From c7718d2023a47cf79b571b260099bf4742320944 Mon Sep 17 00:00:00 2001
From: Alexandru Fikl <alexfikl@gmail.com>
Date: Sat, 6 Aug 2022 09:20:28 +0300
Subject: [PATCH] use np.random.default_rng with a seed consistently
---
doc/algorithm.rst | 3 +-
examples/ipython-demo.ipynb | 397 ++++++++++++++++++------------------
test/test_algorithm.py | 6 +-
test/test_array.py | 60 +++---
test/test_clmath.py | 12 +-
test/test_enqueue_copy.py | 22 +-
test/test_wrapper.py | 32 ++-
7 files changed, 275 insertions(+), 257 deletions(-)
diff --git a/doc/algorithm.rst b/doc/algorithm.rst
index a3cad802..aa21fbeb 100644
--- a/doc/algorithm.rst
+++ b/doc/algorithm.rst
@@ -241,7 +241,8 @@ Here's a usage example::
knl = InclusiveScanKernel(context, np.int32, "a+b")
n = 2**20-2**18+5
- host_data = np.random.randint(0, 10, n).astype(np.int32)
+ rng = np.random.default_rng(seed=42)
+ host_data = rng.integers(0, 10, size=n, dtype=np.int32)
dev_data = cl_array.to_device(queue, host_data)
knl(dev_data)
diff --git a/examples/ipython-demo.ipynb b/examples/ipython-demo.ipynb
index bf646511..b27617ed 100644
--- a/examples/ipython-demo.ipynb
+++ b/examples/ipython-demo.ipynb
@@ -1,202 +1,203 @@
{
- "metadata": {
- "name": "",
- "signature": "sha256:81f3deed7cdc26b0fc756b3ee1eb6e8f9b1be96304ddfc6ff484d223c2b8a942"
- },
- "nbformat": 3,
- "nbformat_minor": 0,
- "worksheets": [
+ "cells": [
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "cc7d0709",
+ "metadata": {
+ "collapsed": false,
+ "jupyter": {
+ "outputs_hidden": false
+ }
+ },
+ "outputs": [],
+ "source": [
+ "from __future__ import division\n",
+ "import numpy as np\n",
+ "import pyopencl as cl\n",
+ "import pyopencl.array"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "8ac8d7bb",
+ "metadata": {},
+ "source": [
+ "Load the PyOpenCL IPython extension:"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "7023ca2f",
+ "metadata": {
+ "collapsed": false,
+ "jupyter": {
+ "outputs_hidden": false
+ }
+ },
+ "outputs": [],
+ "source": [
+ "%load_ext pyopencl.ipython_ext"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "9544b53c",
+ "metadata": {},
+ "source": [
+ "Create an OpenCL context and a command queue:"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "fac17999",
+ "metadata": {
+ "collapsed": false,
+ "jupyter": {
+ "outputs_hidden": false
+ }
+ },
+ "outputs": [],
+ "source": [
+ "ctx = cl.create_some_context(interactive=True)\n",
+ "queue = cl.CommandQueue(ctx)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "a29daf04",
+ "metadata": {},
+ "source": [
+ "-----\n",
+ "\n",
+ "Define an OpenCL kernel using the `%%cl_kernel` magic:"
+ ]
+ },
{
- "cells": [
- {
- "cell_type": "code",
- "collapsed": false,
- "input": [
- "from __future__ import division\n",
- "import numpy as np\n",
- "import pyopencl as cl\n",
- "import pyopencl.array"
- ],
- "language": "python",
- "metadata": {},
- "outputs": [
- {
- "output_type": "stream",
- "stream": "stderr",
- "text": [
- "/usr/lib/python2.7/pkgutil.py:186: ImportWarning: Not importing directory '/usr/lib/python2.7/dist-packages/enthought': missing __init__.py\n",
- " file, filename, etc = imp.find_module(subname, path)\n"
- ]
- }
- ],
- "prompt_number": 1
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "Load the PyOpenCL IPython extension:"
- ]
- },
- {
- "cell_type": "code",
- "collapsed": false,
- "input": [
- "%load_ext pyopencl.ipython_ext"
- ],
- "language": "python",
- "metadata": {},
- "outputs": [],
- "prompt_number": 3
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "Create an OpenCL context and a command queue:"
- ]
- },
- {
- "cell_type": "code",
- "collapsed": false,
- "input": [
- "ctx = cl.create_some_context(interactive=True)\n",
- "queue = cl.CommandQueue(ctx)"
- ],
- "language": "python",
- "metadata": {},
- "outputs": [
- {
- "output_type": "stream",
- "stream": "stdout",
- "text": [
- "Choose platform:\n",
- "[0] <pyopencl.Platform 'AMD Accelerated Parallel Processing' at 0x7fc14f1b0080>\n",
- "[1] <pyopencl.Platform 'Intel(R) OpenCL' at 0x32aed00>\n"
- ]
- },
- {
- "name": "stdout",
- "output_type": "stream",
- "stream": "stdout",
- "text": [
- "Choice [0]:0\n"
- ]
- },
- {
- "output_type": "stream",
- "stream": "stdout",
- "text": [
- "Set the environment variable PYOPENCL_CTX='0' to avoid being asked again.\n"
- ]
- }
- ],
- "prompt_number": 5
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "-----\n",
- "\n",
- "Define an OpenCL kernel using the `%%cl_kernel` magic:"
- ]
- },
- {
- "cell_type": "code",
- "collapsed": false,
- "input": [
- "%%cl_kernel -o \"-cl-fast-relaxed-math\"\n",
- "\n",
- "__kernel void sum_vector(__global const float *a,\n",
- "__global const float *b, __global float *c)\n",
- "{\n",
- " int gid = get_global_id(0);\n",
- " c[gid] = a[gid] + b[gid];\n",
- "}"
- ],
- "language": "python",
- "metadata": {},
- "outputs": [],
- "prompt_number": 6
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "This looks for `cl_ctx` or `ctx` in the user namespace to find a PyOpenCL context.\n",
- "\n",
- "Kernel names are automatically injected into the user namespace, so we can just use `sum_vector` from Python below.\n",
- "\n",
- "Now create some data to work on:"
- ]
- },
- {
- "cell_type": "code",
- "collapsed": false,
- "input": [
- "n = 10000\n",
- "\n",
- "a = cl.array.empty(queue, n, dtype=np.float32)\n",
- "a.fill(15)\n",
- "\n",
- "b_host = np.random.randn(n).astype(np.float32)\n",
- "b = cl.array.to_device(queue, b_host)\n",
- "\n",
- "c = cl.array.empty_like(a)"
- ],
- "language": "python",
- "metadata": {},
- "outputs": [],
- "prompt_number": 7
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "Run the kernel:"
- ]
- },
- {
- "cell_type": "code",
- "collapsed": false,
- "input": [
- "sum_vector(queue, (n,), None, a.data, b.data, c.data)"
- ],
- "language": "python",
- "metadata": {},
- "outputs": [
- {
- "metadata": {},
- "output_type": "pyout",
- "prompt_number": 8,
- "text": [
- "<pyopencl._cl.Event at 0x7fc14f3fdf30>"
- ]
- }
- ],
- "prompt_number": 8
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "Check the result using `numpy` operations:"
- ]
- },
- {
- "cell_type": "code",
- "collapsed": false,
- "input": [
- "assert (c.get() == b_host + 15).all()"
- ],
- "language": "python",
- "metadata": {},
- "outputs": [],
- "prompt_number": 9
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "65c7e6c9",
+ "metadata": {
+ "collapsed": false,
+ "jupyter": {
+ "outputs_hidden": false
}
- ],
- "metadata": {}
+ },
+ "outputs": [],
+ "source": [
+ "%%cl_kernel -o \"-cl-fast-relaxed-math\"\n",
+ "\n",
+ "__kernel void sum_vector(__global const float *a,\n",
+ "__global const float *b, __global float *c)\n",
+ "{\n",
+ " int gid = get_global_id(0);\n",
+ " c[gid] = a[gid] + b[gid];\n",
+ "}"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "cfb57357",
+ "metadata": {},
+ "source": [
+ "This looks for `cl_ctx` or `ctx` in the user namespace to find a PyOpenCL context.\n",
+ "\n",
+ "Kernel names are automatically injected into the user namespace, so we can just use `sum_vector` from Python below.\n",
+ "\n",
+ "Now create some data to work on:"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "1d80ff38",
+ "metadata": {
+ "collapsed": false,
+ "jupyter": {
+ "outputs_hidden": false
+ }
+ },
+ "outputs": [],
+ "source": [
+ "n = 10000\n",
+ "\n",
+ "a = cl.array.empty(queue, n, dtype=np.float32)\n",
+ "a.fill(15)\n",
+ "\n",
+ "b_host = np.random.randn(n).astype(np.float32)\n",
+ "b = cl.array.to_device(queue, b_host)\n",
+ "\n",
+ "c = cl.array.empty_like(a)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "61fccb61",
+ "metadata": {},
+ "source": [
+ "Run the kernel:"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "2ba991b3",
+ "metadata": {
+ "collapsed": false,
+ "jupyter": {
+ "outputs_hidden": false
+ }
+ },
+ "outputs": [],
+ "source": [
+ "sum_vector(queue, (n,), None, a.data, b.data, c.data)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "11a55b38",
+ "metadata": {},
+ "source": [
+ "Check the result using `numpy` operations:"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "ee3560c1",
+ "metadata": {
+ "collapsed": false,
+ "jupyter": {
+ "outputs_hidden": false
+ }
+ },
+ "outputs": [],
+ "source": [
+ "assert (c.get() == b_host + 15).all()"
+ ]
}
- ]
-}
\ No newline at end of file
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "Python 3 (ipykernel)",
+ "language": "python",
+ "name": "python3"
+ },
+ "language_info": {
+ "codemirror_mode": {
+ "name": "ipython",
+ "version": 3
+ },
+ "file_extension": ".py",
+ "mimetype": "text/x-python",
+ "name": "python",
+ "nbconvert_exporter": "python",
+ "pygments_lexer": "ipython3",
+ "version": "3.10.5"
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}
diff --git a/test/test_algorithm.py b/test/test_algorithm.py
index baee3c66..40fe854b 100644
--- a/test/test_algorithm.py
+++ b/test/test_algorithm.py
@@ -571,8 +571,9 @@ def test_scan(ctx_factory, dtype, scan_cls):
knl = scan_cls(context, dtype, "a+b", "0")
+ rng = np.random.default_rng(seed=42)
for n in scan_test_counts:
- host_data = np.random.randint(0, 10, n).astype(dtype)
+ host_data = rng.integers(0, 10, n, dtype=dtype)
dev_data = cl_array.to_device(queue, host_data)
# /!\ fails on Nv GT2?? for some drivers
@@ -617,7 +618,8 @@ def test_scan_with_vectorargs_with_offsets(ctx_factory, scan_cls):
n = 20
- host_data = np.random.randint(0, 10, n).astype(float)
+ rng = np.random.default_rng(seed=42)
+ host_data = rng.integers(0, 10, n).astype(np.float64)
dev_data = cl.array.to_device(queue, host_data)
segment_data = np.zeros(n, dtype=int)
dev_segment_data = cl.array.to_device(queue, segment_data)
diff --git a/test/test_array.py b/test/test_array.py
index db59902f..ffb0714c 100644
--- a/test/test_array.py
+++ b/test/test_array.py
@@ -757,8 +757,6 @@ def test_bitwise(ctx_factory):
for a_dtype, b_dtype in product(dtypes, dtypes):
ary_len = 16
- np.random.seed(10)
-
int32_min = np.iinfo(np.int32).min
int32_max = np.iinfo(np.int32).max
@@ -974,7 +972,9 @@ def test_stride_preservation(ctx_factory):
context = ctx_factory()
queue = cl.CommandQueue(context)
- a = np.random.rand(3, 3)
+ rng = np.random.default_rng(seed=42)
+ a = rng.random(size=(3, 3))
+
at = a.T
print(at.flags.f_contiguous, at.flags.c_contiguous)
at_gpu = cl_array.to_device(queue, at)
@@ -989,10 +989,11 @@ def test_stride_preservation(ctx_factory):
def test_nan_arithmetic(ctx_factory):
context = ctx_factory()
queue = cl.CommandQueue(context)
+ rng = np.random.default_rng(seed=42)
def make_nan_contaminated_vector(size):
- shape = (size,)
- a = np.random.randn(*shape).astype(np.float32)
+ a = rng.standard_normal(size=(size,), dtype=np.float32)
+
from random import randrange
for _i in range(size // 10):
a[randrange(0, size)] = float("nan")
@@ -1468,9 +1469,10 @@ def test_newaxis(ctx_factory):
def test_squeeze(ctx_factory):
context = ctx_factory()
queue = cl.CommandQueue(context)
+ rng = np.random.default_rng(seed=42)
shape = (40, 2, 5, 100)
- a_cpu = np.random.random(size=shape)
+ a_cpu = rng.random(size=shape)
a_gpu = cl_array.to_device(queue, a_cpu)
# Slice with length 1 on dimensions 0 and 1
@@ -1532,11 +1534,12 @@ def test_fancy_indexing(ctx_factory):
pytest.xfail("numpypy: multi value setting is not supported")
context = ctx_factory()
queue = cl.CommandQueue(context)
+ rng = np.random.default_rng(seed=42)
n = 2 ** 20 + 2**18 + 22
numpy_dest = np.zeros(n, dtype=np.int32)
numpy_idx = np.arange(n, dtype=np.int32)
- np.random.shuffle(numpy_idx)
+ rng.shuffle(numpy_idx)
numpy_src = 20000+np.arange(n, dtype=np.int32)
cl_dest = cl_array.to_device(queue, numpy_dest)
@@ -1598,7 +1601,8 @@ def test_get_async(ctx_factory):
pytest.xfail("the async get test fails on POCL + Nvidia,"
"at least the K40, as of pocl 1.6, 2021-01-20")
- a = np.random.rand(10**6).astype(np.dtype("float32"))
+ rng = np.random.default_rng(seed=42)
+ a = rng.random(10**6, dtype=np.float32)
a_gpu = cl_array.to_device(queue, a)
b = a + a**5 + 1
b_gpu = a_gpu + a_gpu**5 + 1
@@ -1631,7 +1635,9 @@ def test_outoforderqueue_get(ctx_factory):
properties=cl.command_queue_properties.OUT_OF_ORDER_EXEC_MODE_ENABLE)
except Exception:
pytest.skip("out-of-order queue not available")
- a = np.random.rand(10**6).astype(np.dtype("float32"))
+
+ rng = np.random.default_rng(seed=42)
+ a = rng.random(10**6, dtype=np.float32)
a_gpu = cl_array.to_device(queue, a)
b_gpu = a_gpu + a_gpu**5 + 1
b1 = b_gpu.get() # testing that this waits for events
@@ -1650,7 +1656,9 @@ def test_outoforderqueue_copy(ctx_factory):
properties=cl.command_queue_properties.OUT_OF_ORDER_EXEC_MODE_ENABLE)
except Exception:
pytest.skip("out-of-order queue not available")
- a = np.random.rand(10**6).astype(np.dtype("float32"))
+
+ rng = np.random.default_rng(seed=42)
+ a = rng.random(10**6, dtype=np.float32)
a_gpu = cl_array.to_device(queue, a)
c_gpu = a_gpu**2 - 7
b_gpu = c_gpu.copy() # testing that this waits for and creates events
@@ -1673,8 +1681,11 @@ def test_outoforderqueue_indexing(ctx_factory):
properties=cl.command_queue_properties.OUT_OF_ORDER_EXEC_MODE_ENABLE)
except Exception:
pytest.skip("out-of-order queue not available")
- a = np.random.rand(10**6).astype(np.dtype("float32"))
- i = (8e5 + 1e5 * np.random.rand(10**5)).astype(np.dtype("int32"))
+
+ rng = np.random.default_rng(seed=42)
+ a = rng.random(10**6, dtype=np.float32)
+ i = (8e5 + 1e5 * rng.random(10**5)).astype(np.int32)
+
a_gpu = cl_array.to_device(queue, a)
i_gpu = cl_array.to_device(queue, i)
c_gpu = (a_gpu**2)[i_gpu - 10000]
@@ -1701,7 +1712,9 @@ def test_outoforderqueue_reductions(ctx_factory):
except Exception:
pytest.skip("out-of-order queue not available")
# 0/1 values to avoid accumulated rounding error
- a = (np.random.rand(10**6) > 0.5).astype(np.dtype("float32"))
+ rng = np.random.default_rng(seed=42)
+ a = (rng.random(10**6) > 0.5).astype(np.float32)
+
a[800000] = 10 # all<5 looks true until near the end
a_gpu = cl_array.to_device(queue, a)
b1 = cl_array.sum(a_gpu).get()
@@ -1792,8 +1805,7 @@ def test_stack(ctx_factory, input_dims, order):
shape = (2, 2, 2)[:input_dims]
axis = -1 if order == "F" else 0
- from numpy.random import default_rng
- rng = default_rng()
+ rng = np.random.default_rng(seed=42)
x_in = rng.random(size=shape)
y_in = rng.random(size=shape)
x_in = x_in if order == "C" else np.asfortranarray(x_in)
@@ -1827,9 +1839,10 @@ def test_assign_different_strides(ctx_factory):
# {{{ test_branch_operations_on_pure_scalars
def test_branch_operations_on_pure_scalars():
- x = np.random.rand()
- y = np.random.rand()
- cond = np.random.choice([False, True])
+ rng = np.random.default_rng(seed=42)
+ x = rng.random()
+ y = rng.random()
+ cond = rng.choice([False, True])
np.testing.assert_allclose(np.maximum(x, y),
cl_array.maximum(x, y))
@@ -1892,7 +1905,8 @@ def test_slice_copy(ctx_factory):
cl_ctx = ctx_factory()
queue = cl.CommandQueue(cl_ctx)
- x = cl.array.to_device(queue, np.random.rand(96, 27))
+ rng = np.random.default_rng(seed=42)
+ x = cl.array.to_device(queue, rng.random(size=(96, 27)))
y = x[::8, ::3]
with pytest.raises(RuntimeError):
y.copy()
@@ -1907,7 +1921,8 @@ def test_ravel(ctx_factory, order):
ctx = ctx_factory()
cq = cl.CommandQueue(ctx)
- x = np.random.randn(10, 4)
+ rng = np.random.default_rng(seed=42)
+ x = rng.standard_normal(size=(10, 4))
if order == "F":
x = np.asfortranarray(x)
@@ -1951,12 +1966,11 @@ def test_arithmetic_on_non_scalars(ctx_factory):
@pytest.mark.parametrize("which", ("add", "sub", "mul", "truediv"))
def test_arithmetic_with_device_scalars(ctx_factory, which):
import operator
- from numpy.random import default_rng
ctx = ctx_factory()
cq = cl.CommandQueue(ctx)
- rng = default_rng()
+ rng = np.random.default_rng(seed=42)
ndim = rng.integers(1, 5)
shape = tuple(rng.integers(2, 7) for i in range(ndim))
@@ -1982,7 +1996,7 @@ def test_if_positive_with_scalars(ctx_factory, then_type, else_type):
ctx = ctx_factory()
cq = cl.CommandQueue(ctx)
- rng = np.random.default_rng()
+ rng = np.random.default_rng(seed=42)
shape = (512,)
criterion_np = rng.random(shape)
diff --git a/test/test_clmath.py b/test/test_clmath.py
index 409875f8..e5032dc1 100644
--- a/test/test_clmath.py
+++ b/test/test_clmath.py
@@ -334,10 +334,10 @@ def test_complex_bessel(ctx_factory, ref_src):
v = 40
n = 10**5
- np.random.seed(13)
+ rng = np.random.default_rng(seed=13)
z = (
np.logspace(-5, 2, n)
- * np.exp(1j * 2 * np.pi * np.random.rand(n)))
+ * np.exp(1j * 2 * np.pi * rng.random(n)))
if ref_src == "pyfmmlib":
pyfmmlib = pytest.importorskip("pyfmmlib")
@@ -404,11 +404,11 @@ def test_hankel_01_complex(ctx_factory, ref_src):
from pytest import skip
skip("no double precision support--cannot test complex bessel function")
+ rng = np.random.default_rng(seed=11)
n = 10**6
- np.random.seed(11)
z = (
np.logspace(-5, 2, n)
- * np.exp(1j * 2 * np.pi * np.random.rand(n)))
+ * np.exp(1j * 2 * np.pi * rng.random(n)))
def get_err(check, ref):
return np.max(np.abs(check-ref)) / np.max(np.abs(ref))
@@ -454,7 +454,9 @@ def test_outoforderqueue_clmath(ctx_factory):
properties=cl.command_queue_properties.OUT_OF_ORDER_EXEC_MODE_ENABLE)
except Exception:
pytest.skip("out-of-order queue not available")
- a = np.random.rand(10**6).astype(np.dtype("float32"))
+
+ rng = np.random.default_rng(seed=42)
+ a = rng.random(10**6, dtype=np.float32)
a_gpu = cl_array.to_device(queue, a)
# testing that clmath functions wait for and create events
b_gpu = clmath.fabs(clmath.sin(a_gpu * 5))
diff --git a/test/test_enqueue_copy.py b/test/test_enqueue_copy.py
index 402bc8b9..9e2ca278 100644
--- a/test/test_enqueue_copy.py
+++ b/test/test_enqueue_copy.py
@@ -73,15 +73,8 @@ def test_enqueue_copy_rect_2d(ctx_factory, honor_skip=True):
buf_out_shp = 300, 400 # shape of 2nd device buffer
# Create host array of random values.
- h_ary_in = \
- np.array(
- np.random.randint(
- 0,
- 256,
- np.product(ary_in_shp)
- ),
- dtype=np.uint8
- ).reshape(ary_in_shp)
+ rng = np.random.default_rng(seed=42)
+ h_ary_in = rng.integers(0, 256, ary_in_shp, dtype=np.uint8)
# Create device buffers
d_in_buf = cl.Buffer(ctx, cl.mem_flags.READ_ONLY, size=np.product(buf_in_shp))
@@ -169,15 +162,8 @@ def test_enqueue_copy_rect_3d(ctx_factory, honor_skip=True):
buf_out_shp = 300, 400, 40 # shape of 2nd device buffer
# Create host array of random values.
- h_ary_in = \
- np.array(
- np.random.randint(
- 0,
- 256,
- np.product(ary_in_shp)
- ),
- dtype=np.uint8
- ).reshape(ary_in_shp)
+ rng = np.random.default_rng(seed=42)
+ h_ary_in = rng.integers(0, 256, ary_in_shp, dtype=np.uint8)
# Create device buffers
d_in_buf = cl.Buffer(ctx, cl.mem_flags.READ_ONLY, size=np.product(buf_in_shp))
diff --git a/test/test_wrapper.py b/test/test_wrapper.py
index e7d87b29..22f27f74 100644
--- a/test/test_wrapper.py
+++ b/test/test_wrapper.py
@@ -353,7 +353,8 @@ def test_that_python_args_fail(ctx_factory):
{ a[get_global_id(0)] *= (b+c); }
""").build()
- a = np.random.rand(50000)
+ rng = np.random.default_rng(seed=42)
+ a = rng.random(50000)
queue = cl.CommandQueue(context)
mf = cl.mem_flags
a_buf = cl.Buffer(context, mf.READ_WRITE | mf.COPY_HOST_PTR, hostbuf=a)
@@ -413,7 +414,9 @@ def test_image_2d(ctx_factory):
""").build()
num_channels = 1
- a = np.random.rand(1024, 512, num_channels).astype(np.float32)
+
+ rng = np.random.default_rng(seed=42)
+ a = rng.random((1024, 512, num_channels), dtype=np.float32)
if num_channels == 1:
a = a[:, :, 0]
@@ -489,7 +492,9 @@ def test_image_3d(ctx_factory):
num_channels = 2
shape = (3, 4, 2)
- a = np.random.random(shape + (num_channels,)).astype(np.float32)
+
+ rng = np.random.default_rng(seed=42)
+ a = rng.random(size=shape + (num_channels,), dtype=np.float32)
queue = cl.CommandQueue(context)
try:
@@ -532,7 +537,8 @@ def test_copy_buffer(ctx_factory):
queue = cl.CommandQueue(context)
mf = cl.mem_flags
- a = np.random.rand(50000).astype(np.float32)
+ rng = np.random.default_rng(seed=42)
+ a = rng.random(50000, dtype=np.float32)
b = np.empty_like(a)
buf1 = cl.Buffer(context, mf.READ_ONLY | mf.COPY_HOST_PTR, hostbuf=a)
@@ -811,7 +817,8 @@ def test_buffer_get_host_array(ctx_factory):
ctx = ctx_factory()
mf = cl.mem_flags
- host_buf = np.random.rand(25).astype(np.float32)
+ rng = np.random.default_rng(seed=42)
+ host_buf = rng.random(25, dtype=np.float32)
buf = cl.Buffer(ctx, mf.READ_WRITE | mf.USE_HOST_PTR, hostbuf=host_buf)
host_buf2 = buf.get_host_array(25, np.float32)
assert (host_buf == host_buf2).all()
@@ -827,7 +834,7 @@ def test_buffer_get_host_array(ctx_factory):
except cl.LogicError:
pass
- host_buf = np.random.rand(25).astype(np.float32)
+ host_buf = rng.random(25, dtype=np.float32)
buf = cl.Buffer(ctx, mf.READ_WRITE | mf.COPY_HOST_PTR, hostbuf=host_buf)
try:
host_buf2 = buf.get_host_array(25, np.float32)
@@ -867,8 +874,9 @@ def test_event_set_callback(ctx_factory):
if ctx._get_cl_version() < (1, 1):
pytest.skip("OpenCL 1.1 or newer required for set_callback")
- a_np = np.random.rand(50000).astype(np.float32)
- b_np = np.random.rand(50000).astype(np.float32)
+ rng = np.random.default_rng(seed=42)
+ a_np = rng.random(50000, dtype=np.float32)
+ b_np = rng.random(50000, dtype=np.float32)
got_called = []
@@ -926,7 +934,9 @@ def test_global_offset(ctx_factory):
""").build()
n = 50
- a = np.random.rand(n).astype(np.float32)
+
+ rng = np.random.default_rng(seed=42)
+ a = rng.random(n, dtype=np.float32)
queue = cl.CommandQueue(context)
mf = cl.mem_flags
@@ -954,7 +964,9 @@ def test_sub_buffers(ctx_factory):
queue = cl.CommandQueue(ctx)
n = 30000
- a = (np.random.rand(n) * 100).astype(np.uint8)
+
+ rng = np.random.default_rng(seed=42)
+ a = (rng.random(n) * 100).astype(np.uint8)
mf = cl.mem_flags
a_buf = cl.Buffer(ctx, mf.READ_WRITE | mf.COPY_HOST_PTR, hostbuf=a)
--
GitLab