faiss/gpu/test/TestGpuSelect.cu

/**
 * Copyright (c) 2015-present, Facebook, Inc.
 * All rights reserved.
 *
 * This source code is licensed under the CC-by-NC license found in the
 * LICENSE file in the root directory of this source tree.
 */

// Copyright 2004-present Facebook. All Rights Reserved.

#include "../utils/DeviceUtils.h"
#include "../utils/BlockSelectKernel.cuh"
#include "../utils/WarpSelectKernel.cuh"
#include "../utils/HostTensor.cuh"
#include "../utils/DeviceTensor.cuh"
#include "../test/TestUtils.h"
#include <algorithm>
#include <gtest/gtest.h>
#include <sstream>
#include <unordered_map>
#include <vector>

int main(int argc, char** argv) {
  testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
}

void testForSize(int rows, int cols, int k, bool dir, bool warp) {
  std::vector<float> v = faiss::gpu::randVecs(rows, cols);
  faiss::gpu::HostTensor<float, 2, true> hostVal({rows, cols});

  for (int r = 0; r < rows; ++r) {
    for (int c = 0; c < cols; ++c) {
      hostVal[r][c] = v[r * cols + c];
    }
  }

  // row -> (val -> idx)
  std::unordered_map<int, std::vector<std::pair<int, float>>> hostOutValAndInd;
  for (int r = 0; r < rows; ++r) {
    std::vector<std::pair<int, float>> closest;

    for (int c = 0; c < cols; ++c) {
      closest.emplace_back(c, (float) hostVal[r][c]);
    }

    auto dirFalseFn =
      [](std::pair<int, float>& a, std::pair<int, float>& b) {
      return a.second < b.second;
    };
    auto dirTrueFn =
      [](std::pair<int, float>& a, std::pair<int, float>& b) {
      return a.second > b.second;
    };

    std::sort(closest.begin(), closest.end(), dir ? dirTrueFn : dirFalseFn);
    hostOutValAndInd.emplace(r, closest);
  }

  // Select top-k on GPU
  faiss::gpu::DeviceTensor<float, 2, true> gpuVal(hostVal, 0);
  faiss::gpu::DeviceTensor<float, 2, true> gpuOutVal({rows, k});
  faiss::gpu::DeviceTensor<int, 2, true> gpuOutInd({rows, k});

  if (warp) {
    faiss::gpu::runWarpSelect(gpuVal, gpuOutVal, gpuOutInd, dir, k, 0);
  } else {
    faiss::gpu::runBlockSelect(gpuVal, gpuOutVal, gpuOutInd, dir, k, 0);
  }

  // Copy back to CPU
  faiss::gpu::HostTensor<float, 2, true> outVal(gpuOutVal, 0);
  faiss::gpu::HostTensor<int, 2, true> outInd(gpuOutInd, 0);

  for (int r = 0; r < rows; ++r) {
    std::unordered_map<int, int> seenIndices;

    for (int i = 0; i < k; ++i) {
      float gpuV = outVal[r][i];
      float cpuV = hostOutValAndInd[r][i].second;

      EXPECT_EQ(gpuV, cpuV) <<
        "rows " << rows << " cols " << cols << " k " << k << " dir " << dir
                << " row " << r << " ind " << i;

      // If there are identical elements in a row that should be
      // within the top-k, then it is possible that the index can
      // differ, because the order in which the GPU will see the
      // equivalent values is different than the CPU (and will remain
      // unspecified, since this is affected by the choice of
      // k-selection algorithm that we use)
      int gpuInd = outInd[r][i];
      int cpuInd = hostOutValAndInd[r][i].first;

      // We should never see duplicate indices, however
      auto itSeenIndex = seenIndices.find(gpuInd);

      EXPECT_EQ(itSeenIndex, seenIndices.end()) <<
        "Row " << r << " user index " << gpuInd << " was seen at both " <<
        itSeenIndex->second << " and " << i;

      seenIndices[gpuInd] = i;

      if (gpuInd != cpuInd) {
        // Gather the values from the original data via index; the
        // values should be the same
        float gpuGatherV = hostVal[r][gpuInd];
        float cpuGatherV = hostVal[r][cpuInd];

        EXPECT_EQ(gpuGatherV, cpuGatherV) <<
          "rows " << rows << " cols " << cols << " k " << k << " dir " << dir
                  << " row " << r << " ind " << i << " source ind "
                  << gpuInd << " " << cpuInd;
      }
    }
  }
}

// General test
TEST(TestGpuSelect, test) {
  for (int i = 0; i < 10; ++i) {
    int rows = faiss::gpu::randVal(10, 100);
    int cols = faiss::gpu::randVal(1, 30000);
    int k = std::min(cols, faiss::gpu::randVal(1, 1024));
    bool dir = faiss::gpu::randBool();

    testForSize(rows, cols, k, dir, false);
  }
}

// Test for k = 1
TEST(TestGpuSelect, test1) {
  for (int i = 0; i < 5; ++i) {
    int rows = faiss::gpu::randVal(10, 100);
    int cols = faiss::gpu::randVal(1, 30000);
    bool dir = faiss::gpu::randBool();

    testForSize(rows, cols, 1, dir, false);
  }
}

// Test for where k = #cols exactly (we are returning all the values,
// just sorted)
TEST(TestGpuSelect, testExact) {
  for (int i = 0; i < 5; ++i) {
    int rows = faiss::gpu::randVal(10, 100);
    int cols = faiss::gpu::randVal(1, 1024);
    bool dir = faiss::gpu::randBool();

    testForSize(rows, cols, cols, dir, false);
  }
}

// General test
TEST(TestGpuSelect, testWarp) {
  for (int i = 0; i < 10; ++i) {
    int rows = faiss::gpu::randVal(10, 100);
    int cols = faiss::gpu::randVal(1, 30000);
    int k = std::min(cols, faiss::gpu::randVal(1, 1024));
    bool dir = faiss::gpu::randBool();

    testForSize(rows, cols, k, dir, true);
  }
}

// Test for k = 1
TEST(TestGpuSelect, test1Warp) {
  for (int i = 0; i < 5; ++i) {
    int rows = faiss::gpu::randVal(10, 100);
    int cols = faiss::gpu::randVal(1, 30000);
    bool dir = faiss::gpu::randBool();

    testForSize(rows, cols, 1, dir, true);
  }
}

// Test for where k = #cols exactly (we are returning all the values,
// just sorted)
TEST(TestGpuSelect, testExactWarp) {
  for (int i = 0; i < 5; ++i) {
    int rows = faiss::gpu::randVal(10, 100);
    int cols = faiss::gpu::randVal(1, 1024);
    bool dir = faiss::gpu::randBool();

    testForSize(rows, cols, cols, dir, true);
  }
}