Skip to content

[SYCL][ESIMD][E2E] Query max workgroups in group barrier test #18161

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 1 commit into from
May 1, 2025
Merged

[SYCL][ESIMD][E2E] Query max workgroups in group barrier test #18161

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
60 changes: 38 additions & 22 deletions sycl/test-e2e/ESIMD/group_barrier.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,40 +14,56 @@
#include "esimd_test_utils.hpp"
#include <sycl/ext/oneapi/experimental/root_group.hpp>
#include <sycl/group_barrier.hpp>
#include <sycl/kernel_bundle.hpp>

namespace syclex = sycl::ext::oneapi::experimental;

static constexpr int WorkGroupSize = 16;

static constexpr int VL = 16;

template <int Val> class MyKernel;

template <bool UseThisWorkItemAPI> bool test(sycl::queue &q) {
bool Pass = true;
const auto MaxWGs = 8;
size_t WorkItemCount = MaxWGs * WorkGroupSize * VL;
std::cout << "Test case UseThisWorkItemAPI="
<< std::to_string(UseThisWorkItemAPI) << std::endl;
const auto Props = sycl::ext::oneapi::experimental::properties{
sycl::ext::oneapi::experimental::use_root_sync};
sycl::buffer<int> DataBuf{sycl::range{WorkItemCount}};
const auto Range = sycl::nd_range<1>{MaxWGs * WorkGroupSize, WorkGroupSize};
auto Bundle =
sycl::get_kernel_bundle<sycl::bundle_state::executable>(q.get_context());
auto Kernel = Bundle.template get_kernel<MyKernel<UseThisWorkItemAPI>>();
sycl::range<1> LocalRange{WorkGroupSize};
auto MaxWGs = Kernel.template ext_oneapi_get_info<
syclex::info::kernel_queue_specific::max_num_work_groups>(q, LocalRange,
0);
auto GlobalRange = LocalRange;
GlobalRange[0] *= MaxWGs / VL;
size_t WorkItemCount = GlobalRange.size() * VL;
sycl::buffer<int> DataBuf{WorkItemCount};
const auto Range = sycl::nd_range<1>{GlobalRange, LocalRange};

q.submit([&](sycl::handler &h) {
sycl::accessor Data{DataBuf, h};
h.parallel_for(Range, Props, [=](sycl::nd_item<1> it) SYCL_ESIMD_KERNEL {
int ID = it.get_global_linear_id();
__ESIMD_NS::simd<int, VL> V(ID, 1);
// Write data to another kernel's data to verify the barrier works.
__ESIMD_NS::block_store(
Data, (WorkItemCount * sizeof(int)) - (ID * sizeof(int) * VL), V);
if constexpr (UseThisWorkItemAPI) {
auto Root =
sycl::ext::oneapi::experimental::this_work_item::get_root_group<
1>();
sycl::group_barrier(Root);
} else {
auto Root = it.ext_oneapi_get_root_group();
sycl::group_barrier(Root);
}
__ESIMD_NS::simd<int, VL> VOther(ID * VL, 1);
__ESIMD_NS::block_store(Data, ID * sizeof(int) * VL, VOther);
});
h.parallel_for<MyKernel<UseThisWorkItemAPI>>(
Range, Props, [=](sycl::nd_item<1> it) SYCL_ESIMD_KERNEL {
int ID = it.get_global_linear_id();
__ESIMD_NS::simd<int, VL> V(ID, 1);
// Write data to another kernel's data to verify the barrier works.
__ESIMD_NS::block_store(
Data, (WorkItemCount * sizeof(int)) - (ID * sizeof(int) * VL),
V);
if constexpr (UseThisWorkItemAPI) {
auto Root = sycl::ext::oneapi::experimental::this_work_item::
get_root_group<1>();
sycl::group_barrier(Root);
} else {
auto Root = it.ext_oneapi_get_root_group();
sycl::group_barrier(Root);
}
__ESIMD_NS::simd<int, VL> VOther(ID * VL, 1);
__ESIMD_NS::block_store(Data, ID * sizeof(int) * VL, VOther);
});
}).wait();
sycl::host_accessor Data{DataBuf};
int ErrCnt = 0;
Expand Down