Skip to content
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.

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

[GPU] Num threads per eu update #7823

Merged
Show file tree
Hide file tree
Changes from 1 commit
Commits
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
Original file line number Diff line number Diff line change
Expand Up @@ -30,8 +30,6 @@ struct gfx_version {
struct device_info {
uint32_t execution_units_count; ///< Number of available execution units.
uint32_t gpu_frequency; ///< Clock frequency in MHz.
uint32_t max_threads_per_execution_unit; ///< Number of available HW threads on EU.
uint32_t max_threads_per_device; ///< Maximum number of HW threads on device.

uint64_t max_work_group_size; ///< Maximum number of work-items in a work-group executing a kernel using the data parallel execution model.
uint64_t max_local_mem_size; ///< Maximum size of local memory arena in bytes.
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -223,12 +223,6 @@ device_info init_device_info(const cl::Device& device) {
info.supports_subgroups_short = extensions.find("cl_intel_subgroups_short") != std::string::npos;
info.supports_subgroups_char = extensions.find("cl_intel_subgroups_char") != std::string::npos;

info.supports_imad = get_imad_support(device);
info.supports_immad = false;
lznamens marked this conversation as resolved.
Show resolved Hide resolved

info.max_threads_per_execution_unit = 7;
info.max_threads_per_device = static_cast<uint32_t>(info.execution_units_count * info.max_threads_per_execution_unit);

info.supports_usm = extensions.find("cl_intel_unified_shared_memory") != std::string::npos;

info.supports_local_block_io = extensions.find("cl_intel_subgroup_local_block_io") != std::string::npos &&
Expand Down Expand Up @@ -256,6 +250,9 @@ device_info init_device_info(const cl::Device& device) {
info.num_sub_slices_per_slice = 0;
info.num_eus_per_sub_slice = 0;
info.num_threads_per_eu = 0;

info.supports_imad = get_imad_support(device);
info.supports_immad = false;
}

return info;
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -832,8 +832,8 @@ void set_params(const program_node& node, kernel_selector::params& params) {
params.engineInfo.maxImage2dWidth = device_info.max_image2d_width;
params.engineInfo.maxImage2dHeight = device_info.max_image2d_height;
params.engineInfo.computeUnitsCount = device_info.execution_units_count;
params.engineInfo.maxThreadsPerExecutionUnit = device_info.max_threads_per_execution_unit;
params.engineInfo.maxThreadsPerDevice = device_info.max_threads_per_device;
params.engineInfo.maxThreadsPerExecutionUnit = device_info.num_threads_per_eu > 0 ? device_info.num_threads_per_eu : 7;
params.engineInfo.maxThreadsPerDevice = params.engineInfo.maxThreadsPerDevice * device_info.execution_units_count;
lznamens marked this conversation as resolved.
Show resolved Hide resolved
params.engineInfo.deviceCache = program.get_tuning_cache();
params.engineInfo.driverVersion = device_info.driver_version;

Expand Down