Skip to content

Use group_load_store compiler extension #2123

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 8 commits into from
Nov 13, 2024
Merged

Use group_load_store compiler extension #2123

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
db1e4f0
Use group_load_store compiler extension
antonwolfy Oct 23, 2024
82c269b
Merge branch 'master' into resolve-cmpl-warning
antonwolfy Oct 23, 2024
61af54d
Merge branch 'master' into resolve-cmpl-warning
antonwolfy Nov 7, 2024
cf44a17
Merge branch 'master' into resolve-cmpl-warning
antonwolfy Nov 8, 2024
2995796
Merge branch 'master' into resolve-cmpl-warning
antonwolfy Nov 9, 2024
f43ede0
Merge branch 'master' into resolve-cmpl-warning
antonwolfy Nov 13, 2024
dc956e4
Tune description of dpnp.bincount
antonwolfy Nov 13, 2024
7acf58c
Merge branch 'master' into resolve-cmpl-warning
antonwolfy Nov 13, 2024
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
33 changes: 22 additions & 11 deletions dpnp/backend/kernels/dpnp_krnl_elemwise.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,6 +41,9 @@
using dpctl::tensor::kernels::alignment_utils::is_aligned;
using dpctl::tensor::kernels::alignment_utils::required_alignment;

using sycl::ext::oneapi::experimental::group_load;
using sycl::ext::oneapi::experimental::group_store;

template <typename T>
constexpr T dispatch_erf_op(T elem)
{
Expand Down Expand Up @@ -523,41 +526,49 @@ static void func_map_init_elemwise_1arg_1type(func_map_t &fmap)
_DataType_input2, \
_DataType_output>) \
{ \
sycl::vec<_DataType_input1, vec_sz> x1 = \
sg.load<vec_sz>(input1_multi_ptr); \
sycl::vec<_DataType_input2, vec_sz> x2 = \
sg.load<vec_sz>(input2_multi_ptr); \
sycl::vec<_DataType_input1, vec_sz> x1{}; \
sycl::vec<_DataType_input2, vec_sz> x2{}; \
\
group_load(sg, input1_multi_ptr, x1); \
group_load(sg, input2_multi_ptr, x2); \
\
res_vec = __vec_operation__; \
} \
else /* input types don't match result type, so \
explicit casting is required */ \
{ \
sycl::vec<_DataType_input1, vec_sz> tmp_x1{}; \
sycl::vec<_DataType_input2, vec_sz> tmp_x2{}; \
\
group_load(sg, input1_multi_ptr, tmp_x1); \
group_load(sg, input2_multi_ptr, tmp_x2); \
\
sycl::vec<_DataType_output, vec_sz> x1 = \
dpnp_vec_cast<_DataType_output, \
_DataType_input1, vec_sz>( \
sg.load<vec_sz>(input1_multi_ptr)); \
tmp_x1); \
sycl::vec<_DataType_output, vec_sz> x2 = \
dpnp_vec_cast<_DataType_output, \
_DataType_input2, vec_sz>( \
sg.load<vec_sz>(input2_multi_ptr)); \
tmp_x2); \
\
res_vec = __vec_operation__; \
} \
} \
else { \
sycl::vec<_DataType_input1, vec_sz> x1 = \
sg.load<vec_sz>(input1_multi_ptr); \
sycl::vec<_DataType_input2, vec_sz> x2 = \
sg.load<vec_sz>(input2_multi_ptr); \
sycl::vec<_DataType_input1, vec_sz> x1{}; \
sycl::vec<_DataType_input2, vec_sz> x2{}; \
\
group_load(sg, input1_multi_ptr, x1); \
group_load(sg, input2_multi_ptr, x2); \
\
for (size_t k = 0; k < vec_sz; ++k) { \
const _DataType_output input1_elem = x1[k]; \
const _DataType_output input2_elem = x2[k]; \
res_vec[k] = __operation__; \
} \
} \
sg.store<vec_sz>(result_multi_ptr, res_vec); \
group_store(sg, res_vec, result_multi_ptr); \
} \
else { \
for (size_t k = start + sg.get_local_id()[0]; \
Expand Down
4 changes: 2 additions & 2 deletions dpnp/dpnp_iface_histograms.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -325,7 +325,7 @@ def bincount(x, weights=None, minlength=None):
-------
out : dpnp.ndarray of ints
The result of binning the input array.
The length of `out` is equal to ``np.amax(x) + 1``.
The length of `out` is equal to ``dpnp.max(x) + 1``.

See Also
--------
Expand Down Expand Up @@ -353,7 +353,7 @@ def bincount(x, weights=None, minlength=None):
...
TypeError: x must be an integer array

A possible use of ``bincount`` is to perform sums over
A possible use of :obj:`dpnp.bincount` is to perform sums over
variable-size chunks of an array, using the `weights` keyword.

>>> w = np.array([0.3, 0.5, 0.2, 0.7, 1., -0.6], dtype=np.float32) # weights
Expand Down
Loading