VM Maker: VMMaker.oscog-eem.2302.mcz

 
Eliot Miranda uploaded a new version of VMMaker to project VM Maker:
http://source.squeak.org/VMMaker/VMMaker.oscog-eem.2302.mcz

==================== Summary ====================

Name: VMMaker.oscog-eem.2302
Author: eem
Time: 21 December 2017, 10:26:07.699491 pm
UUID: 55ec8f63-cdbe-4e79-8f22-48fdea585b88
Ancestors: VMMaker.oscog-eem.2301

Fix a slip in the new ThreadedFFIPlugin preambleCCode (the atput's need an extra parameter).

Comment the reason genPrimitiveAt can't be a CompletePrimitive (the same reason genPrimitiveAtPut can't be).

=============== Diff against VMMaker.oscog-eem.2301 ===============

Item was changed:
  ----- Method: CogObjectRepresentationFor64BitSpur>>genPrimitiveAt (in category 'primitive generators') -----
  genPrimitiveAt
  | formatReg nSlotsOrElementsReg convertToIntAndReturn methodInBounds
   jumpNotIndexable jumpImmediate jumpBadIndex
   jumpIsBytes jumpIsShorts jumpIsWords jumpIsLongs jumpIsMethod jumpIsArray jumpIsContext
   jumpHasFixedFields jumpArrayOutOfBounds jumpFixedFieldsOutOfBounds
   jumpBytesOutOfBounds jumpShortsOutOfBounds jumpWordsOutOfBounds jumpLongsOutOfBounds
   jumpFailAlloc jumpNotSmallInteger |
  "c.f. StackInterpreter>>stSizeOf: SpurMemoryManager>>lengthOf:format: fixedFieldsOf:format:length:"
  <var: #jumpIsBytes type: #'AbstractInstruction *'>
  <var: #jumpIsLongs type: #'AbstractInstruction *'>
  <var: #jumpFailAlloc type: #'AbstractInstruction *'>
  <var: #jumpIsShorts type: #'AbstractInstruction *'>
  <var: #jumpIsMethod type: #'AbstractInstruction *'>
  <var: #jumpBadIndex type: #'AbstractInstruction *'>
  <var: #jumpIsContext type: #'AbstractInstruction *'>
  <var: #jumpImmediate type: #'AbstractInstruction *'>
  <var: #methodInBounds type: #'AbstractInstruction *'>
  <var: #jumpNotIndexable type: #'AbstractInstruction *'>
  <var: #jumpHasFixedFields type: #'AbstractInstruction *'>
  <var: #jumpNotSmallInteger type: #'AbstractInstruction *'>
  <var: #convertToIntAndReturn type: #'AbstractInstruction *'>
  <var: #jumpArrayOutOfBounds type: #'AbstractInstruction *'>
  <var: #jumpBytesOutOfBounds type: #'AbstractInstruction *'>
  <var: #jumpLongsOutOfBounds type: #'AbstractInstruction *'>
  <var: #jumpShortsOutOfBounds type: #'AbstractInstruction *'>
  <var: #jumpWordsOutOfBounds type: #'AbstractInstruction *'>
  <var: #jumpFixedFieldsOutOfBounds type: #'AbstractInstruction *'>
 
  nSlotsOrElementsReg := ClassReg.
 
  cogit genLoadArgAtDepth: 0 into: Arg0Reg.
  jumpImmediate := self genJumpImmediate: ReceiverResultReg.
  cogit MoveR: Arg0Reg R: Arg1Reg.
  jumpBadIndex := self genJumpNotSmallInteger: Arg0Reg scratchReg: TempReg.
  self genConvertSmallIntegerToIntegerInReg: Arg1Reg.
  cogit SubCq: 1 R: Arg1Reg. "1-rel => 0-rel"
 
  "formatReg := self formatOf: ReceiverResultReg"
  self genGetFormatOf: ReceiverResultReg
  into: (formatReg := SendNumArgsReg)
  leastSignificantHalfOfBaseHeaderIntoScratch: TempReg.
 
  self genGetNumSlotsOf: ReceiverResultReg into: nSlotsOrElementsReg.
 
  "dispatch on format in a combination of highest dynamic frequency order first and convenience.
   0 = 0 sized objects (UndefinedObject True False et al)
   1 = non-indexable objects with inst vars (Point et al)
   2 = indexable objects with no inst vars (Array et al)
   3 = indexable objects with inst vars (MethodContext AdditionalMethodState et al)
   4 = weak indexable objects with inst vars (WeakArray et al)
   5 = weak non-indexable objects with inst vars (ephemerons) (Ephemeron)
   6 unused, reserved for exotic pointer objects?
   7 Forwarded Object, 1st field is pointer, rest of fields are ignored
   8 unused, reserved for exotic non-pointer objects?
   9 64-bit indexable
  10 - 11 32-bit indexable
  12 - 15 16-bit indexable
  16 - 23 byte indexable
  24 - 31 compiled method"
  cogit CmpCq: objectMemory firstByteFormat R: formatReg.
  jumpIsBytes := cogit JumpAboveOrEqual: 0.
  cogit CmpCq: objectMemory arrayFormat R: formatReg.
  jumpIsArray := cogit JumpZero: 0.
  jumpNotIndexable := cogit JumpBelow: 0.
  cogit CmpCq: objectMemory weakArrayFormat R: formatReg.
  jumpHasFixedFields := cogit JumpBelowOrEqual: 0.
  cogit CmpCq: objectMemory firstShortFormat R: formatReg.
  jumpIsShorts := cogit JumpAboveOrEqual: 0.
  cogit CmpCq: objectMemory firstLongFormat R: formatReg.
  jumpIsWords := cogit JumpAboveOrEqual: 0.
  cogit CmpCq: objectMemory sixtyFourBitIndexableFormat R: formatReg.
  jumpIsLongs := cogit JumpZero: 0.
  jumpNotIndexable jmpTarget: cogit Label.
  jumpNotIndexable := cogit Jump: 0.
 
  jumpIsArray jmpTarget:
  (cogit CmpR: Arg1Reg R: nSlotsOrElementsReg).
  jumpArrayOutOfBounds := cogit JumpBelowOrEqual: 0.
  cogit AddCq: objectMemory baseHeaderSize >> objectMemory shiftForWord R: Arg1Reg.
  cogit MoveXwr: Arg1Reg R: ReceiverResultReg R: ReceiverResultReg.
  cogit genPrimReturn.
 
  jumpIsBytes jmpTarget:
  (cogit LogicalShiftLeftCq: objectMemory shiftForWord R: nSlotsOrElementsReg).
  cogit AndCq: 7 R: formatReg R: TempReg.
  cogit SubR: TempReg R: nSlotsOrElementsReg;
  CmpR: Arg1Reg R: nSlotsOrElementsReg.
  jumpBytesOutOfBounds := cogit JumpBelowOrEqual: 0.
  cogit CmpCq: objectMemory firstCompiledMethodFormat R: formatReg.
  jumpIsMethod := cogit JumpAboveOrEqual: 0.
  methodInBounds :=
  (cogit AddCq: objectMemory baseHeaderSize R: Arg1Reg).
  cogit backEnd byteReadsZeroExtend
  ifTrue:
  [cogit MoveXbr: Arg1Reg R: ReceiverResultReg R: ReceiverResultReg]
  ifFalse:
  [cogit "formatReg already contains a value <= 16r1f, so no need to zero it"
  MoveXbr: Arg1Reg R: ReceiverResultReg R: formatReg;
  MoveR: formatReg R: ReceiverResultReg].
  convertToIntAndReturn := cogit Label.
  self genConvertIntegerToSmallIntegerInReg: ReceiverResultReg.
  cogit genPrimReturn.
 
  jumpIsShorts jmpTarget:
  (cogit LogicalShiftLeftCq: objectMemory shiftForWord - 1 R: nSlotsOrElementsReg).
  cogit AndCq: 3 R: formatReg.
  cogit SubR: formatReg R: nSlotsOrElementsReg;
  CmpR: Arg1Reg R: nSlotsOrElementsReg.
  jumpShortsOutOfBounds := cogit JumpBelowOrEqual: 0.
  cogit AddR: Arg1Reg R: ReceiverResultReg.
  cogit AddR: Arg1Reg R: ReceiverResultReg.
  cogit MoveM16: objectMemory baseHeaderSize r: ReceiverResultReg R: ReceiverResultReg.
  cogit Jump: convertToIntAndReturn.
 
  jumpIsWords jmpTarget:
  (cogit LogicalShiftLeftCq: objectMemory shiftForWord - 2 R: nSlotsOrElementsReg).
  cogit AndCq: 1 R: formatReg.
  cogit SubR: formatReg R: nSlotsOrElementsReg;
  CmpR: Arg1Reg R: nSlotsOrElementsReg.
  jumpWordsOutOfBounds := cogit JumpBelowOrEqual: 0.
  cogit AddCq: objectMemory baseHeaderSize >>  (objectMemory shiftForWord - 1) R: Arg1Reg.
  cogit MoveX32r: Arg1Reg R: ReceiverResultReg R: TempReg.
  cogit MoveR: TempReg R: ReceiverResultReg.
  cogit Jump: convertToIntAndReturn.
 
  jumpHasFixedFields jmpTarget:
  (cogit AndCq: objectMemory classIndexMask R: TempReg).
  cogit MoveR: TempReg R: formatReg.
  cogit CmpCq: ClassMethodContextCompactIndex R: TempReg.
  jumpIsContext := cogit JumpZero: 0.
  self genGetClassObjectOfClassIndex: formatReg into: Extra0Reg scratchReg: TempReg.
  self genLoadSlot: InstanceSpecificationIndex sourceReg: Extra0Reg destReg: formatReg.
  self genConvertSmallIntegerToIntegerInReg: formatReg.
  cogit
  AndCq: objectMemory fixedFieldsOfClassFormatMask R: formatReg;
  SubR: formatReg R: nSlotsOrElementsReg;
  CmpR: Arg1Reg R: nSlotsOrElementsReg.
  jumpFixedFieldsOutOfBounds := cogit JumpBelowOrEqual: 0.
  "index is (formatReg (fixed fields) + Arg1Reg (0-rel index)) * wordSize + baseHeaderSize"
  cogit AddR: formatReg R: Arg1Reg.
  cogit AddCq: objectMemory baseHeaderSize >> objectMemory shiftForWord R: Arg1Reg.
  cogit MoveXwr: Arg1Reg R: ReceiverResultReg R: ReceiverResultReg.
  cogit genPrimReturn.
 
  jumpIsLongs jmpTarget:
  (cogit CmpR: Arg1Reg R: nSlotsOrElementsReg).
  jumpLongsOutOfBounds := cogit JumpBelowOrEqual: 0.
  cogit AddCq: objectMemory baseHeaderSize >> objectMemory shiftForWord R: Arg1Reg.
  cogit MoveXwr: Arg1Reg R: ReceiverResultReg R: ClassReg.
  cogit MoveR: ClassReg R: TempReg.
  cogit LogicalShiftRightCq: self numSmallIntegerBits - 1 R: TempReg. "If in range this is now 0"
  (cogit lastOpcode setsConditionCodesFor: JumpZero) ifFalse:
  [cogit CmpCq: 0 R: TempReg]. "N.B. FLAGS := ClassReg - 0"
  jumpNotSmallInteger := cogit JumpNonZero: 0.
  cogit MoveR: ClassReg R: ReceiverResultReg.
  cogit Jump: convertToIntAndReturn.
  jumpNotSmallInteger jmpTarget: cogit Label.
  jumpFailAlloc := self genAlloc64BitIntegerValue: ClassReg into: SendNumArgsReg scratchReg: Extra0Reg scratchReg: TempReg.
  cogit MoveR: SendNumArgsReg R: ReceiverResultReg.
  cogit genPrimReturn.
 
  jumpIsMethod jmpTarget: cogit Label.
  "Now check that the index is beyond the method's literals..."
  self getLiteralCountOf: ReceiverResultReg plusOne: true inBytes: true into: nSlotsOrElementsReg scratch: TempReg.
  cogit CmpR: Arg1Reg R: nSlotsOrElementsReg.
  cogit JumpBelow: methodInBounds.
 
  jumpFailAlloc jmpTarget:
  (jumpLongsOutOfBounds jmpTarget:
  (jumpFixedFieldsOutOfBounds jmpTarget:
  (jumpArrayOutOfBounds jmpTarget:
  (jumpBytesOutOfBounds jmpTarget:
  (jumpShortsOutOfBounds jmpTarget:
  (jumpWordsOutOfBounds jmpTarget:
  (jumpNotIndexable jmpTarget:
  (jumpIsContext jmpTarget:
  (jumpBadIndex jmpTarget:
  (jumpImmediate jmpTarget: cogit Label)))))))))).
 
+ ^0 "Can't be complete because of contexts."!
- ^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))
  # elif defined(__GNUC__) && (defined(__amd64__) || defined(__x86_64__) ||  defined(__amd64) || defined(__x86_64))
  # define setsp(sp) asm volatile ("movq %0,%%rsp" : : "m"(sp))
  # elif defined(__GNUC__) && (defined(__arm__))
  # define setsp(sp) asm volatile ("ldr %%sp, %0" : : "m"(sp))
  #endif
  #if !!defined(getsp)
  # define getsp() 0
  #endif
  #if !!defined(setsp)
  # define setsp(ignored) 0
  #endif
 
  #if !!defined(STACK_ALIGN_BYTES)
  #  define STACK_ALIGN_BYTES 0
  #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(_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__))
      /*
       * 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.
       * Grab the actual stack pointer to correct.
       */
  # define ALLOCA_LIES_SO_SETSP_BEFORE_CALL 1
  # else
  # define ALLOCA_LIES_SO_SETSP_BEFORE_CALL 0
  # endif
  #endif /* !!defined(ALLOCA_LIES_SO_SETSP_BEFORE_CALL) */
 
  #if !!defined(PLATFORM_API_USES_CALLEE_POPS_CONVENTION)
  # define PLATFORM_API_USES_CALLEE_POPS_CONVENTION 0
  #endif
 
  /* This alignment stuff is a hack for integerAt:put:size:signed:/primitiveFFIIntegerAt[Put].
   * The assumption right now is that all processors suppoprt unaligned access.  That only
   * holds true for x86, x86-64 & ARMv6 & later.  But this keeps us going until we can addresws it properly.
   */
  #define unalignedShortAt(a) shortAt(a)
+ #define unalignedShortAtput(a,v) shortAtput(a,v)
- #define unalignedShortAtput(a) shortAtput(a)
  #define unalignedLong32At(a) long32At(a)
+ #define unalignedLong32Atput(a,v) long32Atput(a,v)
- #define unalignedLong32Atput(a) long32Atput(a)
  #define unalignedLong64At(a) long64At(a)
+ #define unalignedLong64Atput(a,v) long64Atput(a,v)
- #define unalignedLong64Atput(a) long64Atput(a)
 
  /* 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
  '!