erfc for arbitrary precision real field

asked 2020-07-17 13:22:20 +0100

WhatsUp
11 ●1

The following code raised a runtime error:

R = RealField(500)
print(erfc(R(28.0)))

Output:

/home/sc_serv/sage/local/lib/python3.7/site-packages/sage/rings/real_mpfr.pyx in sage.rings.real_mpfr.RealNumber.erfc (build/cythonized/sage/rings/real_mpfr.c:31648)()
   5047         """
   5048         cdef RealNumber x = self._new()
-> 5049         sig_on()
   5050         mpfr_erfc(x.value, self.value, (<RealField_class>self._parent).rnd)
   5051         sig_off()

RuntimeError: Aborted

However, it would work if the RealField is removed:

print(erfc(28.0))

Output:

6.56321584032878e-343

Is there a way to make erfc work also for arbitrary precision real fields?

edit retag flag offensive close merge delete

add a comment

2 Answers

Sort by » oldest newest most voted

answered 2020-07-17 22:39:22 +0100

jaydfox
91 ●5

updated 2020-07-17 22:47:27 +0100

The issue is not the RealField. The issue seems to have something to do with the precision. 28.0 is simply a RealNumber with the default precision of 53 bits. Here are my results for various precisions:

sage: erfc(RealField(53)(28))
6.56321584032878e-343
sage: erfc(RealField(64)(28))
6.56321584032878415e-343
sage: erfc(RealField(128)(28))
6.5632158403287841523809104816062747741e-343
sage: erfc(RealField(192)(28))
6.56321584032878415238091048160627477413416268994828261802e-343
sage: erfc(RealField(256)(28))
uceil_log2.c:40: MPFR assertion failed: exp < 1023
---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)

The real error is the "MPRF assertion failed: exp < 1023". This probably has something to do with the C implementation of mpfr_erfc, which is a third party library. I.e., it is used by SAGE, but it was not written by the SAGE team. I believe the MPFR library is still actively maintained, so maybe the SAGE team can reach out to the MPFR developers?

Edit: By the way, if you want to test with a standard double precision floating point number, put a letter 'r' after the number:

sage: isinstance(28.0, sage.rings.real_mpfr.RealNumber)
True
sage: isinstance(28.0r, float)
True
sage: erfc(28.0)
6.56321584032878e-343
sage: erfc(28.0r)
0.0

edit flag offensive delete link

add a comment

answered 2020-07-19 12:04:03 +0100

tmonteil
27423 ●31 ●204 ●517 http://wiki.sagemath.o...

It works for me on 9.2.beta5 compiled from source, but with MPFR coming from my distro (libmpfr-dev package from Debian buster x86_64):

sage: R = RealField(5000)
....: print(erfc(R(28.0)))
....: 
6.563215840328784152380910481606274774134162689948282618021777806342324947794861698508585865279431011096752124531902419799309548263163678405299165421468926529756254284140867535233615012322334166170928503904086662017717248546748850127772332145411464185469489280260339088060235989228965843585936313865164734112442590831250266796576150016384641751358842396622115815414226304418079260475584875218573787585295929224096136714786361975866293997811039127970371335072820687621548403207278280764251340150092695582025750001434924105792823313203560201160341102473911930337624932555374585331265969965105269933130458163745578151352533927015070036231688942582148509848578635195809075408109589600332886026100202493893872822063883882953532315845612170517048543977112601393051614315880119177110350110356515266819875606915674560735226077346306998422763124176864165134035565227108933089306529302102108301994852596238985987401162654832997613615932852294726716250359138015072257484970655748380496068667053379552628609104926576396476543030043423989862535086552019022314368469618455141719577505798386449635616469359569975605353831944590150026622030636379223666200115938783888320717911446524435737548074875506351487946003131182599893046311540255956839538210349124618319107602941518963879471841876064339644999968291611090263995185161031176788592846481694313759144524067300523377786375902794975290499947567163622335460211299103648351220548274818081256066819963433322071061349340927987304921883890192373464108935170754385292484798247e-343

So there might be a compilation issue somewhere in your build.

Note that mpmath and gmpy2 might be responsible too.

Could you please (both) give us some informations so that someone can try to understand:

which version of Sage did you use ?
which OS ?
did you install Sage from the distribution, and which package ?
did you install Sage from the binaries, and which ones ?
did you compile Sage yourself ?
if you compiled Sage, was a libmpfr-dev library installed on your ditstro ?
... ?

edit flag offensive delete link

Comments

I'm running version 9.0 on Windows. The name of the installer file was SageMath-9.0-Installer-v0.6.0.exe. Below is the output I see in sage, when I check the version:

sage: version()
'SageMath version 9.0, Release Date: 2020-01-01'

I have not recompiled, and it's running on top of cygwin, so I'm not sure if the MPFR library is part of cygwin or part of the sagemath installation.

jaydfox ( 2020-07-20 03:15:00 +0100 )edit

I'm running on the sage cell server https://sagecell.sagemath.org/ and the version() command gives 'SageMath version 9.1, Release Date: 2020-05-20'. Still it doesn't work at this moment. If the issue is solved in version 9.2, then I would be happy to wait for the sage cell server to update to 9.2.

WhatsUp ( 2020-08-01 17:53:27 +0100 )edit

add a comment

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

erfc for arbitrary precision real field

2 Answers

Comments

Your Answer

Question Tools

Stats

Related questions

erfc for arbitrary precision real field edit

2 Answers

Comments

Your Answer

Question Tools

Stats

Related questions

erfc for arbitrary precision real field