saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-07-29 20:14:01 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Ifdef cpuinfo code on platforms we do not set affinity (#13486)

Browse source 
### Description
Remove code that invokes cpuinfo library on platforms we do not set
affinity.

### Motivation and Context
`cpuinfo` library increases binary size.
This commit is contained in:
Dmitri Smirnov 2022-11-28 13:44:16 -08:00 committed by GitHub
parent 2d523c507e
commit 4fbe16e493
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 13 additions and 17 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

30
onnxruntime/core/platform/posix/env.cc
View file

@ -36,8 +36,12 @@ limitations under the License.
#include <utility> // for std::forward
#include <vector>
#ifdef CPUINFO_SUPPORTED
// We can not use CPUINFO if it is not supported and we do not want to used
// it on certain platforms because of the binary size increase.
// We could use it to find out the number of physical cores for certain supported platforms
#if defined(CPUINFO_SUPPORTED) && !defined(__APPLE__) && !defined(__ANDROID__) && !defined(__wasm__) && !defined(_AIX)
#include <cpuinfo.h>
#define ORT_USE_CPUINFO
#endif
#include "core/common/common.h"
@ -271,36 +275,31 @@ class PosixEnv : public Env {
return new PosixThread(name_prefix, index, start_address, param, thread_options);
}
// we are guessing the number of phys cores based on a popular HT case.
// we are guessing the number of phys cores based on a popular HT case (2 logical proc per core)
static int DefaultNumCores() {
return std::max(1, static_cast<int>(std::thread::hardware_concurrency() / 2));
}
// Return the number of physical cores
int GetNumPhysicalCpuCores() const override {
#ifdef CPUINFO_SUPPORTED
#ifdef ORT_USE_CPUINFO
if(cpuinfo_available_) {
return narrow<int>(cpuinfo_get_cores_count());
}
#endif
// We guess the number of cores
#endif // ORT_USE_CPUINFO
return DefaultNumCores();
}
std::vector<LogicalProcessors> GetThreadAffinityMasks() const override {
std::vector<LogicalProcessors> ret;
#ifdef CPUINFO_SUPPORTED
#ifdef ORT_USE_CPUINFO
if (cpuinfo_available_) {
#if !defined(__APPLE__) && !defined(__ANDROID__) && !defined(__wasm__) && !defined(_AIX)
// We currently do not implement affinity on more than 64 cores.
auto num_phys_cores = cpuinfo_get_cores_count();
ret.reserve(num_phys_cores);
for (uint32_t i = 0; i < num_phys_cores; ++i) {
const auto* core = cpuinfo_get_core(i);
LogicalProcessors th_aff;
// Processor count will never exceed 64 in a given group.
// TBD: Processor groups are currently not taken into account.
th_aff.reserve(core->processor_count);
auto log_proc_idx = core->processor_start;
for (uint32_t count = 0; count < core->processor_count; count++, ++log_proc_idx) {
@ -309,9 +308,8 @@ class PosixEnv : public Env {
}
ret.push_back(std::move(th_aff));
}
#endif
}
#endif // CPUINFO_SUPPORTED
#endif
// Just the size of the thread-pool
if(ret.empty()) {
ret.resize(GetNumPhysicalCpuCores());
@ -581,18 +579,16 @@ class PosixEnv : public Env {
}
private:
Telemetry telemetry_provider_;
#ifdef ORT_USE_CPUINFO
PosixEnv() {
#ifdef CPUINFO_SUPPORTED
cpuinfo_available_ = cpuinfo_initialize();
if(!cpuinfo_available_) {
LOGS_DEFAULT(INFO) << "cpuinfo_initialize failed";
}
#endif
}
Telemetry telemetry_provider_;
#ifdef CPUINFO_SUPPORTED
bool cpuinfo_available_{false};
#endif
#endif // ORT_USE_CPUINFO
};
} // namespace