On Mon, Oct 28, 2002 at 12:07:49PM -0800, David Hanley wrote:
> (deftype hash-type () '(unsigned-byte 32))
>
>
> (defparameter hash-code 0)
> (declaim (type hash-type hash-code))
>
> (defun foo( x )
> (declare (type hash-type x))
> (setf hash-code (logxor hash-code x)))
>
> this always gives me a warning :
>
> note: doing unsigned work to integer conversion (cost
> 20) for:
> the second argument of SET
>
> it doesn't seem to matter if i make hash-type fixnum
> or not.
> [...]
> fine, then the result is being unboxed for some
> reason?!?
That's right. The result will have to be boxed to be stored in
hash-code, since it can be accessed from anywhere, and must look like a
lisp object.
If you require no boxing, you could store your hash-code in a
specialized vector of length one [created by (MAKE-ARRAY 1 :ELEMENT-TYPE
'(UNSIGNED-BYTE 32))]; that said, I don't see any boxing when HASH-TYPE
is FIXNUM; the disassemble output (from DISASSEMBLE) says:
; 090519F6: MOV EAX, [#x90519D0] ; 'HASH-CODE
; no-arg-parsing entry point
; 9FC: MOV EAX, [EAX-3]
; 9FF: XOR EAX, EDX
; A01: MOV ECX, [#x90519D0] ; 'HASH-CODE
; A07: MOV [ECX-3], EAX
; [...]
Cheers,
Christophe
--
http://www-jcsu.jesus.cam.ac.uk/~csr21/ +44 1223 510 299/+44 7729 383 757
(set-pprint-dispatch 'number (lambda (s o) (declare (special b)) (format s b)))
(defvar b "~&Just another Lisp hacker~%") (pprint #36rJesusCollegeCambridge)
