summaryrefslogtreecommitdiff
path: root/numpy/core/setup.py
diff options
context:
space:
mode:
Diffstat (limited to 'numpy/core/setup.py')
-rw-r--r--numpy/core/setup.py10
1 files changed, 10 insertions, 0 deletions
diff --git a/numpy/core/setup.py b/numpy/core/setup.py
index 3bf6f6250..100f2fc80 100644
--- a/numpy/core/setup.py
+++ b/numpy/core/setup.py
@@ -1003,6 +1003,7 @@ def configuration(parent_package='',top_path=None):
join('src', 'umath', 'loops_trigonometric.dispatch.c.src'),
join('src', 'umath', 'loops_umath_fp.dispatch.c.src'),
join('src', 'umath', 'loops_exponent_log.dispatch.c.src'),
+ join('src', 'umath', 'loops_hyperbolic.dispatch.c.src'),
join('src', 'umath', 'matmul.h.src'),
join('src', 'umath', 'matmul.c.src'),
join('src', 'umath', 'clip.h'),
@@ -1033,8 +1034,17 @@ def configuration(parent_package='',top_path=None):
svml_path = join('numpy', 'core', 'src', 'umath', 'svml')
svml_objs = []
+ # we have converted the following into universal intrinsics
+ # so we can bring the benefits of performance for all platforms
+ # not just for avx512 on linux without performance/accuracy regression,
+ # actually the other way around, better performance and
+ # after all maintainable code.
+ svml_filter = (
+ 'svml_z0_tanh_d_la.s', 'svml_z0_tanh_s_la.s'
+ )
if can_link_svml() and check_svml_submodule(svml_path):
svml_objs = glob.glob(svml_path + '/**/*.s', recursive=True)
+ svml_objs = [o for o in svml_objs if not o.endswith(svml_filter)]
config.add_extension('_multiarray_umath',
# Forcing C language even though we have C++ sources.