Am 15.04.2013 11:40, schrieb Igor Stasenko:

On 15 April 2013 11:31, Helmut Rohregger helmut.rohregger@gmail.com wrote:

...

Hi all,

i am interested in the implementations of primitives in the VMMaker package and found some code, which i think may be implemented incorrectly. Maybe some of you can help me out and correct me if i am wrong.

Its located in InterpreterPrimitives>>#primitiveBitShift (VMMaker-dtl.319): the code: "self success: integerArgument <= 31" checks, if we lose bits by shifting to the left. This is because of processing SmallIntegers here, isn't it? But a "1 bitShift: 31" will produce a LargePositiveInteger instead of a SmallInteger. In fact a left shift of 30 will produce a LargePositiveInteger too.

I know the code works, because the resulting object gets checked being a SmallInteger and fails if not. But the correct check would be "self success: integerArgument <= 29", isn't it? What do you think?

hmm., how do you know if shifted result will fit into smallint, if you don't check both operands?

"1 << 29 " fits "SmallInteger maxVal << 29 " doesnt. even "SmallInteger maxVal << 1 " doesnt.

but " x << 30 " does never fit, no matter which value 'x' has.

Am 15.04.2013 11:55, schrieb Igor Stasenko:

On 15 April 2013 11:46, Helmut Rohregger helmut.rohregger@gmail.com wrote:

...

Am 15.04.2013 11:40, schrieb Igor Stasenko:

...

On 15 April 2013 11:31, Helmut Rohregger helmut.rohregger@gmail.com wrote:

...

Hi all,

i am interested in the implementations of primitives in the VMMaker package and found some code, which i think may be implemented incorrectly. Maybe some of you can help me out and correct me if i am wrong.

Its located in InterpreterPrimitives>>#primitiveBitShift (VMMaker-dtl.319): the code: "self success: integerArgument <= 31" checks, if we lose bits by shifting to the left. This is because of processing SmallIntegers here, isn't it? But a "1 bitShift: 31" will produce a LargePositiveInteger instead of a SmallInteger. In fact a left shift of 30 will produce a LargePositiveInteger too.

I know the code works, because the resulting object gets checked being a SmallInteger and fails if not. But the correct check would be "self success: integerArgument <= 29", isn't it? What do you think?

hmm., how do you know if shifted result will fit into smallint, if you don't check both operands?

"1 << 29 " fits "SmallInteger maxVal << 29 " doesnt. even "SmallInteger maxVal << 1 " doesnt.

but " x << 30 " does never fit, no matter which value 'x' has.

0 << 30 = 0. fits.

hehe i know. and "0 << SmallInteger maxVal" fits too, but will fail with the implemented primitive.

Hi Nicolas,

Am 15.04.2013 19:22, schrieb Nicolas Cellier:

If you look at the terminal instructions in primitiveBitShift you'll see:

self push: (self positive32BitIntegerFor: shifted)

If you inquire a bit, you will see that positive32BitIntegerFor: will create a LargePositiveInteger instead of failing the primitive.

ok, i have assumed, that "positive32BitIntegerFor" will return a small integer because of "32 bit" in its name :-) well, it does return a 32 bit integer but if to big for small integer, a largePositiveInteger gets built...

The balance between purity and efficiency I guess...

this makes sense! thank you!

Nicolas

2013/4/15 Helmut Rohregger <helmut.rohregger@gmail.com mailto:helmut.rohregger@gmail.com>

Am 15.04.2013 11:55, schrieb Igor Stasenko:

      On 15 April 2013 11:46, Helmut Rohregger
    <helmut.rohregger@gmail.com
    <mailto:helmut.rohregger@gmail.com>> wrote:

        Am 15.04.2013 11:40, schrieb Igor Stasenko:

               On 15 April 2013 11:31, Helmut Rohregger
            <helmut.rohregger@gmail.com
            <mailto:helmut.rohregger@gmail.com>>
            wrote:

                Hi all,

                i am interested in the implementations of
                primitives in the VMMaker
                package
                and found some code, which i think may be
                implemented incorrectly.
                Maybe some of you can help me out and correct me
                if i am wrong.

                Its located in
                InterpreterPrimitives>>#primitiveBitShift
                (VMMaker-dtl.319):
                the code: "self success: integerArgument <= 31"
                checks, if we lose bits
                by
                shifting to the left. This is because of
                processing SmallIntegers here,
                isn't it? But a "1 bitShift: 31" will produce a
                LargePositiveInteger
                instead
                of a SmallInteger. In fact a left shift of 30 will
                produce a
                LargePositiveInteger too.

                I know the code works, because the resulting
                object gets checked being a
                SmallInteger and fails if not. But the correct
                check would be "self
                success:
                integerArgument <= 29", isn't it?
                What do you think?

            hmm., how do you know if shifted result will fit into
            smallint, if you
            don't check both operands?

            "1 << 29 " fits
            "SmallInteger maxVal << 29 " doesnt.
            even
            "SmallInteger maxVal << 1 " doesnt.


        but " x << 30 " does never fit, no matter which value 'x' has.

    0 << 30 = 0.
    fits.


hehe i know.
and "0 << SmallInteger maxVal" fits too, but will fail with the
implemented primitive.