The number of hardware threads available to a ThreadPool can be limited if setting an affinity mask.
For example:
> start /B /AFFINITY 0xF lld-link.exe ...
Would let LLD only use 4 hyper-threads.
Previously, there was an outstanding issue on Windows Server 2019 on dual-CPU machines, which was preventing from using both CPU sockets. In normal conditions, when no affinity mask was set, ProcessorGroup::AllThreads was different from ProcessorGroup::UsableThreads. The previous code in llvm/lib/Support/Windows/Threading.inc L201 was improperly assuming those two values to be equal, and consequently was limiting the execution to only one CPU socket.
Differential Revision: https://reviews.llvm.org/D92419
The number of hardware threads available to a ThreadPool can be limited if setting an affinity mask.
For example:
> start /B /AFFINITY 0xF lld-link.exe ...
Would let LLD only use 4 hyper-threads.
Previously, there was an outstanding issue on Windows Server 2019 on dual-CPU machines, which was preventing from using both CPU sockets. In normal conditions, when no affinity mask was set, ProcessorGroup::AllThreads was different from ProcessorGroup::UsableThreads. The previous code in llvm/lib/Support/Windows/Threading.inc L201 was improperly assuming those two values to be equal, and consequently was limiting the execution to only one CPU socket.
Differential Revision: https://reviews.llvm.org/D92419
This reverts commit d9c8b0256cfc673c2413b13993c9440be598818f.
Some MSVC std::packaged_task implementations are not compatible with move-only types.
This caused failures on some of the Windows builders (e.g.
http://lab.llvm.org:8011/builders/sanitizer-windows/builds/69412).
Reverting until I can come up with a workaround.
This will allow non-copyable function objects (e.g. lambdas that capture
unique_ptrs) to be used with ThreadPool.
Differential Revision: https://reviews.llvm.org/D87467
The goal of this patch is to maximize CPU utilization on multi-socket or high core count systems, so that parallel computations such as LLD/ThinLTO can use all hardware threads in the system. Before this patch, on Windows, a maximum of 64 hardware threads could be used at most, in some cases dispatched only on one CPU socket.
== Background ==
Windows doesn't have a flat cpu_set_t like Linux. Instead, it projects hardware CPUs (or NUMA nodes) to applications through a concept of "processor groups". A "processor" is the smallest unit of execution on a CPU, that is, an hyper-thread if SMT is active; a core otherwise. There's a limit of 32-bit processors on older 32-bit versions of Windows, which later was raised to 64-processors with 64-bit versions of Windows. This limit comes from the affinity mask, which historically is represented by the sizeof(void*). Consequently, the concept of "processor groups" was introduced for dealing with systems with more than 64 hyper-threads.
By default, the Windows OS assigns only one "processor group" to each starting application, in a round-robin manner. If the application wants to use more processors, it needs to programmatically enable it, by assigning threads to other "processor groups". This also means that affinity cannot cross "processor group" boundaries; one can only specify a "preferred" group on start-up, but the application is free to allocate more groups if it wants to.
This creates a peculiar situation, where newer CPUs like the AMD EPYC 7702P (64-cores, 128-hyperthreads) are projected by the OS as two (2) "processor groups". This means that by default, an application can only use half of the cores. This situation could only get worse in the years to come, as dies with more cores will appear on the market.
== The problem ==
The heavyweight_hardware_concurrency() API was introduced so that only *one hardware thread per core* was used. Once that API returns, that original intention is lost, only the number of threads is retained. Consider a situation, on Windows, where the system has 2 CPU sockets, 18 cores each, each core having 2 hyper-threads, for a total of 72 hyper-threads. Both heavyweight_hardware_concurrency() and hardware_concurrency() currently return 36, because on Windows they are simply wrappers over std:🧵:hardware_concurrency() -- which can only return processors from the current "processor group".
== The changes in this patch ==
To solve this situation, we capture (and retain) the initial intention until the point of usage, through a new ThreadPoolStrategy class. The number of threads to use is deferred as late as possible, until the moment where the std::threads are created (ThreadPool in the case of ThinLTO).
When using hardware_concurrency(), setting ThreadCount to 0 now means to use all the possible hardware CPU (SMT) threads. Providing a ThreadCount above to the maximum number of threads will have no effect, the maximum will be used instead.
The heavyweight_hardware_concurrency() is similar to hardware_concurrency(), except that only one thread per hardware *core* will be used.
When LLVM_ENABLE_THREADS is OFF, the threading APIs will always return 1, to ensure any caller loops will be exercised at least once.
Differential Revision: https://reviews.llvm.org/D71775
to reflect the new license.
We understand that people may be surprised that we're moving the header
entirely to discuss the new license. We checked this carefully with the
Foundation's lawyer and we believe this is the correct approach.
Essentially, all code in the project is now made available by the LLVM
project under our new license, so you will see that the license headers
include that license only. Some of our contributors have contributed
code under our old license, and accordingly, we have retained a copy of
our old license notice in the top-level files in each project and
repository.
llvm-svn: 351636
This is failing to compile when LLVM_ENABLE_THREADS is false,
and the fix is not immediately obvious, so reverting while I look
into it.
llvm-svn: 334658
Previously ThreadPool could only queue async "jobs", i.e. work
that was done for its side effects and not for its result. It's
useful occasionally to queue async work that returns a value.
From an API perspective, this is very intuitive. The previous
API just returned a shared_future<void>, so all we need to do is
make it return a shared_future<T>, where T is the type of value
that the operation returns.
Making this work required a little magic, but ultimately it's not
too bad. Instead of keeping a shared queue<packaged_task<void()>>
we just keep a shared queue<unique_ptr<TaskBase>>, where TaskBase
is a class with a pure virtual execute() method, then have a
templated derived class that stores a packaged_task<T()>. Everything
else works out pretty cleanly.
Differential Revision: https://reviews.llvm.org/D48115
llvm-svn: 334643
Remove all checks that required main thread to run faster than tasks in
ThreadPool, and yields which are now unnecessary. This should fix some
bot failures.
llvm-svn: 256056
This is a very simple implementation of a thread pool using C++11
thread. It accepts any std::function<void()> for asynchronous
execution. Individual task can be synchronize using the returned
future, or the client can block on the full queue completion.
In case LLVM is configured with Threading disabled, it falls back
to sequential execution using std::async with launch:deferred.
This is intended to support parallelism for ThinLTO processing in
linker plugin, but is generic enough for any other uses.
This is a recommit of r255444 ; trying to workaround a bug in the
MSVC 2013 standard library. I think I was hit by:
http://connect.microsoft.com/VisualStudio/feedbackdetail/view/791185/std-packaged-task-t-where-t-is-void-or-a-reference-class-are-not-movable
Recommit of r255589, trying to please g++ as well.
Differential Revision: http://reviews.llvm.org/D15464
From: mehdi_amini <mehdi_amini@91177308-0d34-0410-b5e6-96231b3b80d8>
llvm-svn: 255593
This is a very simple implementation of a thread pool using C++11
thread. It accepts any std::function<void()> for asynchronous
execution. Individual task can be synchronize using the returned
future, or the client can block on the full queue completion.
In case LLVM is configured with Threading disabled, it falls back
to sequential execution using std::async with launch:deferred.
This is intended to support parallelism for ThinLTO processing in
linker plugin, but is generic enough for any other uses.
This is a recommit of r255444 ; trying to workaround a bug in the
MSVC 2013 standard library. I think I was hit by:
http://connect.microsoft.com/VisualStudio/feedbackdetail/view/791185/std-packaged-task-t-where-t-is-void-or-a-reference-class-are-not-movable
Differential Revision: http://reviews.llvm.org/D15464
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 255589
This is a very simple implementation of a thread pool using C++11
thread. It accepts any std::function<void()> for asynchronous
execution. Individual task can be synchronize using the returned
future, or the client can block on the full queue completion.
In case LLVM is configured with Threading disabled, it falls back
to sequential execution using std::async with launch:deferred.
This is intended to support parallelism for ThinLTO processing in
linker plugin, but is generic enough for any other uses.
Differential Revision: http://reviews.llvm.org/D15464
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 255444