Align UR_API fields (8 byte) for optimization create/move/copy structs on x64 cpus #2747
Conversation
…s on x64 cpus Affected API structures: - ur_image_desc_t 80 -> 72 bytes - ur_exp_command_buffer_update_value_arg_desc_t 48 -> 40 bytes - ur_exp_command_buffer_update_memobj_arg_desc_t 40 -> 32 bytes - ur_exp_command_buffer_update_pointer_arg_desc_t 40 -> 32 bytes - ur_sampler_desc_t 32 -> 24 bytes - ur_program_properties_t 32 -> 24 bytes - ur_exp_sampler_addr_modes_t 32 -> 24 bytes - ur_platform_native_properties_t 24 -> 16 bytes - ur_device_native_properties_t 24 -> 16 bytes - ur_context_properties_t 24 -> 16 bytes - ur_context_native_properties_t 24 -> 16 bytes - ur_buffer_channel_properties_t 24 -> 16 bytes - ur_buffer_alloc_location_properties_t 24 -> 16 bytes - ur_mem_native_properties_t 24 -> 16 bytes - ur_sampler_native_properties_t 24 -> 16 bytes - ur_usm_host_desc_t 24 -> 16 bytes - ur_usm_device_desc_t 24 -> 16 bytes - ur_usm_alloc_location_desc_t 24 -> 16 bytes - ur_usm_pool_desc_t 24 -> 16 bytes - ur_physical_mem_properties_t 24 -> 16 bytes - ur_program_native_properties_t 24 -> 16 bytes - ur_kernel_arg_mem_obj_properties_t 24 -> 16 bytes - ur_kernel_native_properties_t 24 -> 16 bytes - ur_queue_properties_t 24 -> 16 bytes - ur_queue_index_properties_t 24 -> 16 bytes - ur_queue_native_properties_t 24 -> 16 bytes - ur_event_native_properties_t 24 -> 16 bytes - ur_exp_async_usm_alloc_properties_t 24 -> 16 bytes - ur_exp_file_descriptor_t 24 -> 16 bytes - ur_exp_sampler_cubemap_properties_t 24 -> 16 bytes - ur_exp_command_buffer_desc_t 24 -> 16 bytes - ur_exp_enqueue_ext_properties_t 24 -> 16 bytes - ur_exp_enqueue_native_command_properties_t 24 -> 16 bytes
github-actions
bot
added
loader
Unified Runtime -> intel/llvm Repo Move NoticeInformationThe source code of Unified Runtime has been moved to intel/llvm under the unified-runtime top-level directory, The code will be mirrored to oneapi-src/unified-runtime and the specification will continue to be hosted at oneapi-src.github.io/unified-runtime. The contribution guide will be updated with new instructions for contributing to Unified Runtime. PR MigrationAll open PRs including this one will be marked with the Should you wish to continue with your PR you will need to migrate it to intel/llvm. If your PR should remain open and not be closed automatically, you can remove the This is an automated comment. |
@RossBrunton, @aarongreig,
Would you like to organize migration to aligned structures your Unified Runtime API for modern x64 processors? This should be guaranteed to lead to more frequent structs entry into CPU cache, which can greatly affect performance if aligned structures are used frequently. I hope that you (as Intel employees) know advantage this optimization method at the architectural level codebase and its inconveniences as stylistic ABI breakdown.
Very briefly, your API is not badly broken, I have changed only first and second field in all aligned structures.
(pNext and sType)
More info about technique:
https://stackoverflow.com/a/20882083
https://zijishi.xyz/post/optimization-technique/learning-to-use-data-alignment/
https://en.wikipedia.org/wiki/Data_structure_alignment
Affected API structures: