DagScheduler::Config

Config

#include <xalgospp/scheduling/dag_scheduler.hh>

struct Config

Defined in src/lib/xalgospp/scheduling/dag_scheduler.hh:177

Name	Kind	Owner
`num_numa_nodes`	`variable`	Declared here
`threads_per_node`	`variable`	Declared here
`enable_pinning`	`variable`	Declared here
`max_concurrent_high_mem`	`variable`	Declared here
`max_concurrency_multiplier`	`variable`	Declared here
`enable_dynamic_backpressure`	`variable`	Declared here
`enable_autotuning`	`variable`	Declared here
`raw_frame_size_bytes`	`variable`	Declared here
`warmup_submissions`	`variable`	Declared here
`node_memory_bandwidth_limit_gbps`	`variable`	Declared here
`percent_bandwidth_is_high_mem`	`variable`	Declared here

Return	Name	Description
`std::size_t`	`num_numa_nodes`	The total number of NUMA nodes available. 0 means to auto-detect.
`std::size_t`	`threads_per_node`	Threads to allocate per NUMA node. 0 means to map to physical cores.
`bool`	`enable_pinning`	Whether to allow pinning of tasks to specific cores.
`std::size_t`	`max_concurrent_high_mem`	The threshold at which throttling will begin for high-memory-bandwidth tasks.
`std::size_t`	`max_concurrency_multiplier`	A generalized concurrency limit for suspension of generator Tasks (if enabled).
`bool`	`enable_dynamic_backpressure`	Whether back pressure monitoring for generator Tasks should be enabled.
`bool`	`enable_autotuning`	Allow the Scheduler to automatically determine resource and queue parameters.
`std::size_t`	`raw_frame_size_bytes`	Raw frame size used for auto-tuning. Auto-determined if set to 0.
`std::size_t`	`warmup_submissions`	When autotuning, the number of DAG submissions used for warmup before timing.
`double`	`node_memory_bandwidth_limit_gbps`	The peak memory bandwidth, per NUMA node.
`double`	`percent_bandwidth_is_high_mem`	Percentage of total bandwidth usage to classify as “high memory” (memory-bound).

std::size_t num_numa_nodes { 0 }

Defined in src/lib/xalgospp/scheduling/dag_scheduler.hh:181

The total number of NUMA nodes available. 0 means to auto-detect.

std::size_t threads_per_node { 0 }

Defined in src/lib/xalgospp/scheduling/dag_scheduler.hh:185

Threads to allocate per NUMA node. 0 means to map to physical cores.

bool enable_pinning { true }

Defined in src/lib/xalgospp/scheduling/dag_scheduler.hh:189

Whether to allow pinning of tasks to specific cores.

std::size_t max_concurrent_high_mem { 2 }

Defined in src/lib/xalgospp/scheduling/dag_scheduler.hh:193

The threshold at which throttling will begin for high-memory-bandwidth tasks.

std::size_t max_concurrency_multiplier { 8 }

Defined in src/lib/xalgospp/scheduling/dag_scheduler.hh:197

A generalized concurrency limit for suspension of generator Tasks (if enabled).

bool enable_dynamic_backpressure { true }

Defined in src/lib/xalgospp/scheduling/dag_scheduler.hh:201

Whether back pressure monitoring for generator Tasks should be enabled.

bool enable_autotuning { true }

Defined in src/lib/xalgospp/scheduling/dag_scheduler.hh:205

Allow the Scheduler to automatically determine resource and queue parameters.

std::size_t raw_frame_size_bytes { 0 }

Defined in src/lib/xalgospp/scheduling/dag_scheduler.hh:209

Raw frame size used for auto-tuning. Auto-determined if set to 0.

std::size_t warmup_submissions { 5 }

Defined in src/lib/xalgospp/scheduling/dag_scheduler.hh:213

When autotuning, the number of DAG submissions used for warmup before timing.

double node_memory_bandwidth_limit_gbps { 50.0 }

Defined in src/lib/xalgospp/scheduling/dag_scheduler.hh:217

The peak memory bandwidth, per NUMA node.

double percent_bandwidth_is_high_mem { 0.25 }

Defined in src/lib/xalgospp/scheduling/dag_scheduler.hh:221

Percentage of total bandwidth usage to classify as “high memory” (memory-bound).