Abdulaziz Ghuloum wrote:
> On Sep 27, 2007, at 5:06 PM, William D Clinger wrote:
>
>
>> Why, pray tell, must the third argument to fxcopy-bit-field
>> be less than (fixnum-width)?
>>
>
> So that your code would still work when you do
> (declare unsafe non-r6rs full-speed-forward)
> and without the compiler coercing the value of
> the shift to some small value.
You must be quoting from the confession statements.
What you say is that a restriction was added in support
of a particular theory of what strange compiler-related
features it might be nice to pile on. Damn if that
doesn't describe a lot of R6.
What if, instead, the focus of the question was on
"when a compiler can optimize away the coercing of the
value of the shift to some small value?"
I like the idea of compiler optimizations that are
easy to implement (so I don't expect heroic type
inference efforts to eliminate the coercion) but
that don't detract from the generality and simplicity
of the language and that simply can give warnings
when the optimization can't be applied. That would
be a tool. What you are describing is more like a
religion.
> This also forces
> implementors to implement these operations in a
> portable, predictable, reliable, and safe manner.
>
Because, clearly, they don't understand anything but
force and, obviously, foreclosing debate on the most
convenient semantics of those weird edge cases is
the path to truth.
-t
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://lists.r6rs.org/pipermail/r6rs-discuss/attachments/20070927/62ea92f2/attachment.htm
Received on Thu Sep 27 2007 - 20:26:24 UTC