Home My Page Projects MPFR
Summary Activity Forums Tracker Lists Tasks Docs News SCM Files

[#3353] Incorrect behavior with x86 FPU configured in single precision

Date:
2007-05-31 08:59
Priority:
3
State:
Open
Submitted by:
Vincent Lefèvre (vlefevre)
Assigned to:
Nobody (None)
Category:
general
Resolution:
None
Target Version:
None
 
Summary:
Incorrect behavior with x86 FPU configured in single precision

Detailed description
Under Linux/x86 with the traditional FPU, some functions do not work if the FPU rounding precision has been changed to single (this is a bad practice and should be useless, but one never knows what other software will do).

This can be tested with:
make check CFLAGS="-g -O2 -ffloat-store -DMPFR_FPU_PREC=_FPU_SINGLE"

This bug has partially been fixed in the trunk (changesets r4333 and r4426).

Followup

Message
Date: 2008-09-18 12:57
Sender: Paul Zimmermann

All tests seem to pass on a 64-bit machine (Opteron) with
revision 5636.

On a 32-bit computer (Pentium 3) only tset_d fails:
Seed GMP_CHECK_RANDOMIZE=1222666232 (include this in bug reports)
-0.10101000111100000010101100000000000000000000000000000E499
Error: get_d o set_d <> identity for d = -1.08007920352320089721e+150 -1.08007924528214274330e+150
FAIL: tset_d

Attached Files:

Changes:

No Changes Have Been Made to This Item