On Fri, Sep 30, 2011 at 10:10 PM, Alaric Snell-Pym
<alaric@snell-pym.org.uk> wrote:
>
> On 09/30/2011 02:07 PM, Alex Shinn wrote:
>> On Fri, Sep 30, 2011 at 5:34 PM, Alaric Snell-Pym
>> <alaric@snell-pym.org.uk> wrote:
>>>
>>> On 09/29/2011 05:25 PM, Andre van Tonder wrote:
>>>
>>>> On a system that that invokes a compiler on the argument of EVAL, the compiler
>>>> may depend on the textual representation of the code '(cons 1 2).  It may even
>>>> do certain optimizations and rewritings based on the textual representation.
>>>> Again, having procedure objects in here confuses levels and can cause problems
>>>> for such a compiler.
>>>
>>> Does it really? Can anyone point to an implementation of EVAL that can't
>>> cope with this?
>>
>> Chicken's inline egg would not support this.  That's intended
>> for inline C code, but could be modified and used as an EVAL
>> replacement.
>
> I'll discount that one, then, as it merely "could be" :-)

From the latest trunk of inline egg:

#;1> (define fib-src '(lambda (n) (let fib ((n n)) (if (< n 2) 1 (+
(fib (- n 1)) (fib (- n 2)))))))
#;2> (define f1 (eval fib-src))
#;3> (define f2 (inline-eval fib-src))
#;4> (time (f1 35))
14.805s CPU time, 3.301s GC time (major), 1048714 mutations,
1734/275003 GCs (major/minor)
14930352
#;5> (time (f2 35))
1.132s CPU time, 0.139s GC time (major), 107560 mutations, 72/29328
GCs (major/minor)
14930352

--
Alex

_______________________________________________
Scheme-reports mailing list
Scheme-reports@scheme-reports.org
http://lists.scheme-reports.org/cgi-bin/mailman/listinfo/scheme-reports