>>Viv,
>>
>>It's debatable. But do confess you're cheating a bit
>
><g>
>
>>
>>And it doesn't know you're cheating. It stores the decremented value and does not refetch it
>>
>>I suspect it's by design and for performance reasons [ a dup will be faster than a load ]
>
>Guess so. I looked at your ildasm code and your test below confirms the performance benefit . But I'm not sure whether it's me or MS that's 'cheating' . It could certainly be argued that the decrement operation on ScaleFactor has not been completed when the value is assigned to i........


Oh, the assignment has been completed - see ildasm code - it just doesn't expect (hence the dup) it will be different.
You put a white pigeon in the hat, and transform it into a rabbit

>Regards,
>Viv