def test_cuda():
try:
from pycuda.tools import mark_cuda_test
except ImportError:
pass
yield "SP CUDA Maxwell", mark_cuda_test(do_test_maxwell_cavities_cuda), numpy.float32
yield "DP CUDA Maxwell", mark_cuda_test(do_test_maxwell_cavities_cuda), numpy.float64
def test_cuda():
try:
from pycuda.tools import mark_cuda_test
except ImportError:
pass
yield "SP CUDA Maxwell", mark_cuda_test(
do_test_maxwell_cavities_cuda), np.float32
yield "DP CUDA Maxwell", mark_cuda_test(
do_test_maxwell_cavities_cuda), np.float64
def make_unary_function_test(name, a=0, b=1, threshold=0, complex=False):
def test():
gpu_func = getattr(cumath, name)
cpu_func = getattr(np, numpy_func_names.get(name, name))
if complex:
_dtypes = complex_dtypes
else:
_dtypes = dtypes
for s in sizes:
for dtype in _dtypes:
np.random.seed(1)
A = (np.random.random(s) * (b - a) + a).astype(dtype)
if complex:
A += (np.random.random(s) * (b - a) + a) * 1j
args = gpuarray.to_gpu(A)
gpu_results = gpu_func(args).get()
cpu_results = cpu_func(A)
max_err = np.max(np.abs(cpu_results - gpu_results))
assert (max_err <= threshold).all(), \
(max_err, name, dtype)
gpu_results2 = gpuarray.empty_like(args)
gr2 = gpu_func(args, out=gpu_results2)
assert gpu_results2 is gr2
gr2 = gr2.get()
max_err = np.max(np.abs(cpu_results - gr2))
assert (max_err <= threshold).all(), \
(max_err, name, dtype)
return mark_cuda_test(test)
def make_unary_function_test(name, a=0, b=1, threshold=0, complex=False):
def test():
gpu_func = getattr(cumath, name)
cpu_func = getattr(np, numpy_func_names.get(name, name))
if complex:
_dtypes = complex_dtypes
else:
_dtypes = dtypes
for s in sizes:
for dtype in _dtypes:
np.random.seed(1)
A = (np.random.random(s)*(b-a) + a).astype(dtype)
if complex:
A += (np.random.random(s)*(b-a) + a)*1j
args = gpuarray.to_gpu(A)
gpu_results = gpu_func(args).get()
cpu_results = cpu_func(A)
max_err = np.max(np.abs(cpu_results - gpu_results))
assert (max_err <= threshold).all(), \
(max_err, name, dtype)
return mark_cuda_test(test)
def make_unary_function_test(name, a=0, b=1, threshold=0):
def test():
gpu_func = getattr(cumath, name)
cpu_func = getattr(np, numpy_func_names.get(name, name))
for s in sizes:
for dtype in dtypes:
args = gpuarray.arange(a, b, (b - a) / s, dtype=np.float32)
gpu_results = gpu_func(args).get()
cpu_results = cpu_func(args.get())
max_err = np.max(np.abs(cpu_results - gpu_results))
assert (max_err <= threshold).all(), \
(max_err, name, dtype)
return mark_cuda_test(test)
def make_unary_function_test(name, a=0, b=1, threshold=0):
def test():
gpu_func = getattr(cumath, name)
cpu_func = getattr(np, numpy_func_names.get(name, name))
for s in sizes:
for dtype in dtypes:
args = gpuarray.arange(a, b, (b-a)/s, dtype=np.float32)
gpu_results = gpu_func(args).get()
cpu_results = cpu_func(args.get())
max_err = np.max(np.abs(cpu_results - gpu_results))
assert (max_err <= threshold).all(), \
(max_err, name, dtype)
return mark_cuda_test(test)
def make_unary_function_test(name, (a, b)=(0, 1), threshold=0):
def test():
gpu_func = getattr(cumath, name)
cpu_func = getattr(np, numpy_func_names.get(name, name))
for s in sizes:
for dtype in dtypes:
args = gpuarray.arange(a, b, (b - a) / s, dtype=np.float32)
gpu_results = gpu_func(args).get()
cpu_results = cpu_func(args.get())
max_err = np.max(np.abs(cpu_results - gpu_results))
assert (max_err <= threshold).all(), \
(max_err, name, dtype)
return mark_cuda_test(test)
if have_pycuda():
test_ceil = make_unary_function_test("ceil", (-10, 10))
test_floor = make_unary_function_test("ceil", (-10, 10))
test_fabs = make_unary_function_test("fabs", (-10, 10))
test_exp = make_unary_function_test("exp", (-3, 3), 1e-5)
test_log = make_unary_function_test("log", (1e-5, 1), 5e-7)
test_log10 = make_unary_function_test("log10", (1e-5, 1), 3e-7)
test_sqrt = make_unary_function_test("sqrt", (1e-5, 1), 2e-7)
test_sin = make_unary_function_test("sin", (-10, 10), 1e-7)
test_cos = make_unary_function_test("cos", (-10, 10), 1e-7)
test_asin = make_unary_function_test("asin", (-0.9, 0.9), 5e-7)
test_acos = make_unary_function_test("acos", (-0.9, 0.9), 5e-7)
def make_unary_function_test(name, (a, b)=(0, 1), threshold=0):
def test():
gpu_func = getattr(cumath, name)
cpu_func = getattr(np, numpy_func_names.get(name, name))
for s in sizes:
for dtype in dtypes:
args = gpuarray.arange(a, b, (b-a)/s, dtype=np.float32)
gpu_results = gpu_func(args).get()
cpu_results = cpu_func(args.get())
max_err = np.max(np.abs(cpu_results - gpu_results))
assert (max_err <= threshold).all(), \
(max_err, name, dtype)
return mark_cuda_test(test)
if have_pycuda():
test_ceil = make_unary_function_test("ceil", (-10, 10))
test_floor = make_unary_function_test("ceil", (-10, 10))
test_fabs = make_unary_function_test("fabs", (-10, 10))
test_exp = make_unary_function_test("exp", (-3, 3), 1e-5)
test_log = make_unary_function_test("log", (1e-5, 1), 5e-7)
test_log10 = make_unary_function_test("log10", (1e-5, 1), 3e-7)
test_sqrt = make_unary_function_test("sqrt", (1e-5, 1), 2e-7)
test_sin = make_unary_function_test("sin", (-10, 10), 1e-7)
test_cos = make_unary_function_test("cos", (-10, 10), 1e-7)
