tree_d_3d.kernels.tests.test_metal_kernels

The Metal versions of the kernels on the GPU of an Apple-silicon Mac, against the CPU.

Each test calls a <name>_metal.py function and the Pyccel (or NumPy) version with the same arrays. The GPU computes in float32, so the results agree to about 1e-5 of the largest value. Skipped without MLX and a Metal GPU.

Attributes

Name	Description
SOFTENING	No description.
TOL	No description.
pytestmark	No description.

Functions

Name	Description
system	No description.
test_direct_sums	No description.
test_direct_sums_report_coincident_charges	No description.
test_pair_energy	No description.
test_tree_evaluation	No description.

SOFTENINGattributemodule attribute#

View source

SOFTENING = 0.01

TOLattributemodule attribute#

View source

TOL = {'rtol': 0.0, 'atol': 3e-05}

pytestmarkattributemodule attribute#

View source

pytestmark = pytest.mark.skipif(not metal_available(), reason='MLX with a Metal GPU is not available')

systemfunction#

View source

def system()

test_direct_sumsfunction#

View source

def test_direct_sums(system, name, metal)

test_direct_sums_report_coincident_chargesfunction#

View source

def test_direct_sums_report_coincident_charges(name, metal)

test_pair_energyfunction#

View source

def test_pair_energy(system)

test_tree_evaluationfunction#

View source

def test_tree_evaluation(system, kernel)