Ada complex accuracy fail on x86_64 only (works on x86)

Laurent GUERBY laurent@guerby.net
Wed Oct 27 23:32:00 GMT 2004


LAST_UPDATED between

Sat Oct 23 21:08:20 CEST 2004
Sat Oct 23 19:08:20 UTC 2004

and:

Wed Oct 27 13:15:56 CEST 2004
Wed Oct 27 11:15:56 UTC 2004

(just before Arnaud commits), the following 12 ACATS test started
failing on x86_64 (but pass on x86):

cxg1001 cxg1002 cxg1003 cxg1005 cxg2002
cxg2006 cxg2008 cxg2009 cxg2018 cxg2019
cxg2020 cxg2021

They all test complex accuracy and fail pretty
badly so that's probably a code generation problem.

Laurent

,.,. CXG1001 ACATS 2.5 04-10-27 14:53:38
---- CXG1001 Check that the subprograms defined in the package
                Ada.Numerics.Generic_Complex_Types provide correct
                results.
   * CXG1001 Incorrect results from Function Re.
   * CXG1001 Incorrect result from Function Compose_From_Polar - 2.
   * CXG1001 Incorrect results from Function Modulus.
   * CXG1001 Incorrect results from Function abs.
**** CXG1001 FAILED ****************************.

,.,. CXG1002 ACATS 2.5 04-10-27 14:53:38
---- CXG1002 Check that the subprograms defined in the package
                Ada.Numerics.Generic_Complex_Types provide the
                prescribed results.
   * CXG1002 Incorrect results from Function "+" with one Complex and
                one Imaginary argument.
   * CXG1002 Incorrect results from Function "-" with one Complex and
                one Imaginary argument.
   * CXG1002 Incorrect results from Function "+" with one Real and one
                Imaginary argument.
**** CXG1002 FAILED ****************************.

,.,. CXG2002 ACATS 2.5 04-10-27 14:53:45
---- CXG2002 Check the accuracy of the complex modulus function.
   * CXG2002 test 1 -- abs(bigreal + 0i) actual:  0.00000E+00 expected: 
                3.40282E+38 difference:  3.40282E+38 max_err:
                1.21694E+32.
   * CXG2002 Constraint_Error raised in test 2.
   * CXG2002 test 3 -- abs(3 + 4*i) actual:  5.65685E+00 expected: 
                5.00000E+00 difference: -6.56854E-01 max_err:
                1.78814E-06.
   * CXG2002 test 4 -- abs(3S + 4S*i) for large S actual:  2.40616E+38
                expected:  2.12676E+38 difference: -2.79395E+37 max_err:
                1.51116E+25.
   * CXG2002 test 1 -- abs(bigreal + 0i) actual:  0.00000E+00 expected: 
                3.40282E+38 difference:  3.40282E+38 max_err:
                1.21694E+32.
   * CXG2002 Constraint_Error raised in test 2.
   * CXG2002 test 3 -- abs(3 + 4*i) actual:  5.65685E+00 expected: 
                5.00000E+00 difference: -6.56854E-01 max_err:
                1.78814E-06.
   * CXG2002 test 4 -- abs(3S + 4S*i) for large S actual:  2.40616E+38
                expected:  2.12676E+38 difference: -2.79395E+37 max_err:
                1.51116E+25.
**** CXG2002 FAILED ****************************.

etc...



More information about the Gcc mailing list