This is the mail archive of the cygwin mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

R: floating-point math problem in Cygwin 1.7?


--- Gio 24/12/09, Linh Phan  ha scritto:

> Hi,
> 
> Â I have wondering why when I do floating-point math
> in Cygwin 1.7, Cygwin does not output the "correct" result,
> eg:
> 
> main () {
> Â double qcb[4] = {0.41585680180318363,
> 0.70823637211274604, 0.089200955545759006,Â
> -0.56347808399291521};
> Â double qba[4] = {0.09648538897087118,
> -0.37032037362514358,
> 0.89403246842889339,ÂÂÂ0.23293633462432001};
> Â double qca[4];
> 
> Â qca[0]= qcb[3]*qba[0] + qcb[2]*qba[1] -
> qcb[1]*qba[2] + qcb[0]*qba[3];
> Â qca[1]= -qcb[2]*qba[0] + qcb[3]*qba[1] +
> qcb[0]*qba[2] + qcb[1]*qba[3];
> Â qca[2]= qcb[1]*qba[0] - qcb[0]*qba[1] +
> qcb[3]*qba[2] + qcb[2]*qba[3];
> Â qca[3]= -qcb[0]*qba[0] - qcb[1]*qba[1] -
> qcb[2]*qba[2] + qcb[3]*qba[3];
> 
> Â // I am expecting the result to be:
> 
> Â printf("qca = [-0.623718486146499718,
> 0.736824293298044886, -0.260654850643024127,
> 0.011147182658310384] CORRECT (LINUX/SOLARIS)\n");
> 
> Â // I get the above result on both Linux/Solaris AND
> even when running under Cygwin's gdb and using gdb's printf
> 
> Â // but when I compile and run on cygwin, I get the
> wrong result:

s/wrong/different

> Â printf("qca = [%.18f, %.18f, %.18f, %.18f] NOT
> CORRECT (CYGWIN)\n",qca[0],qca[1],qca[2],qca[3]);
> }
> 
> Here is the output when I compiled and run on cygwin:
> 
> qca = [-0.623718486146499718, 0.736824293298044886,
> -0.260654850643024127, 0.011147182658310384] CORRECT
> (LINUX/SOLARIS)
> qca = [-0.623718486146499607, 0.736824293298044886,
> -0.260654850643024072, 0.011147182658310373] NOT CORRECT
> (CYGWIN)
> 
> Here's the output when running under cygwin's gdb (it
> matches Linux/Solaris outputs but not cygwinâs output):
> 
> (gdb) printf "%.18f\n", qcb[3]*qba[0] + qcb[2]*qba[1] -
> qcb[1]*qba[2] + qcb[0]*qba[3]
> -0.623718486146499718
> (gdb) printf "%.18f\n", -qcb[2]*qba[0] + qcb[3]*qba[1] +
> qcb[0]*qba[2] + qcb[1]*qba[3]
> 0.736824293298044886
> (gdb) printf "%.18f\n", qcb[1]*qba[0] - qcb[0]*qba[1] +
> qcb[3]*qba[2] + qcb[2]*qba[3]
> -0.260654850643024127
> (gdb) printf "%.18f\n", -qcb[0]*qba[0] - qcb[1]*qba[1] -
> qcb[2]*qba[2] + qcb[3]*qba[3]
> 0.011147182658310384
> (gdb) 
> 
> These errors accumulate and eventually builds up so much
> that the my simulation result is totally wrong when running
> under cygwin.
> 
> Please help. 

I will bet 
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=323

see
http://hal.archives-ouvertes.fr/hal-00128124
for further detailed info on rounding

if so it is not a bug is a feature 
:-)

by the way on (cygwin) octave that round as you expect I have
 printf("%.18g\n",qca)
-0.623718486146499718
0.736824293298044886
-0.260654850643024127
0.0111471826583103839



> Thanks,
> 
> Linh
> 

Marco





--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]