This is the mail archive of the gcc-bugs@gcc.gnu.org mailing list for the GCC 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]

[Bug c/59089] New: sin and/or cos produce bogus results with -O2


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59089

            Bug ID: 59089
           Summary: sin and/or cos produce bogus results with -O2
           Product: gcc
           Version: 4.7.1
            Status: UNCONFIRMED
          Severity: major
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: terra at gnome dot org

Created attachment 31199
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=31199&action=edit
sc.c

welinder@sherwood:~> gcc --version
gcc (SUSE Linux) 4.7.1 20120723 [gcc-4_7-branch revision 189773]

When sin(x) and cos(x) are used near each other and -O2 is active,
results are utterly unreliable:

welinder@sherwood:~> gcc -Wall -O0 sc.c -lm
welinder@sherwood:~> ./a.out 1e19
cot(1e+19) = 0.40440089012342833596
welinder@sherwood:~> gcc -Wall -O2 sc.c -lm
welinder@sherwood:~> ./a.out 1e19
cot(1e+19) = 0.60484784464511465174



Reference value from
http://www.wolframalpha.com/input/?i=Cot[10^19]&a=TrigRD_R
--> 0.404400890123428300526065555419647330231956673028985010736178...

Further info: -O2 uses sincos, not sin and cos:

welinder@sherwood:~> gcc -Wall -O2 sc.c
/tmp/ccouebOR.o: In function `gnm_cot':
sc.c:(.text+0xd): undefined reference to `sincos'
collect2: error: ld returned 1 exit status
welinder@sherwood:~> gcc -Wall -O0 sc.c
/tmp/ccR0uAFj.o: In function `gnm_cot':
sc.c:(.text+0x1b): undefined reference to `sin'
sc.c:(.text+0x3a): undefined reference to `cos'
collect2: error: ld returned 1 exit status


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