New primitive to include (please?) Re: [Vm-dev] Method executing but not sent to the receiver
So...if everybody is happy with this new primitive, can we include it in the main distribution of VMMaker and VMMaker-oscog ? I mean...I can do it, I have commit access. But I would live to have the accept. Andreas and Levente have already agreed. Eliot ? can I commit it to VMMaker-oscog ? David ? can I commit it to VMMaker ? is this number 190 okay?
Thanks
Mariano
I copy past the code here:
CompiledMethod >> executeWithReceiver: aReceiver arguments: anArray "Execute the compiledMethod against the aReceiver and args in argArray."
<primitive: 190> self primitiveFailed
And this is the primitive.
StackInterpreterPrimitives >> primitiveExecuteWithReceiverAr guments "method, recevier, and the array of arguments are on top of stack. Execute method against receiver and args. Set primitiveFunctionPointer because no cache lookup has been done for the method, and hence primitiveFunctionPointer is stale." | receiverMethod argCnt argumentArray primitiveIndex receiverObject | receiverMethod := self stackValue: 2. receiverObject := self stackValue: 1. argumentArray := self stackTop. ((objectMemory isOopCompiledMethod: receiverMethod) and: [objectMemory isArray: argumentArray]) ifFalse: [^self primitiveFailFor: PrimErrBadArgument]. argCnt := self argumentCountOf: receiverMethod. argCnt = (objectMemory fetchWordLengthOf: argumentArray) ifFalse: [^self primitiveFailFor: PrimErrBadNumArgs]. self pop: 3. self push: receiverObject. 0 to: argCnt - 1 do: [:i| self push: (objectMemory fetchPointer: i ofObject: argumentArray)]. newMethod := receiverMethod. primitiveIndex := self primitiveIndexOf: newMethod. primitiveFunctionPointer := self functionPointerFor: primitiveIndex inClass: nil. argumentCount := argCnt. "We set the messageSelector for executeMethod below since things like the at cache read messageSelector and so it cannot be left stale." messageSelector := objectMemory nilObject. self executeNewMethod. "Recursive xeq affects primErrorCode" self initPrimCall
On Wed, Apr 20, 2011 at 2:42 PM, Andreas Raab andreas.raab@gmx.de wrote:
On 4/20/2011 14:29, Mariano Martinez Peck wrote:
CompiledMethod >> valueWithReceiver: aReceiver arguments: anArray "Execute compiledMethod against the receiver and args in argArray. The receiver is the CompiledMethod"
<primitive: 666> self primitiveFailed
Where the primitive is send directly, and not a message to the receiver.
Is something like this available or I would need to create a new primitive?
Which by the way, is the primitive that should exist, and Object>>withArgs:executeMethod: should be nuked ASAP. There is no reason to pollute the Object namespace any more than absolutely necessary.
Cheers,
- Andreas
On Mon, Apr 25, 2011 at 04:59:08PM +0200, Mariano Martinez Peck wrote:
So...if everybody is happy with this new primitive, can we include it in the main distribution of VMMaker and VMMaker-oscog ? I mean...I can do it, I have commit access. But I would live to have the accept. Andreas and Levente have already agreed. Eliot ? can I commit it to VMMaker-oscog ? David ? can I commit it to VMMaker ?
If Andreas and Levente like it, then it is fine with me :) When you commit to VMMaker, can you please also update VMMaker class>>versionString to the next minor version number? Thanks!
Note that #primitiveFailFor: is not yet available for the standard interpreter, so you will need to use #primitiveFail. As luck would have it I was just looking at that last night in hopes of adopting the oscog implementation into Interpreter, but I think I spotted a slang generation bug that needs to be addressed first :-/
is this number 190 okay?
I am away and cannot check right now, but Eliot can probably answer.
Dave
Thanks
Mariano
I copy past the code here:
CompiledMethod >> executeWithReceiver: aReceiver arguments: anArray "Execute the compiledMethod against the aReceiver and args in argArray."
<primitive: 190> self primitiveFailed
And this is the primitive.
StackInterpreterPrimitives >> primitiveExecuteWithReceiverAr guments "method, recevier, and the array of arguments are on top of stack. Execute method against receiver and args. Set primitiveFunctionPointer because no cache lookup has been done for the method, and hence primitiveFunctionPointer is stale." | receiverMethod argCnt argumentArray primitiveIndex receiverObject | receiverMethod := self stackValue: 2. receiverObject := self stackValue: 1. argumentArray := self stackTop. ((objectMemory isOopCompiledMethod: receiverMethod) and: [objectMemory isArray: argumentArray]) ifFalse: [^self primitiveFailFor: PrimErrBadArgument]. argCnt := self argumentCountOf: receiverMethod. argCnt = (objectMemory fetchWordLengthOf: argumentArray) ifFalse: [^self primitiveFailFor: PrimErrBadNumArgs]. self pop: 3. self push: receiverObject. 0 to: argCnt - 1 do: [:i| self push: (objectMemory fetchPointer: i ofObject: argumentArray)]. newMethod := receiverMethod. primitiveIndex := self primitiveIndexOf: newMethod. primitiveFunctionPointer := self functionPointerFor: primitiveIndex inClass: nil. argumentCount := argCnt. "We set the messageSelector for executeMethod below since things like the at cache read messageSelector and so it cannot be left stale." messageSelector := objectMemory nilObject. self executeNewMethod. "Recursive xeq affects primErrorCode" self initPrimCall
On Wed, Apr 20, 2011 at 2:42 PM, Andreas Raab andreas.raab@gmx.de wrote:
On 4/20/2011 14:29, Mariano Martinez Peck wrote:
CompiledMethod >> valueWithReceiver: aReceiver arguments: anArray "Execute compiledMethod against the receiver and args in argArray. The receiver is the CompiledMethod"
<primitive: 666> self primitiveFailed
Where the primitive is send directly, and not a message to the receiver.
Is something like this available or I would need to create a new primitive?
Which by the way, is the primitive that should exist, and Object>>withArgs:executeMethod: should be nuked ASAP. There is no reason to pollute the Object namespace any more than absolutely necessary.
Cheers,
- Andreas
-- Mariano http://marianopeck.wordpress.com
vm-dev@lists.squeakfoundation.org
-
David T. Lewis -
Mariano Martinez Peck