This patch moves tests of fma to auto-libm-test-in, adding the
required support to gen-auto-libm-tests.
Because fma can have exact zero results depending on the rounding
mode, results of fma cannot always be determined from a single value
computed in higher precision with a sticky bit. Thus, this patch adds
support for recomputing results with the original MPFR/MPC function in
the case where an exact zero is involved. (This also affects some
results for cpow; when we start testing cpow in all rounding modes, I
think it will be most appropriate to make those tests use
IGNORE_ZERO_INF_SIGN, since ISO C does not attempt to determine signs
of zero results, or special caes in general, for cpow, and I think
signs of zero for cpow are beyond the scope of glibc's accuracy
goals.)
Simply treating the existing test inputs for fma like those for other
functions (i.e., as representing the given value rounded up or down to
any of the supported floating-point formats) increases the size of
auto-libm-test-out by about 16MB (i.e., about half the file is fma
test data). While rounded versions of tests are perfectly reasonable
test inputs for fma, in this case having them seems excessive, so this
patch allows functions to specify in gen-auto-libm-tests that the
given test inputs are only to be interpreted exactly, not as
corresponding to values rounded up and down. This reduces the size of
the generated test data for fma to a more reasonable 2MB.
A consequence of this patch is that fma is now tested for correct
presence or absence of "inexact" exceptions, where previously this
wasn't tested because I didn't want to try to add that test coverage
manually to all the existing tests. As far as I know, the existing
fma implementations are already correct in this regard.
This patch provides the first cases where the gen-auto-libm-tests
support for distinguishing before-rounding/after-rounding underflow
actually produces separate entries in auto-libm-test-out (for
functions without exactly determined results, the affected cases are
all considered underflow-optional, so this only affects functions like
fma with exactly determined results). I didn't see any signs of
problems with this logic in the output.
Tested x86_64 and x86.
* math/auto-libm-test-in: Add tests of fma.
* math/auto-libm-test-out: Regenerated.
* math/libm-test.inc (fma_test_data): Use AUTO_TESTS_fff_f.
(fma_towardzero_test_data): Likewise.
(fma_downward_test_data): Likewise.
(fma_upward_test_data): Likewise.
* math/gen-auto-libm-tests.c (rounding_mode_desc): Add field
mpc_mode.
(rounding_modes): Add values for new field.
(func_calc_method): Add value mpfr_fff_f.
(func_calc_desc): Add mpfr_fff_f union field.
(test_function): Add field exact_args.
(FUNC): Add macro argument EXACT_ARGS.
(FUNC_mpfr_f_f): Update call to FUNC.
(FUNC_mpfr_f_f): Likewise.
(FUNC_mpfr_ff_f): Likewise.
(FUNC_mpfr_if_f): Likewise.
(FUNC_mpc_c_f): Likewise.
(FUNC_mpc_c_c): Likewise.
(test_functions): Add fma. Update calls to FUNC.
(handle_input_arg): Add argument exact_args.
(add_test): Update call to handle_input_arg.
(calc_generic_results): Add argument mode. Handle mpfr_fff_f.
(output_for_one_input_case): Update call to calc_generic_results.
Recalculate exact zero results in each rounding mode.
README for libm-test math test suite
====================================
The libm-test math test suite tests a number of function points of
math functions in the GNU C library. The following sections contain a
brief overview. Please note that the test drivers and the Perl script
"gen-libm-test.pl" have some options. A full list of options is
available with --help (for the test drivers) and -h for
"gen-libm-test.pl".
What is tested?
===============
The tests just evaluate the functions at specified points and compare
the results with precomputed values and the requirements of the ISO
C99 standard.
Besides testing the special values mandated by IEEE 754 (infinities,
NaNs and minus zero), some more or less random values are tested.
Files that are part of libm-test
================================
The main file is "libm-test.inc". It is platform and floating point
format independent. The file must be preprocessed by the Perl script
"gen-libm-test.pl". The results are "libm-test.c" and a file
"libm-test-ulps.h" with platform specific deltas.
The test drivers test-double.c, test-float.c, test-ldouble.c test the
normal double, float and long double implementation of libm. The test
drivers with an i in it (test-idouble.c, test-ifloat.c,
test-ildoubl.c) test the corresponding inline functions (where
available - otherwise they also test the real functions in libm).
"gen-libm-test.pl" needs a platform specific files with ULPs (Units of
Last Precision). The file is called "libm-test-ulps" and lives in
platform specific sysdep directory.
How can I generate "libm-test-ulps"?
====================================
To automatically generate a new "libm-test-ulps" run "make regen-ulps".
This generates the file "math/NewUlps" in the build directory. The file
contains the sorted results of all the tests. You can use the "NewUlps"
file as the machine's updated "libm-test-ulps" file. Copy "NewUlps" to
"libm-test-ulps" in the appropriate machine sysdep directory. Verify
the changes, post your patch, and check it in after review.
To manually generate a new "libm-test-ulps" file, first remove "ULPs"
file in the current directory, then you can execute for example:
./testrun.sh math/test-double -u --ignore-max-ulp=yes
This generates a file "ULPs" with all double ULPs in it, ignoring any
previously calculated ULPs, and running with the newly built dynamic
loader and math library (assumes you didn't install your build). Now
generate the ULPs for all other formats, the tests will be appending the
data to the "ULPs" file. As final step run "gen-libm-test.pl" with the
file as input and ask to generate a pretty printed output in the file
"NewUlps":
gen-libm-test.pl -u ULPs -n
Copy "NewUlps" to "libm-test-ulps" in the appropriate machine sysdep
directory.
Note that the test drivers have an option "-u" to output an unsorted
list of all epsilons that the functions have. The output can be read
in directly but it's better to pretty print it first.
"gen-libm-test.pl" has an option to generate a pretty-printed and
sorted new ULPs file from the output of the test drivers.
Contents of libm-test-ulps
==========================
Since libm-test-ulps can be generated automatically, just a few
notes. The file contains lines for single tests, like:
Test "cos (pi/2) == 0":
float: 1
and lines for maximal errors of single functions, like:
Function "yn":
idouble: 6.0000
The keywords are float, ifloat, double, idouble, ldouble and ildouble
(the prefix i stands for inline).
Adding tests to libm-test.inc
=============================
The tests are evaluated by a set of special test macros. The macros
start with "TEST_" followed by a specification the input values, an
underscore and a specification of the output values. As an example,
the test macro for a function with input of type FLOAT (FLOAT is
either float, double, long double) and output of type FLOAT is
"TEST_f_f". The macro's parameter are the name of the function, the
input parameter, output parameter and optionally one exception
parameter.
The accepted parameter types are:
- "f" for FLOAT
- "b" for boolean - just tests if the output parameter evaluates to 0
or 1 (only for output).
- "c" for complex. This parameter needs two values, first the real,
then the imaginary part.
- "i" for int.
- "l" for long int.
- "L" for long long int.
- "F" for the address of a FLOAT (only as input parameter)
- "I" for the address of an int (only as input parameter)
c6af2d896c