Activity for REDUCE
Rainer Schöpf committed [r4697]
Sun, 22 Jul 2018 12:27:13 -0000
Slightly different way of detecting PSL during run of regresion test file.
Rainer Schöpf committed [r4696]
Sun, 22 Jul 2018 11:00:22 -0000
Floating point arithmetic: Make sure that the implementation of the elementary functions handle non-finite IEEE floats correctly (or at least gracefully).
Rainer Schöpf committed [r4695]
Sun, 22 Jul 2018 07:33:37 -0000
Define a couple of functions even if our floating point model is not IEEE - probably superfluous, but better safe than sorry.
Rainer Schöpf committed [r4694]
Wed, 11 Jul 2018 05:32:21 -0000
Debian package build process: change the dependency on gnuplot-x11 to (gnuplot-x11 | gnuplot-qt).
Rainer Schöpf committed [r4693]
Sun, 08 Jul 2018 20:14:02 -0000
Correction to floatfix: signal an error for non-finite IEEE floats.
Rainer Schöpf committed [r4692]
Sun, 08 Jul 2018 11:41:25 -0000
Forgot to commit revision_history.txt
Rainer Schöpf committed [r4691]
Sun, 08 Jul 2018 11:39:44 -0000
Fixes for IEEE floating point in PSL (Linux x86_64):
Rainer Schöpf committed [r4690]
Sun, 08 Jul 2018 11:11:59 -0000
IEEE floating point for PSL:
Arthur Norman committed [r4689]
Sun, 08 Jul 2018 10:47:46 -0000
CSL: change behaviour to match the Standrad Lisp Report, specifically
Rainer Schöpf modified ticket #81
Sat, 07 Jul 2018 09:22:26 -0000
num_odesolve does not handle intervals with pi. Here is the fix.
Rainer Schöpf posted a comment on ticket #81
Sat, 07 Jul 2018 09:22:26 -0000
Correction committed.
Arthur Norman committed [r4688]
Sat, 07 Jul 2018 07:58:26 -0000
CSL: this has more changes by way of adding comments that explain what I
Arthur Norman committed [r4687]
Wed, 04 Jul 2018 14:56:57 -0000
CSL: I believe that in the "conservative" world I can not both create
Arthur Norman committed [r4686]
Tue, 03 Jul 2018 21:31:18 -0000
CSL: The allocation code for my conservative collector is quite a lot
Marduk Bolaños posted a comment on ticket #81
Tue, 03 Jul 2018 18:44:22 -0000
Now I can't reproduce it either. It is very strange.
Rainer Schöpf posted a comment on ticket #81
Tue, 03 Jul 2018 18:22:51 -0000
I committed a correction for the the problem with pi as part of the interval specification. I'm afraid that I cannot reproduce the car of nil error you mention. I suspect this is caused by rounding. Can you please supply a complete example? Rainer
Rainer Schöpf modified ticket #81
Tue, 03 Jul 2018 18:20:06 -0000
num_odesolve does not handle intervals with pi. Here is the fix.
Rainer Schöpf committed [r4685]
Tue, 03 Jul 2018 18:19:06 -0000
Fix for bug report
Marduk Bolaños posted a comment on ticket #81
Tue, 03 Jul 2018 13:09:46 -0000
By the way, the default accuracy, defined in the beginning of rungekuttaeval u := accuracycontrol(u,20,6); is larger than machine precision and very likely unnecessarily large. My experiments with the harmonic oscillator show that with low accuracy (e.g. 6) after 500 oscillations and using 14,000 steps the energy is conserved. Which might also suggest that this method is sympletic. If someone knows the algorithm behind the code it would be good to explicitly mention it in the documentation.
Marduk Bolaños posted a comment on ticket #81
Tue, 03 Jul 2018 12:52:47 -0000
It turns out that rungekuttares was not written carefully. Here is the corrected version, that checks the size of the list before applying car and cdr. symbolic procedure rungekuttares(l,st); % eliminate intermediate points. if st=iterations!* then l else << for i:=1:iterations!* collect < 1) then l:= cdr l; car l>> >> where m=st/iterations!*;

rssfeedwidget.com