Re: GCL slow reading files
David Scherfgen via Maxima-discuss <[email protected]>
| Newsgroups | gmane.comp.mathematics.maxima.general |
|---|---|
| Message-ID | <CAMTHLKgnGy8s5Q9SAt5DD3QjiTTawUBuJdn+1=A7V1tnW90bbQ@mail.gmail.com> |
It seems that this has been fixed by Camm recently. I now get 0.000 seconds and 0 bytes allocated. Loading the "draw" package is still slow, but only when it's loaded for the first time. After that, it just loads the binary. Maybe we could add an option to the configure script to tell Maxima to compile all share packages? Best regards David Scherfgen Am Fr., 10. Apr. 2026 um 17:27 Uhr schrieb Raymond Toy < [email protected]>: > On 4/10/26 12:27 AM, David Scherfgen wrote: > > Raymond Toy <[email protected]> schrieb am Fr., 10. Apr. 2026, 00:27: > >> On 4/9/26 8:33 AM, Leo Butler wrote: >> >> To see what I mean, start a fresh maxima+gcl build and do: >> >> showtime:true $ >> load(draw); >> >> On my fairly fast machine, ecl takes about 3.3 sec. gcl takes 65 sec; >> ccl64, 0.57 sec; cmucl, 2.2 sec. Ccl is known for having a very, very fast >> compiler. >> > On my machine, GCL takes 2.4 seconds to load "draw". But it uses an older > version of GCL, namely 2.6.14. Maybe the "slowness" was introduced with > 2.7.x. > > I could confirm that GCL 2.6.14 too is quite slow in reading files. > Reading a 50 MB file took more than 4 seconds on a lightning fast NVMe SSD > with ZFS and a huge cache on top. > > Wow, it’s much worse with 2.7.2: > > > (with-open-file (s "doc/info/maxima.info-1" :direction :input :element-type '(unsigned-byte 8)) > (let* ((len (file-length s)) > (buf (make-array len :element-type '(unsigned-byte 8)))) > (time (read-sequence buf s)))) > > real time : 23.040 secs > run-gbc time : 19.840 secs > child run time : 0.000 secs > gbc time : 0.500 secs > allocation : 1357 Mbytes > > 1005136 > > maxima.info-1 is just a million bytes. Does that mean gcl spent 19.8 out > of 23 sec doing GC? But even 4 sec seems really, really slow. (I’m using an > external SSD for this. But cmucl does the same test in less than a > millisecond or so.) > ​ > _______________________________________________ Maxima-discuss mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/maxima-discuss