Skip to content

Fixed regression in iteration space simplifier, fixed arguments of numpy to usm_ndarray copying function #1202

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
May 8, 2023
Merged

Fixed regression in iteration space simplifier, fixed arguments of numpy to usm_ndarray copying function #1202

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
eb97326
Simplify_iteration_space was doing it incorrectly since gh-1198
oleksandr-pavlyk May 8, 2023
bd60be8
corrected use of src_offset in copy_from_numpy_into_usm_ndarray
oleksandr-pavlyk May 8, 2023
8e43c77
Adding test for gh-1201
oleksandr-pavlyk May 8, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions dpctl/tensor/libtensor/source/copy_numpy_ndarray_into_usm_ndarray.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -211,8 +211,8 @@ void copy_numpy_ndarray_into_usm_ndarray(
}

// Minumum and maximum element offsets for source np.ndarray
py::ssize_t npy_src_min_nelem_offset(0);
py::ssize_t npy_src_max_nelem_offset(0);
py::ssize_t npy_src_min_nelem_offset(src_offset);
py::ssize_t npy_src_max_nelem_offset(src_offset);
for (int i = 0; i < nd; ++i) {
if (simplified_src_strides[i] < 0) {
npy_src_min_nelem_offset +=
Expand Down
111 changes: 57 additions & 54 deletions dpctl/tensor/libtensor/source/simplify_iteration_space.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -141,19 +141,22 @@ void simplify_iteration_space(int &nd,
assert(simplified_shape.size() == static_cast<size_t>(nd));

simplified_src_strides.reserve(nd);
simplified_src_strides.push_back(
(src_strides[0] >= 0) ? src_strides[0] : -src_strides[0]);
if ((src_strides[0] < 0) && (shape[0] > 1)) {
src_offset += (shape[0] - 1) * src_strides[0];
}
assert(simplified_src_strides.size() == static_cast<size_t>(nd));

simplified_dst_strides.reserve(nd);
simplified_dst_strides.push_back(
(dst_strides[0] >= 0) ? dst_strides[0] : -dst_strides[0]);
if ((dst_strides[0] < 0) && (shape[0] > 1)) {
dst_offset += (shape[0] - 1) * dst_strides[0];

if (src_strides[0] < 0 && dst_strides[0] < 0) {
simplified_src_strides.push_back(-src_strides[0]);
simplified_dst_strides.push_back(-dst_strides[0]);
if (shape[0] > 1) {
src_offset += (shape[0] - 1) * src_strides[0];
dst_offset += (shape[0] - 1) * dst_strides[0];
}
}
else {
simplified_src_strides.push_back(src_strides[0]);
simplified_dst_strides.push_back(dst_strides[0]);
}

assert(simplified_src_strides.size() == static_cast<size_t>(nd));
assert(simplified_dst_strides.size() == static_cast<size_t>(nd));
}
}
Expand Down Expand Up @@ -226,27 +229,28 @@ void simplify_iteration_space_3(
assert(simplified_shape.size() == static_cast<size_t>(nd));

simplified_src1_strides.reserve(nd);
simplified_src1_strides.push_back(
(src1_strides[0] >= 0) ? src1_strides[0] : -src1_strides[0]);
if ((src1_strides[0] < 0) && (shape[0] > 1)) {
src1_offset += src1_strides[0] * (shape[0] - 1);
}
assert(simplified_src1_strides.size() == static_cast<size_t>(nd));

simplified_src2_strides.reserve(nd);
simplified_src2_strides.push_back(
(src2_strides[0] >= 0) ? src2_strides[0] : -src2_strides[0]);
if ((src2_strides[0] < 0) && (shape[0] > 1)) {
src2_offset += src2_strides[0] * (shape[0] - 1);
}
assert(simplified_src2_strides.size() == static_cast<size_t>(nd));

simplified_dst_strides.reserve(nd);
simplified_dst_strides.push_back(
(dst_strides[0] >= 0) ? dst_strides[0] : -dst_strides[0]);
if ((dst_strides[0] < 0) && (shape[0] > 1)) {
dst_offset += dst_strides[0] * (shape[0] - 1);

if ((src1_strides[0] < 0) && (src2_strides[0] < 0) &&
(dst_strides[0] < 0)) {
simplified_src1_strides.push_back(-src1_strides[0]);
simplified_src2_strides.push_back(-src2_strides[0]);
simplified_dst_strides.push_back(-dst_strides[0]);
if (shape[0] > 1) {
src1_offset += src1_strides[0] * (shape[0] - 1);
src2_offset += src2_strides[0] * (shape[0] - 1);
dst_offset += dst_strides[0] * (shape[0] - 1);
}
}
else {
simplified_src1_strides.push_back(src1_strides[0]);
simplified_src2_strides.push_back(src2_strides[0]);
simplified_dst_strides.push_back(dst_strides[0]);
}

assert(simplified_src1_strides.size() == static_cast<size_t>(nd));
assert(simplified_src2_strides.size() == static_cast<size_t>(nd));
assert(simplified_dst_strides.size() == static_cast<size_t>(nd));
}
}
Expand Down Expand Up @@ -333,35 +337,34 @@ void simplify_iteration_space_4(
assert(simplified_shape.size() == static_cast<size_t>(nd));

simplified_src1_strides.reserve(nd);
simplified_src1_strides.push_back(
(src1_strides[0] >= 0) ? src1_strides[0] : -src1_strides[0]);
if ((src1_strides[0] < 0) && (shape[0] > 1)) {
src1_offset += src1_strides[0] * (shape[0] - 1);
}
assert(simplified_src1_strides.size() == static_cast<size_t>(nd));

simplified_src2_strides.reserve(nd);
simplified_src2_strides.push_back(
(src2_strides[0] >= 0) ? src2_strides[0] : -src2_strides[0]);
if ((src2_strides[0] < 0) && (shape[0] > 1)) {
src2_offset += src2_strides[0] * (shape[0] - 1);
}
assert(simplified_src2_strides.size() == static_cast<size_t>(nd));

simplified_src3_strides.reserve(nd);
simplified_src3_strides.push_back(
(src3_strides[0] >= 0) ? src3_strides[0] : -src3_strides[0]);
if ((src3_strides[0] < 0) && (shape[0] > 1)) {
src3_offset += src3_strides[0] * (shape[0] - 1);
}
assert(simplified_src3_strides.size() == static_cast<size_t>(nd));

simplified_dst_strides.reserve(nd);
simplified_dst_strides.push_back(
(dst_strides[0] >= 0) ? dst_strides[0] : -dst_strides[0]);
if ((dst_strides[0] < 0) && (shape[0] > 1)) {
dst_offset += dst_strides[0] * (shape[0] - 1);

if ((src1_strides[0] < 0) && (src2_strides[0] < 0) &&
(src3_strides[0] < 0) && (dst_strides[0] < 0))
{
simplified_src1_strides.push_back(-src1_strides[0]);
simplified_src2_strides.push_back(-src2_strides[0]);
simplified_src3_strides.push_back(-src3_strides[0]);
simplified_dst_strides.push_back(-dst_strides[0]);
if (shape[0] > 1) {
src1_offset += src1_strides[0] * (shape[0] - 1);
src2_offset += src2_strides[0] * (shape[0] - 1);
src3_offset += src3_strides[0] * (shape[0] - 1);
dst_offset += dst_strides[0] * (shape[0] - 1);
}
}
else {
simplified_src1_strides.push_back(src1_strides[0]);
simplified_src2_strides.push_back(src2_strides[0]);
simplified_src3_strides.push_back(src3_strides[0]);
simplified_dst_strides.push_back(dst_strides[0]);
}

assert(simplified_src1_strides.size() == static_cast<size_t>(nd));
assert(simplified_src2_strides.size() == static_cast<size_t>(nd));
assert(simplified_src3_strides.size() == static_cast<size_t>(nd));
assert(simplified_dst_strides.size() == static_cast<size_t>(nd));
}
}
Expand Down
13 changes: 13 additions & 0 deletions dpctl/tests/test_usm_ndarray_ctor.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2059,3 +2059,16 @@ def test_byte_bounds():
y = x[::-1, ::2]
lo, hi = y._byte_bounds
assert hi - lo == (n0 * n1 - 1) * x.itemsize


def test_gh_1201():
n = 100
a = np.flipud(np.arange(n, dtype="i4"))
try:
b = dpt.asarray(a)
except dpctl.SyclDeviceCreationError:
pytest.skip("No SYCL devices available")
assert (dpt.asnumpy(b) == a).all()
c = dpt.flip(dpt.empty(a.shape, dtype=a.dtype))
c[:] = a
assert (dpt.asnumpy(c) == a).all()