Vm-dev April 2016

vm-dev@lists.squeakfoundation.org

35 participants
202 discussions

VM Maker: VMMaker.oscog-nice.1841.mcz
by commits＠source.squeak.org 24 Apr '16

24 Apr '16

Nicolas Cellier uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-nice.1841.mcz ==================== Summary ==================== Name: VMMaker.oscog-nice.1841 Author: nice Time: 24 April 2016, 11:38:02.212 pm UUID: 0d40f497-289c-46c5-ace2-b3280a5b096f Ancestors: VMMaker.oscog-nice.1840 Revert type inference for shifts << >> because it prevents cog v3 to start (until we find why). Fix positive32BitIntegerFor: for Spur64 which did work only because the constant was generated ULL. Now the status of kernel tests is back to something normal at least on mac for Squeak. =============== Diff against VMMaker.oscog-nice.1840 =============== Item was changed: ----- Method: CCodeGenerator>>returnTypeForSend:in: (in category 'type inference') ----- returnTypeForSend: sendNode in: aTMethod "Answer the return type for a send. Absent sends default to #sqInt. The inferred type should match as closely as possible the C type of generated expessions so that inlining would not change the expression." | sel methodOrNil | methodOrNil := self anyMethodNamed: (sel := sendNode selector). (methodOrNil notNil and: [methodOrNil returnType notNil]) ifTrue: [^self baseTypeForType: methodOrNil returnType]. ^kernelReturnTypes at: sel ifAbsent: [sel caseOf: { [#integerValueOf:] -> [#sqInt]. [#isIntegerObject:] -> [#int]. [#negated] -> [self promoteArithmeticTypes: (self typeFor: sendNode receiver in: aTMethod) and: #int]. [#+] -> [self typeForArithmetic: sendNode in: aTMethod]. [#-] -> [self typeForArithmetic: sendNode in: aTMethod]. [#*] -> [self typeForArithmetic: sendNode in: aTMethod]. [#/] -> [self typeForArithmetic: sendNode in: aTMethod]. [#//] -> [self typeForArithmetic: sendNode in: aTMethod]. [#\\] -> [self typeForArithmetic: sendNode in: aTMethod]. + " [#>>] -> [self - [#>>] -> [self promoteArithmeticTypes: (self unsignedTypeForIntegralType: (self typeFor: sendNode receiver in: aTMethod)) and: (self typeFor: sendNode args first in: aTMethod)]. + [#<<] -> [self typeForArithmetic: sendNode in: aTMethod]." - [#<<] -> [self typeForArithmetic: sendNode in: aTMethod]. [#rem:] -> [self typeForArithmetic: sendNode in: aTMethod]. [#quo:] -> [self typeForArithmetic: sendNode in: aTMethod]. [#addressOf:] -> [(self typeFor: sendNode receiver in: aTMethod) ifNil: [#sqInt] ifNotNil: [:type| type, (type last isLetter ifTrue: [' *'] ifFalse: ['*'])]]. [#at:] -> [self typeForDereference: sendNode in: aTMethod]. [#bitAnd:] -> [self typeForArithmetic: sendNode in: aTMethod]. [#bitOr:] -> [self typeForArithmetic: sendNode in: aTMethod]. [#bitXor:] -> [self typeForArithmetic: sendNode in: aTMethod]. [#bitClear:] -> [self typeForArithmetic: sendNode in: aTMethod]. [#bitInvert32] -> [#'unsigned int']. [#bitInvert64] -> [self promoteArithmeticTypes: (self typeFor: sendNode receiver in: aTMethod) and: #int]. [#byteSwap32] -> [#'unsigned int']. [#byteSwap64] -> [#'unsigned long long']. [#byteSwapped32IfBigEndian:] -> [#'unsigned int']. [#byteSwapped64IfBigEndian:] -> [#'unsigned long long']. [#=] -> [#int]. [#~=] -> [#int]. [#==] -> [#int]. [#~~] -> [#int]. [#<] -> [#int]. [#<=] -> [#int]. [#>] -> [#int]. [#>=] -> [#int]. [#between:and:] -> [#int]. [#anyMask:] -> [#int]. [#allMask:] -> [#int]. [#noMask:] -> [#int]. [#isNil] -> [#int]. [#notNil] -> [#int]. [#&] -> [#int]. [#|] -> [#int]. [#not] -> [#int]. [#asFloat] -> [#double]. [#atan] -> [#double]. [#exp] -> [#double]. [#log] -> [#double]. [#sin] -> [#double]. [#sqrt] -> [#double]. [#asLong] -> [#long]. [#asInteger] -> [#sqInt]. [#asUnsignedInteger] -> [#usqInt]. [#asUnsignedLong] -> [#'unsigned long']. [#asVoidPointer] -> [#'void *']. [#signedIntToLong] -> [#usqInt]. "c.f. generateSignedIntToLong:on:indent:" [#signedIntToShort] -> [#usqInt]. "c.f. generateSignedIntToShort:on:indent:" [#cCoerce:to:] -> [sendNode args last value]. [#cCoerceSimple:to:] -> [sendNode args last value]. [#sizeof:] -> [#'unsigned long']. "Technically it's a size_t but it matches unsigned long on target architectures so far..." [#ifTrue:ifFalse:] -> [self typeForConditional: sendNode in: aTMethod]. [#ifFalse:ifTrue:] -> [self typeForConditional: sendNode in: aTMethod]. [#ifTrue:] -> [self typeForConditional: sendNode in: aTMethod]. [#ifFalse:] -> [self typeForConditional: sendNode in: aTMethod] } otherwise: "If there /is/ a method for sel but its retrn type is as yet unknown we /mustn't/ default it. We can only default unbound selectors." [methodOrNil ifNotNil: [nil] ifNil: [#sqInt]]]! Item was changed: ----- Method: StackInterpreter>>positive32BitIntegerFor: (in category 'primitive support') ----- positive32BitIntegerFor: integerValue "integerValue is interpreted as POSITIVE, e.g. as the result of Bitmap>at:. N.B. Returning in each arm separately enables Slang inlining. /Don't/ return the ifTrue:ifFalse: unless Slang inlining of conditionals is fixed." <inline: true> <var: 'integerValue' type: #'unsigned int'> objectMemory hasSixtyFourBitImmediates ifTrue: + [^objectMemory integerObjectOf: (integerValue asUnsignedLong bitAnd: 16rFFFFFFFF)] - [^objectMemory integerObjectOf: (integerValue bitAnd: 16rFFFFFFFF)] ifFalse: [^self maybeInlinePositive32BitIntegerFor: integerValue]!

1 0

Re: [Vm-dev] VM Maker: VMMaker.oscog-nice.1840.mcz
by Nicolas Cellier 24 Apr '16

24 Apr '16

If things are not going better this evening, I'll retract some of these changes. 2016-04-24 22:31 GMT+02:00 <commits(a)source.squeak.org>: > > Nicolas Cellier uploaded a new version of VMMaker to project VM Maker: > http://source.squeak.org/VMMaker/VMMaker.oscog-nice.1840.mcz > > ==================== Summary ==================== > > Name: VMMaker.oscog-nice.1840 > Author: nice > Time: 24 April 2016, 10:30:31.797 pm > UUID: 218875e3-8515-4804-a608-9e3228ffb843 > Ancestors: VMMaker.oscog-nice.1839 > > Oops, correct the inversion of type inference for left/right shift. > Protect the type of most used integerValueOf:/isIntegerObject: > > This restore good health for CogSpur32 tests, but not yet for CogSpur64 > which has a problem with SmallFloat, and definitely not for CogV3 which now > fails to start. > > =============== Diff against VMMaker.oscog-nice.1839 =============== > > Item was changed: > ----- Method: CCodeGenerator>>returnTypeForSend:in: (in category 'type > inference') ----- > returnTypeForSend: sendNode in: aTMethod > "Answer the return type for a send. Absent sends default to > #sqInt. > The inferred type should match as closely as possible the C type > of > generated expessions so that inlining would not change the > expression." > | sel methodOrNil | > methodOrNil := self anyMethodNamed: (sel := sendNode selector). > (methodOrNil notNil and: [methodOrNil returnType notNil]) ifTrue: > [^self baseTypeForType: methodOrNil returnType]. > ^kernelReturnTypes > at: sel > ifAbsent: > [sel > caseOf: { > + [#integerValueOf:] -> > [#sqInt]. > + [#isIntegerObject:] -> > [#int]. > [#negated] > -> [self promoteArithmeticTypes: (self typeFor: sendNode receiver in: > aTMethod) and: #int]. > [#+] > -> [self typeForArithmetic: sendNode in: aTMethod]. > [#-] > -> [self typeForArithmetic: sendNode in: aTMethod]. > [#*] > -> [self typeForArithmetic: sendNode in: aTMethod]. > [#/] > -> [self typeForArithmetic: sendNode in: aTMethod]. > [#//] > -> [self typeForArithmetic: sendNode in: aTMethod]. > [#\\] > -> [self typeForArithmetic: sendNode in: aTMethod]. > + [#>>] > -> [self > - [#>>] > -> [self typeForArithmetic: sendNode in: aTMethod]. > - [#<<] > -> [self > > promoteArithmeticTypes: (self > unsignedTypeForIntegralType: (self typeFor: sendNode receiver in: aTMethod)) > > and: (self typeFor: sendNode args first in: > aTMethod)]. > + [#<<] > -> [self typeForArithmetic: sendNode in: aTMethod]. > [#rem:] > -> [self typeForArithmetic: sendNode in: aTMethod]. > [#quo:] > -> [self typeForArithmetic: sendNode in: aTMethod]. > [#addressOf:] -> > [(self typeFor: sendNode receiver in: aTMethod) > > ifNil: [#sqInt] > > ifNotNil: [:type| type, (type last isLetter ifTrue: > [' *'] ifFalse: ['*'])]]. > [#at:] > -> [self typeForDereference: sendNode in: aTMethod]. > [#bitAnd:] > -> [self typeForArithmetic: sendNode in: aTMethod]. > [#bitOr:] > -> [self typeForArithmetic: sendNode in: aTMethod]. > [#bitXor:] > -> [self typeForArithmetic: sendNode in: aTMethod]. > [#bitClear:] > -> [self typeForArithmetic: sendNode in: aTMethod]. > [#bitInvert32] -> > [#'unsigned int']. > [#bitInvert64] -> > [self promoteArithmeticTypes: (self typeFor: sendNode receiver in: > aTMethod) and: #int]. > [#byteSwap32] -> > [#'unsigned int']. > [#byteSwap64] -> > [#'unsigned long long']. > [#byteSwapped32IfBigEndian:] -> > [#'unsigned int']. > [#byteSwapped64IfBigEndian:] -> > [#'unsigned long long']. > [#=] > -> [#int]. > [#~=] > -> [#int]. > [#==] > -> [#int]. > [#~~] > -> [#int]. > [#<] > -> [#int]. > [#<=] > -> [#int]. > [#>] > -> [#int]. > [#>=] > -> [#int]. > [#between:and:] -> [#int]. > [#anyMask:] > -> [#int]. > [#allMask:] > -> [#int]. > [#noMask:] > -> [#int]. > [#isNil] > -> [#int]. > [#notNil] > -> [#int]. > [#&] > -> [#int]. > [#|] > -> [#int]. > [#not] > -> [#int]. > [#asFloat] > -> [#double]. > [#atan] > -> [#double]. > [#exp] > -> [#double]. > [#log] > -> [#double]. > [#sin] > -> [#double]. > [#sqrt] > -> [#double]. > [#asLong] > -> [#long]. > [#asInteger] -> > [#sqInt]. > [#asUnsignedInteger] -> [#usqInt]. > [#asUnsignedLong] -> > [#'unsigned long']. > [#asVoidPointer] -> > [#'void *']. > [#signedIntToLong] -> > [#usqInt]. "c.f. generateSignedIntToLong:on:indent:" > [#signedIntToShort] -> [#usqInt]. > "c.f. generateSignedIntToShort:on:indent:" > [#cCoerce:to:] -> > [sendNode args last value]. > [#cCoerceSimple:to:] -> [sendNode > args last value]. > [#sizeof:] > -> [#'unsigned long']. "Technically it's a size_t but it matches > unsigned long on target architectures so far..." > [#ifTrue:ifFalse:] -> > [self typeForConditional: sendNode in: aTMethod]. > [#ifFalse:ifTrue:] -> > [self typeForConditional: sendNode in: aTMethod]. > [#ifTrue:] > -> [self typeForConditional: sendNode in: aTMethod]. > [#ifFalse:] > -> [self typeForConditional: sendNode in: aTMethod] } > otherwise: "If there /is/ a method for sel > but its retrn type is as yet unknown we /mustn't/ default it. > We can only > default unbound selectors." > [methodOrNil ifNotNil: [nil] > ifNil: [#sqInt]]]! > >

1 0

VM Maker: VMMaker.oscog-nice.1840.mcz
by commits＠source.squeak.org 24 Apr '16

24 Apr '16

Nicolas Cellier uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-nice.1840.mcz ==================== Summary ==================== Name: VMMaker.oscog-nice.1840 Author: nice Time: 24 April 2016, 10:30:31.797 pm UUID: 218875e3-8515-4804-a608-9e3228ffb843 Ancestors: VMMaker.oscog-nice.1839 Oops, correct the inversion of type inference for left/right shift. Protect the type of most used integerValueOf:/isIntegerObject: This restore good health for CogSpur32 tests, but not yet for CogSpur64 which has a problem with SmallFloat, and definitely not for CogV3 which now fails to start. =============== Diff against VMMaker.oscog-nice.1839 =============== Item was changed: ----- Method: CCodeGenerator>>returnTypeForSend:in: (in category 'type inference') ----- returnTypeForSend: sendNode in: aTMethod "Answer the return type for a send. Absent sends default to #sqInt. The inferred type should match as closely as possible the C type of generated expessions so that inlining would not change the expression." | sel methodOrNil | methodOrNil := self anyMethodNamed: (sel := sendNode selector). (methodOrNil notNil and: [methodOrNil returnType notNil]) ifTrue: [^self baseTypeForType: methodOrNil returnType]. ^kernelReturnTypes at: sel ifAbsent: [sel caseOf: { + [#integerValueOf:] -> [#sqInt]. + [#isIntegerObject:] -> [#int]. [#negated] -> [self promoteArithmeticTypes: (self typeFor: sendNode receiver in: aTMethod) and: #int]. [#+] -> [self typeForArithmetic: sendNode in: aTMethod]. [#-] -> [self typeForArithmetic: sendNode in: aTMethod]. [#*] -> [self typeForArithmetic: sendNode in: aTMethod]. [#/] -> [self typeForArithmetic: sendNode in: aTMethod]. [#//] -> [self typeForArithmetic: sendNode in: aTMethod]. [#\\] -> [self typeForArithmetic: sendNode in: aTMethod]. + [#>>] -> [self - [#>>] -> [self typeForArithmetic: sendNode in: aTMethod]. - [#<<] -> [self promoteArithmeticTypes: (self unsignedTypeForIntegralType: (self typeFor: sendNode receiver in: aTMethod)) and: (self typeFor: sendNode args first in: aTMethod)]. + [#<<] -> [self typeForArithmetic: sendNode in: aTMethod]. [#rem:] -> [self typeForArithmetic: sendNode in: aTMethod]. [#quo:] -> [self typeForArithmetic: sendNode in: aTMethod]. [#addressOf:] -> [(self typeFor: sendNode receiver in: aTMethod) ifNil: [#sqInt] ifNotNil: [:type| type, (type last isLetter ifTrue: [' *'] ifFalse: ['*'])]]. [#at:] -> [self typeForDereference: sendNode in: aTMethod]. [#bitAnd:] -> [self typeForArithmetic: sendNode in: aTMethod]. [#bitOr:] -> [self typeForArithmetic: sendNode in: aTMethod]. [#bitXor:] -> [self typeForArithmetic: sendNode in: aTMethod]. [#bitClear:] -> [self typeForArithmetic: sendNode in: aTMethod]. [#bitInvert32] -> [#'unsigned int']. [#bitInvert64] -> [self promoteArithmeticTypes: (self typeFor: sendNode receiver in: aTMethod) and: #int]. [#byteSwap32] -> [#'unsigned int']. [#byteSwap64] -> [#'unsigned long long']. [#byteSwapped32IfBigEndian:] -> [#'unsigned int']. [#byteSwapped64IfBigEndian:] -> [#'unsigned long long']. [#=] -> [#int]. [#~=] -> [#int]. [#==] -> [#int]. [#~~] -> [#int]. [#<] -> [#int]. [#<=] -> [#int]. [#>] -> [#int]. [#>=] -> [#int]. [#between:and:] -> [#int]. [#anyMask:] -> [#int]. [#allMask:] -> [#int]. [#noMask:] -> [#int]. [#isNil] -> [#int]. [#notNil] -> [#int]. [#&] -> [#int]. [#|] -> [#int]. [#not] -> [#int]. [#asFloat] -> [#double]. [#atan] -> [#double]. [#exp] -> [#double]. [#log] -> [#double]. [#sin] -> [#double]. [#sqrt] -> [#double]. [#asLong] -> [#long]. [#asInteger] -> [#sqInt]. [#asUnsignedInteger] -> [#usqInt]. [#asUnsignedLong] -> [#'unsigned long']. [#asVoidPointer] -> [#'void *']. [#signedIntToLong] -> [#usqInt]. "c.f. generateSignedIntToLong:on:indent:" [#signedIntToShort] -> [#usqInt]. "c.f. generateSignedIntToShort:on:indent:" [#cCoerce:to:] -> [sendNode args last value]. [#cCoerceSimple:to:] -> [sendNode args last value]. [#sizeof:] -> [#'unsigned long']. "Technically it's a size_t but it matches unsigned long on target architectures so far..." [#ifTrue:ifFalse:] -> [self typeForConditional: sendNode in: aTMethod]. [#ifFalse:ifTrue:] -> [self typeForConditional: sendNode in: aTMethod]. [#ifTrue:] -> [self typeForConditional: sendNode in: aTMethod]. [#ifFalse:] -> [self typeForConditional: sendNode in: aTMethod] } otherwise: "If there /is/ a method for sel but its retrn type is as yet unknown we /mustn't/ default it. We can only default unbound selectors." [methodOrNil ifNotNil: [nil] ifNil: [#sqInt]]]!

1 0

Re: [Vm-dev] VM Maker: VMMaker.oscog-nice.1838.mcz
by Nicolas Cellier 24 Apr '16

24 Apr '16

2016-04-24 13:48 GMT+02:00 <commits(a)source.squeak.org>: > > Nicolas Cellier uploaded a new version of VMMaker to project VM Maker: > http://source.squeak.org/VMMaker/VMMaker.oscog-nice.1838.mcz > > ==================== Summary ==================== > > Name: VMMaker.oscog-nice.1838 > Author: nice > Time: 24 April 2016, 1:47:02.874 pm > UUID: 73cf9fda-4e5a-4e0c-a1b9-92ace5f8f3c9 > Ancestors: VMMaker.oscog-nice.1837 > > Use a type inference that better fits C rules (C99 standard section 6.3): > - in (float op int), int is promoted to double not float > - in (unsigned short op short) both operands are promoted to #int due to > int promotion rules. > > The idea behind this is that behaviour of generated code for inlined and > non inlined code should be the same. We can guaranty such property only if > we adhere to C rules. > > =============== Diff against VMMaker.oscog-nice.1837 =============== > > Item was changed: > ----- Method: CCodeGenerator>>computeKernelReturnTypes (in category > 'public') ----- > computeKernelReturnTypes > | dictionary | > dictionary := Dictionary newFromPairs: > #(oopAt: #sqInt oopAt:put: #sqInt > oopAtPointer: #sqInt oopAtPointer:put: #sqInt > byteAt: #sqInt byteAt:put: #sqInt > byteAtPointer: #sqInt byteAtPointer:put: #sqInt > shortAt: #sqInt shortAt:put: #sqInt > shortAtPointer: #sqInt shortAtPointer:put: #sqInt > intAt: #sqInt intAt:put: #sqInt > intAtPointer: #sqInt intAtPointer:put: #sqInt > longAt: #sqInt longAt:put: #sqInt > longAtPointer: #sqInt longAtPointer:put: #sqInt > + long32At: #int long32At:put: #int > - long32At: #sqInt long32At:put: #sqInt > unalignedLongAt: #sqInt > unalignedLongAt:put: #sqInt > + unalignedLong32At: #int > unalignedLong32At:put: #int > - unalignedLong32At: #sqInt > unalignedLong32At:put: #sqInt > > Ah, I forgot to comment the purpose of #computeKernelReturnTypes change: the idea is that we use a type inference that works both for 32 and 64 bits VM. Why? Because we generate single source code for both 32 and 64 bits plugins. So we should have type inference as much independent of BytesPerWord as possible. long64At: #sqLong long64At:put: #sqLong > > fetchFloatAt:into: #void storeFloatAt:from: #void > fetchFloatAtPointer:into: #void > storeFloatAtPointer:from: #void > fetchSingleFloatAt:into: #void storeSingleFloatAt:from: > #void > fetchSingleFloatAtPointer:into: #void > storeSingleFloatAtPointer:from: #void > > pointerForOop: #'char *' oopForPointer: #sqInt). > - BytesPerWord = 8 ifTrue: > - [#(long32At: long32At:put: unalignedLong32At: > unalignedLong32At:put:) do: > - [:accessor| > - dictionary at: accessor put: #int]]. > ^dictionary! > > Item was changed: > ----- Method: CCodeGenerator>>promoteArithmeticTypes:and: (in category > 'type inference') ----- > promoteArithmeticTypes: firstType and: secondType > "Answer the return type for an arithmetic send. This is so that > the inliner can still inline > + simple expressions. Deal with pointer arithmetic, floating point > arithmetic and promotion. > + It is important to choose deterministically to get stable source > generation. > + Also, the behaviour of inlined and non inlined code should be > unchanged." > - simple expressions. Deal with pointer arithmetic, floating point > arithmetic and promotion." > - | firstSize secondSize | > ((#(#double float) includes: firstType) > or: [#(#double float) includes: secondType]) ifTrue: > + [^(firstType = #double or: [secondType = #double]) > + ifTrue: [#double] > + ifFalse: [#float] "in C99 6.3.1.8, float+int, int > is converted as a float"]. > - [^(firstType = #float and: [secondType = #float]) > - ifTrue: [#float] > - ifFalse: [#double]]. > "deal with unknowns, answering nil." > (firstType isNil or: [secondType isNil]) ifTrue: > [^nil]. > + "Deal with integer promotion and arithmetic conversion" > + ^self promoteIntegerArithmeticTypes: firstType and: secondType! > - "Deal with promotion; answer the longest type, defaulting to the > recever if they're the same. > - See e.g. section 6.3.1.8 Usual arithmetic conversions, from the > C99 standard: > - Otherwise, the integer promotions are performed on both > operands > - Then the following rules are applied to the promoted > operands: > - > - If both operands have the same type, then no > further conversion is needed. > - > - Otherwise, if both operands have signed integer > types or both have unsigned integer > - types, the operand with the type of lesser integer > conversion rank is converted to the > - type of the operand with greater rank. > - > - Otherwise, if the operand that has unsigned > integer type has rank greater or equal to > - the rank of the type of the other operand, then > the operand with signed integer type > - is converted to the type of the operand with > unsigned integer type. > - > - Otherwise, if the type of the operand with signed > integer type can represent all of the > - values of the type of the operand with unsigned > integer type, then the operand with > - unsigned integer type is converted to the type of > the operand with signed integer type. > - > - Otherwise, both operands are converted to the > unsigned integer type corresponding to > - the type of the operand with signed integer type. > - > - It is important to choose deterministically to get stable source > generation. So if the types have > - the same size but differ in signedness we choose the unsigned > type, which is in partial agreement > - with the above" > - ^(firstSize := self sizeOfIntegralCType: firstType) = (secondSize > := self sizeOfIntegralCType: secondType) > - ifTrue: > - [(firstType first = $u) > - ifTrue: [firstType] > - ifFalse: [(secondType first = $u) ifTrue: > [secondType] ifFalse: [firstType]]] > - ifFalse: > - [firstSize > secondSize ifTrue: [firstType] > ifFalse: [secondType]]! > > Item was added: > + ----- Method: CCodeGenerator>>promoteIntegerArithmeticTypes:and: (in > category 'type inference') ----- > + promoteIntegerArithmeticTypes: firstType and: secondType > + "Answer the return type for an arithmetic send. > + Deal with integer promotion rules of C99. > + See section 6.3 Conversions of the standard. > + > + 6.3.1.1 ...snip... > + If an int can represent all values of the original type, > the value is converted to an int; > + otherwise, it is converted to an unsigned int. These are > called the integer promotions. > + All other types are unchanged by the inte ger promotions > + > + 6.3.1.8 ...snip... > + Otherwise, the integer promotions are performed on both > operands > + Then the following rules are applied to the promoted > operands: > + > + If both operands have the same type, then no > further conversion is needed. > + > + Otherwise, if both operands have signed integer > types or both have unsigned integer > + types, the operand with the type of lesser integer > conversion rank is converted to the > + type of the operand with greater rank. > + > + Otherwise, if the operand that has unsigned > integer type has rank greater or equal to > + the rank of the type of the other operand, then > the operand with signed integer type > + is converted to the type of the operand with > unsigned integer type. > + > + Otherwise, if the type of the operand with signed > integer type can represent all of the > + values of the type of the operand with unsigned > integer type, then the operand with > + unsigned integer type is converted to the type of > the operand with signed integer type. > + > + Otherwise, both operands are converted to the > unsigned integer type corresponding to > + the type of the operand with signed integer type. > + > + This is so that the generated code behaviour is unsensitive to > inlining." > + | length1 length2 intSize | > + length1 := self sizeOfIntegralCType: firstType. > + length2 := self sizeOfIntegralCType: secondType. > + intSize := self sizeOfIntegralCType: #int. > + (length1 < intSize and: [length2 < intSize]) ifTrue: [^#int]. > "Integer promotion" > + length1 > length2 ifTrue: [^firstType]. > + length2 > length1 ifTrue: [^secondType]. > + firstType first = $u ifTrue: [^firstType]. > + secondType first = $u ifTrue: [^secondType]. > + ^firstType! > >

1 0

VM Maker: VMMaker.oscog-nice.1839.mcz
by commits＠source.squeak.org 24 Apr '16

24 Apr '16

Nicolas Cellier uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-nice.1839.mcz ==================== Summary ==================== Name: VMMaker.oscog-nice.1839 Author: nice Time: 24 April 2016, 2:57:19.464 pm UUID: 9f8b7677-00f1-4c43-bae8-2ab899cf0de4 Ancestors: VMMaker.oscog-nice.1838 Unify type inference used to #inferReturnTypeFromReturnsIn: with that based on AST. That means many more type inference for message sends (returnTypeForSend:in:). Ideally any message generated specially thru initializeCTranslationDictionary should have a corresponding entry in returnTypeForSend:in:, because type inference cannot apply to special tricks applied by CCodeGenerator. There is still possible ambiguity related to different types in different path (if else/ ()?: ). This was the whole point of #addTypesFor:to:in:. IMO we cannot resolve automatically and maybe the best will be to raise a Warning, optionnally ignore it and let expert user manually fix it.. Now that type inference is a bit more thorough, release type restrictions in #isFunctional in order to obtain much more aggressive inlining. With those, a gain of 10% is possible on the LargeIntegersPlugin dominated bench: [ArbitraryPrecisionFloatTest suite run] timeToRun. =============== Diff against VMMaker.oscog-nice.1838 =============== Item was changed: ----- Method: CCodeGenerator>>returnTypeForSend:in: (in category 'type inference') ----- returnTypeForSend: sendNode in: aTMethod "Answer the return type for a send. Absent sends default to #sqInt. + The inferred type should match as closely as possible the C type of + generated expessions so that inlining would not change the expression." - The bitwise operators answer unsigned versions of their argument types, at least in gcc - although this author can't find that in the C99 spec. If you can find this, please let me know." | sel methodOrNil | methodOrNil := self anyMethodNamed: (sel := sendNode selector). (methodOrNil notNil and: [methodOrNil returnType notNil]) ifTrue: [^self baseTypeForType: methodOrNil returnType]. ^kernelReturnTypes at: sel ifAbsent: [sel caseOf: { + [#negated] -> [self promoteArithmeticTypes: (self typeFor: sendNode receiver in: aTMethod) and: #int]. [#+] -> [self typeForArithmetic: sendNode in: aTMethod]. [#-] -> [self typeForArithmetic: sendNode in: aTMethod]. [#*] -> [self typeForArithmetic: sendNode in: aTMethod]. [#/] -> [self typeForArithmetic: sendNode in: aTMethod]. + [#//] -> [self typeForArithmetic: sendNode in: aTMethod]. + [#\\] -> [self typeForArithmetic: sendNode in: aTMethod]. + [#>>] -> [self typeForArithmetic: sendNode in: aTMethod]. + [#<<] -> [self + promoteArithmeticTypes: (self unsignedTypeForIntegralType: (self typeFor: sendNode receiver in: aTMethod)) + and: (self typeFor: sendNode args first in: aTMethod)]. + [#rem:] -> [self typeForArithmetic: sendNode in: aTMethod]. + [#quo:] -> [self typeForArithmetic: sendNode in: aTMethod]. [#addressOf:] -> [(self typeFor: sendNode receiver in: aTMethod) ifNil: [#sqInt] ifNotNil: [:type| type, (type last isLetter ifTrue: [' *'] ifFalse: ['*'])]]. [#at:] -> [self typeForDereference: sendNode in: aTMethod]. [#bitAnd:] -> [self typeForArithmetic: sendNode in: aTMethod]. [#bitOr:] -> [self typeForArithmetic: sendNode in: aTMethod]. [#bitXor:] -> [self typeForArithmetic: sendNode in: aTMethod]. + [#bitClear:] -> [self typeForArithmetic: sendNode in: aTMethod]. + [#bitInvert32] -> [#'unsigned int']. + [#bitInvert64] -> [self promoteArithmeticTypes: (self typeFor: sendNode receiver in: aTMethod) and: #int]. + [#byteSwap32] -> [#'unsigned int']. + [#byteSwap64] -> [#'unsigned long long']. + [#byteSwapped32IfBigEndian:] -> [#'unsigned int']. + [#byteSwapped64IfBigEndian:] -> [#'unsigned long long']. + [#=] -> [#int]. + [#~=] -> [#int]. + [#==] -> [#int]. + [#~~] -> [#int]. + [#<] -> [#int]. + [#<=] -> [#int]. + [#>] -> [#int]. + [#>=] -> [#int]. + [#between:and:] -> [#int]. + [#anyMask:] -> [#int]. + [#allMask:] -> [#int]. + [#noMask:] -> [#int]. + [#isNil] -> [#int]. + [#notNil] -> [#int]. + [#&] -> [#int]. + [#|] -> [#int]. + [#not] -> [#int]. [#asFloat] -> [#double]. [#atan] -> [#double]. [#exp] -> [#double]. [#log] -> [#double]. [#sin] -> [#double]. [#sqrt] -> [#double]. [#asLong] -> [#long]. + [#asInteger] -> [#sqInt]. [#asUnsignedInteger] -> [#usqInt]. [#asUnsignedLong] -> [#'unsigned long']. [#asVoidPointer] -> [#'void *']. [#signedIntToLong] -> [#usqInt]. "c.f. generateSignedIntToLong:on:indent:" [#signedIntToShort] -> [#usqInt]. "c.f. generateSignedIntToShort:on:indent:" [#cCoerce:to:] -> [sendNode args last value]. [#cCoerceSimple:to:] -> [sendNode args last value]. + [#sizeof:] -> [#'unsigned long']. "Technically it's a size_t but it matches unsigned long on target architectures so far..." [#ifTrue:ifFalse:] -> [self typeForConditional: sendNode in: aTMethod]. [#ifFalse:ifTrue:] -> [self typeForConditional: sendNode in: aTMethod]. [#ifTrue:] -> [self typeForConditional: sendNode in: aTMethod]. [#ifFalse:] -> [self typeForConditional: sendNode in: aTMethod] } otherwise: "If there /is/ a method for sel but its retrn type is as yet unknown we /mustn't/ default it. We can only default unbound selectors." [methodOrNil ifNotNil: [nil] ifNil: [#sqInt]]]! Item was added: + ----- Method: CCodeGenerator>>unsignedTypeForIntegralType: (in category 'type inference') ----- + unsignedTypeForIntegralType: aCTypeString + ^aCTypeString first = $u + ifTrue: [aCTypeString] + ifFalse: + [(aCTypeString beginsWith: 'sq') + ifTrue: ['u' , aCTypeString] + ifFalse: ['unsigned ' , aCTypeString]]! Item was changed: ----- Method: TMethod>>addTypesFor:to:in: (in category 'type inference') ----- addTypesFor: node to: typeSet in: aCodeGen + "Add the value types for the node to typeSet. - "Add the value tupes for the node to typeSet. Answer if any type was derived from an as-yet-untyped method, which allows us to abort inferReturnTypeFromReturnsIn: if the return type depends on a yet-to-be-typed method." | expr | expr := node. [expr isAssignment or: [expr isStmtList]] whileTrue: [expr isAssignment ifTrue: [expr := expr variable]. expr isStmtList ifTrue: [expr := expr statements last]]. expr isSend ifTrue: [(#(ifTrue: ifFalse: ifTrue:ifFalse: ifFalse:ifTrue:) includes: expr selector) ifTrue: [^expr args inject: false into: [:asYetUntyped :block| asYetUntyped | (self addTypesFor: block to: typeSet in: aCodeGen)]]. + (expr typeOrNilFrom: aCodeGen in: self) + ifNil: [^(aCodeGen methodNamed: expr selector) notNil and: [expr selector ~~ selector]] + ifNotNil: + [:type | + typeSet add: type. + ^false].]. - (#(= ~= == ~~ < > <= >= anyMask: allMask: noMask:) includes: expr selector) ifTrue: - [typeSet add: #sqInt. ^false]. - (#(+ - * / // \\ rem: quo: bitAnd: bitClear: bitOr: bitXor: bitShift:) includes: expr selector) ifTrue: - [| types | - types := Set new. - self addTypesFor: expr receiver to: types in: aCodeGen. - (types size = 1 and: [types anyOne last = $*]) ifTrue: "pointer arithmetic" - [typeSet add: types anyOne. ^false]. - self addTypesFor: expr args first to: types in: aCodeGen. - types := aCodeGen harmonizeReturnTypesIn: types. - types size = 2 ifTrue: - [(types includes: #double) ifTrue: - [typeSet add: #double. ^false]. - (types includes: #float) ifTrue: - [typeSet add: #float. ^false]. - ^false]. "don't know; leave unspecified." - types notEmpty ifTrue: - [typeSet add: types anyOne]. - ^false]. - "Abort only for untyped methods that will be typed, but don't be phased by recursion." - ^(aCodeGen returnTypeForSend: expr in: self) - ifNotNil: [:type| typeSet add: type. false] - ifNil: [(aCodeGen methodNamed: expr selector) notNil and: [expr selector ~~ selector]]]. expr isVariable ifTrue: [(aCodeGen typeOfVariable: expr name) ifNotNil: [:type| typeSet add: type] ifNil: [typeSet add: (expr name = 'self' ifTrue: [#void] ifFalse: [#sqInt])]]. expr isConstant ifTrue: + [(expr typeOrNilFrom: aCodeGen in: self) + ifNotNil: [:type | typeSet add: type]].. - [| val | - val := expr value. - val isInteger ifTrue: - [typeSet add: ((val >= 0 ifTrue: [val] ifFalse: [-1 - val]) highBit <= 32 - ifTrue: [#sqInt] - ifFalse: [#sqLong])]. - (#(nil true false) includes: val) ifTrue: - [typeSet add: #sqInt]. - val isFloat ifTrue: - [typeSet add: #float]]. ^false! Item was changed: ----- 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 sqLong usqLong + #'int *' #'unsigned int *' #'sqInt *' #'usqInt *' #'sqLong *' #'usqLong *' #'CogMethod *' #'char *') includes: returnType! - ^#(sqInt usqInt sqLong usqLong #'sqInt *' #'CogMethod *' #'char *') includes: returnType!

1 0

VM Maker: VMMaker.oscog-nice.1838.mcz
by commits＠source.squeak.org 24 Apr '16

24 Apr '16

Nicolas Cellier uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-nice.1838.mcz ==================== Summary ==================== Name: VMMaker.oscog-nice.1838 Author: nice Time: 24 April 2016, 1:47:02.874 pm UUID: 73cf9fda-4e5a-4e0c-a1b9-92ace5f8f3c9 Ancestors: VMMaker.oscog-nice.1837 Use a type inference that better fits C rules (C99 standard section 6.3): - in (float op int), int is promoted to double not float - in (unsigned short op short) both operands are promoted to #int due to int promotion rules. The idea behind this is that behaviour of generated code for inlined and non inlined code should be the same. We can guaranty such property only if we adhere to C rules. =============== Diff against VMMaker.oscog-nice.1837 =============== Item was changed: ----- Method: CCodeGenerator>>computeKernelReturnTypes (in category 'public') ----- computeKernelReturnTypes | dictionary | dictionary := Dictionary newFromPairs: #(oopAt: #sqInt oopAt:put: #sqInt oopAtPointer: #sqInt oopAtPointer:put: #sqInt byteAt: #sqInt byteAt:put: #sqInt byteAtPointer: #sqInt byteAtPointer:put: #sqInt shortAt: #sqInt shortAt:put: #sqInt shortAtPointer: #sqInt shortAtPointer:put: #sqInt intAt: #sqInt intAt:put: #sqInt intAtPointer: #sqInt intAtPointer:put: #sqInt longAt: #sqInt longAt:put: #sqInt longAtPointer: #sqInt longAtPointer:put: #sqInt + long32At: #int long32At:put: #int - long32At: #sqInt long32At:put: #sqInt unalignedLongAt: #sqInt unalignedLongAt:put: #sqInt + unalignedLong32At: #int unalignedLong32At:put: #int - unalignedLong32At: #sqInt unalignedLong32At:put: #sqInt long64At: #sqLong long64At:put: #sqLong fetchFloatAt:into: #void storeFloatAt:from: #void fetchFloatAtPointer:into: #void storeFloatAtPointer:from: #void fetchSingleFloatAt:into: #void storeSingleFloatAt:from: #void fetchSingleFloatAtPointer:into: #void storeSingleFloatAtPointer:from: #void pointerForOop: #'char *' oopForPointer: #sqInt). - BytesPerWord = 8 ifTrue: - [#(long32At: long32At:put: unalignedLong32At: unalignedLong32At:put:) do: - [:accessor| - dictionary at: accessor put: #int]]. ^dictionary! Item was changed: ----- Method: CCodeGenerator>>promoteArithmeticTypes:and: (in category 'type inference') ----- promoteArithmeticTypes: firstType and: secondType "Answer the return type for an arithmetic send. This is so that the inliner can still inline + simple expressions. Deal with pointer arithmetic, floating point arithmetic and promotion. + It is important to choose deterministically to get stable source generation. + Also, the behaviour of inlined and non inlined code should be unchanged." - simple expressions. Deal with pointer arithmetic, floating point arithmetic and promotion." - | firstSize secondSize | ((#(#double float) includes: firstType) or: [#(#double float) includes: secondType]) ifTrue: + [^(firstType = #double or: [secondType = #double]) + ifTrue: [#double] + ifFalse: [#float] "in C99 6.3.1.8, float+int, int is converted as a float"]. - [^(firstType = #float and: [secondType = #float]) - ifTrue: [#float] - ifFalse: [#double]]. "deal with unknowns, answering nil." (firstType isNil or: [secondType isNil]) ifTrue: [^nil]. + "Deal with integer promotion and arithmetic conversion" + ^self promoteIntegerArithmeticTypes: firstType and: secondType! - "Deal with promotion; answer the longest type, defaulting to the recever if they're the same. - See e.g. section 6.3.1.8 Usual arithmetic conversions, from the C99 standard: - Otherwise, the integer promotions are performed on both operands - Then the following rules are applied to the promoted operands: - - If both operands have the same type, then no further conversion is needed. - - Otherwise, if both operands have signed integer types or both have unsigned integer - types, the operand with the type of lesser integer conversion rank is converted to the - type of the operand with greater rank. - - Otherwise, if the operand that has unsigned integer type has rank greater or equal to - the rank of the type of the other operand, then the operand with signed integer type - is converted to the type of the operand with unsigned integer type. - - Otherwise, if the type of the operand with signed integer type can represent all of the - values of the type of the operand with unsigned integer type, then the operand with - unsigned integer type is converted to the type of the operand with signed integer type. - - Otherwise, both operands are converted to the unsigned integer type corresponding to - the type of the operand with signed integer type. - - It is important to choose deterministically to get stable source generation. So if the types have - the same size but differ in signedness we choose the unsigned type, which is in partial agreement - with the above" - ^(firstSize := self sizeOfIntegralCType: firstType) = (secondSize := self sizeOfIntegralCType: secondType) - ifTrue: - [(firstType first = $u) - ifTrue: [firstType] - ifFalse: [(secondType first = $u) ifTrue: [secondType] ifFalse: [firstType]]] - ifFalse: - [firstSize > secondSize ifTrue: [firstType] ifFalse: [secondType]]! Item was added: + ----- Method: CCodeGenerator>>promoteIntegerArithmeticTypes:and: (in category 'type inference') ----- + promoteIntegerArithmeticTypes: firstType and: secondType + "Answer the return type for an arithmetic send. + Deal with integer promotion rules of C99. + See section 6.3 Conversions of the standard. + + 6.3.1.1 ...snip... + If an int can represent all values of the original type, the value is converted to an int; + otherwise, it is converted to an unsigned int. These are called the integer promotions. + All other types are unchanged by the inte ger promotions + + 6.3.1.8 ...snip... + Otherwise, the integer promotions are performed on both operands + Then the following rules are applied to the promoted operands: + + If both operands have the same type, then no further conversion is needed. + + Otherwise, if both operands have signed integer types or both have unsigned integer + types, the operand with the type of lesser integer conversion rank is converted to the + type of the operand with greater rank. + + Otherwise, if the operand that has unsigned integer type has rank greater or equal to + the rank of the type of the other operand, then the operand with signed integer type + is converted to the type of the operand with unsigned integer type. + + Otherwise, if the type of the operand with signed integer type can represent all of the + values of the type of the operand with unsigned integer type, then the operand with + unsigned integer type is converted to the type of the operand with signed integer type. + + Otherwise, both operands are converted to the unsigned integer type corresponding to + the type of the operand with signed integer type. + + This is so that the generated code behaviour is unsensitive to inlining." + | length1 length2 intSize | + length1 := self sizeOfIntegralCType: firstType. + length2 := self sizeOfIntegralCType: secondType. + intSize := self sizeOfIntegralCType: #int. + (length1 < intSize and: [length2 < intSize]) ifTrue: [^#int]. "Integer promotion" + length1 > length2 ifTrue: [^firstType]. + length2 > length1 ifTrue: [^secondType]. + firstType first = $u ifTrue: [^firstType]. + secondType first = $u ifTrue: [^secondType]. + ^firstType!

1 0

VM Maker: VMMaker.oscog-nice.1837.mcz
by commits＠source.squeak.org 24 Apr '16

24 Apr '16

Nicolas Cellier uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker.oscog-nice.1837.mcz ==================== Summary ==================== Name: VMMaker.oscog-nice.1837 Author: nice Time: 24 April 2016, 1:37:15.458 pm UUID: 7c7bd77c-ff12-4085-bbe1-9ba0962b5709 Ancestors: VMMaker.oscog-eem.1834 Extend type inference capabilities from the AST. Use minimum generality for generated integer constants: - use int if variable fits in int - else unsigned int - else long long (long would be enough for 64bits versions, but long long works for both 32 and 64) - else unsigned long long This is to avoid spurious unsigned promotion due to improved type inference. Protect #pathTo:followWeak: which can currently only work with a signed int, from possibly future spurious unsigned inference (I didn't bissect at which stage exactly this happens, but it happened). =============== Diff against VMMaker.oscog-eem.1834 =============== Item was changed: ----- Method: CCodeGenerator>>cLiteralForInteger:hex: (in category 'C code generator') ----- cLiteralForInteger: anInteger hex: aBoolean + "Answer the string for generating a literal integer. + Use hexadecimal notation as prescribed by aBoolean. + Use long long suffix (LL) if the integer does not fit on 32 bits. + Use unsigned suffix (U) if the integer does not fit on a signed integer (resp. long long). + Correctly generate INT_MIN and LONG_LONG_MIN. + Indeed -0x8000000 is parsed as - (0x8000000) by C Compiler. + 0x8000000 does not fit on a signed int, it is interpreted as unsigned. + That makes INT_MIN unsigned which is badly broken..." + | printString | printString := aBoolean ifTrue: [anInteger positive ifTrue: ['0x' , (anInteger printStringBase: 16)] ifFalse: ['-0x' , (anInteger negated printStringBase: 16)]] ifFalse: [anInteger printString]. + ^anInteger positive + ifTrue: [anInteger > 16r7FFFFFFF "INT_MAX" + ifTrue: [anInteger > 16rFFFFFFFF "UINT_MAX" + ifTrue: [anInteger > 16r7FFFFFFFFFFFFFFF "LONG_LONG_MAX" + ifTrue: [printString , 'ULL'] + ifFalse: [printString , 'LL']] + ifFalse: [printString , 'U']] + ifFalse: [printString]] + ifFalse: [anInteger < -16r8000000 + ifTrue: [anInteger = -16r800000000000000 "LONG_LONG_MIN" + ifTrue: ['(-0x7FFFFFFFFFFFFFFFLL-1)'] + ifFalse: [printString , 'LL']] + ifFalse: [anInteger = -16r8000000 "INT_MIN" + ifTrue: ['(-0x7FFFFFFF-1)'] + ifFalse: [printString]]]! - ^anInteger > 16rFFFFFFFF - ifTrue: [printString, ObjectMemory unsignedLongLongSuffix] - ifFalse: [anInteger < 16r7FFFFFFF - ifTrue: [printString] - ifFalse: [printString, ObjectMemory unsignedIntegerSuffix]]! Item was changed: ----- Method: CoInterpreterPrimitives>>pathTo:using:followWeak: (in category 'object access primitives') ----- pathTo: goal using: stack followWeak: followWeak "Trace objects and frames from the root, marking visited objects, pushing the current path on stack, until goal is found. If found, unmark, leaving path in stack, and answer 0. Otherwise answer an error: PrimErrBadArgument if stack is not an Array PrimErrBadIndex if search overflows stack PrimErrNotFound if goal cannot be found" | current index next stackSize stackp freeStartAtStart | + <var: #index type: #sqInt> "Force the sign because typeInference does not seem to work" (objectMemory isArray: stack) ifFalse: [^PrimErrBadArgument]. self assert: objectMemory allObjectsUnmarked. freeStartAtStart := objectMemory freeStart. "check no allocations during search" objectMemory beRootIfOld: stack. "so no store checks are necessary on stack" stackSize := objectMemory lengthOf: stack. objectMemory mark: stack. "no need. the current context is not reachable from the active process (suspendedContext is nil)" "objectMemory mark: self activeProcess." current := objectMemory specialObjectsOop. objectMemory mark: current. index := objectMemory lengthOf: current. stackp := 0. [[(index := index - 1) >= -1] whileTrue: [(stackPages couldBeFramePointer: current) ifTrue: [next := index >= 0 ifTrue: [self field: index ofFrame: (self cCoerceSimple: current to: #'char *')] ifFalse: [objectMemory nilObject]] ifFalse: [index >= 0 ifTrue: [next := (objectMemory isContextNonImm: current) ifTrue: [self fieldOrSenderFP: index ofContext: current] ifFalse: [objectMemory fetchPointer: index ofObject: current]] ifFalse: [next := objectMemory fetchClassOfNonImm: current]]. (stackPages couldBeFramePointer: next) ifTrue: [self assert: (self isFrame: (self cCoerceSimple: next to: #'char *') onPage: (stackPages stackPageFor: (self cCoerceSimple: next to: #'char *')))] ifFalse: [next >= heapBase ifTrue: "exclude Cog methods" [self assert: (self checkOkayOop: next)]]. next = goal ifTrue: [self assert: freeStartAtStart = objectMemory freeStart. self unmarkAfterPathTo. objectMemory storePointer: stackp ofObject: stack withValue: current. self pruneStack: stack stackp: stackp. ^0]. ((objectMemory isNonIntegerObject: next) and: [(stackPages couldBeFramePointer: next) ifTrue: [(self frameIsMarked: next) not] ifFalse: [next >= heapBase "exclude Cog methods" and: [(objectMemory isMarked: next) not and: [((objectMemory isPointers: next) or: [objectMemory isCompiledMethod: next]) and: [followWeak or: [(objectMemory isWeakNonImm: next) not]]]]]]) ifTrue: [stackp + 2 > stackSize ifTrue: [self assert: freeStartAtStart = objectMemory freeStart. self unmarkAfterPathTo. objectMemory nilFieldsOf: stack. ^PrimErrBadIndex]. "PrimErrNoMemory ?" objectMemory storePointerUnchecked: stackp ofObject: stack withValue: current; storePointerUnchecked: stackp + 1 ofObject: stack withValue: (objectMemory integerObjectOf: index). stackp := stackp + 2. (stackPages couldBeFramePointer: (self cCoerceSimple: next to: #'char *')) ifTrue: [self markFrame: next. index := self fieldsInFrame: (self cCoerceSimple: next to: #'char *')] ifFalse: [objectMemory mark: next. (objectMemory isCompiledMethod: next) ifTrue: [index := (objectMemory literalCountOf: next) + LiteralStart] ifFalse: [index := objectMemory lengthOf: next]]. current := next]]. current = objectMemory specialObjectsOop ifTrue: [self assert: freeStartAtStart = objectMemory freeStart. self unmarkAfterPathTo. objectMemory nilFieldsOf: stack. ^PrimErrNotFound]. index := objectMemory integerValueOf: (objectMemory fetchPointer: stackp - 1 ofObject: stack). current := objectMemory fetchPointer: stackp - 2 ofObject: stack. stackp := stackp - 2] repeat! Item was changed: ----- Method: StackInterpreterPrimitives>>pathTo:using:followWeak: (in category 'object access primitives') ----- pathTo: goal using: stack followWeak: followWeak "Trace objects and frames from the root, marking visited objects, pushing the current path on stack, until goal is found. If found, unmark, leaving path in stack, and answer 0. Otherwise answer an error: PrimErrBadArgument if stack is not an Array PrimErrBadIndex if search overflows stack PrimErrNotFound if goal cannot be found" + <var: #index type: #sqInt> "beware, must be signed" | current index next stackSize stackp freeStartAtStart | (objectMemory isArray: stack) ifFalse: [^PrimErrBadArgument]. self assert: objectMemory allObjectsUnmarked. freeStartAtStart := objectMemory freeStart. "check no allocations during search" objectMemory beRootIfOld: stack. "so no store checks are necessary on stack" stackSize := objectMemory lengthOf: stack. objectMemory mark: stack. "no need. the current context is not reachable from the active process (suspendedContext is nil)" "objectMemory mark: self activeProcess." current := objectMemory specialObjectsOop. objectMemory mark: current. index := objectMemory lengthOf: current. stackp := 0. [[(index := index - 1) >= -1] whileTrue: [(stackPages couldBeFramePointer: current) ifTrue: [next := index >= 0 ifTrue: [self field: index ofFrame: (self cCoerceSimple: current to: #'char *')] ifFalse: [objectMemory nilObject]] ifFalse: [index >= 0 ifTrue: [next := (objectMemory isContextNonImm: current) ifTrue: [self fieldOrSenderFP: index ofContext: current] ifFalse: [objectMemory fetchPointer: index ofObject: current]] ifFalse: [next := objectMemory fetchClassOfNonImm: current]]. (stackPages couldBeFramePointer: next) ifTrue: [self assert: (self isFrame: (self cCoerceSimple: next to: #'char *') onPage: (stackPages stackPageFor: (self cCoerceSimple: next to: #'char *')))] ifFalse: [self assert: (self checkOkayOop: next)]. next = goal ifTrue: [self assert: freeStartAtStart = objectMemory freeStart. self unmarkAfterPathTo. objectMemory storePointer: stackp ofObject: stack withValue: current. self pruneStack: stack stackp: stackp. ^0]. ((objectMemory isNonIntegerObject: next) and: [(stackPages couldBeFramePointer: next) ifTrue: [(self frameIsMarked: next) not] ifFalse: [(objectMemory isMarked: next) not and: [((objectMemory isPointers: next) or: [objectMemory isCompiledMethod: next]) and: [followWeak or: [(objectMemory isWeakNonImm: next) not]]]]]) ifTrue: [stackp + 2 > stackSize ifTrue: [self assert: freeStartAtStart = objectMemory freeStart. self unmarkAfterPathTo. objectMemory nilFieldsOf: stack. ^PrimErrBadIndex]. "PrimErrNoMemory ?" objectMemory storePointerUnchecked: stackp ofObject: stack withValue: current; storePointerUnchecked: stackp + 1 ofObject: stack withValue: (objectMemory integerObjectOf: index). stackp := stackp + 2. (stackPages couldBeFramePointer: (self cCoerceSimple: next to: #'char *')) ifTrue: [self markFrame: next. index := self fieldsInFrame: (self cCoerceSimple: next to: #'char *')] ifFalse: [objectMemory mark: next. (objectMemory isCompiledMethod: next) ifTrue: [index := (objectMemory literalCountOf: next) + LiteralStart] ifFalse: [index := objectMemory lengthOf: next]]. current := next]]. current = objectMemory specialObjectsOop ifTrue: [self assert: freeStartAtStart = objectMemory freeStart. self unmarkAfterPathTo. objectMemory nilFieldsOf: stack. ^PrimErrNotFound]. index := objectMemory integerValueOf: (objectMemory fetchPointer: stackp - 1 ofObject: stack). current := objectMemory fetchPointer: stackp - 2 ofObject: stack. stackp := stackp - 2] repeat! Item was added: + ----- Method: TAssignmentNode>>typeOrNilFrom:in: (in category 'type inference') ----- + typeOrNilFrom: aCodeGenerator in: aTMethod + "This is the default type in case of doubt" + ^variable typeOrNilFrom: aCodeGenerator in: aTMethod! Item was added: + ----- Method: TConstantNode>>typeOrNilFrom:in: (in category 'type inference') ----- + typeOrNilFrom: aCodeGenerator in: aTMethod + "For integers, answer int unless the value does not fit into a 32bits signed int. + In that case, answer the shortest architecture independant integer type that could hold the constant. + This method must be consistent with CCodeGenerator>>cLiteralFor:" + | hb | + value isInteger + ifTrue: + [value positive + ifTrue: + [hb := value highBit. + hb < 32 ifTrue: [^#int]. + hb = 32 ifTrue: [^#'unsigned int']. + hb = 64 ifTrue: [^#'unsigned long long']. + ^#'long long'] + ifFalse: + [hb := value bitInvert highBit. + hb < 32 ifTrue: [^#int]. + ^#'long long']]. + value isFloat ifTrue: [^#double]. + (#(nil true false) includes: value) ifTrue: [^#int]. + ^nil! Item was added: + ----- Method: TInlineNode>>typeOrNilFrom:in: (in category 'type inference') ----- + typeOrNilFrom: aCodeGenerator in: aTMethod + ^method returnType! Item was added: + ----- Method: TReturnNode>>typeOrNilFrom:in: (in category 'type inference') ----- + typeOrNilFrom: aCodeGenerator in: aTMethod + ^expression typeOrNilFrom: aCodeGenerator in: aTMethod! Item was added: + ----- Method: TStmtListNode>>typeOrNilFrom:in: (in category 'type inference') ----- + typeOrNilFrom: aCodeGenerator in: aTMethod + statements isEmpty ifTrue: [^nil]. + ^statements last typeOrNilFrom: aCodeGenerator in: aTMethod!

1 0

bitXor: is slower than - on Spur32 (MacOSX) r3684
by Nicolas Cellier 24 Apr '16

24 Apr '16

Hi, for large integers (> 64 bits), a bitClear: implementation using bitXor: is faster than by using - indeed we don't have to care of carry. { [(12398767868759866578644 bitOr: 567812564128976768553) - 567812564128976768553] bench. [(12398767868759866578644 bitOr: 567812564128976768553) bitXor: 567812564128976768553] bench. } #('3,480,000 per second. 288 nanoseconds per run.' '4,950,000 per second. 202 nanoseconds per run.') For large integers < 64 bits or small integers, I would expect sort of similar speed, but bitXor: is slower: { [(1234 bitOr: 5678) - 5678] bench. [(1234 bitOr: 5678) bitXor: 5678] bench. } #('170,000,000 per second. 5.88 nanoseconds per run.' '133,000,000 per second. 7.53 nanoseconds per run.') { [(12398767868754 bitOr: 5678125641253) - 5678125641253] bench. [(12398767868754 bitOr: 5678125641253) bitXor: 5678125641253] bench. } #('7,100,000 per second. 141 nanoseconds per run.' '5,110,000 per second. 196 nanoseconds per run.') OK, (Smalltalk specialSelectors includes: #bitXor:) is false, but is it the sole explanation?

3 4

[commit][3685] Fix missing newline at end of sqWin32File.h
by commits＠squeakvm.org 23 Apr '16

23 Apr '16

Revision: 3685 Author: eliot Date: 2016-04-23 11:06:55 -0700 (Sat, 23 Apr 2016) Log Message: ----------- Fix missing newline at end of sqWin32File.h Modified Paths: -------------- trunk/platforms/win32/plugins/FilePlugin/sqWin32File.h Modified: trunk/platforms/win32/plugins/FilePlugin/sqWin32File.h =================================================================== --- trunk/platforms/win32/plugins/FilePlugin/sqWin32File.h 2016-04-23 04:32:43 UTC (rev 3684) +++ trunk/platforms/win32/plugins/FilePlugin/sqWin32File.h 2016-04-23 18:06:55 UTC (rev 3685) @@ -55,4 +55,4 @@ in_out_wide_path[in_size] = 0; \ } -#endif \ No newline at end of file +#endif /* __SQ_WIN32_FILE_H */

1 0

[commit][3682] Remove double backslashes at end of default implementation of new swap macros .
by commits＠squeakvm.org 23 Apr '16

23 Apr '16

Revision: 3682 Author: eliot Date: 2016-04-22 20:23:49 -0700 (Fri, 22 Apr 2016) Log Message: ----------- Remove double backslashes at end of default implementation of new swap macros. Modified Paths: -------------- branches/Cog/platforms/Cross/vm/sqMemoryAccess.h Property Changed: ---------------- branches/Cog/platforms/Cross/vm/sqSCCSVersion.h Modified: branches/Cog/platforms/Cross/vm/sqMemoryAccess.h =================================================================== --- branches/Cog/platforms/Cross/vm/sqMemoryAccess.h 2016-04-23 00:59:25 UTC (rev 3681) +++ branches/Cog/platforms/Cross/vm/sqMemoryAccess.h 2016-04-23 03:23:49 UTC (rev 3682) @@ -218,18 +218,18 @@ # define SQ_SWAP_8_BYTES(x) _byteswap_uint64(x) #else # define SQ_SWAP_4_BYTES(x) \ - (((unsigned int)(x) << 24) | \\ - (((unsigned int)(x) << 8) & 0xff0000U) | \\ - (((unsigned int)(x) >> 8) & 0xff00U) | \\ + (((unsigned int)(x) << 24) | \ + (((unsigned int)(x) << 8) & 0xff0000U) | \ + (((unsigned int)(x) >> 8) & 0xff00U) | \ ( (unsigned int)(x) >> 24)) # define SQ_SWAP_8_BYTES(x) \ - (((unsigned long long)(x) << 56) | \\ - (((unsigned long long)(x) << 40) & 0xff000000000000ULL) | \\ - (((unsigned long long)(x) << 24) & 0xff0000000000ULL) | \\ - (((unsigned long long)(x) << 8) & 0xff00000000ULL) | \\ - (((unsigned long long)(x) >> 8) & 0xff000000ULL) | \\ - (((unsigned long long)(x) >> 24) & 0xff0000ULL) | \\ - (((unsigned long long)(x) >> 40) & 0xff00ULL) | \\ + (((unsigned long long)(x) << 56) | \ + (((unsigned long long)(x) << 40) & 0xff000000000000ULL) | \ + (((unsigned long long)(x) << 24) & 0xff0000000000ULL) | \ + (((unsigned long long)(x) << 8) & 0xff00000000ULL) | \ + (((unsigned long long)(x) >> 8) & 0xff000000ULL) | \ + (((unsigned long long)(x) >> 24) & 0xff0000ULL) | \ + (((unsigned long long)(x) >> 40) & 0xff00ULL) | \ ( (unsigned long long)(x) >> 56)) #endif Property changes on: branches/Cog/platforms/Cross/vm/sqSCCSVersion.h ___________________________________________________________________ Modified: checkindate - Fri Apr 22 17:58:08 PDT 2016 + Fri Apr 22 20:22:44 PDT 2016

3 2

← Newer
1
2
3
4
5
6
7
8
...
21
Older →

Jump to page:

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

Vm-dev April 2016