Skip to content

[SYCL] Deprecate parallel_for and single_task overloads with properties #16726

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

Closed
HPS-1 wants to merge 1 commit into from
Closed

[SYCL] Deprecate parallel_for and single_task overloads with properties #16726

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
2 changes: 2 additions & 0 deletions devops/cts_exclude_filter_compfails
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,2 +1,4 @@
# Please use "#" to add comments here.
# Do not delete the file even if it's empty.
# See https://github.com/intel/llvm/pull/16615
handler
2 changes: 1 addition & 1 deletion devops/dependencies.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@
},
"igc": {
"github_tag": "v2.5.6",
"version": "2.5.6",
"version": "v2.5.6",
"url": "https://github.com/intel/intel-graphics-compiler/releases/tag/v2.5.6",
"root": "{DEPS_ROOT}/opencl/runtime/linux/oclgpu"
},
Expand Down
2 changes: 1 addition & 1 deletion llvm/docs/requirements-hashed.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -151,7 +151,7 @@ imagesize==1.4.1 \
--hash=sha256:0d8d18d08f840c19d0ee7ca1fd82490fdc3729b7ac93f49870406ddde8ef8d8b \
--hash=sha256:69150444affb9cb0d5cc5a92b3676f0b2fb7cd9ae39e947a5e11a36b4497cd4a
# via sphinx
jinja2==3.1.4 \
jinja2==3.1.5 \
--hash=sha256:4a3aee7acbbe7303aede8e9648d13b8bf88a429282aa6122a993f0ac800cb369 \
--hash=sha256:bc5dd2abb727a5319567b7a813e6a2e7318c39f4f487cfe6c89c6f9c7d25197d
# via
Expand Down
15 changes: 8 additions & 7 deletions sycl/cmake/modules/UnifiedRuntimeTag.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,8 @@
# commit 8b7a99578966eb691a961d9620ea38d235196b2f
# Merge: ed095412 7b0e3b19
# Author: Martin Grant <[email protected]>
# Date: Mon Jan 20 09:27:22 2025 +0000
# Merge pull request #2582 from przemektmalon/przemek/intel-host-usm-support
# Enable Host USM backed images on Level Zero
set(UNIFIED_RUNTIME_TAG 8b7a99578966eb691a961d9620ea38d235196b2f)
# commit b074893e854d28141cd67bc5935ed87e47eb3bb6
# Merge: 71a5eab0 128ea023
# Author: Ross Brunton <[email protected]>
# Date: Tue Jan 21 11:21:50 2025 +0000
# Merge pull request #2539 from RossBrunton/ross/specconst
#
# Added `DEVICE_INFO_PROGRAM_SET_SPECIALIZATION_CONSTANTS`
set(UNIFIED_RUNTIME_TAG b074893e854d28141cd67bc5935ed87e47eb3bb6)
10 changes: 10 additions & 0 deletions sycl/doc/extensions/experimental/sycl_ext_oneapi_bindless_images.asciidoc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -428,6 +428,15 @@ The second way to allocate image memory is to use USM allocations. SYCL already
provides a number of USM allocation functions. This proposal would add another,
pitched memory allocation, through `pitched_alloc_device`.

Bindless images can be backed by device, host, or shared USM memory allocations.

[NOTE]
====
Image memory backed by USM device and host allocations is generally supported,
whereas shared USM allocations depend on the SYCL backend as well as the device
capabilities.
====

```cpp
namespace sycl::ext::oneapi::experimental {

Expand Down Expand Up @@ -2328,4 +2337,5 @@ These features still need to be handled:
|6.4|2024-10-15| - Fix bindless spec examples and include examples in bindless
spec using asciidoc include.
|6.5|2024-10-22| - Allow 3-channel image formats on some backends.
|6.6|2025-01-20| - Clarify support for the specific types of USM allocations.
|======================
115 changes: 76 additions & 39 deletions sycl/include/sycl/handler.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2292,61 +2292,80 @@ class __SYCL_EXPORT handler {

template <typename KernelName = detail::auto_name, typename KernelType,
typename PropertiesT>
std::enable_if_t<
ext::oneapi::experimental::is_property_list<PropertiesT>::value>
single_task(PropertiesT Props, _KERNELFUNCPARAM(KernelFunc)) {
__SYCL_DEPRECATED(
"Use sycl::ext::oneapi::experimental::single_task (provided in the "
"sycl_ext_oneapi_enqueue_functions extension) instead.")
std::enable_if_t<ext::oneapi::experimental::is_property_list<
PropertiesT>::value> single_task(PropertiesT Props,
_KERNELFUNCPARAM(KernelFunc)) {
single_task_lambda_impl<KernelName, KernelType, PropertiesT>(Props,
KernelFunc);
}

template <typename KernelName = detail::auto_name, typename KernelType,
typename PropertiesT>
std::enable_if_t<
ext::oneapi::experimental::is_property_list<PropertiesT>::value>
parallel_for(range<1> NumWorkItems, PropertiesT Props,
_KERNELFUNCPARAM(KernelFunc)) {
__SYCL_DEPRECATED(
"Use sycl::ext::oneapi::experimental::single_task (provided in the "
"sycl_ext_oneapi_enqueue_functions extension) instead.")
std::enable_if_t<ext::oneapi::experimental::is_property_list<
PropertiesT>::value> parallel_for(range<1> NumWorkItems,
PropertiesT Props,
_KERNELFUNCPARAM(KernelFunc)) {
parallel_for_lambda_impl<KernelName, KernelType, 1, PropertiesT>(
NumWorkItems, Props, std::move(KernelFunc));
}

template <typename KernelName = detail::auto_name, typename KernelType,
typename PropertiesT>
std::enable_if_t<
ext::oneapi::experimental::is_property_list<PropertiesT>::value>
parallel_for(range<2> NumWorkItems, PropertiesT Props,
_KERNELFUNCPARAM(KernelFunc)) {
__SYCL_DEPRECATED(
"Use sycl::ext::oneapi::experimental::single_task (provided in the "
"sycl_ext_oneapi_enqueue_functions extension) instead.")
std::enable_if_t<ext::oneapi::experimental::is_property_list<
PropertiesT>::value> parallel_for(range<2> NumWorkItems,
PropertiesT Props,
_KERNELFUNCPARAM(KernelFunc)) {
parallel_for_lambda_impl<KernelName, KernelType, 2, PropertiesT>(
NumWorkItems, Props, std::move(KernelFunc));
}

template <typename KernelName = detail::auto_name, typename KernelType,
typename PropertiesT>
std::enable_if_t<
ext::oneapi::experimental::is_property_list<PropertiesT>::value>
parallel_for(range<3> NumWorkItems, PropertiesT Props,
_KERNELFUNCPARAM(KernelFunc)) {
__SYCL_DEPRECATED(
"Use sycl::ext::oneapi::experimental::single_task (provided in the "
"sycl_ext_oneapi_enqueue_functions extension) instead.")
std::enable_if_t<ext::oneapi::experimental::is_property_list<
PropertiesT>::value> parallel_for(range<3> NumWorkItems,
PropertiesT Props,
_KERNELFUNCPARAM(KernelFunc)) {
parallel_for_lambda_impl<KernelName, KernelType, 3, PropertiesT>(
NumWorkItems, Props, std::move(KernelFunc));
}

template <typename KernelName = detail::auto_name, typename KernelType,
typename PropertiesT, int Dims>
std::enable_if_t<
ext::oneapi::experimental::is_property_list<PropertiesT>::value>
parallel_for(nd_range<Dims> Range, PropertiesT Properties,
_KERNELFUNCPARAM(KernelFunc)) {
__SYCL_DEPRECATED(
"Use sycl::ext::oneapi::experimental::single_task (provided in the "
"sycl_ext_oneapi_enqueue_functions extension) instead.")
std::enable_if_t<ext::oneapi::experimental::is_property_list<
PropertiesT>::value> parallel_for(nd_range<Dims> Range,
PropertiesT Properties,
_KERNELFUNCPARAM(KernelFunc)) {
parallel_for_impl<KernelName>(Range, Properties, std::move(KernelFunc));
}

/// Reductions @{

template <typename KernelName = detail::auto_name, typename PropertiesT,
typename... RestT>
std::enable_if_t<
(sizeof...(RestT) > 1) &&
detail::AreAllButLastReductions<RestT...>::value &&
ext::oneapi::experimental::is_property_list<PropertiesT>::value>
parallel_for(range<1> Range, PropertiesT Properties, RestT &&...Rest) {
__SYCL_DEPRECATED(
"Use sycl::ext::oneapi::experimental::parallel_for (provided in the "
"sycl_ext_oneapi_enqueue_functions extension) instead.")
std::enable_if_t<(sizeof...(RestT) > 1) &&
detail::AreAllButLastReductions<RestT...>::value &&
ext::oneapi::experimental::is_property_list<
PropertiesT>::value> parallel_for(range<1> Range,
PropertiesT Properties,
RestT &&...Rest) {
#ifndef __SYCL_DEVICE_ONLY__
throwIfGraphAssociated<ext::oneapi::experimental::detail::
UnsupportedGraphFeatures::sycl_reductions>();
Expand All @@ -2357,11 +2376,15 @@ class __SYCL_EXPORT handler {

template <typename KernelName = detail::auto_name, typename PropertiesT,
typename... RestT>
std::enable_if_t<
(sizeof...(RestT) > 1) &&
detail::AreAllButLastReductions<RestT...>::value &&
ext::oneapi::experimental::is_property_list<PropertiesT>::value>
parallel_for(range<2> Range, PropertiesT Properties, RestT &&...Rest) {
__SYCL_DEPRECATED(
"Use sycl::ext::oneapi::experimental::parallel_for (provided in the "
"sycl_ext_oneapi_enqueue_functions extension) instead.")
std::enable_if_t<(sizeof...(RestT) > 1) &&
detail::AreAllButLastReductions<RestT...>::value &&
ext::oneapi::experimental::is_property_list<
PropertiesT>::value> parallel_for(range<2> Range,
PropertiesT Properties,
RestT &&...Rest) {
#ifndef __SYCL_DEVICE_ONLY__
throwIfGraphAssociated<ext::oneapi::experimental::detail::
UnsupportedGraphFeatures::sycl_reductions>();
Expand All @@ -2372,11 +2395,15 @@ class __SYCL_EXPORT handler {

template <typename KernelName = detail::auto_name, typename PropertiesT,
typename... RestT>
std::enable_if_t<
(sizeof...(RestT) > 1) &&
detail::AreAllButLastReductions<RestT...>::value &&
ext::oneapi::experimental::is_property_list<PropertiesT>::value>
parallel_for(range<3> Range, PropertiesT Properties, RestT &&...Rest) {
__SYCL_DEPRECATED(
"Use sycl::ext::oneapi::experimental::parallel_for (provided in the "
"sycl_ext_oneapi_enqueue_functions extension) instead.")
std::enable_if_t<(sizeof...(RestT) > 1) &&
detail::AreAllButLastReductions<RestT...>::value &&
ext::oneapi::experimental::is_property_list<
PropertiesT>::value> parallel_for(range<3> Range,
PropertiesT Properties,
RestT &&...Rest) {
#ifndef __SYCL_DEVICE_ONLY__
throwIfGraphAssociated<ext::oneapi::experimental::detail::
UnsupportedGraphFeatures::sycl_reductions>();
Expand Down Expand Up @@ -2411,11 +2438,15 @@ class __SYCL_EXPORT handler {

template <typename KernelName = detail::auto_name, int Dims,
typename PropertiesT, typename... RestT>
std::enable_if_t<
(sizeof...(RestT) > 1) &&
detail::AreAllButLastReductions<RestT...>::value &&
ext::oneapi::experimental::is_property_list<PropertiesT>::value>
parallel_for(nd_range<Dims> Range, PropertiesT Properties, RestT &&...Rest) {
__SYCL_DEPRECATED(
"Use sycl::ext::oneapi::experimental::parallel_for (provided in the "
"sycl_ext_oneapi_enqueue_functions extension) instead.")
std::enable_if_t<(sizeof...(RestT) > 1) &&
detail::AreAllButLastReductions<RestT...>::value &&
ext::oneapi::experimental::is_property_list<
PropertiesT>::value> parallel_for(nd_range<Dims> Range,
PropertiesT Properties,
RestT &&...Rest) {
#ifndef __SYCL_DEVICE_ONLY__
throwIfGraphAssociated<ext::oneapi::experimental::detail::
UnsupportedGraphFeatures::sycl_reductions>();
Expand All @@ -2437,6 +2468,9 @@ class __SYCL_EXPORT handler {

template <typename KernelName = detail::auto_name, typename KernelType,
int Dims, typename PropertiesT>
__SYCL_DEPRECATED(
"Use sycl::ext::oneapi::experimental::parallel_for (provided in the "
"sycl_ext_oneapi_enqueue_functions extension) instead.")
void parallel_for_work_group(range<Dims> NumWorkGroups, PropertiesT Props,
_KERNELFUNCPARAM(KernelFunc)) {
parallel_for_work_group_lambda_impl<KernelName, KernelType, Dims,
Expand All @@ -2446,6 +2480,9 @@ class __SYCL_EXPORT handler {

template <typename KernelName = detail::auto_name, typename KernelType,
int Dims, typename PropertiesT>
__SYCL_DEPRECATED(
"Use sycl::ext::oneapi::experimental::parallel_for (provided in the "
"sycl_ext_oneapi_enqueue_functions extension) instead.")
void parallel_for_work_group(range<Dims> NumWorkGroups,
range<Dims> WorkGroupSize, PropertiesT Props,
_KERNELFUNCPARAM(KernelFunc)) {
Expand Down
Loading
Loading