Initialize variables (pytorch#4447)

Summary:
Pull Request resolved: pytorch#4447

X-link: facebookresearch/FBGEMM#1510

There are some changes:
1. ensure that variables are initialised by `cppcoreguidelines-init-variables` check.
2. enable "Wunused-variable" and "Wmaybe-uninitialized" warnings.
3. fix all warnings.

Pull Request resolved: pytorch#4420

Reviewed By: gchalump

Differential Revision: D77812012

Pulled By: q10

fbshipit-source-id: 9addd48d713dc88c2399a540e335e3d7bc748460

Author: cyyever

.clang-tidy

@@ -7,6 +7,7 @@ InheritParentConfig: true
 # @nolint
 Checks: '
 -*,
+cppcoreguidelines-init-variables,
 bugprone-argument-comment,
 misc-use-internal-linkage,
 modernize*,

CMakeLists.txt

@@ -208,8 +208,11 @@ else(MSVC)
   string(APPEND CMAKE_CXX_FLAGS " -Wunknown-pragmas")
   string(APPEND CMAKE_CXX_FLAGS " -Wimplicit-fallthrough")
   string(APPEND CMAKE_CXX_FLAGS " -Wno-strict-aliasing")
+  string(APPEND CMAKE_CXX_FLAGS " -Wunused-variable")
   if("${CMAKE_CXX_COMPILER_ID}" MATCHES "Clang" AND CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 17.0.0)
     string(APPEND CMAKE_CXX_FLAGS " -Wno-vla-cxx-extension")
+  elseif(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
+    string(APPEND CMAKE_CXX_FLAGS " -Wmaybe-uninitialized")
   endif()
   target_compile_options(fbgemm_avx2 PRIVATE
     "-m64" "-mavx2" "-mf16c" "-mfma")

bench/AlignedVec.h

@@ -107,10 +107,9 @@ class aligned_allocator {
 
     // Mallocator wraps malloc().
     void* pv = nullptr;
-    int ret;
+    int ret = 0;
 #ifdef _MSC_VER
     pv = _aligned_malloc(n * sizeof(T), Alignment);
-    ret = 0;
 #else
     ret = posix_memalign(&pv, Alignment, n * sizeof(T));
 #endif

bench/BenchUtils.cc

@@ -150,9 +150,8 @@ aligned_vector<float> getRandomSparseVector(
   std::sort(sorted_res.begin(), sorted_res.end());
   int32_t numZeros =
       size - static_cast<int32_t>(std::round(size * fractionNonZeros));
-  float thr;
   if (numZeros) {
-    thr = sorted_res[numZeros - 1];
+    float thr = sorted_res[numZeros - 1];
 
     for (auto& f : res) {
       if (f <= thr) {

bench/BenchUtils.h

@@ -28,6 +28,7 @@
 
 #ifdef _OPENMP
 #include <omp.h>
+#include <cmath>
 #endif
 
 #ifdef USE_MKL
@@ -136,8 +137,6 @@ double measureWithWarmup(
   {
 #endif
     for (int i = 0; i < measuredIterations; ++i) {
-      std::chrono::time_point<std::chrono::high_resolution_clock> start, end;
-
       const auto thread_id = useOpenMP ? fbgemm_get_thread_num() : 0;
 
       if (thread_id == 0) {
@@ -149,7 +148,7 @@ double measureWithWarmup(
 #pragma omp barrier
       }
 #endif
-      start = std::chrono::high_resolution_clock::now();
+      auto start = std::chrono::high_resolution_clock::now();
 
       fn();
 
@@ -159,7 +158,7 @@ double measureWithWarmup(
       }
 #endif
 
-      end = std::chrono::high_resolution_clock::now();
+      auto end = std::chrono::high_resolution_clock::now();
       auto dur =
           std::chrono::duration_cast<std::chrono::nanoseconds>(end - start);
 
@@ -256,7 +255,6 @@ void performance_test(
 #endif
 
   std::string type;
-  double gflops, gbs, ttot;
   for (auto s : shapes) {
     int m = s[0];
     int n = s[1];
@@ -266,6 +264,7 @@ void performance_test(
     aligned_vector<int> Aint(m * k);
     randFill(Aint, 0, 4);
     std::vector<aligned_vector<float>> A;
+    A.reserve(num_instances);
     for (int i = 0; i < num_instances; ++i) {
       A.emplace_back(Aint.begin(), Aint.end());
     }
@@ -321,6 +320,7 @@ void performance_test(
 
     double nflops = 2.0 * m * n * k;
     double nbytes = 4.0 * m * k + sizeof(btype) * 1.0 * k * n + 4.0 * m * n;
+    double gflops = 0, gbs = 0, ttot = 0.0;
 
     // warm up MKL and fbgemm
     // check correctness at the same time

bench/EmbeddingIndexRemappingBenchmark.cc

@@ -135,9 +135,9 @@ static int run_benchmark(
 }
 
 int main() {
-  int batch_size;
-  int num_rows;
-  int average_len;
+  int batch_size = 0;
+  int num_rows = 0;
+  int average_len = 0;
 
   vector<vector<int>> inputs(GetInputs_());
 

bench/EmbeddingSpMDMBenchmark.cc

@@ -12,7 +12,6 @@
 #endif
 #include <algorithm>
 #include <cassert>
-#include <chrono>
 #include <cmath>
 #include <cstdint>
 #include <iomanip>

bench/EmbeddingSpMDMNBit2Benchmark.cc

@@ -12,7 +12,6 @@
 #endif
 #include <algorithm>
 #include <cassert>
-#include <chrono>
 #include <cmath>
 #include <cstdint>
 #include <iomanip>

bench/EmbeddingSpMDMNBitBenchmark.cc

@@ -492,10 +492,10 @@ static int run_benchmark(
 }
 
 int main() {
-  int batch_size;
-  int num_rows;
-  int embedding_dim;
-  int average_len;
+  int batch_size = 0;
+  int num_rows = 0;
+  int embedding_dim = 0;
+  int average_len = 0;
 
   vector<vector<int>> inputs(GetInputs_());
 

bench/EmbeddingSpMDMNBitRowWiseSparseBenchmark.cc

@@ -325,10 +325,10 @@ static int run_benchmark(
 }
 
 int main() {
-  int batch_size;
-  int num_rows;
-  int embedding_dim;
-  int average_len;
+  int batch_size = 0;
+  int num_rows = 0;
+  int embedding_dim = 0;
+  int average_len = 0;
 
   vector<vector<int>> inputs(GetInputs_());