On Fri, May 11, 2018 at 9:47 PM, Waldek Hebisch
<[email protected]> wrote:
> oldk1331 wrote:
>>
>> On Fri, Mar 9, 2018 at 9:46 PM, Waldek Hebisch <[email protected]>
>> wrote:
>> > Now all things planned for this release are commited. There
>> > is now time for testing. Qian says that 'sbcl_make_sized_vector'
>> > is buggy -- I will look into this.
>>
>> Any update on 'sbcl_make_sized_vector'?
>
> Earlier:
>
>> It is a bug to have "`(ceiling (* ,n ,n-bits) sb-vm:n-word-bits)"
>> in 'sbcl_make_sized_vector', we should fix it to
>> "`(ceiling (* ,n ,nb) sb-vm:n-word-bits)".
>
> The difference between 'n-bits' and 'nb' is that we want 'nb'
> but 'n-bits' is what sbcl thinks that we should use. It
> is safer to use value that sbcl wants to have. So no
> bug.
What does this mean? Can you elaborate it?
I think there is still a bug in 'sbcl_make_sized_vector':
(6) -> )lisp (sbcl_make_sized_vector 32 5)
Value = #(0 0 0 0 229)
The vector is not initialized to 0.
> BTW: The code was changed to current form because simpler
> version did not work. I do not remember if I tested
> 'nb' but first formula that I tried did not work, so
> I changed computation to exactly match what sbcl was
> doing.
My patch posted on March 8 works and is much simpler.
But it depends on ALLOCATE-VECTOR-WITH-WIDETAG.
--- a/src/lisp/primitives.lisp
+++ b/src/lisp/primitives.lisp
@@ -101,22 +101,14 @@
#+:sbcl
(defmacro sbcl_make_sized_vector(nb n)
- (let ((get-tag (find-symbol "%VECTOR-WIDETAG-AND-N-BITS" "SB-IMPL"))
- (length-sym nil))
+ (let ((get-tag (find-symbol "%VECTOR-WIDETAG-AND-N-BITS" "SB-IMPL")))
(if (null get-tag)
- (progn
- (setf get-tag
- (find-symbol "%VECTOR-WIDETAG-AND-N-BITS-SHIFT"
- "SB-IMPL"))
- (setf length-sym (find-symbol "VECTOR-LENGTH-IN-WORDS"
- "SB-IMPL"))))
+ (setf get-tag
+ (find-symbol "%VECTOR-WIDETAG-AND-N-BITS-SHIFT"
+ "SB-IMPL")))
(multiple-value-bind (typetag n-bits)
(FUNCALL get-tag `(unsigned-byte ,nb))
- (let ((length-form
- (if length-sym
- `(,length-sym ,n ,n-bits)
- `(ceiling (* ,n ,n-bits) sb-vm:n-word-bits))))
- `(SB-KERNEL:ALLOCATE-VECTOR ,typetag ,n ,length-form)))))
+ `(SB-IMPL::ALLOCATE-VECTOR-WITH-WIDETAG ,typetag ,n ,n-bits))))
(defmacro DEF_SIZED_UOPS(nb)
>> Also I think it is better to use ALLOCATE-VECTOR-WITH-WIDETAG
>> than ALLOCATE-VECTOR.
>
> We can not use ALLOCATE-VECTOR-WITH-WIDETAG because it is not
> present in earlier versions of sbcl.
>
> --
> Waldek Hebisch
ALLOCATE-VECTOR-WITH-WIDETAG is introduced in
2012-12-19 and in sbcl-1.1.3.
Our binary is built by sbcl-1.1.1, can we consider to version
bump sbcl to 1.1.3?
--
You received this message because you are subscribed to the Google Groups
"FriCAS - computer algebra system" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/fricas-devel.
For more options, visit https://groups.google.com/d/optout.
