Add C++ examples, C++ CI (#10)

* Add C++ examples, C++ CI

* Remove extraneous entry from CMake

* Use g++-14 for C++ CI

* Minor fix

* Verbose build for debugging

* Use gcc 13

* GCC 12

.github/workflows/ci.yml

@@ -33,9 +33,9 @@ jobs:
     - name: CMake
       run: |
         sudo apt-get update
-        sudo apt-get install libhdf5-dev
-        cmake -B build
+        sudo apt-get install libhdf5-dev g++-12
+        CXX=g++-12 cmake -B build
     - name: Build
-      run: make -C build -j `nproc`
+      run: VERBOSE=true make -C build -j `nproc`
     - name: Test
       run: ctest --test-dir ./build/test/bash

CMakeLists.txt

@@ -5,6 +5,8 @@ cmake_policy(SET CMP0079 NEW)
 
 set(CMAKE_C_STANDARD 11)
 
+set(CMAKE_CXX_STANDARD 20)
+
 set(CMAKE_C_FLAGS "-O3 -march=native")
 
 add_subdirectory(include)

examples/CMakeLists.txt

@@ -14,3 +14,5 @@ add_example(bsp-ls)
 add_example(benchmark_read)
 add_example(benchmark_read_parallel)
 add_example(benchmark_write)
+
+add_subdirectory(cpp)

examples/cpp/CMakeLists.txt

@@ -0,0 +1,15 @@
+function(add_example example_name)
+  add_executable(${example_name}-cpp ${example_name}.cpp)
+  target_link_libraries(${example_name}-cpp binsparse-rc)
+endfunction()
+
+add_example(simple_matrix_write)
+add_example(simple_matrix_read)
+add_example(simple_read)
+add_example(simple_write)
+add_example(mtx2bsp)
+add_example(bsp2mtx)
+add_example(check_equivalence)
+add_example(bsp-ls)
+add_example(benchmark_read)
+add_example(benchmark_write)

examples/cpp/benchmark_read.cpp

@@ -0,0 +1,123 @@
+#include <binsparse/binsparse.h>
+#include <stdlib.h>
+#include <time.h>
+
+double gettime() {
+  struct timespec time;
+  clock_gettime(CLOCK_MONOTONIC, &time);
+  return ((double) time.tv_sec) + ((double) 1e-9) * time.tv_nsec;
+}
+
+int compar(const void* a, const void* b) {
+  double x = *((const double*) a);
+  double y = *((const double*) b);
+
+  double diff = x - y;
+
+  if (diff > 0) {
+    return 1;
+  } else if (diff < 0) {
+    return -1;
+  } else {
+    return 0;
+  }
+}
+
+double compute_variance(double* x, size_t n) {
+  double sum = 0;
+
+  for (size_t i = 0; i < n; i++) {
+    sum += x[i];
+  }
+
+  double mean = sum / n;
+
+  double sum_of_squares = 0;
+  for (size_t i = 0; i < n; i++) {
+    sum_of_squares += (x[i] - mean) * (x[i] - mean);
+  }
+
+  return sum_of_squares / (n - 1);
+}
+
+void flush_cache() {
+#ifdef __APPLE__
+  system("bash -c \"sync && sudo purge\"");
+#elif __linux__
+  system("bash -c \"sync\" && sudo echo 3 > /proc/sys/vm/drop_caches");
+#else
+  static_assert(false);
+#endif
+}
+
+int main(int argc, char** argv) {
+  if (argc < 2) {
+    fprintf(stderr, "usage: ./benchmark_read [file_name.h5]\n");
+    return 1;
+  }
+
+  char* file_name = argv[1];
+
+  printf("Opening %s\n", file_name);
+
+  const int num_trials = 1;
+
+  double durations[num_trials];
+
+  size_t nbytes = 0;
+
+  // To flush the filesystem cache before each trial, change to `true`.
+  bool cold_cache = false;
+
+  for (size_t i = 0; i < num_trials; i++) {
+    if (cold_cache) {
+      flush_cache();
+    }
+    double begin = gettime();
+    bsp_matrix_t mat = bsp_read_matrix(file_name, NULL);
+    double end = gettime();
+    durations[i] = end - begin;
+    nbytes = bsp_matrix_nbytes(mat);
+    bsp_destroy_matrix_t(mat);
+  }
+
+  printf("[");
+  for (size_t i = 0; i < num_trials; i++) {
+    printf("%lf", durations[i]);
+    if (i + 1 < num_trials) {
+      printf(", ");
+    }
+  }
+  printf("]\n");
+
+  qsort(durations, num_trials, sizeof(double), compar);
+
+  double variance = compute_variance(durations, num_trials);
+
+  double median_time = durations[num_trials / 2];
+
+  printf("Read file in %lf seconds\n", median_time);
+
+  if (num_trials > 1) {
+    printf("Variance is %lf seconds, standard devication is %lf seconds\n",
+           variance, sqrt(variance));
+  }
+
+  double gbytes = ((double) nbytes) / 1024 / 1024 / 1024;
+  double gbytes_s = gbytes / median_time;
+
+  printf("Achieved %lf GiB/s\n", gbytes_s);
+
+  printf("[");
+  for (size_t i = 0; i < num_trials; i++) {
+    printf("%lf", durations[i]);
+    if (i + 1 < num_trials) {
+      printf(", ");
+    }
+  }
+  printf("]\n");
+
+  printf("FORPARSER: %s,%lf,%lf\n", file_name, median_time, gbytes_s);
+
+  return 0;
+}

examples/cpp/benchmark_write.cpp

@@ -0,0 +1,170 @@
+#include <binsparse/binsparse.h>
+#include <stdlib.h>
+#include <time.h>
+
+double gettime() {
+  struct timespec time;
+  clock_gettime(CLOCK_MONOTONIC, &time);
+  return ((double) time.tv_sec) + ((double) 1e-9) * time.tv_nsec;
+}
+
+int compar(const void* a, const void* b) {
+  double x = *((const double*) a);
+  double y = *((const double*) b);
+
+  double diff = x - y;
+
+  if (diff > 0) {
+    return 1;
+  } else if (diff < 0) {
+    return -1;
+  } else {
+    return 0;
+  }
+}
+
+double compute_variance(double* x, size_t n) {
+  double sum = 0;
+
+  for (size_t i = 0; i < n; i++) {
+    sum += x[i];
+  }
+
+  double mean = sum / n;
+
+  double sum_of_squares = 0;
+  for (size_t i = 0; i < n; i++) {
+    sum_of_squares += (x[i] - mean) * (x[i] - mean);
+  }
+
+  return sum_of_squares / (n - 1);
+}
+
+void flush_cache() {
+#ifdef __APPLE__
+  system("bash -c \"sync && sudo purge\"");
+#elif __linux__
+  system("bash -c \"sync\" && sudo echo 3 > /proc/sys/vm/drop_caches");
+#else
+  static_assert(false);
+#endif
+}
+
+void flush_writes() {
+#ifdef __APPLE__
+  system("bash -c \"sync\"");
+#elif __linux__
+  system("bash -c \"sync\"");
+#else
+  static_assert(false);
+#endif
+}
+
+void delete_file(const char* file_name) {
+  char command[2048];
+  snprintf(command, 2047, "rm %s", file_name);
+  system(command);
+}
+
+int main(int argc, char** argv) {
+  if (argc < 2) {
+    fprintf(stderr,
+            "usage: ./benchmark_read [file_name.h5] [scratch_space] [optional: "
+            "compression_level]\n");
+    return 1;
+  }
+
+  char* file_name = argv[1];
+  char* scratch_space = argv[2];
+
+  int compression_level = 0;
+
+  if (argc >= 4) {
+    compression_level = atoi(argv[3]);
+  }
+
+  printf("Opening %s\n", file_name);
+
+  const int num_trials = 1;
+
+  double durations[num_trials];
+
+  bsp_matrix_t mat = bsp_read_matrix(file_name, NULL);
+  size_t nbytes = bsp_matrix_nbytes(mat);
+
+  char output_filename[2048];
+  strncpy(output_filename, scratch_space, 2047);
+  strncpy(output_filename + strlen(scratch_space), "/benchmark_write_file_n.h5",
+          2047 - strlen(scratch_space));
+
+  // Current output name logic does not do much.
+  assert(num_trials <= 10);
+
+  // To flush the filesystem cache before each trial, change to `true`.
+  bool cold_cache = false;
+
+  // To flush each write to the filesystem and include this in the timing,
+  // change to `true`.
+  bool flush_each_write = true;
+
+  for (size_t i = 0; i < num_trials; i++) {
+    if (cold_cache) {
+      flush_cache();
+    }
+
+    output_filename[strlen(scratch_space) + 21] = '0' + i;
+    printf("Writing to file %s\n", output_filename);
+
+    double begin = gettime();
+    bsp_write_matrix(output_filename, mat, NULL, NULL, compression_level);
+
+    if (flush_each_write) {
+      flush_writes();
+    }
+
+    double end = gettime();
+    durations[i] = end - begin;
+
+    delete_file(output_filename);
+  }
+
+  printf("[");
+  for (size_t i = 0; i < num_trials; i++) {
+    printf("%lf", durations[i]);
+    if (i + 1 < num_trials) {
+      printf(", ");
+    }
+  }
+  printf("]\n");
+
+  qsort(durations, num_trials, sizeof(double), compar);
+
+  double variance = compute_variance(durations, num_trials);
+
+  double median_time = durations[num_trials / 2];
+
+  printf("Wrote file in %lf seconds\n", median_time);
+
+  if (num_trials > 1) {
+    printf("Variance is %lf seconds, standard devication is %lf seconds\n",
+           variance, sqrt(variance));
+  }
+
+  double gbytes = ((double) nbytes) / 1024 / 1024 / 1024;
+  double gbytes_s = gbytes / median_time;
+
+  printf("Achieved %lf GiB/s\n", gbytes_s);
+
+  printf("[");
+  for (size_t i = 0; i < num_trials; i++) {
+    printf("%lf", durations[i]);
+    if (i + 1 < num_trials) {
+      printf(", ");
+    }
+  }
+  printf("]\n");
+
+  printf("FORPARSER: %s,%lf,%lf\n", file_name, median_time, gbytes_s);
+
+  return 0;
+}