Barton Willis wrote:
> -----maxima-bounces at math.utexas.edu wrote: -----
>
>
>> Andreas Eder wrote:
>>
>>> changed memq to member with :test '#eq to save function call overhead
>>>
> and consing;
>
> Running the test suite with the SBCL profiler shows that (for a longer
> list,
> see my 3 April post 'modifying great'; see also Raymond's post also
> on 3 April):
>
> seconds | consed | calls | sec/call | name
> ------------------------------------------------------------------
> 36.387 | 447,354,856 | 134,280 | 0.000271 | NEW-FILE-SEARCH1
> 0.000 | 1,621,632 | 33,784,015 | 0.000000 | MEMQ
>
> Oh sure, every little improvement helps. But given this data, why do we
> worry so much about memq? Is memq a big problem for other Lisp versions?
>
> Assuming the profiling data is accurate and that I'm not
> missinterperting it, maybe a re-write of new-file-search1 would make
> SBCL Maxima as fast as GCL Maxima.
>
>
>
Somehow that doesn't make any sense to me. How many files can maxima be
actually looking for during the whole test suite?
Note, however, that for the first few (5?) tests, CMUCL also said
new-file-search1 was the top user. If I let the test suite run longer,
new-file-search1 is much less.
I'm thoroughly confused....
Ray