Skip to content

Commit

Permalink
Allow for NUMA memory replication for NNUE weights. Bind threads to e…
Browse files Browse the repository at this point in the history 
…nsure execution on a specific NUMA node.

This patch introduces NUMA memory replication, currently only utilized for the NNUE weights. Along with it comes all machinery required to identify NUMA nodes and bind threads to specific processors/nodes. It also comes with small changes to Thread and ThreadPool to allow easier execution of custom functions on the designated thread. Old thread binding (WinProcGroup) machinery is removed because it's incompatible with this patch. Small changes to unrelated parts of the code were made to ensure correctness, like some classes being made unmovable, raw pointers replaced with unique_ptr. etc.

Windows 7 and Windows 10 is partially supported. Windows 11 is fully supported. Linux is fully supported, with explicit exclusion of Android. No additional dependencies.

-----------------

A new UCI option `NumaPolicy` is introduced. It can take the following values:
```
system - gathers NUMA node information from the system (lscpu or windows api), for each threads binds it to a single NUMA node
none - assumes there is 1 NUMA node, never binds threads
auto - this is the default value, depends on the number of set threads and NUMA nodes, will only enable binding on multinode systems and when the number of threads reaches a threshold (dependent on node size and count)
[[custom]] - 
  // ':'-separated numa nodes
  // ','-separated cpu indices
  // supports "first-last" range syntax for cpu indices, 
  for example '0-15,32-47:16-31,48-63'
```

Setting `NumaPolicy` forces recreation of the threads in the ThreadPool, which in turn forces the recreation of the TT.

The threads are distributed among NUMA nodes in a round-robin fashion based on fill percentage (i.e. it will strive to fill all NUMA nodes evenly). Threads are bound to NUMA nodes, not specific processors, because that's our only requirement and the OS can schedule them better.

Special care is made that maximum memory usage on systems that do not require memory replication stays as previously, that is, unnecessary copies are avoided.

On linux the process' processor affinity is respected. This means that if you for example use taskset to restrict Stockfish to a single NUMA node then the `system` and `auto` settings will only see a single NUMA node (more precisely, the processors included in the current affinity mask) and act accordingly.

-----------------

We can't ensure that a memory allocation takes place on a given NUMA node without using libnuma on linux, or using appropriate custom allocators on windows (https://learn.microsoft.com/en-us/windows/win32/memory/allocating-memory-from-a-numa-node), so to avoid complications the current implementation relies on first-touch policy. Due to this we also rely on the memory allocator to give us a new chunk of untouched memory from the system. This appears to work reliably on linux, but results may vary.

MacOS is not supported, because AFAIK it's not affected, and implementation would be problematic anyway.

Windows is supported since Windows 7 (https://learn.microsoft.com/en-us/windows/win32/api/processtopologyapi/nf-processtopologyapi-setthreadgroupaffinity). Until Windows 11/Server 2022 NUMA nodes are split such that they cannot span processor groups. This is because before Windows 11/Server 2022 it's not possible to set thread affinity spanning processor groups. The splitting is done manually in some cases (required after Windows 10 Build 20348). Since Windows 11/Server 2022 we can set affinites spanning processor group so this splitting is not done, so the behaviour is pretty much like on linux.

Linux is supported, **without** libnuma requirement. `lscpu` is expected.

-----------------

Passed SMP STC: https://tests.stockfishchess.org/tests/view/6654fc74a86388d5e27db1cd
```
LLR: 2.95 (-2.94,2.94) <-1.75,0.25>
Total: 67152 W: 17354 L: 17177 D: 32621
Ptnml(0-2): 64, 7428, 18408, 7619, 57
```

Passed STC: https://tests.stockfishchess.org/tests/view/6654fb27a86388d5e27db15c
```
LLR: 2.94 (-2.94,2.94) <-1.75,0.25>
Total: 131648 W: 34155 L: 34045 D: 63448
Ptnml(0-2): 426, 13878, 37096, 14008, 416
```

fixes #5253
  • Loading branch information
@Sopel97
Sopel97 committed May 28, 2024
1 parent 4759764 commit a027ef9
Show file tree
Hide file tree
Showing 18 changed files with 1,399 additions and 289 deletions.
2 changes: 1 addition & 1 deletion src/Makefile
View file
Original file line number Diff line number Diff line change
Expand Up @@ -63,7 +63,7 @@ HEADERS = benchmark.h bitboard.h evaluate.h misc.h movegen.h movepick.h \
nnue/layers/sqr_clipped_relu.h nnue/nnue_accumulator.h nnue/nnue_architecture.h \
nnue/nnue_common.h nnue/nnue_feature_transformer.h position.h \
search.h syzygy/tbprobe.h thread.h thread_win32_osx.h timeman.h \
tt.h tune.h types.h uci.h ucioption.h perft.h nnue/network.h engine.h score.h
tt.h tune.h types.h uci.h ucioption.h perft.h nnue/network.h engine.h score.h numa.h

OBJS = $(notdir $(SRCS:.cpp=.o))

Expand Down
88 changes: 70 additions & 18 deletions src/engine.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -18,15 +18,15 @@

#include "engine.h"

#include <cassert>
#include <deque>
#include <iosfwd>
#include <memory>
#include <ostream>
#include <sstream>
#include <string_view>
#include <utility>
#include <vector>
#include <sstream>
#include <iosfwd>
#include <cassert>

#include "evaluate.h"
#include "misc.h"
Expand All @@ -48,10 +48,14 @@ constexpr auto StartFEN = "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq -

Engine::Engine(std::string path) :
binaryDirectory(CommandLine::get_binary_directory(path)),
numaContext(NumaConfig::from_system()),
states(new std::deque<StateInfo>(1)),
networks(NN::Networks(
NN::NetworkBig({EvalFileDefaultNameBig, "None", ""}, NN::EmbeddedNNUEType::BIG),
NN::NetworkSmall({EvalFileDefaultNameSmall, "None", ""}, NN::EmbeddedNNUEType::SMALL))) {
threads(),
networks(
numaContext,
NN::Networks(
NN::NetworkBig({EvalFileDefaultNameBig, "None", ""}, NN::EmbeddedNNUEType::BIG),
NN::NetworkSmall({EvalFileDefaultNameSmall, "None", ""}, NN::EmbeddedNNUEType::SMALL))) {
pos.set(StartFEN, false, &states->back());
capSq = SQ_NONE;
}
Expand All @@ -74,7 +78,7 @@ void Engine::stop() { threads.stop = true; }
void Engine::search_clear() {
wait_for_search_finished();

tt.clear(options["Threads"]);
tt.clear(threads);
threads.clear();

// @TODO wont work with multiple instances
Expand Down Expand Up @@ -124,40 +128,71 @@ void Engine::set_position(const std::string& fen, const std::vector<std::string>

// modifiers

void Engine::resize_threads() { threads.set({options, threads, tt, networks}, updateContext); }
void Engine::set_numa_config_from_option(const std::string& o) {
if (o == "auto" || o == "system")
{
numaContext.set_numa_config(NumaConfig::from_system());
}
else if (o == "none")
{
numaContext.set_numa_config(NumaConfig{});
}
else
{
numaContext.set_numa_config(NumaConfig::from_string(o));
}

// Force reallocation of threads in case affinities need to change.
resize_threads();
}

void Engine::resize_threads() {
threads.wait_for_search_finished();
threads.set(numaContext.get_numa_config(), {options, threads, tt, networks}, updateContext);

// Reallocate the hash with the new threadpool size
set_tt_size(options["Hash"]);
}

void Engine::set_tt_size(size_t mb) {
wait_for_search_finished();
tt.resize(mb, options["Threads"]);
tt.resize(mb, threads);
}

void Engine::set_ponderhit(bool b) { threads.main_manager()->ponder = b; }

// network related

void Engine::verify_networks() const {
networks.big.verify(options["EvalFile"]);
networks.small.verify(options["EvalFileSmall"]);
networks->big.verify(options["EvalFile"]);
networks->small.verify(options["EvalFileSmall"]);
}

void Engine::load_networks() {
load_big_network(options["EvalFile"]);
load_small_network(options["EvalFileSmall"]);
networks.modify_and_replicate([this](NN::Networks& networks_) {
networks_.big.load(binaryDirectory, options["EvalFile"]);
networks_.small.load(binaryDirectory, options["EvalFileSmall"]);
});
threads.clear();
}

void Engine::load_big_network(const std::string& file) {
networks.big.load(binaryDirectory, file);
networks.modify_and_replicate(
[this, &file](NN::Networks& networks_) { networks_.big.load(binaryDirectory, file); });
threads.clear();
}

void Engine::load_small_network(const std::string& file) {
networks.small.load(binaryDirectory, file);
networks.modify_and_replicate(
[this, &file](NN::Networks& networks_) { networks_.small.load(binaryDirectory, file); });
threads.clear();
}

void Engine::save_network(const std::pair<std::optional<std::string>, std::string> files[2]) {
networks.big.save(files[0].first);
networks.small.save(files[1].first);
networks.modify_and_replicate([&files](NN::Networks& networks_) {
networks_.big.save(files[0].first);
networks_.small.save(files[1].first);
});
}

// utility functions
Expand All @@ -169,7 +204,7 @@ void Engine::trace_eval() const {

verify_networks();

sync_cout << "\n" << Eval::trace(p, networks) << sync_endl;
sync_cout << "\n" << Eval::trace(p, *networks) << sync_endl;
}

OptionsMap& Engine::get_options() { return options; }
Expand All @@ -184,4 +219,21 @@ std::string Engine::visualize() const {
return ss.str();
}

std::vector<std::pair<size_t, size_t>> Engine::get_bound_thread_count_by_numa_node() const {
auto counts = threads.get_bound_thread_count_by_numa_node();
const NumaConfig& cfg = numaContext.get_numa_config();
std::vector<std::pair<size_t, size_t>> ratios;
NumaIndex n = 0;
for (; n < counts.size(); ++n)
ratios.emplace_back(counts[n], cfg.num_cpus_in_numa_node(n));
if (!counts.empty())
for (; n < cfg.num_numa_nodes(); ++n)
ratios.emplace_back(0, cfg.num_cpus_in_numa_node(n));
return ratios;
}

std::string Engine::get_numa_config_as_string() const {
return numaContext.get_numa_config().to_string();
}

}
31 changes: 22 additions & 9 deletions src/engine.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -35,6 +35,7 @@
#include "thread.h"
#include "tt.h"
#include "ucioption.h"
#include "numa.h"

namespace Stockfish {

Expand All @@ -47,6 +48,13 @@ class Engine {
using InfoIter = Search::InfoIteration;

Engine(std::string path = "");

// Can't be movable due to components holding backreferences to fields
Engine(const Engine&) = delete;
Engine(Engine&&) = delete;
Engine& operator=(const Engine&) = delete;
Engine& operator=(Engine&&) = delete;

~Engine() { wait_for_search_finished(); }

std::uint64_t perft(const std::string& fen, Depth depth, bool isChess960);
Expand All @@ -63,6 +71,7 @@ class Engine {

// modifiers

void set_numa_config_from_option(const std::string& o);
void resize_threads();
void set_tt_size(size_t mb);
void set_ponderhit(bool);
Expand All @@ -83,23 +92,27 @@ class Engine {

// utility functions

void trace_eval() const;
OptionsMap& get_options();
std::string fen() const;
void flip();
std::string visualize() const;
void trace_eval() const;
OptionsMap& get_options();
std::string fen() const;
void flip();
std::string visualize() const;
std::vector<std::pair<size_t, size_t>> get_bound_thread_count_by_numa_node() const;
std::string get_numa_config_as_string() const;

private:
const std::string binaryDirectory;

NumaReplicationContext numaContext;

Position pos;
StateListPtr states;
Square capSq;

OptionsMap options;
ThreadPool threads;
TranspositionTable tt;
Eval::NNUE::Networks networks;
OptionsMap options;
ThreadPool threads;
TranspositionTable tt;
NumaReplicated<Eval::NNUE::Networks> networks;

Search::SearchManager::UpdateContext updateContext;
};
Expand Down
134 changes: 11 additions & 123 deletions src/misc.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -48,6 +48,7 @@ using fun8_t = bool (*)(HANDLE, BOOL, PTOKEN_PRIVILEGES, DWORD, PTOKEN_PRIVILEGE
#endif

#include <atomic>
#include <charconv>
#include <cmath>
#include <cstdlib>
#include <fstream>
Expand All @@ -56,6 +57,7 @@ using fun8_t = bool (*)(HANDLE, BOOL, PTOKEN_PRIVILEGES, DWORD, PTOKEN_PRIVILEGE
#include <mutex>
#include <sstream>
#include <string_view>
#include <system_error>

#include "types.h"

Expand Down Expand Up @@ -592,129 +594,6 @@ void aligned_large_pages_free(void* mem) { std_aligned_free(mem); }
#endif


namespace WinProcGroup {

#ifndef _WIN32

void bind_this_thread(size_t) {}

#else

namespace {
// Retrieves logical processor information using Windows-specific
// API and returns the best node id for the thread with index idx. Original
// code from Texel by Peter Österlund.
int best_node(size_t idx) {

int threads = 0;
int nodes = 0;
int cores = 0;
DWORD returnLength = 0;
DWORD byteOffset = 0;

// Early exit if the needed API is not available at runtime
HMODULE k32 = GetModuleHandle(TEXT("Kernel32.dll"));
auto fun1 = (fun1_t) (void (*)()) GetProcAddress(k32, "GetLogicalProcessorInformationEx");
if (!fun1)
return -1;

// First call to GetLogicalProcessorInformationEx() to get returnLength.
// We expect the call to fail due to null buffer.
if (fun1(RelationAll, nullptr, &returnLength))
return -1;

// Once we know returnLength, allocate the buffer
SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX *buffer, *ptr;
ptr = buffer = (SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX*) malloc(returnLength);

// Second call to GetLogicalProcessorInformationEx(), now we expect to succeed
if (!fun1(RelationAll, buffer, &returnLength))
{
free(buffer);
return -1;
}

while (byteOffset < returnLength)
{
if (ptr->Relationship == RelationNumaNode)
nodes++;

else if (ptr->Relationship == RelationProcessorCore)
{
cores++;
threads += (ptr->Processor.Flags == LTP_PC_SMT) ? 2 : 1;
}

assert(ptr->Size);
byteOffset += ptr->Size;
ptr = (SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX*) (((char*) ptr) + ptr->Size);
}

free(buffer);

std::vector<int> groups;

// Run as many threads as possible on the same node until the core limit is
// reached, then move on to filling the next node.
for (int n = 0; n < nodes; n++)
for (int i = 0; i < cores / nodes; i++)
groups.push_back(n);

// In case a core has more than one logical processor (we assume 2) and we
// still have threads to allocate, spread them evenly across available nodes.
for (int t = 0; t < threads - cores; t++)
groups.push_back(t % nodes);

// If we still have more threads than the total number of logical processors
// then return -1 and let the OS to decide what to do.
return idx < groups.size() ? groups[idx] : -1;
}
}


// Sets the group affinity of the current thread
void bind_this_thread(size_t idx) {

// Use only local variables to be thread-safe
int node = best_node(idx);

if (node == -1)
return;

// Early exit if the needed API are not available at runtime
HMODULE k32 = GetModuleHandle(TEXT("Kernel32.dll"));
auto fun2 = fun2_t((void (*)()) GetProcAddress(k32, "GetNumaNodeProcessorMaskEx"));
auto fun3 = fun3_t((void (*)()) GetProcAddress(k32, "SetThreadGroupAffinity"));
auto fun4 = fun4_t((void (*)()) GetProcAddress(k32, "GetNumaNodeProcessorMask2"));
auto fun5 = fun5_t((void (*)()) GetProcAddress(k32, "GetMaximumProcessorGroupCount"));

if (!fun2 || !fun3)
return;

if (!fun4 || !fun5)
{
GROUP_AFFINITY affinity;
if (fun2(node, &affinity)) // GetNumaNodeProcessorMaskEx
fun3(GetCurrentThread(), &affinity, nullptr); // SetThreadGroupAffinity
}
else
{
// If a numa node has more than one processor group, we assume they are
// sized equal and we spread threads evenly across the groups.
USHORT elements, returnedElements;
elements = fun5(); // GetMaximumProcessorGroupCount
GROUP_AFFINITY* affinity = (GROUP_AFFINITY*) malloc(elements * sizeof(GROUP_AFFINITY));
if (fun4(node, affinity, elements, &returnedElements)) // GetNumaNodeProcessorMask2
fun3(GetCurrentThread(), &affinity[idx % returnedElements],
nullptr); // SetThreadGroupAffinity
free(affinity);
}
}

#endif

} // namespace WinProcGroup

#ifdef _WIN32
#include <direct.h>
#define GETCWD _getcwd
Expand All @@ -723,6 +602,15 @@ void bind_this_thread(size_t idx) {
#define GETCWD getcwd
#endif

size_t str_to_size_t(const std::string& s) {
size_t value;
auto result = std::from_chars(s.data(), s.data() + s.size(), value);

if (result.ec != std::errc())
std::exit(EXIT_FAILURE);

return value;
}

std::string CommandLine::get_binary_directory(std::string argv0) {
std::string pathSeparator;
Expand Down
Loading

0 comments on commit a027ef9

Please sign in to comment.