxy floatObjectOf: floatRet] ifFalse: [oop := self ffiCreateIntegralResultOop: intRet a ofAtomicType: atomicType in: calloutState]. ^interpreterProxy methodReturnValue: oop! Item was changed: ----- Method: ThreadedX64Win64FFIPlugin>>ffiCalloutTo:SpecOnStack:in: (in category 'callout support') ----- ffiCalloutTo: procAddr SpecOnStack: specOnStack in: calloutState <var: #procAddr type: #'void *'> <var: #calloutState type: #'CalloutState *'> <var: #loadFloatRegs declareC: 'extern void loadFloatRegs(double, double, double, double)'> "Go out, call this guy and create the return value. This *must* be inlined because of the alloca of the outgoing stack frame in ffiCall:WithFlags:NumArgs:Args:AndTypes:" | myThreadIndex atomicType floatRet intRet loadFloatRegs oop | <var: #floatRet type: #double> <var: #intRet type: #usqLong> <inline: true> self cCode: '' inSmalltalk: [loadFloatRegs := #used. loadFloatRegs class]. self cppIf: COGMTVM ifTrue: [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: [myThreadIndex := interpreterProxy disownVM: 0]]. - self registerArgsSlop + self cStackAlignment > 0 ifTrue: - [self setsp: calloutState argVector]. - calloutState floatRegisterSignature > 0 ifTrue: [self load: (calloutState floatRegisters at: 0) Flo: (calloutState floatRegisters at: 1) at: (calloutState floatRegisters at: 2) Re: (calloutState floatRegisters at: 3) gs: (calloutState floatRegisters at: 4)]. + (self allocaLiesSoSetSpBeforeCall or: [self mustAlignStack]) ifTrue: + [self setsp: calloutState argVector]. + atomicType := self atomicTypeOf: calloutState ffiRetHeader. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [atomicType = FFITypeSingleFloat ifTrue: [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'float (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3)] ifFalse: "atomicType = FFITypeDoubleFloat" [floatRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'double (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3)]] ifFalse: [intRet := self dispatchFunctionPointer: (self cCoerceSimple: procAddr to: 'usqIntptr_t (*)(sqIntptr_t, sqIntptr_t, sqIntptr_t, sqIntptr_t)') with: (calloutState integerRegisters at: 0) with: (calloutState integerRegisters at: 1) with: (calloutState integerRegisters at: 2) with: (calloutState integerRegisters at: 3)]. "undo any callee argument pops because it may confuse stack management with the alloca." (self isCalleePopsConvention: calloutState callFlags) ifTrue: [self setsp: calloutState argVector]. self cppIf: COGMTVM ifTrue: [(calloutState callFlags anyMask: FFICallFlagThreaded) ifTrue: [interpreterProxy ownVM: myThreadIndex]]. (calloutState ffiRetHeader anyMask: FFIFlagPointer+FFIFlagStructure) ifTrue: ["Note: Order is important here since FFIFlagPointer + FFIFlagStructure is used to represent 'typedef void* VoidPointer' and VoidPointer must be returned as pointer *not* as struct." (calloutState ffiRetHeader anyMask: FFIFlagPointer) ifTrue: [oop := self ffiReturnPointer: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState] ifFalse: [oop := self ffiReturnStruct: intRet ofType: (self ffiReturnType: specOnStack) in: calloutState]. ^oop]. (atomicType >> 1) = (FFITypeSingleFloat >> 1) ifTrue: [oop := interpreterProxy floatObjectOf: floatRet] ifFalse: [oop := self ffiCreateIntegralResultOop: intRet ofAtomicType: atomicType in: calloutState]. ^interpreterProxy methodReturnValue: oop!

3 2

[OpenSmalltalk/opensmalltalk-vm] e9ff2f: Regenerate FFI plugin source code as per VMMaker.o...
by GitHub 13 Dec '16

13 Dec '16

Branch: refs/heads/Cog Home: https://github.com/OpenSmalltalk/opensmalltalk-vm Commit: e9ff2f20501ee86801747f06f653fd669e138de9 https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/e9ff2f20501ee86801… Author: nicolas-cellier-aka-nice <nicolas.cellier.aka.nice(a)gmail.com> Date: 2016-12-13 (Tue, 13 Dec 2016) Changed paths: M platforms/Cross/vm/sqCogStackAlignment.h M src/plugins/SqueakFFIPrims/ARM32FFIPlugin.c M src/plugins/SqueakFFIPrims/IA32FFIPlugin.c M src/plugins/SqueakFFIPrims/SqueakFFIPrims.c M src/plugins/SqueakFFIPrims/X64SysVFFIPlugin.c M src/plugins/SqueakFFIPrims/X64Win64FFIPlugin.c Log Message: ----------- Regenerate FFI plugin source code as per VMMaker.oscog-nice.2038 This does remove the SEGV problems related to stack_alignment and related to reserved stack space below alloca'ed return value on mingw32. Rescue the STACK_ALIGN_BYTES knowledge for exotic arch (powerpc & sparc) from FFI preamble code => to sqCogStackAlignment.h TODO: clean-up the temporary hack for compiling pharo (macro overwritten thru compiler -D options)

1 0

[OpenSmalltalk/opensmalltalk-vm] 1acd02: Merge commit '5a949459043392bbc093d78574c2d7c07b10...
by GitHub 13 Dec '16

13 Dec '16

Branch: refs/heads/Cog Home: https://github.com/OpenSmalltalk/opensmalltalk-vm Commit: 1acd02683a9e25936615e0573c712b0ecfd8d06e https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/1acd02683a9e259366… Author: Esteban Lorenzano <estebanlm(a)gmail.com> Date: 2016-12-09 (Fri, 09 Dec 2016) Changed paths: M .gitignore M build.macos32x86/pharo.stack.spur/plugins.ext M build.macos32x86/pharo.stack.spur/plugins.int M nsspur64src/vm/cogit.h M nsspur64src/vm/cogitX64.c M nsspur64src/vm/cointerp.c M nsspur64src/vm/cointerp.h M nsspur64src/vm/gcc3x-cointerp.c M nsspursrc/vm/cogit.h M nsspursrc/vm/cogitARMv5.c M nsspursrc/vm/cogitIA32.c M nsspursrc/vm/cogitMIPSEL.c M nsspursrc/vm/cointerp.c M nsspursrc/vm/cointerp.h M nsspursrc/vm/gcc3x-cointerp.c M nsspurstack64src/vm/gcc3x-interp.c M nsspurstack64src/vm/interp.c M nsspurstacksrc/vm/gcc3x-interp.c M nsspurstacksrc/vm/interp.c M platforms/iOS/vm/Common/Classes/sqSqueakMainApp.m M platforms/unix/vm/sqUnixMain.c M platforms/win32/vm/sqWin32Backtrace.c M spur64src/vm/cogit.h M spur64src/vm/cogitX64.c M spur64src/vm/cointerp.c M spur64src/vm/cointerp.h M spur64src/vm/gcc3x-cointerp.c M spursista64src/vm/cogit.h M spursista64src/vm/cogitX64.c M spursista64src/vm/cointerp.c M spursista64src/vm/cointerp.h M spursista64src/vm/gcc3x-cointerp.c M spursistasrc/vm/cogit.h M spursistasrc/vm/cogitARMv5.c M spursistasrc/vm/cogitIA32.c M spursistasrc/vm/cogitMIPSEL.c M spursistasrc/vm/cointerp.c M spursistasrc/vm/cointerp.h M spursistasrc/vm/gcc3x-cointerp.c M spursrc/vm/cogit.h M spursrc/vm/cogitARMv5.c M spursrc/vm/cogitIA32.c M spursrc/vm/cogitMIPSEL.c M spursrc/vm/cointerp.c M spursrc/vm/cointerp.h M spursrc/vm/gcc3x-cointerp.c M spurstack64src/vm/gcc3x-interp.c M spurstack64src/vm/interp.c M spurstacksrc/vm/gcc3x-interp.c M spurstacksrc/vm/interp.c M src/plugins/AsynchFilePlugin/AsynchFilePlugin.c M src/plugins/B2DPlugin/B2DPlugin.c M src/plugins/BitBltPlugin/BitBltPlugin.c M src/plugins/FilePlugin/FilePlugin.c M src/plugins/MiscPrimitivePlugin/MiscPrimitivePlugin.c M src/plugins/SqueakFFIPrims/ARM32FFIPlugin.c M src/plugins/SqueakFFIPrims/IA32FFIPlugin.c M src/plugins/SqueakFFIPrims/X64SysVFFIPlugin.c M src/plugins/SqueakFFIPrims/X64Win64FFIPlugin.c M src/vm/cogit.h M src/vm/cogitARMv5.c M src/vm/cogitIA32.c M src/vm/cogitMIPSEL.c M src/vm/cointerp.c M src/vm/cointerp.h M src/vm/cointerpmt.c M src/vm/cointerpmt.h M src/vm/gcc3x-cointerp.c M src/vm/gcc3x-cointerpmt.c M stacksrc/vm/gcc3x-interp.c M stacksrc/vm/interp.c Log Message: ----------- Merge commit '5a949459043392bbc093d78574c2d7c07b10fcb2' Commit: 7a7283139b72e56572f53caa167cb735a53ba949 https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/7a7283139b72e56572… Author: Esteban Lorenzano <estebanlm(a)gmail.com> Date: 2016-12-09 (Fri, 09 Dec 2016) Changed paths: M third-party/libpng.spec.win Log Message: ----------- fix appveyor build script to fail when… it fails :) fix png download url. Commit: c8d42164e557e5107f4452f0d1e6267f6e65feaa https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/c8d42164e557e5107f… Author: Esteban Lorenzano <estebanlm(a)gmail.com> Date: 2016-12-13 (Tue, 13 Dec 2016) Changed paths: M nsspur64src/vm/cogit.h M nsspur64src/vm/cogitX64.c M nsspur64src/vm/cointerp.c M nsspur64src/vm/cointerp.h M nsspur64src/vm/gcc3x-cointerp.c M nsspursrc/vm/cogit.h M nsspursrc/vm/cogitARMv5.c M nsspursrc/vm/cogitIA32.c M nsspursrc/vm/cogitMIPSEL.c M nsspursrc/vm/cointerp.c M nsspursrc/vm/cointerp.h M nsspursrc/vm/gcc3x-cointerp.c M nsspurstack64src/vm/gcc3x-interp.c M nsspurstack64src/vm/interp.c M nsspurstacksrc/vm/gcc3x-interp.c M nsspurstacksrc/vm/interp.c M spur64src/vm/cogit.h M spur64src/vm/cogitX64.c M spur64src/vm/cointerp.c M spur64src/vm/cointerp.h M spur64src/vm/gcc3x-cointerp.c M spursista64src/vm/cogit.h M spursista64src/vm/cogitX64.c M spursista64src/vm/cointerp.c M spursista64src/vm/cointerp.h M spursista64src/vm/gcc3x-cointerp.c M spursistasrc/vm/cogit.h M spursistasrc/vm/cogitARMv5.c M spursistasrc/vm/cogitIA32.c M spursistasrc/vm/cogitMIPSEL.c M spursistasrc/vm/cointerp.c M spursistasrc/vm/cointerp.h M spursistasrc/vm/gcc3x-cointerp.c M spursrc/vm/cogit.h M spursrc/vm/cogitARMv5.c M spursrc/vm/cogitIA32.c M spursrc/vm/cogitMIPSEL.c M spursrc/vm/cointerp.c M spursrc/vm/cointerp.h M spursrc/vm/gcc3x-cointerp.c M spurstack64src/vm/gcc3x-interp.c M spurstack64src/vm/interp.c M spurstacksrc/vm/gcc3x-interp.c M spurstacksrc/vm/interp.c M src/vm/cogit.h M src/vm/cogitARMv5.c M src/vm/cogitIA32.c M src/vm/cogitMIPSEL.c M src/vm/cointerp.c M src/vm/cointerp.h M src/vm/cointerpmt.c M src/vm/cointerpmt.h M src/vm/gcc3x-cointerp.c M src/vm/gcc3x-cointerpmt.c M stacksrc/vm/gcc3x-interp.c M stacksrc/vm/interp.c Log Message: ----------- Merge commit '8a77b2517d9ec3d467cb0f1ae427e94c4a5eeb10' Commit: 945fbc37d97f2db5ddcb1abfd14e9a1e7cb55c9c https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/945fbc37d97f2db5dd… Author: Esteban Lorenzano <estebanlm(a)gmail.com> Date: 2016-12-13 (Tue, 13 Dec 2016) Changed paths: M platforms/win32/vm/sqWin32Main.c M platforms/win32/vm/sqWin32Window.c Log Message: ----------- add double dash for Pharo on windows (the only remaining) Commit: 99fdc442dc59f3bbe3fc48b2a037c873b9529dcb https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/99fdc442dc59f3bbe3… Author: Esteban Lorenzano <estebanlm(a)gmail.com> Date: 2016-12-13 (Tue, 13 Dec 2016) Changed paths: M platforms/win32/vm/sqWin32Main.c M platforms/win32/vm/sqWin32Window.c M third-party/libpng.spec.win Log Message: ----------- Merge pull request #107 from estebanlm/Cog add double dash for Pharo on windows (the only remaining) Compare: https://github.com/OpenSmalltalk/opensmalltalk-vm/compare/8a77b2517d9e...99…

1 0

← Newer
1
2
3
4
5
6
7
8
9
10
11
Older →

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

Vm-dev December 2016