PerfIVFFlat.cu

/**
 * Copyright (c) Facebook, Inc. and its affiliates.
 *
 * This source code is licensed under the MIT license found in the
 * LICENSE file in the root directory of this source tree.
 */


#include "../../IndexIVFFlat.h"
#include "../../index_io.h"
#include "../../utils.h"

#include "../GpuIndexIVFFlat.h"
#include "IndexWrapper.h"
#include "../test/TestUtils.h"
#include "../utils/DeviceTensor.cuh"
#include "../utils/DeviceUtils.h"
#include "../utils/HostTensor.cuh"
#include "../utils/Timer.h"
#include <gflags/gflags.h>
#include <map>
#include <memory>
#include <vector>

#include <cuda_profiler_api.h>

DEFINE_int32(nprobe, 5, "number of coarse centroids to probe");
DEFINE_int32(k, 3, "final number of closest results returned");
DEFINE_int32(num_queries, 3, "number of query vectors");
DEFINE_string(in, "/home/jhj/local/index.out", "index file for input");
DEFINE_bool(diff, true, "show exact distance + index output discrepancies");
DEFINE_bool(use_float16, false, "use encodings in float16");
DEFINE_bool(use_float16_coarse, false, "coarse quantizer in float16");
DEFINE_int64(seed, -1, "specify random seed");
DEFINE_int32(num_gpus, 1, "number of gpus to use");
DEFINE_int32(index, 2, "0 = no indices on GPU; 1 = 32 bit, 2 = 64 bit on GPU");

using namespace faiss::gpu;

int main(int argc, char** argv) {
  gflags::ParseCommandLineFlags(&argc, &argv, true);

  cudaProfilerStop();

  auto seed = FLAGS_seed != -1L ? FLAGS_seed : time(nullptr);
  printf("using seed %ld\n", seed);

  auto numQueries = FLAGS_num_queries;

  auto index = std::unique_ptr<faiss::IndexIVFFlat>(
    dynamic_cast<faiss::IndexIVFFlat*>(faiss::read_index(FLAGS_in.c_str())));
  FAISS_ASSERT((bool) index);
  index->nprobe = FLAGS_nprobe;

  auto dim = index->d;

  printf("Database: dim %d num vecs %ld\n", dim, index->ntotal);
  printf("Coarse centroids: %ld\n", index->quantizer->ntotal);
  printf("L2 lookup: %d queries, nprobe %d, total k %d\n",
         numQueries, FLAGS_nprobe, FLAGS_k);
  printf("float16 coarse quantizer %s\n",
         FLAGS_use_float16_coarse ? "enabled" : "disabled");
  printf("float16 encoding %s\n",
         FLAGS_use_float16 ? "enabled" : "disabled");

  // Convert to GPU index
  printf("Copying index to %d GPU(s)...\n", FLAGS_num_gpus);

  auto initFn = [&index](faiss::gpu::GpuResources* res, int dev) ->
    std::unique_ptr<faiss::gpu::GpuIndexIVFFlat> {
    GpuIndexIVFFlatConfig config;
    config.device = dev;
    config.indicesOptions = (faiss::gpu::IndicesOptions) FLAGS_index;
    config.flatConfig.useFloat16 = FLAGS_use_float16_coarse;
    config.useFloat16IVFStorage = FLAGS_use_float16;

    auto p = std::unique_ptr<faiss::gpu::GpuIndexIVFFlat>(
      new faiss::gpu::GpuIndexIVFFlat(res,
                                      index->d,
                                      index->nlist,
                                      index->metric_type,
                                      config));
    p->copyFrom(index.get());
    return p;
  };

  IndexWrapper<faiss::gpu::GpuIndexIVFFlat> gpuIndex(FLAGS_num_gpus, initFn);
  gpuIndex.setNumProbes(FLAGS_nprobe);
  printf("copy done\n");

  // Build query vectors
  HostTensor<float, 2, true> cpuQuery({numQueries, dim});
  faiss::float_rand(cpuQuery.data(), cpuQuery.numElements(), seed);

  // Time faiss CPU
  HostTensor<float, 2, true> cpuDistances({numQueries, FLAGS_k});
  HostTensor<faiss::Index::idx_t, 2, true> cpuIndices({numQueries, FLAGS_k});

  float cpuTime = 0.0f;

  {
    CpuTimer timer;
    index->search(numQueries,
                  cpuQuery.data(),
                  FLAGS_k,
                  cpuDistances.data(),
                  cpuIndices.data());

    cpuTime = timer.elapsedMilliseconds();
  }

  printf("CPU time %.3f ms\n", cpuTime);

  HostTensor<float, 2, true> gpuDistances({numQueries, FLAGS_k});
  HostTensor<faiss::Index::idx_t, 2, true> gpuIndices({numQueries, FLAGS_k});

  CUDA_VERIFY(cudaProfilerStart());
  faiss::gpu::synchronizeAllDevices();

  float gpuTime = 0.0f;

  // Time GPU
  {
    CpuTimer timer;

    gpuIndex.getIndex()->search(cpuQuery.getSize(0),
                                     cpuQuery.data(),
                                     FLAGS_k,
                                     gpuDistances.data(),
                                     gpuIndices.data());

    // There is a device -> host copy above, so no need to time
    // additional synchronization with the GPU
    gpuTime = timer.elapsedMilliseconds();
  }

  CUDA_VERIFY(cudaProfilerStop());
  printf("GPU time %.3f ms\n", gpuTime);

  compareLists(cpuDistances.data(), cpuIndices.data(),
               gpuDistances.data(), gpuIndices.data(),
               numQueries, FLAGS_k,
               "", true, FLAGS_diff, false);

  CUDA_VERIFY(cudaDeviceSynchronize());
  // printf("\ncudaMalloc usage %zd\n",
  //        resources.getMemoryManager().getHighWaterCudaMalloc());

  return 0;
}