[Vm-dev] Simulating the BalloonEnginePlugin, FloatArrayPlugin & Matrix2x3Plugin primitives

[David T. Lewis]

Just in case this does not get addressed quickly, I opened a Mantis issue
for it so we do not lose track of the request:

  http://bugs.squeak.org/view.php?id=7805

Dave

On Mon, Jan 06, 2014 at 09:30:17AM -0800, Eliot Miranda wrote:
>  
> Hi All,
> 
>     I'm just revising plugin treatment in Spur and came across this old
> snippet of mysterious code:
> 
> InterpreterSimulator>>loadNewPlugin: pluginString
> | plugin simClass |
> transcript cr; show:'Looking for module ', pluginString.
> (#('FloatArrayPlugin' 'Matrix2x3Plugin')
> includes: pluginString) ifTrue:
> [transcript show: ' ... defeated'. ^ nil].
> 
> In the past I got as far as rewriting it to read...
> 
> InterpreterSimulator>>loadNewPlugin: pluginString
> | plugin plugins simulatorClasses |
> transcript cr; show: 'Looking for module ', pluginString.
> "but *why*??"
> (#('FloatArrayPlugin' 'Matrix2x3Plugin') includes: pluginString) ifTrue:
> [transcript show: ' ... defeated'. ^nil].
> plugins := InterpreterPlugin allSubclasses select: [:psc| psc moduleName
> asString = pluginString asString].
> 
> In revising the code for Spur I removed the defeat code and found out more.
>  It's essentially because the BalloonPlugin has difficulty simulating
> accesses of 32-bit floats.  If you simply defeat the code and let things
> run soon you get failures in FloatArrayPlugin & Matrix2x3Plugin primitives.
>  These can be fixed by implementing the following in the FloatArrayPlugin &
> Matrix2x3Plugin:
> 
> cCoerce: value to: cType
> ^cType = 'float'
> ifTrue: [value asIEEE32BitWord]
> ifFalse: [value]
> 
> But soon you hit more difficult failures in the BalloonEnginePlugin, e.g.
> in
> 
> BalloonEngineBase>>transformPointX: xValue y: yValue into: dstPoint
> "Transform srcPoint into dstPoint by using the currently loaded matrix"
> "Note: This should be rewritten so that inlining works (e.g., removing
> the declarations and adding argument coercions at the appropriate points)"
> | x y transform |
> <inline: true>
> <var: #dstPoint type:'int *'>
> <var: #xValue type: 'double '>
> <var: #yValue type: 'double '>
> <var: #transform type:'float *'>
> transform := self edgeTransform.
> x := ((((transform at: 0) * xValue) +
> ((transform at: 1) * yValue) +
> (transform at: 2)) * self aaLevelGet asFloat) asInteger.
> y := ((((transform at: 3) * xValue) +
> ((transform at: 4) * yValue) +
> (transform at: 5)) * self aaLevelGet asFloat) asInteger.
> dstPoint at: 0 put: x.
> dstPoint at: 1 put: y.
> 
> where x and y end up being the integer representation of 64-bit floats
> while dstPoint accepts the integer representation of 32-bit floats.  At
> least I think that's what's going on.
> 
> In any case I need to focus on Spur and can't spare the time to fix this.
>  But I find it unsatisfactory.  It means the VM simulation isn't accurate.
>  In the simulation the primitives fail and Smalltalk code is run.  In the
> real VM the primitives work.  And that's deeply unsatisfying.
> 
> So if there's anyone itching for a VM challenge try and make
> the BalloonEnginePlugin, FloatArrayPlugin & Matrix2x3Plugin primitives
> simulate correctly, removing the defeat code above.  That would be a great
> new year's gift.
> -- 
> best,
> Eliot