BLI_math 'compare' cleanup & enhancements.
authorBastien Montagne <montagne29@wanadoo.fr>
Fri, 10 Jul 2015 12:32:35 +0000 (14:32 +0200)
committerBastien Montagne <montagne29@wanadoo.fr>
Fri, 10 Jul 2015 13:02:43 +0000 (15:02 +0200)
commit7837f0e8332f3726e0322b0c48b0da4d7c2d5813
tree4e0b9d6fb5faa6bacd080484d7ef625566d32699
parentbbcbd2eed9081e0c1127d32e41592a8dc6b5604b
BLI_math 'compare' cleanup & enhancements.

This commit:
* Adds a 'compare_ff' function for absolute 'almost equal' comparison of floats.
* Makes 'compare_vxvx' functions use that new 'compare_ff' one.
* Adds a 'compare_ff_relative' function for secured ulp-based relative comparison of floats.
* Adds matching 'compare_vxvx_relative' functions.
* Adds some basic tests for compare_ff_relative.

See https://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/

Note that we could replace our python/mathutils' EXPP_FloatsAreEqual() by BLI's compare_ff_relative
(using a very small absolute max_diff), but these do not have exact same behavior...
Left a comment there for now, we can do it later if/when we are sure it won't break anything!
source/blender/blenlib/BLI_math_base.h
source/blender/blenlib/BLI_math_vector.h
source/blender/blenlib/intern/math_base_inline.c
source/blender/blenlib/intern/math_vector_inline.c
source/blender/python/mathutils/mathutils.c
tests/gtests/blenlib/BLI_math_base_test.cc [new file with mode: 0644]
tests/gtests/blenlib/CMakeLists.txt