octave-bug-tracker
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Octave-bug-tracker] [bug #59830] Architecture-dependent results for ari


From: Rik
Subject: [Octave-bug-tracker] [bug #59830] Architecture-dependent results for arithmetic operations involving NA
Date: Sat, 13 Feb 2021 14:24:42 -0500 (EST)
User-agent: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/87.0.4280.66 Safari/537.36

Update of bug #59830 (project octave):

                  Status:               Need Info => Ready For Test         

    _______________________________________________________

Follow-up Comment #18:

I pushed a change to the stable branch that isn't perfect
(http://hg.savannah.gnu.org/hgweb/octave/rev/e9b7ac7031fb), but does allow NA
to work by default on MIPS architectures.

Ideally, the configure test OCTAVE_MIPS_NAN would check the actual value of
the high word in an IEEE quiet NaN and see whether it is the non-standard
value.  Right now, it just assumes that if your target is MIPS then you are
using the non-standard option.  This is the default, but someone might build
with the '-mnan=2008' option at some point.

The issue of binary compatibility of files saved with NA values could be
addressed by changing the __lo_ieee_replace_old_NA() function in
liboctave/utils/lo-ieee.cc.  Right now that function is called when loading
double values from a binary file and updates any old values of NA to the new
value of NA.  It could be extended to also search for any MIPS NA value and
replace with the architecture-specific value of NA.

These are definitely edge cases, and I don't use MIPS so I don't have an
incentive to work on this.  As a workaround, you could save data using
Octave's -text format along with the '-z' flag to compress the resulting file.

    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?59830>

_______________________________________________
  Message sent via Savannah
  https://savannah.gnu.org/




reply via email to

[Prev in Thread] Current Thread [Next in Thread]