Records
Overview
- struct accessor::ctor_internal_tag
- struct accessor::ctor_internal_tag
- class accessor<DataT, Dims, Mode, celerity::target::device> - Celerity wrapper around SYCL accessors.
- class accessor<DataT, Dims, Mode, celerity::target::host_task>
- struct all
- struct allow_by_ref_t
- class buffer
- struct buffer::tracker - A `tacker` instance is shared by all shallow copies of this `bu...
- class buffer_allocation_window - In addition to the usual per-item access through the subscript ...
- class buffer_snapshot - Owned representation of buffer contents as captured by celerity...
- struct chunk
- class collective_group - Each collective host task is executed within a collective group...
- class collective_partition - A one-dimensional partition, additionally carrying the MPI comm...
- class collective_tag - Tag type marking a `handler::host_task` as a collective task. D...
- class collective_tag_factory - Tag type construction helper. Do not construct this type direct...
- class distr_queue
- struct distr_queue::ctor_internal_tag
- struct distr_queue::tracker - A `tacker` instance is shared by all copies of this `distr_queu...
- class even_split - For a 1D kernel, splits an nD-buffer evenly along its slowest d...
- struct fixed
- class group
- class handler
- class host_object - A `host_object` wraps state that exists separately on each work...
- struct host_object::instance - Host objects that own their instance (i.e. not host_object <T &...
- struct host_object::tracker - A `tacker` instance is shared by all copies of any `host_object...
- struct host_object::tracker - A `tacker` instance is shared by all copies of any `host_object...
- struct host_object::tracker - A `tacker` instance is shared by all copies of any `host_object...
- class host_object<T>
- class host_object<void>
- class id
- class item
- class local_accessor
- class nd_item
- class nd_range
- struct neighborhood
- class on_master_node_tag - Tag type marking a `handler::host_task` as a master-node task. ...
- struct one_to_one
- class oversubscribe - Suggests that more than chunks of the execution space assigned ...
- class partition - Represents the sub-range of the iteration space handled by each...
- class range
- class side_effect - Provides access to a `host_object` through capture in a `host_t...
- struct side_effect::ctor_internal_tag
- class side_effect<void, Order>
- struct slice
- class split_1d - Suggests that the task should be split into 1D chunks. This is ...
- class split_2d - Suggests that the task should be split into 2D chunks.
- struct subrange
- class abstract_command
- class abstract_scheduler
- struct abstract_scheduler::event_buffer_created
- struct abstract_scheduler::event_buffer_debug_name_changed
- struct abstract_scheduler::event_buffer_destroyed
- struct abstract_scheduler::event_epoch_reached
- struct abstract_scheduler::event_host_object_created
- struct abstract_scheduler::event_host_object_destroyed
- struct abstract_scheduler::event_task_available
- struct abstract_scheduler::event_test_inspect
- struct abstract_scheduler::policy_set
- class access_front - Maintains a set of concurrent instructions that are accessing a...
- struct access_record
- struct access_tag
- class accessor_base
- class alloc_instruction - Allocates a contiguous range of memory, either for use as a bac...
- struct alloc_instruction_record - IDAG record type for `alloc_instruction`.
- class allocation_id - Uniquely identifies an allocation across all memories on the lo...
- struct allocation_with_offset - Instruction-graph equivalent of a USM pointer that permits poin...
- struct as_decimal_size - Wrap a byte count in this to auto-format it as KB / MB / etc.
- struct as_decimal_throughput - Wrap a byte-per-second ratio in this to auto-format it as KB/s,...
- struct as_sub_second - Wrap a `std::chrono::duration` in this to auto-format it as sec...
- struct assert_host_object_ctor_param_is_rvalue
- struct assigned_state - Instruction is assigned and waiting for a `complete_assigned()`...
- class async_event - Type-erased event signalling completion of events at the execut...
- class async_event_impl - Abstract base class for `async_event` implementations.
- struct async_instruction_state
- struct auto_select_devices
- class await_push_command
- class await_receive_instruction - Waits on the receive arbiter to complete part of the receive. T...
- struct await_receive_instruction_record - IDAG record type for a `await_receive_instruction`.
- class backend - The backend is responsible for asynchronously allocating device...
- struct batch - We submit the set of instructions and pilots generated within a...
- struct boundary_check_info
- struct boundary_check_info::accessor_info
- class box - An arbitrary-dimensional box described by its minimum and maxim...
- struct box_coordinate_order - Comparison operator (similar to std::less) that orders boxes by...
- struct buffer_access_allocation - Description of an accessor or a reduction write in terms of the...
- struct buffer_access_allocation_record - IDAG combined record for a device-kernel / host-task buffer acc...
- class buffer_access_map
- struct buffer_allocation_record
- struct buffer_allocation_state - Per-allocation state for a single buffer. This is where we trac...
- class buffer_fence_promise
- struct buffer_id
- struct buffer_memory_record - IDAG debug info for device-kernel / host-task access to a singl...
- struct buffer_memory_state - Per-memory state for a single buffer. Dependencies and last wri...
- struct buffer_reduction_allocation_record - IDAG combined record for a device-kernel access to a reduction-...
- struct buffer_reduction_record - IDAG debug info for a device-kernel access to a reduction outpu...
- struct buffer_state - State for a single buffer.
- struct buffer_state::gather_receive - Tracks a pending reduction await-push, which will be compiled i...
- struct buffer_state::region_receive - Tracks a pending non-reduction await-push that will be compiled...
- class cgf_diagnostics
- class clone_collective_group_instruction - Creates a new (MPI) collective group by cloning an existing one...
- struct clone_collective_group_instruction_record - IDAG record type for `clone_collective_group_instruction`.
- class closure_hydrator - The closure hydrator is used to inject information into objects...
- struct closure_hydrator::accessor_info
- struct collective_clone_from_tag - Constructor tag for mpi_communicator
- struct collective_group_id
- struct collective_group_state
- class command_graph
- class command_graph_generator
- struct command_graph_generator::buffer_state
- struct command_graph_generator::host_object_state
- struct command_graph_generator::policy_set
- struct command_hash_by_id - Hash function for `unordered_sets/maps` of `command *` that is ...
- struct command_id
- struct command_record
- class command_recorder
- class communicator - Interface for peer-to-peer and collective communication across ...
- struct communicator::stride - Addresses a 1/2/3-dimensional subrange of a type-erased (buffer...
- class complete_event - `async_event` implementation that is immediately complete. Used...
- struct conditional_eagerly_assignable_state - Instruction has been inserted into `assignment_queue`, but has ...
- class config
- class coordinate
- struct coordinate_storage
- struct coordinate_storage<Interface, 0>
- class copy_instruction - Copies one or more subranges of elements from one allocation to...
- struct copy_instruction_record - IDAG record type for `copy_instruction`.
- class dense_map - Like a simple std::unordered_map, but implemented by indexing i...
- struct dependency_record
- class destroy_host_object_instruction - Host object instances are owned by the instruction executor, so...
- struct destroy_host_object_instruction_record - IDAG record type for a `destroy_host_object_instruction`.
- struct device_id
- struct device_info - Information about a single device in the local system.
- class device_kernel_instruction - Launches a SYCL device kernel on a single device. Bound accesso...
- struct device_kernel_instruction_record - IDAG record type for a `device_kernel_instruction`.
- class double_buffered_queue - (Thread-safe) multi-producer single-consumer queue that uses do...
- struct double_buffered_queue::read_end - Aligned group 2: Reader thread can continue to concurrently acc...
- struct double_buffered_queue::write_end - Aligned group 1: The write-queue and its associated synchroniza...
- class dry_run_executor - Executor implementation selected when Celerity performs a dry r...
- struct engine_impl
- class epoch_command
- class epoch_instruction - Instruction-graph equivalent of an epoch task or command.
- struct epoch_instruction_record - IDAG record type for a `epoch_instruction`.
- class epoch_monitor
- class execution_command
- class executor - An executor processes receives and processes an instruction str...
- class executor::delegate - Implement this as the owner of an executor to receive callbacks...
- struct executor_impl
- class fence_command
- class fence_instruction - Fulfills a fence promise. Issued directly after a copy_instruct...
- struct fence_instruction_record - IDAG record type for a `fence_instruction`.
- struct fence_instruction_record::buffer_variant
- struct fence_instruction_record::host_object_variant
- class fence_promise
- class fill_identity_instruction - Fills an allocation with the identity value of a reduction. Use...
- struct fill_identity_instruction_record - IDAG record type for a `fill_identity_instruction`.
- class filter_iterator
- class free_instruction - Returns an allocation made with alloc_instruction to the system.
- struct free_instruction_record - IDAG record type for `free_instruction`.
- class gather_receive_event - Event for `receive_arbiter::gather_receive`, which waits for in...
- class gather_receive_instruction - A special type of receive instruction used for global reduction...
- struct gather_receive_instruction_record - IDAG record type for a `gather_receive_instruction`.
- struct gather_request - State for a single incomplete `gather_receive` operation.
- struct gather_transfer - A transfer originating through `gather_receive`. It is fully de...
- class generator_impl
- struct has_member_find
- struct has_member_find<Container, Key, void>
- class hint_base
- class horizon_command
- class horizon_instruction - Instruction-graph equivalent of a horizon task or command.
- struct horizon_instruction_record - IDAG record type for a `horizon_instruction`.
- struct host_config
- class host_object_fence_promise
- struct host_object_id
- struct host_object_instance - Host objects that own their instance (i.e. not host_object <T &...
- struct host_object_state
- struct host_object_tracker - A `tacker` instance is shared by all copies of any `host_object...
- struct host_object_transfer
- class host_task_instruction - Launches a host task in a thread pool. Bound accessors are hydr...
- struct host_task_instruction_record - IDAG record type for a `host_task_instruction`.
- struct hydration_id
- struct impl::device_state
- struct impl::host_state
- struct inbound_pilot - A pilot message as packaged on the receiver side.
- struct incoming_gather_chunk - A single chunk in a `gather_request` that is currently being re...
- struct incoming_region_fragment - A single box received by the communicator, as described earlier...
- struct incomplete_instruction_state - State maintained for every instruction between its submission a...
- class inner_node
- class instruction - A node in the `instruction_graph`. This is not implemented as a...
- struct instruction_dependency_record
- struct instruction_garbage - The executor will maintain runtime state about entities that th...
- class instruction_graph - The instruction graph (IDAG) provides a static, parallel schedu...
- struct instruction_graph::instruction_epoch
- class instruction_graph_generator - Tracks the node-local state of buffers and host objects, receiv...
- class instruction_graph_generator::delegate - Implement this as the owner of instruction_graph_generator to r...
- struct instruction_graph_generator::policy_set
- struct instruction_id
- struct instruction_id_less - Orders instruction pointers by instruction id.
- struct instruction_pilot_batch
- struct instruction_priority_less - Comparison operator to make priority_queue <instruction *> retu...
- struct instruction_record - IDAG base record type for `detail::instruction`.
- class instruction_recorder - Records instructions and outbound pilots on instruction-graph g...
- class intrusive_graph_node
- struct intrusive_graph_node::dependency
- class iterable_range
- struct kernel_flavor_traits<celerity::detail::nd_range_kernel_flavor, Dims>
- struct kernel_flavor_traits<celerity::detail::simple_kernel_flavor, Dims>
- struct lane_state - State maintained per host thread queue or device in-order queue.
- class live_executor - Executor implementation for a normal (non-dry) run of a Celerit...
- struct live_executor::policy_set
- struct local_reduction - Transient state for a node-local eager reduction that is emitte...
- struct localized_chunk - A chunk of a task's execution space that will be assigned to a ...
- struct make_from_t
- struct memory_id
- struct memory_info - Information about a single memory in the local system.
- struct memory_state - `allocation_id`s are "namespaced" to their memory ID, so we mai...
- struct message_id
- struct mode_traits
- class mpi_communicator - MPI implementation of the `communicator` interface. Wraps an `M...
- struct mpi_communicator::datatype_deleter
- struct mpi_communicator::in_flight_pilot - Keeps a stable pointer to a `pilot_message` alive during an asy...
- class mpi_event - async_event wrapper around an MPI_Request.
- struct multi_region_transfer - A (non-gather) transfer that has been mentioned in one or more ...
- struct nd_copy_layout - Describes a box-shaped copy operation between two box-shaped al...
- struct nd_copy_layout::stride_dimension
- struct nd_range_kernel_flavor
- struct no_local_size
- struct node_id
- struct non_empty_t
- struct normalized_t
- struct ones_t
- struct oob_bounding_box
- class out_of_order_engine - State machine controlling when and in what manner instructions ...
- struct out_of_order_engine::assignment - Directions on how a single (ready) instruction is to be dispatc...
- struct outbound_pilot - A pilot message as packaged on the sender side.
- struct pair_hash
- struct pilot_message - Metadata exchanged in preparation for a peer-to-peer data trans...
- struct plot
- class push_command
- class range_mapper
- class range_mapper_base
- struct raw_allocation_id
- class receive_arbiter - Matches receive instructions to inbound pilots and triggers in-...
- class receive_instruction - Requests the receive of a single buffer region. The receive can...
- class receive_instruction_impl - Common implementation mixin for receive_instruction and split_r...
- struct receive_instruction_record - IDAG record type for a `receive_instruction`.
- struct receive_instruction_record_impl - Base implementation for IDAG record types of `receive_instructi...
- class reduce_instruction - Performs an out-of-memory reduction by reading from a gather al...
- struct reduce_instruction_record - IDAG record type for a `reduce_instruction`.
- class reducer - Type-erased runtime reduction operation. Used to prepare and co...
- class reducer_impl
- struct reducer_transfer
- class reduction_command
- class reduction_descriptor<DataT, Dims, BinaryOperation, false>
- class reduction_descriptor<DataT, Dims, BinaryOperation, true>
- struct reduction_id
- struct reduction_info - Graph-level metadata on reductions.
- struct reduction_record
- class region - An arbitrary-dimensional set of points described by a normalize...
- class region_map - The region_map is a spatial data structure for storing values w...
- class region_map_impl - The region map is implemented as a customized R-Tree [Guttman 1...
- class region_map_impl<ValueType, 0>
- class region_map_types - Convenience types shared by inner_node and region_map_impl.
- struct region_map_types::erase_node_action
- struct region_map_types::insert_node_action
- struct region_map_types::insert_result
- class region_receive_event - Event for `receive_arbiter::receive`, which immediately awaits ...
- struct region_request - State for a single incomplete `receive` operation or a `begin_s...
- class runtime
- class scheduler
- class send_instruction - (MPI_) sends a subrange of an allocation to a single remote nod...
- struct send_instruction_record - IDAG record type for a `send_instruction`.
- class side_effect_map
- struct simple_kernel_flavor
- class split_receive_instruction - Informs the receive arbiter about the bounding box allocation f...
- struct split_receive_instruction_record - IDAG record type for a `split_receive_instruction`.
- class subregion_receive_event - Event for `receive_arbiter::await_split_receive_subregion`, whi...
- class subscript_proxy
- class sycl_backend - Backend implementation which sources all allocations from SYCL ...
- struct sycl_backend::impl
- struct sycl_backend_enumerator - Enumerates the SYCL backends devices are compatible with and th...
- class sycl_cuda_backend - CUDA specialized implementation of `sycl_backend` that uses nat...
- class sycl_event - Abstract base class for `async_event` implementations.
- class sycl_generic_backend - Generic implementation of `sycl_backend` providing a fallback i...
- struct system_info - All information about the local system that influences the gene...
- struct target_state - State maintained for every "multi-lane" target, i.e., `target::...
- class task
- class task_command
- struct task_geometry
- struct task_id
- class task_manager
- struct task_manager::buffer_state
- struct task_manager::host_object_state
- struct task_manager::policy_set
- struct task_record
- class task_recorder
- class task_ring_buffer
- class task_ring_buffer::reservation
- class task_ring_buffer::task_buffer_iterator
- class thread_queue - A single-thread job queue accepting functors and returning even...
- struct thread_queue::completion - The object passed through std::future from queue thread to owne...
- class thread_queue::event - Abstract base class for `async_event` implementations.
- struct thread_queue::impl
- struct thread_queue::job
- struct tracy_integration
- struct tracy_integration::async_lane_cursor - References a position in an `async_lane_state::zone_queue` from...
- struct tracy_integration::async_lane_id - Unique identifier for an `async_lane_state`.
- struct tracy_integration::async_lane_state - State for an async (fiber) lane. Keeps the active (suspended) z...
- struct tracy_integration::async_zone
- struct tracy_integration::instruction_info
- struct transfer_id - Uniquely identifies one version of a buffer's (distributed) dat...
- class transform_iterator
- struct type_switch - See `utils::type_switch_t`.
- struct type_switch<MatchingKey, <...>>
- struct type_switch<NonMatching, <...>>
- struct unassigned_state - Instruction is not eligible for assignment yet. Implies `num_in...
- struct unassigned_transfer - A transfer that is only known through inbound pilots so far, bu...
- struct unconditional_assignable_state - Instruction is inserted into `assignment_queue` and will be unc...
- struct unnamed_kernel
- struct user_allocation_transfer
- class write_command_state - write_command_state is a command_id with two bits of additional...
- struct zeros_t