Smalltalk › Squeak › Squeak VM

CArray arithmetic methods...

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

5 messages Options

Eliot Miranda-2

CArray arithmetic methods...

Hi All,

now I'm successfully simulating the LargeIntegers plugin I'm hitting what I consider a bug in CArray, used to model C pointers in the simulator. The first problem is that CArray>>#+ & #- return an integer value, not another CArray, (in C (char *)(exp) + n is still of type (char *), not int) and that the addition/subtraction doesn't scale by the unit size (in C (type *)p + n == (type *)((char *)p + n * sizeof(type))). The second problem is that << >> * & // are meaningless on C pointers. Anyone think of a good reason why I shouldn't a) rewrite CArray>>#+ & #- to read

+ n

^ self shallowCopy += (n * unitSize)

- n

^ self shallowCopy -= (n * unitSize)

(since e.g.

+= increment

ptrOffset := ptrOffset + increment)

and b) delete CArray>>#*, #//, #<<, #>>, #bitAnd:, #bitOr:, #bitShift: & #signedBitShift: ?

best

Eliot

Eliot Miranda-2

Re: CArray arithmetic methods...

On Sat, Oct 9, 2010 at 3:01 PM, Eliot Miranda <[hidden email]> wrote:

Hi All,

now I'm successfully simulating the LargeIntegers plugin I'm hitting what I consider a bug in CArray, used to model C pointers in the simulator. The first problem is that CArray>>#+ & #- return an integer value, not another CArray, (in C (char *)(exp) + n is still of type (char *), not int) and that the addition/subtraction doesn't scale by the unit size (in C (type *)p + n == (type *)((char *)p + n * sizeof(type))). The second problem is that << >> * & // are meaningless on C pointers. Anyone think of a good reason why I shouldn't a) rewrite CArray>>#+ & #- to read

+ n
^ self shallowCopy += (n * unitSize)

- n
^ self shallowCopy -= (n * unitSize)

(since e.g.
+= increment

ptrOffset := ptrOffset + increment)

and b) delete CArray>>#*, #//, #<<, #>>, #bitAnd:, #bitOr:, #bitShift: & #signedBitShift: ?

So far so good. The only thing that broke was the use of BitBltSimulator/BitBltSimulation>>#tableLookup:at: which can be deleted and in BitBltSimulation>>#pickSourcePixels:flags:srcMask:destMask:srcShiftInc:dstShiftInc: rewrite

destPix := self tableLookup: cmLookupTable at: (sourcePix bitAnd: cmMask).

destPix := cmLookupTable at: (sourcePix bitAnd: cmMask).

best
Eliot

stephane ducasse-2

Re: CArray arithmetic methods...

In reply to this post by Eliot Miranda-2

Eliot

how do you run the simulator?

Long time ago we had to use an MVC image and after it got broken and we could not use it anymore.
Now in Pharo we do not have MVC so are we talking about the same?

Stef

On Oct 10, 2010, at 12:01 AM, Eliot Miranda wrote:

> Hi All,
>
> now I'm successfully simulating the LargeIntegers plugin I'm hitting what I consider a bug in CArray, used to model C pointers in the simulator. The first problem is that CArray>>#+ & #- return an integer value, not another CArray, (in C (char *)(exp) + n is still of type (char *), not int) and that the addition/subtraction doesn't scale by the unit size (in C (type *)p + n == (type *)((char *)p + n * sizeof(type))). The second problem is that << >> * & // are meaningless on C pointers. Anyone think of a good reason why I shouldn't a) rewrite CArray>>#+ & #- to read
>
> + n
> ^ self shallowCopy += (n * unitSize)
>
> - n
> ^ self shallowCopy -= (n * unitSize)
>
> (since e.g.
> += increment
>
> ptrOffset := ptrOffset + increment)
>
> and b) delete CArray>>#*, #//, #<<, #>>, #bitAnd:, #bitOr:, #bitShift: & #signedBitShift: ?
>
> best
> Eliot

Eliot Miranda-2

Re: CArray arithmetic methods...

On Sat, Oct 9, 2010 at 11:57 PM, stephane ducasse <[hidden email]> wrote:

Eliot

how do you run the simulator?

See http://www.squeakvm.org/svn/squeak/branches/Cog/image/VMMaker-Squeak4.1.image and the VM Simulation Workspace there-in. Substitute StackInterpreterSimulator for CogVMSimulator and you don't need the Bochs plugin.

Long time ago we had to use an MVC image and after it got broken and we could not use it anymore.
Now in Pharo we do not have MVC so are we talking about the same?

One openAsMorph method is all it takes.

Stef

On Oct 10, 2010, at 12:01 AM, Eliot Miranda wrote:

> Hi All,
>
> now I'm successfully simulating the LargeIntegers plugin I'm hitting what I consider a bug in CArray, used to model C pointers in the simulator. The first problem is that CArray>>#+ & #- return an integer value, not another CArray, (in C (char *)(exp) + n is still of type (char *), not int) and that the addition/subtraction doesn't scale by the unit size (in C (type *)p + n == (type *)((char *)p + n * sizeof(type))). The second problem is that << >> * & // are meaningless on C pointers. Anyone think of a good reason why I shouldn't a) rewrite CArray>>#+ & #- to read
>
> + n
> ^ self shallowCopy += (n * unitSize)
>
> - n
> ^ self shallowCopy -= (n * unitSize)
>
> (since e.g.
> += increment
>
> ptrOffset := ptrOffset + increment)
>
> and b) delete CArray>>#*, #//, #<<, #>>, #bitAnd:, #bitOr:, #bitShift: & #signedBitShift: ?
>
> best
> Eliot

stephane ducasse-2

Re: CArray arithmetic methods...

Thanks we will have a look at that :)
>
> See http://www.squeakvm.org/svn/squeak/branches/Cog/image/VMMaker-Squeak4.1.image and the VM Simulation Workspace there-in. Substitute StackInterpreterSimulator for CogVMSimulator and you don't need the Bochs plugin.
>
>
> Long time ago we had to use an MVC image and after it got broken and we could not use it anymore.
> Now in Pharo we do not have MVC so are we talking about the same?
>
> One openAsMorph method is all it takes.
>