From 6205f48994571977d18f8d693fcf4e6461558cd3 Mon Sep 17 00:00:00 2001
From: Nicolas Pinto <pinto@mit.edu>
Date: Sun, 3 Jul 2011 19:43:18 -0400
Subject: [PATCH] Cleaning up
---
test/test_array.py | 212 +++++++++++++++------------------------------
1 file changed, 68 insertions(+), 144 deletions(-)
diff --git a/test/test_array.py b/test/test_array.py
index e9cdf6a9..591b7e7e 100644
--- a/test/test_array.py
+++ b/test/test_array.py
@@ -5,8 +5,6 @@ import sys
import pytools.test
-
-
def have_cl():
try:
import pyopencl
@@ -22,36 +20,31 @@ if have_cl():
from pyopencl.characterize import has_double_support
-
-
@pytools.test.mark_test.opencl
def test_pow_array(ctx_getter):
context = ctx_getter()
queue = cl.CommandQueue(context)
- a = np.array([1,2,3,4,5]).astype(np.float32)
+ a = np.array([1, 2, 3, 4, 5]).astype(np.float32)
a_gpu = cl_array.to_device(queue, a)
- result = pow(a_gpu,a_gpu).get()
- assert (np.abs(a**a - result) < 1e-3).all()
+ result = pow(a_gpu, a_gpu).get()
+ assert (np.abs(a ** a - result) < 1e-3).all()
- result = (a_gpu**a_gpu).get()
+ result = (a_gpu ** a_gpu).get()
assert (np.abs(pow(a, a) - result) < 1e-3).all()
-
-
@pytools.test.mark_test.opencl
def test_pow_number(ctx_getter):
context = ctx_getter()
queue = cl.CommandQueue(context)
- a = np.array([1,2,3,4,5,6,7,8,9,10]).astype(np.float32)
+ a = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9, 10]).astype(np.float32)
a_gpu = cl_array.to_device(queue, a)
result = pow(a_gpu, 2).get()
- assert (np.abs(a**2 - result) < 1e-3).all()
-
+ assert (np.abs(a ** 2 - result) < 1e-3).all()
@pytools.test.mark_test.opencl
@@ -69,7 +62,7 @@ def test_abs(ctx_getter):
res = a.get()
- for i in range (111):
+ for i in range(111):
assert abs(res[i]) >= 0
assert res[i] == i
@@ -79,13 +72,11 @@ def test_len(ctx_getter):
context = ctx_getter()
queue = cl.CommandQueue(context)
- a = np.array([1,2,3,4,5,6,7,8,9,10]).astype(np.float32)
+ a = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9, 10]).astype(np.float32)
a_cpu = cl_array.to_device(queue, a)
assert len(a_cpu) == 10
-
-
@pytools.test.mark_test.opencl
def test_multiply(ctx_getter):
"""Test the muliplication of an array with a scalar. """
@@ -93,11 +84,9 @@ def test_multiply(ctx_getter):
context = ctx_getter()
queue = cl.CommandQueue(context)
-
for sz in [10, 50000]:
for dtype, scalars in [
(np.float32, [2]),
- #(np.complex64, [2, 2j])
]:
for scalar in scalars:
a = np.arange(sz).astype(dtype)
@@ -106,6 +95,7 @@ def test_multiply(ctx_getter):
assert (a * scalar == a_doubled).all()
+
@pytools.test.mark_test.opencl
def test_multiply_array(ctx_getter):
"""Test the multiplication of two arrays."""
@@ -113,16 +103,14 @@ def test_multiply_array(ctx_getter):
context = ctx_getter()
queue = cl.CommandQueue(context)
- a = np.array([1,2,3,4,5,6,7,8,9,10]).astype(np.float32)
+ a = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9, 10]).astype(np.float32)
a_gpu = cl_array.to_device(queue, a)
b_gpu = cl_array.to_device(queue, a)
- a_squared = (b_gpu*a_gpu).get()
-
- assert (a*a == a_squared).all()
-
+ a_squared = (b_gpu * a_gpu).get()
+ assert (a * a == a_squared).all()
@pytools.test.mark_test.opencl
@@ -132,13 +120,11 @@ def test_addition_array(ctx_getter):
context = ctx_getter()
queue = cl.CommandQueue(context)
- a = np.array([1,2,3,4,5,6,7,8,9,10]).astype(np.float32)
+ a = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9, 10]).astype(np.float32)
a_gpu = cl_array.to_device(queue, a)
- a_added = (a_gpu+a_gpu).get()
-
- assert (a+a == a_added).all()
-
+ a_added = (a_gpu + a_gpu).get()
+ assert (a + a == a_added).all()
@pytools.test.mark_test.opencl
@@ -148,21 +134,20 @@ def test_addition_scalar(ctx_getter):
context = ctx_getter()
queue = cl.CommandQueue(context)
- a = np.array([1,2,3,4,5,6,7,8,9,10]).astype(np.float32)
+ a = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9, 10]).astype(np.float32)
a_gpu = cl_array.to_device(queue, a)
- a_added = (7+a_gpu).get()
-
- assert (7+a == a_added).all()
-
+ a_added = (7 + a_gpu).get()
+ assert (7 + a == a_added).all()
@pytools.test.mark_test.opencl
def test_substract_array(ctx_getter):
"""Test the substraction of two arrays."""
#test data
- a = np.array([1,2,3,4,5,6,7,8,9,10]).astype(np.float32)
- b = np.array([10,20,30,40,50,60,70,80,90,100]).astype(np.float32)
+ a = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9, 10]).astype(np.float32)
+ b = np.array([10, 20, 30, 40, 50,
+ 60, 70, 80, 90, 100]).astype(np.float32)
context = ctx_getter()
queue = cl.CommandQueue(context)
@@ -170,13 +155,11 @@ def test_substract_array(ctx_getter):
a_gpu = cl_array.to_device(queue, a)
b_gpu = cl_array.to_device(queue, b)
- result = (a_gpu-b_gpu).get()
- assert (a-b == result).all()
-
- result = (b_gpu-a_gpu).get()
- assert (b-a == result).all()
-
+ result = (a_gpu - b_gpu).get()
+ assert (a - b == result).all()
+ result = (b_gpu - a_gpu).get()
+ assert (b - a == result).all()
@pytools.test.mark_test.opencl
@@ -187,18 +170,16 @@ def test_substract_scalar(ctx_getter):
queue = cl.CommandQueue(context)
#test data
- a = np.array([1,2,3,4,5,6,7,8,9,10]).astype(np.float32)
+ a = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9, 10]).astype(np.float32)
#convert a to a gpu object
a_gpu = cl_array.to_device(queue, a)
- result = (a_gpu-7).get()
- assert (a-7 == result).all()
-
- result = (7-a_gpu).get()
- assert (7-a == result).all()
-
+ result = (a_gpu - 7).get()
+ assert (a - 7 == result).all()
+ result = (7 - a_gpu).get()
+ assert (7 - a == result).all()
@pytools.test.mark_test.opencl
@@ -208,16 +189,14 @@ def test_divide_scalar(ctx_getter):
context = ctx_getter()
queue = cl.CommandQueue(context)
- a = np.array([1,2,3,4,5,6,7,8,9,10]).astype(np.float32)
+ a = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9, 10]).astype(np.float32)
a_gpu = cl_array.to_device(queue, a)
- result = (a_gpu/2).get()
- assert (a/2 == result).all()
-
- result = (2/a_gpu).get()
- assert (np.abs(2/a - result) < 1e-5).all()
-
+ result = (a_gpu / 2).get()
+ assert (a / 2 == result).all()
+ result = (2 / a_gpu).get()
+ assert (np.abs(2 / a - result) < 1e-5).all()
@pytools.test.mark_test.opencl
@@ -228,19 +207,17 @@ def test_divide_array(ctx_getter):
queue = cl.CommandQueue(context)
#test data
- a = np.array([10,20,30,40,50,60,70,80,90,100]).astype(np.float32)
- b = np.array([10,10,10,10,10,10,10,10,10,10]).astype(np.float32)
+ a = np.array([10, 20, 30, 40, 50, 60, 70, 80, 90, 100]).astype(np.float32)
+ b = np.array([10, 10, 10, 10, 10, 10, 10, 10, 10, 10]).astype(np.float32)
a_gpu = cl_array.to_device(queue, a)
b_gpu = cl_array.to_device(queue, b)
- a_divide = (a_gpu/b_gpu).get()
- assert (np.abs(a/b - a_divide) < 1e-3).all()
-
- a_divide = (b_gpu/a_gpu).get()
- assert (np.abs(b/a - a_divide) < 1e-3).all()
-
+ a_divide = (a_gpu / b_gpu).get()
+ assert (np.abs(a / b - a_divide) < 1e-3).all()
+ a_divide = (b_gpu / a_gpu).get()
+ assert (np.abs(b / a - a_divide) < 1e-3).all()
@pytools.test.mark_test.opencl
@@ -262,8 +239,6 @@ def test_random(ctx_getter):
assert (a < 1).all()
-
-
@pytools.test.mark_test.opencl
def test_nan_arithmetic(ctx_getter):
context = ctx_getter()
@@ -272,10 +247,8 @@ def test_nan_arithmetic(ctx_getter):
def make_nan_contaminated_vector(size):
shape = (size,)
a = np.random.randn(*shape).astype(np.float32)
- #for i in range(0, shape[0], 3):
- #a[i] = float('nan')
from random import randrange
- for i in range(size//10):
+ for i in range(size // 10):
a[randrange(0, size)] = float('nan')
return a
@@ -286,14 +259,12 @@ def test_nan_arithmetic(ctx_getter):
b = make_nan_contaminated_vector(size)
b_gpu = cl_array.to_device(queue, b)
- ab = a*b
- ab_gpu = (a_gpu*b_gpu).get()
+ ab = a * b
+ ab_gpu = (a_gpu * b_gpu).get()
assert (np.isnan(ab) == np.isnan(ab_gpu)).all()
-
-
@pytools.test.mark_test.opencl
def test_elwise_kernel(ctx_getter):
context = ctx_getter()
@@ -313,7 +284,7 @@ def test_elwise_kernel(ctx_getter):
c_gpu = cl_array.empty_like(a_gpu)
lin_comb(5, a_gpu, 6, b_gpu, c_gpu)
- assert la.norm((c_gpu - (5*a_gpu+6*b_gpu)).get()) < 1e-5
+ assert la.norm((c_gpu - (5 * a_gpu + 6 * b_gpu)).get()) < 1e-5
@pytools.test.mark_test.opencl
@@ -347,34 +318,6 @@ def test_elwise_kernel_with_options(ctx_getter):
assert la.norm(gv - gt) < 1e-5
-@pytools.test.mark_test.opencl
-def test_elwise_kernel_with_options(ctx_getter):
- context = ctx_getter()
- queue = cl.CommandQueue(context)
-
- from pyopencl.clrandom import rand as clrand
-
- a_gpu = clrand(context, queue, (50,), np.float32)
- b_gpu = clrand(context, queue, (50,), np.float32)
-
- from pyopencl.elementwise import ElementwiseKernel
-
- options = ['-DADD_ONE']
- lin_comb = ElementwiseKernel(
- context,
- "float a, float *x, float b, float *y, float *z",
- "z[i] = a*x[i] + b*y[i]",
- "linear_combination",
- options=options,
- )
-
- c_gpu = cl_array.empty_like(a_gpu)
- lin_comb(5, a_gpu, 6, b_gpu, c_gpu)
-
- assert la.norm((c_gpu - (5*a_gpu+6*b_gpu)).get()) < 1e-5
-
-
-
@pytools.test.mark_test.opencl
def test_take(ctx_getter):
context = ctx_getter()
@@ -383,9 +326,7 @@ def test_take(ctx_getter):
idx = cl_array.arange(queue, 0, 200000, 2, dtype=np.uint32)
a = cl_array.arange(queue, 0, 600000, 3, dtype=np.float32)
result = cl_array.take(a, idx)
- assert ((3*idx).get() == result.get()).all()
-
-
+ assert ((3 * idx).get() == result.get()).all()
@pytools.test.mark_test.opencl
@@ -398,8 +339,6 @@ def test_arange(ctx_getter):
assert (np.arange(n, dtype=np.float32) == a.get()).all()
-
-
@pytools.test.mark_test.opencl
def test_reverse(ctx_getter):
context = ctx_getter()
@@ -414,8 +353,6 @@ def test_reverse(ctx_getter):
assert (a[::-1] == a_gpu.get()).all()
-
-
@pytools.test.mark_test.opencl
def test_sum(ctx_getter):
context = ctx_getter()
@@ -429,9 +366,7 @@ def test_sum(ctx_getter):
sum_a = np.sum(a)
sum_a_gpu = cl_array.sum(a_gpu).get()
- assert abs(sum_a_gpu-sum_a)/abs(sum_a) < 1e-4
-
-
+ assert abs(sum_a_gpu - sum_a) / abs(sum_a) < 1e-4
@pytools.test.mark_test.opencl
@@ -457,8 +392,6 @@ def test_minmax(ctx_getter):
assert op_a_gpu == op_a, (op_a_gpu, op_a, dtype, what)
-
-
@pytools.test.mark_test.opencl
def test_subset_minmax(ctx_getter):
context = ctx_getter()
@@ -499,8 +432,6 @@ def test_subset_minmax(ctx_getter):
assert min_a_gpu == min_a
-
-
@pytools.test.mark_test.opencl
def test_dot(ctx_getter):
context = ctx_getter()
@@ -516,9 +447,7 @@ def test_dot(ctx_getter):
dot_ab_gpu = cl_array.dot(a_gpu, b_gpu).get()
- assert abs(dot_ab_gpu-dot_ab)/abs(dot_ab) < 1e-4
-
-
+ assert abs(dot_ab_gpu - dot_ab) / abs(dot_ab) < 1e-4
if False:
@@ -538,7 +467,8 @@ if False:
a_gpu_slice = a_gpu[start:end]
a_slice = a[start:end]
- assert la.norm(a_gpu_slice.get()-a_slice) == 0
+ assert la.norm(a_gpu_slice.get() - a_slice) == 0
+
@pytools.test.mark_test.opencl
def test_if_positive(ctx_getter):
@@ -559,8 +489,9 @@ def test_if_positive(ctx_getter):
print(max_a_b_gpu)
print(np.maximum(a, b))
- assert la.norm(max_a_b_gpu.get()- np.maximum(a, b)) == 0
- assert la.norm(min_a_b_gpu.get()- np.minimum(a, b)) == 0
+ assert la.norm(max_a_b_gpu.get() - np.maximum(a, b)) == 0
+ assert la.norm(min_a_b_gpu.get() - np.minimum(a, b)) == 0
+
@pytools.test.mark_test.opencl
def test_take_put(ctx_getter):
@@ -570,9 +501,9 @@ def test_take_put(ctx_getter):
for n in [5, 17, 333]:
one_field_size = 8
buf_gpu = cl_array.zeros(queue,
- n*one_field_size, dtype=np.float32)
+ n * one_field_size, dtype=np.float32)
dest_indices = cl_array.to_device(queue,
- np.array([ 0, 1, 2, 3, 32, 33, 34, 35], dtype=np.uint32))
+ np.array([0, 1, 2, 3, 32, 33, 34, 35], dtype=np.uint32))
read_map = cl_array.to_device(queue,
np.array([7, 6, 5, 4, 3, 2, 1, 0], dtype=np.uint32))
@@ -580,9 +511,10 @@ def test_take_put(ctx_getter):
arrays=[buf_gpu for i in range(n)],
dest_indices=dest_indices,
src_indices=read_map,
- src_offsets=[i*one_field_size for i in range(n)],
+ src_offsets=[i * one_field_size for i in range(n)],
dest_shape=(96,))
+
@pytools.test.mark_test.opencl
def test_astype(ctx_getter):
context = ctx_getter()
@@ -607,9 +539,7 @@ def test_astype(ctx_getter):
a2 = a_gpu.astype(np.float32).get()
assert a2.dtype == np.float32
- assert la.norm(a - a2)/la.norm(a) < 1e-7
-
-
+ assert la.norm(a - a2) / la.norm(a) < 1e-7
@pytools.test.mark_test.opencl
@@ -624,13 +554,12 @@ def test_scan(ctx_getter):
knl = cls(context, dtype, "a+b", "0")
for n in [
- 10, 2**10-5, 2**10,
- 2**20-2**18,
- 2**20-2**18+5,
- 2**10+5,
- 2**20+5,
- 2**20, 2**24
- ]:
+ 10, 2 ** 10 - 5, 2 ** 10,
+ 2 ** 20 - 2 ** 18,
+ 2 ** 20 - 2 ** 18 + 5,
+ 2 ** 10 + 5,
+ 2 ** 20 + 5,
+ 2 ** 20, 2 ** 24]:
host_data = np.random.randint(0, 10, n).astype(dtype)
dev_data = cl_array.to_device(queue, host_data)
@@ -645,21 +574,17 @@ def test_scan(ctx_getter):
collect()
-
-
@pytools.test.mark_test.opencl
def test_stride_preservation(ctx_getter):
context = ctx_getter()
queue = cl.CommandQueue(context)
- A = np.random.rand(3,3)
+ A = np.random.rand(3, 3)
AT = A.T
print(AT.flags.f_contiguous, AT.flags.c_contiguous)
AT_GPU = cl_array.to_device(queue, AT)
print(AT_GPU.flags.f_contiguous, AT_GPU.flags.c_contiguous)
- assert np.allclose(AT_GPU.get(),AT)
-
-
+ assert np.allclose(AT_GPU.get(), AT)
@pytools.test.mark_test.opencl
@@ -675,10 +600,9 @@ def test_vector_fill(ctx_getter):
a_gpu = cl_array.zeros(queue, 100, dtype=cl_array.vec.float4)
-
-
if __name__ == "__main__":
- # make sure that import failures get reported, instead of skipping the tests.
+ # make sure that import failures get reported, instead of skipping the
+ # tests.
import pyopencl as cl
import sys
--
GitLab