Re: [Boost-testing] [math] floating point classification - testing help wanted

From: Martin Wille (mw8329_at_[hidden])
Date: 2005-12-01 12:50:47

John Maddock wrote:

> So.... I'm looking for testers to run the attached program on as many 
> different platforms/compilers as possible, basically I want to see if 
> there's any mileage in this or if I'm on to a looser with this code!
> 
> Please build the test program with optimisations turned on: many 
> compilers make non-IEEE arithmetic assumptions once you tell them to 
> optimise the code.

Some results:

Linux-x86_64, glibc-2.3.5:

  gcc 3.3.6 without optimizations: ok
  gcc 3.3.6 -ffast-math breaks (1)
  gcc 3.3.6 with a lot of optimizations, without -ffast-math: ok

  gcc 3.4.4 without optimizations: ok
  gcc 3.4.4 -ffast-math breaks (1)
  gcc 3.4.4 with a lot of optimizations, without -ffast-math: ok

  gcc 4.0.0 without optimizations: ok
  gcc 4.0.0 with a lot of optimizations, without -ffast-math: ok
  gcc 4.0.0 with a lot of optimizations, including -ffast-math: ok

Linux-x86, glibc-2.3.5:

  gcc 3.2.3 without optimizations: ok
  gcc 3.2.3 with a lot of optimizations, including -ffast-math: ok

  gcc 3.3.6 without optimizations: ok
  gcc 3.3.6 -ffast-math breaks (1)
  gcc 3.3.6 with a lot of optimizations, without -ffast-math: ok

  gcc 3.4.4 without optimizations: ok
  gcc 3.4.4 -ffast-math breaks (1)
  gcc 3.4.4 with a lot of optimizations, without -ffast-math: ok

  gcc 4.0.2 without optimizations: ok
  gcc 4.0.2 with a lot of optimizations, without -ffast-math: ok
  gcc 4.0.2 with a lot of optimizations, including -ffast-math: ok

  icc 8.1 without optimizations, using gcc 3.3.6 libs: breaks (1)
  icc 8.1 with -O3, using gcc 3.3.6 libs:  breaks (1)
  icc 8.1 without optimizations, using icc libs: ok
  icc 8.1 with -O3, using icc libs: ok

  icc 9.0 without optimizations, using gcc 3.3.6 libs: breaks (1)
  icc 9.0 with -O3, using gcc 3.3.6 libs:  breaks (1)
  icc 9.0 without optimizations, using icc libs: ok
  icc 9.0 with -O3, using icc libs: ok

I didn't use -ffinite-math-only nor -funsafe-math-optimizations flags.

HTH,
m

1:

Running 1 test case...
Platform has FP_NORMAL macro.
FP_ZERO: 2
FP_NORMAL: 4
FP_INFINITE: 1
FP_NAN: 0
FP_SUBNORMAL: 3
Testing type float
classify.cpp(167): error in "test_main_caller( argc, argv )": check 
(::pending::fpclassify)(t) == (int)FP_NAN failed [1 != 0]
classify.cpp(168): error in "test_main_caller( argc, argv )": check 
(::pending::fpclassify)(-t) == (int)FP_NAN failed [1 != 0]
classify.cpp(173): error in "test_main_caller( argc, argv )": check 
(::pending::fpclassify)(t) == (int)FP_NAN failed [1 != 0]
classify.cpp(174): error in "test_main_caller( argc, argv )": check 
(::pending::fpclassify)(-t) == (int)FP_NAN failed [1 != 0]
Testing type double
classify.cpp(167): error in "test_main_caller( argc, argv )": check 
(::pending::fpclassify)(t) == (int)FP_NAN failed [1 != 0]
classify.cpp(168): error in "test_main_caller( argc, argv )": check 
(::pending::fpclassify)(-t) == (int)FP_NAN failed [1 != 0]
classify.cpp(173): error in "test_main_caller( argc, argv )": check 
(::pending::fpclassify)(t) == (int)FP_NAN failed [1 != 0]
classify.cpp(174): error in "test_main_caller( argc, argv )": check 
(::pending::fpclassify)(-t) == (int)FP_NAN failed [1 != 0]
Testing type long double
classify.cpp(167): error in "test_main_caller( argc, argv )": check 
(::pending::fpclassify)(t) == (int)FP_NAN failed [2 != 0]
classify.cpp(168): error in "test_main_caller( argc, argv )": check 
(::pending::fpclassify)(-t) == (int)FP_NAN failed [2 != 0]
classify.cpp(173): error in "test_main_caller( argc, argv )": check 
(::pending::fpclassify)(t) == (int)FP_NAN failed [2 != 0]
classify.cpp(174): error in "test_main_caller( argc, argv )": check 
(::pending::fpclassify)(-t) == (int)FP_NAN failed [2 != 0]

Send instant messages to your online friends http://au.messenger.yahoo.com