Re: SBCL tells me it is deleting unreachable code that is almost surely reachable

On 4/15/26 10:12 AM, Stavros Macrakis wrote:

> Of course, this:
>> ((memq sgn '($neg _*'*_$zero _*'*_$pos)) sgn)
> should be
>> ((memq sgn '($neg $zero $pos)) sgn)
> and might as well be
>> ((member sgn '($neg $zero $pos)) sgn)
How silly of me! I didn’t notice the args in the wrong order.
> If the compiler is so incompetent that it doesn't recognize it can't 
> use *eq* for a list of symbols, the compiler should be fixed!
> And, as Ray suggests, it will almost certainly inline *member/eq* for 
> a length-3 list: that is both faster and smaller than calling *member 
> *on most (all?) Lisp systems.
> And *memq* is already declaimed *inline* in *clmacs.lisp*, line 232.
Ah, I’m used to seeing the inline declaration just before the function, 
so I didn’t look.
>
> Looking at the generated code, I see that with *optimize speed:0 
> space:3*, SBCL calls *member-eq*, even though that takes more space. 
> Apparently it uses an inaccurate heuristic to estimate space usage.
Surely a simple function call to member-eq takes less space than 
inlining 3 comparisons and branches?
>
>    -s
>
> On Wed, Apr 15, 2026 at 12:10 PM Raymond Toy <[email protected]> 
> wrote:
>
>     On 4/15/26 8:25 AM, Barton Willis via Maxima-discuss wrote:
>
>>     Here is a revision of a function in hayat. The revision
>>     eliminates or helps to eliminate two testsuite failures.
>>
>>     The SBCL (2.4.7)  compiler tells me that
>>
>>     ; file: C:/Users/barton/maxima-code-pure/maxima-code/src/./hayat.lisp
>>     ; in: DEFUN COEF-SIGN
>>     ;     (EQ MAXIMA::SGN 'MAXIMA::$POS)
>>     ;
>>     ; note: deleting unreachable code
>>
>>     I think that is rubbish—likely, I'll commit the (commented out)
>>     memq version.
>
>     Check to see what |$csign| returns? Or rather what csign-impl
>     returns. For me, I see the return type is * (anything).
>
>     |memq| should do exactly the same as testing |(or (eq sgn '$neg)
>     (eq sgn '$zero) (eq sgn '$pos))|. Well, it would if it were
>     declared inline, but it’s not.
>
>     Seems to me that it would be easy to test if that code were
>     actually deleted by calling |coef-sign| with an arg that is known
>     to be |$pos|.
>
>>
>>     (defun coef-sign (coef)
>>       (let ((sgn ($csign coef)))
>>         (cond ((or (eq sgn '$complex) (eq sgn '$imaginary)) '$im)
>>               ;((memq sgn '($neg '$zero '$pos)) sgn)
>>               ((or (eq sgn '$neg) (eq sgn '$zero) (eq sgn '$pos)) sgn)
>>               (t
>>                  (let ((asgn ($asksign coef)))
>>                     ;; When sgn is zero and limit is active, record
>>     this fact as a noun equality
>>                     ;; in *limit-assumptions*.
>>                  (when (and *limit-assumptions* (eq asgn '$zero))
>>                    (push (ftake '$equal coef 0) *limit-assumptions*))
>>                 asgn)))))
>>
>>
>>
>>     _______________________________________________
>>     Maxima-discuss mailing list
>>     [email protected]
>>     https://lists.sourceforge.net/lists/listinfo/maxima-discuss
>     &#8203;
>     _______________________________________________
>     Maxima-discuss mailing list
>     [email protected]
>     https://lists.sourceforge.net/lists/listinfo/maxima-discuss
>
&#8203;

_______________________________________________
Maxima-discuss mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/maxima-discuss