[Vm-dev] VM Maker: VMMaker.oscog-eem.2634.mcz

commits at source.squeak.org commits at source.squeak.org
Wed Dec 25 02:28:22 UTC 2019


Eliot Miranda uploaded a new version of VMMaker to project VM Maker:
http://source.squeak.org/VMMaker/VMMaker.oscog-eem.2634.mcz

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

Name: VMMaker.oscog-eem.2634
Author: eem
Time: 24 December 2019, 6:28:07.401155 pm
UUID: 3a0d5136-0b7c-48e3-a748-0b0831a1dac2
Ancestors: VMMaker.oscog-eem.2633

Oops. Nuke the sender of nilLocalFP.

=============== Diff against VMMaker.oscog-eem.2633 ===============

Item was changed:
  ----- Method: Cogit>>simulateEnilopmart:numArgs: (in category 'simulation only') -----
  simulateEnilopmart: enilopmartAddress numArgs: n
  	<doNotGenerate>
  	"Enter Cog code, popping the class reg and receiver from the stack
  	 and then returning to the address beneath them.
  	 In the actual VM the enilopmart is a function pointer and so senders
  	 of this method end up calling the enilopmart to enter machine code.
  	 In simulation we either need to start simulating execution (if we're in
  	 the interpreter) or return to the simulation (if we're in the run-time
  	 called from machine code. We should also smash the register state
  	 since, being an abnormal entry, no saved registers will be restored."
  	self assert: (coInterpreter isOnRumpCStack: processor sp).
  	self assert: ((coInterpreter stackValue: n) between: guardPageSize and: methodZone freeStart - 1).
- 	"As a convenience for stack printing, nil localFP so we know we're in machine code."
- 	coInterpreter nilLocalFP.
  	(printInstructions or: [printRegisters]) ifTrue:
  		[coInterpreter printExternalHeadFrame].
  	processor
  		smashRegistersWithValuesFrom: 16r80000000 by: objectMemory wordSize;
  		simulateLeafCallOf: enilopmartAddress
  		nextpc: 16rBADF00D
  		memory: coInterpreter memory.
  	"If we're already simulating in the context of machine code then
  	 this will take us back to handleCallSimulationTrap:.  Otherwise
  	 start executing machine code in the simulator."
  	(ReenterMachineCode new returnValue: #continueNoReturn) signal.
  	self simulateCogCodeAt: enilopmartAddress.
  	"We should either longjmp back to the interpreter or
  	 stay in machine code so control should not reach here."
  	self assert: false!



More information about the Vm-dev mailing list