duoram/preprocessing/preprocessing.cpp

#include <type_traits>  // std::is_same<>
#include <limits>       // std::numeric_limits<>
#include <climits>      // CHAR_BIT
#include <cmath>        // std::log2, std::ceil, std::floor
#include <stdexcept>    // std::runtime_error
#include <array>        // std::array<>
#include <iostream>     // std::istream and std::ostream
#include <vector>       // std::vector<>
#include <memory>       // std::shared_ptr<>
#include <utility>      // std::move
#include <algorithm>    // std::copy
#include <cstring>      // std::memcpy
#include <bsd/stdlib.h> // arc4random_buf
#include <x86intrin.h>  // SSE and AVX intrinsics
#include <../boost/asio/thread_pool.hpp>
#include <../boost/lexical_cast.hpp>
#include <../boost/asio.hpp>
#include <fcntl.h>
#include <cstdlib>
#include <chrono>
#include <sys/mman.h>
#include <sys/stat.h>
#include <fstream>
#include <future>
#include <mutex>
 
typedef __m128i node_t;
constexpr size_t leaf_size = 1;
typedef __m128i leaf_type;
typedef std::array<leaf_type, leaf_size> leaf_t;
size_t communication_cost = 0;
#include "bitutils.h"
#include "block.h" 
#include "prg_aes_impl.h"
#include "filesio.h"

using boost::asio::ip::tcp;
using socket_t = boost::asio::ip::tcp::socket;
 
using namespace dpf;  // The namespace is found in bitutils.h

#include "mpc.h"
#include "network.h"
#include "dpfgen.h"
#include "share-conversion.h"

int main(int argc, char * argv[])
{     
   boost::asio::io_context io_context;
   std::string addr = "127.0.0.1";  
   const std::string host1 = (argc < 2) ? "127.0.0.1" : argv[1];
   const std::string host2 = (argc < 3) ? "127.0.0.1" : argv[2];
   const size_t n_threads = atoi(argv[3]);
   const size_t expo = atoi(argv[4]);
   const size_t db_nitems = 1ULL << expo;
   const size_t maxRAM = atoi(argv[5]);
   //std::cout << "n_threads = " << n_threads << std::endl;
 


   const size_t number_of_sockets = 5 * n_threads;

   std::vector<socket_t> socketsPb, socketsP2;
   std::vector<int> ports, ports2_1, ports2_0;
 
   bool party;

    /* The function make_connections appears in network.h */
   make_connections(party, host1, host2,  io_context, socketsPb, socketsP2, ports,  ports2_1, ports2_0, number_of_sockets);
 
      size_t RAM_needed_per_thread = 164 * db_nitems;
      std::cout << "RAM needed = " << n_threads*RAM_needed_per_thread << " bytes = " << n_threads*RAM_needed_per_thread/1073741824 << " GiB" << std::endl;
      std::cout << "RAM needed per thread = " << RAM_needed_per_thread << " bytes = " << (RAM_needed_per_thread>>30) << " GiB" << std::endl;
      size_t thread_per_batch = std::floor(double(maxRAM<<30)/RAM_needed_per_thread);
      if (thread_per_batch > n_threads) {
	thread_per_batch = n_threads;
      }
      std::cout << "thread_per_batch = " << thread_per_batch << std::endl;
      if (thread_per_batch < 1) {
       std::cout << "You need more RAM" << std::endl;
       exit(0);
      }
      size_t n_batches = std::ceil(double(n_threads)/thread_per_batch);
      std::cout << "n_batches = " << n_batches << std::endl;

   uint8_t ** target_share_read = new uint8_t*[thread_per_batch];

   generate_random_targets(target_share_read,  thread_per_batch, party, expo);
   
   AES_KEY aeskey;
   
   auto start = std::chrono::steady_clock::now(); 
   

   __m128i * final_correction_word = (__m128i *) std::aligned_alloc(sizeof(__m256i), thread_per_batch * sizeof(__m128i));
   __m128i ** output = (__m128i ** ) malloc(sizeof(__m128i *) * thread_per_batch);
    
   int8_t  ** flags  = (int8_t ** ) malloc(sizeof(uint8_t *) * thread_per_batch);
     
   for(size_t j = 0; j < thread_per_batch; ++j)
   {
     output[j] = (__m128i *)std::aligned_alloc(sizeof(node_t), db_nitems * sizeof(__m128i));
     flags[j]  = (int8_t *)std::aligned_alloc(sizeof(node_t), db_nitems * sizeof(uint8_t));
   }
     
 
   boost::asio::thread_pool pool_share_conversion(thread_per_batch);
    

    
    // The following function call creates and evaluates DPFs at target_share_read[j] for j \in \{0, \ldots, n_threads}
    // the flag vectors are stored in flags
    // the leaves are stored in output
    // the final correctionword is stored in final_correction_word

  dpfP2 * dpf_instance = (dpfP2 * ) malloc (sizeof(dpfP2) * n_threads);


  cw_construction computecw_array;

 
     boost::asio::read(socketsP2[0], boost::asio::buffer(&computecw_array, sizeof(computecw_array)));
     #ifdef VERBOSE
      std::cout << "computecw_array.rand_b: " << computecw_array.rand_b[0] << " " << computecw_array.rand_b[1] << std::endl;
     #endif

      /* The function create_dpfs appears in dpf-gen.h*/
      bool reading = true;

      

     size_t *thread_communication_costs = new size_t[thread_per_batch];
     for(size_t iter = 0; iter < n_batches; ++iter)
     { 
        if (n_batches > 1) {
          printf("Starting create_dpfs batch %lu / %lu\n", iter+1, n_batches);
        }
        boost::asio::thread_pool pool(thread_per_batch);
        for(size_t j = 0; j < thread_per_batch; ++j)
        {
	  thread_communication_costs[j] = 0; 
	  boost::asio::post(pool,
	    std::bind(create_dpfs, reading,  db_nitems, std::ref(aeskey),
		target_share_read[j], std::ref(socketsPb), std::ref(socketsP2),
		0, db_nitems-1, output[j],  flags[j],
		std::ref(final_correction_word[j]), computecw_array,
		std::ref(dpf_instance), party, 5 * j, j,
		std::ref(thread_communication_costs[j])));
        }    
        pool.join();
        for(size_t j = 0; j < thread_per_batch; ++j) {
	  communication_cost += thread_communication_costs[j];
	}
     }
     delete[] thread_communication_costs;
      
     boost::asio::write(socketsP2[0], boost::asio::buffer(dpf_instance, n_threads * sizeof(dpfP2))); // do this in parallel.
     communication_cost += (n_threads * sizeof(dpfP2));
 
   #ifdef DEBUG

    for(size_t j = 0; j < n_threads; ++j)
    {
      std::cout << "n_threads = " << j << std::endl;
      for(size_t i = 0; i < db_nitems; ++i)
       {
        int8_t flags_reconstruction;
          boost::asio::write(socketsPb[0], boost::asio::buffer(&flags[j][i], sizeof(flags[j][i])));
          boost::asio::read(socketsPb[0], boost::asio::buffer(&flags_reconstruction, sizeof(flags_reconstruction)));
          flags_reconstruction -= flags[j][i];
          if(flags_reconstruction != 0) std::cout << i << " (flag) ---> " << (int) flags_reconstruction << std::endl;

          int64_t output_reconstruction;
          boost::asio::write(socketsPb[0], boost::asio::buffer(&output[j][i][0], sizeof(output[j][i][0])));
          boost::asio::read(socketsPb[0], boost::asio::buffer(&output_reconstruction, sizeof(output_reconstruction)));
          output_reconstruction -= output[j][i][0];
          if(output_reconstruction != 0) std::cout << i << " (output) ---> " << output_reconstruction << std::endl;
        }

        int64_t final_correction_word_reconstruction = 0;
        boost::asio::write(socketsPb[0], boost::asio::buffer(&final_correction_word[j][0], sizeof(final_correction_word[j][0])));
        boost::asio::read(socketsPb[0], boost::asio::buffer(&final_correction_word_reconstruction, sizeof(final_correction_word_reconstruction)));
        final_correction_word_reconstruction = final_correction_word_reconstruction + final_correction_word[j][0];
        std::cout << "final_correction_word_reconstruction = " << final_correction_word_reconstruction << std::endl << std::endl;
     }
    #endif
 
    /* 
     leaves is a additive shares of the outputs (leaves of the DPF)
     leafbits is the additive shares of flag bits of the DPFs
    */
   int64_t ** leaves = (int64_t ** ) malloc(sizeof(int64_t *) * thread_per_batch);
   int64_t ** leafbits  = (int64_t ** ) malloc(sizeof(int64_t *) * thread_per_batch); 
   for(size_t j = 0; j < thread_per_batch; ++j)
   {
    leaves[j] = (int64_t *)std::aligned_alloc(sizeof(node_t), db_nitems * sizeof(int64_t));
    leafbits[j]  = (int64_t *)std::aligned_alloc(sizeof(node_t), db_nitems * sizeof(int64_t));
   }



    /* The function convert_shares appears in share-conversion.h */
   for(size_t j = 0; j < thread_per_batch; ++j)
   {
     boost::asio::post(pool_share_conversion,  std::bind(convert_shares, j, output, flags, n_threads, db_nitems, final_correction_word, 	leaves, leafbits, 
                                                          std::ref(socketsPb), std::ref(socketsP2), party));	 	
   }
    
    pool_share_conversion.join();

    boost::asio::thread_pool pool_xor_to_additive(thread_per_batch); 

    int64_t *additve_shares = new int64_t[thread_per_batch]; 
    for(size_t j = 0; j < thread_per_batch; ++j)
    {
     boost::asio::post(pool_xor_to_additive, std::bind(xor_to_additive, party, target_share_read[j], std::ref(socketsPb[j]), std::ref(socketsP2[j]), expo, std::ref(additve_shares[j])));
    }

    pool_xor_to_additive.join();
    
 
    
    /* For the artifact, don't actually write these in order to not use very
     * large amounts of storage

    for(size_t i = 0; i < thread_per_batch; ++i)
    {
     write_evalfull_outs_into_a_file(party, i, db_nitems, flags[i],  leaves[i], final_correction_word[i], additve_shares[i]); 
    }
    */

    auto end = std::chrono::steady_clock::now();
    std::chrono::duration<double> elapsed_seconds = end-start;
    //std::cout << "time to generate and evaluate " << n_threads << " dpfs of size 2^" << atoi(argv[4]) << " is: " << elapsed_seconds.count() << "s\n";
    std::cout << "WallClockTime: "  << elapsed_seconds.count() << std::endl;
 
    // std::cout << "elapsed_ FIO = " << elapsed_seconds.count() << std::endl;

    std::cout << "CommunicationCost: " << communication_cost << " bytes" << std::endl;
    #ifdef VERBOSE
     for(size_t j = 0; j < n_threads; ++j)
     {
      int64_t add_;
      boost::asio::write(socketsPb[0], boost::asio::buffer(&additve_shares[j], sizeof(additve_shares[j])));
      boost::asio::read(socketsPb[0], boost::asio::buffer(&add_, sizeof(add_)));
      add_ = add_ + additve_shares[j];
      std::cout << "add_ = " << add_ << std::endl;
     }
    #endif
    delete[] additve_shares;
    
    return 0;
}