NodeJS: Native c++ module with multi-threading (openmp) - c++

Question

I have a very basic native module for NodeJS which essentially wraps a more complicated c++ program which includes multithreading using OpenMP (it's a Monte Carlo simulation). This c++ program works fine on its own and when called from python using a cython module to wrap the c++ code. However, when I compile it to a native node module (using /OpenMP as the parameter to the MSBuild compiler) it only uses one thread.
The following is my binding.gyp for reference:
{ "targets": [
{
"target_name": "LossForecast",
"sources": [ "NodeLossForecast.cpp", "src/AutoDiff.cpp" ],
"include_dirs":["src", "<!(node -e \"require('nan')\")"],
'conditions': [
['OS=="win"',
{
'cflags': ["/EHsc", "/O2", "/openmp"]
}
]
]
} ]}

Answer 1

So, I am writing down my files for running a simple node-addon-api pi program using OpenMP. Although node-addon-api is experimental, it will be soon released as an official api for node.js. Its very simple.
This is for Windows only. And I can confirm if it is running in multiple processors.
binding.gyp
{
"targets": [
{
"target_name": "nodeOpenMP",
"defines": [
"VERSION=0.0.1",
"NAPI_DISABLE_CPP_EXCEPTIONS"
],
"sources": [
"module.cc"
],
"include_dirs": [
"<!(echo %cd%/node_modules/node-addon-api)"
],
"conditions": [
[
'OS=="win"',
{
'msvs_settings': {
'VCCLCompilerTool' : {
'AdditionalOptions' : ['/MT','/openmp']
}
}
}
]
]
}
]
}
You have to add VCCLCompilerTool and then under AdditionalOptions add `/openmp' as mentioned above.
Here's a simple PI program that I wrote,
module.cc
#include <napi.h>
#include <omp.h>
#include <iostream>
#include <windows.h> // for GetCurrentProcessorNumber()
int getThreads(){
// Get number of threads
omp_set_num_threads(3);
int n = 0;
#pragma omp parallel reduction(+:n)
n += 1;
return n;
}
double GetPi(short numOfThreads,long numberOfSteps){
long i;
double pi, sum = 0.0;
double step = 1.0/(double) numberOfSteps;
omp_set_num_threads(numOfThreads);
#pragma omp parallel
{
std::cout << "This thread is running on processor: " << GetCurrentProcessorNumber() << "\n";
double x;
#pragma omp for reduction(+:sum)
for(i = 0; i < numberOfSteps; i++) {
x = ( i + 0.5 ) * step;
sum += 4.0 / (1 + x*x);
}
}
std::cout << "Total no. of threads (not processors)" <<getThreads() << std::endl;
pi = step * (double)sum;
return pi;
}
Napi::Value calculatePi(const Napi::CallbackInfo& info ){
Napi::Env env = info.Env();
// check for no. of arguments
if(info.Length()<2){
Napi::TypeError::New(env, "Wrong number of arguments").ThrowAsJavaScriptException();
return env.Null();
}
if (!info[0].IsNumber() || !info[1].IsNumber()) {
Napi::TypeError::New(env, "Wrong arguments").ThrowAsJavaScriptException();
return env.Null();
}
double numThreads = info[0].As<Napi::Number>().DoubleValue();
double numSteps = info[1].As<Napi::Number>().DoubleValue();
double pi = GetPi(numThreads, numSteps);
Napi::Number PI = Napi::Number::New(env, pi);
return PI;
}
Napi::Object Init(Napi::Env env, Napi::Object exports){
// register the functions that are to be exported
exports.Set(Napi::String::New(env, "pi"), Napi::Function::New(env, calculatePi));
return exports;
}
NODE_API_MODULE(nodeOpenMP, Init);
testOMP.js
const omp = require("./build/Release/nodeOpenMP");
const numThreads = 4, numSteps = 1000000;
console.log( numThreads, numSteps );
Simply copy paste the files above accordingly. Don't forget to install node-addon-api(locally) and node-gyp(globally) and then run:
node-gyp configure build && node --no-warnings testOMP.js
You should get an output like this:
This thread is running on processor: 3
This thread is running on processor: 3
This thread is running on processor: 0
This thread is running on processor: 3
Total no. of threads (not processors): 3
I am thinking of soon making a cross-platform npm package node-openmp.
Follow this repo and contribute. I am open to any contributions.
Himujjal/node-openmp