[#llvm-ThreadPoolStrategy]
= xref:llvm.adoc[llvm]::ThreadPoolStrategy
:relfileprefix: ../
:mrdocs:

This tells how a thread pool will be used

== Synopsis

Declared in `<llvm/Support/Threading.h>`

[source,cpp,subs="verbatim,replacements,macros,-callouts"]
----
class ThreadPoolStrategy;
----

== Member Functions

[cols="1,4"]
|===
| Name| Description
| xref:llvm/ThreadPoolStrategy/apply_thread_strategy.adoc[`apply_thread_strategy`] 
| Assign the current thread to an ideal hardware CPU or NUMA node.
| xref:llvm/ThreadPoolStrategy/compute_cpu_socket.adoc[`compute_cpu_socket`] 
| Finds the CPU socket where a thread should go. Returns 'std::nullopt' if the thread shall remain on the actual CPU socket.
| xref:llvm/ThreadPoolStrategy/compute_thread_count.adoc[`compute_thread_count`] 
| Retrieves the max available threads for the current strategy. This accounts for affinity masks and takes advantage of all CPU sockets.
|===


== Data Members

[cols="1,4"]
|===
| Name| Description
| xref:llvm/ThreadPoolStrategy/Limit.adoc[`Limit`] 
| When true, clamp ThreadsRequested to hardware thread or core counts.
| xref:llvm/ThreadPoolStrategy/ThreadsRequested.adoc[`ThreadsRequested`] 
| Suggested upper bound on worker threads; 0 means use all available.
| xref:llvm/ThreadPoolStrategy/UseHyperThreads.adoc[`UseHyperThreads`] 
| If SMT is active, use hyper threads; otherwise one std::thread per core.
| xref:llvm/ThreadPoolStrategy/UseJobserver.adoc[`UseJobserver`] 
| When true, coordinate with a GNU Make jobserver if one is present.
|===


== Non-Member Functions

[cols="1,4"]
|===
| Name| Description
| xref:llvm/get_threadpool_strategy.adoc[`get_threadpool_strategy`]
| Build a strategy from a number of threads as a string provided in `Num.`
| xref:llvm/hardware_concurrency-03.adoc[`hardware_concurrency`]
| Parse `Num` into a hardware_concurrency strategy.
| xref:llvm/hardware_concurrency-09.adoc[`hardware_concurrency`]
| Use all available hardware threads, honoring affinity masks.
| xref:llvm/heavyweight_hardware_concurrency-01.adoc[`heavyweight_hardware_concurrency`]
| Build a strategy that prefers physical cores over SMT.
| xref:llvm/heavyweight_hardware_concurrency-08.adoc[`heavyweight_hardware_concurrency`]
| Parse `Num` into a heavyweight_hardware_concurrency strategy.
| xref:llvm/jobserver_concurrency.adoc[`jobserver_concurrency`]
| Coordinate with a GNU Make jobserver when one is present.
| xref:llvm/optimal_concurrency.adoc[`optimal_concurrency`]
| Limit threads to about `TaskCount` so small jobs avoid oversubscription.
| xref:llvm/lto/createInProcessThinBackend.adoc[`lto::createInProcessThinBackend`]
| Create a ThinBackend that runs individual jobs in‐process.
| xref:llvm/lto/createWriteIndexesThinBackend.adoc[`lto::createWriteIndexesThinBackend`]
| Create a ThinBackend that writes module indexes for distributed builds.
|===


[.small]#Created with https://www.mrdocs.com[MrDocs]#