Skip to content

[SYCL] extends previous Global/LocalSize 0 fix #23065

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

Open
cperkinsintel wants to merge 1 commit into
base: sycl
Choose a base branch
from
Open

[SYCL] extends previous Global/LocalSize 0 fix #23065

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
16 changes: 10 additions & 6 deletions sycl/source/detail/scheduler/commands.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2295,12 +2295,16 @@ std::string_view ExecCGCommand::getTypeString() const {
// for users who need more control.
static void adjustNDRangePerKernel(NDRDescT &NDR, ur_kernel_handle_t Kernel,
const device_impl &DeviceImpl) {
if (NDR.NumWorkGroups[0] == 0)
return; // Not parallel_for_work_group -- nothing to fill in.
// In pfwg mode NumWorkGroups is the only field the user sets; GlobalSize
// and LocalSize must both be zero (see NDRDescT contract in
// ndrange_desc.hpp).
assert(NDR.GlobalSize[0] == 0 && NDR.LocalSize[0] == 0);
if (NDR.GlobalSize[0] != 0)
return; // GlobalSize is set - no need to adjust
if (NDR.LocalSize[0] != 0)
return; // User set LocalSize but GlobalSize is zero (e.g. nd_range with
// zero global, non-zero local). Per SYCL 2020 the kernel is not
// executed; leave the range as-is.
Comment on lines +2298 to +2303
// Zero global and zero local: either parallel_for_work_group (NumWorkGroups
// is set) or plain parallel_for with an empty range. Fill in WGSize so that
// downstream layers (in particular DeviceASAN's preLaunchKernel) always see
// a non-zero local work size even when the launch is a no-op.
Comment on lines +2304 to +2307
// TODO might be good to cache this info together with the kernel info to
// avoid get_kernel_work_group_info on every kernel run
range<3> WGSize = get_kernel_device_specific_info<
Expand Down
Loading