vector<fbgemm::float16> src_mat(resultPtr->matSize());
      fbgemm::float16* pmat = resultPtr->pmat();
      memcpy(
          src_mat.data(), pmat, resultPtr->matSize() * sizeof(fbgemm::float16));
      resultPtr->unpackFromSrc(fbgemm::matrix_op_t::Transpose, src_mat.data());
    }

    auto* Y =

            

Reported by FlawFinder.

                }
}

TEST(BoundsInference, _1) {
  // Verify that bounds inference works for the following example:
  // for i in 0..100:
  //   b[i] = a[i]
  // For this loop bounds inference should yield the following:
  // {{b, kStore, 0, 99}, {a, kLoad, 0, 99}}

            

Reported by Cppcheck.

              
using namespace torch::jit::tensorexpr;

TEST(ATen, _cast_Float) {
  KernelScope kernel_scope;
  const int kTotalSize = 128;
  Placeholder a_buf(BufHandle("A", {ExprHandle(kTotalSize)}, kInt));
  Placeholder b_buf(BufHandle("B", {ExprHandle(kTotalSize)}, kFloat));


            

Reported by Cppcheck.

              // End to end training loop test in C++ so that we can run LSAN on this test to
// catch memory leaks. Enabling LSAN with python multiprocessing has been
// challenging and we don't have a good solution yet.
TEST_F(TestE2ETensorPipe, TestTrainingLoop) {
  runTrainingLoop();
  // Ensure the tensorpipe internal state is cleared up.
  auto tensorpipeAgent = std::static_pointer_cast<TensorPipeAgent>(rpcAgent);

  // Shutdown RPC agent for all RPCs to clean up.

            

Reported by Cppcheck.

              }
} // namespace

TEST(MobileProfiler, ModuleHierarchy) {
  std::string filePath(__FILE__);
  auto testModelFile = filePath.substr(0, filePath.find_last_of("/\\") + 1);
  testModelFile.append("to_be_profiled_module.ptl");

  std::vector<IValue> inputs;

            

Reported by Cppcheck.

              using namespace std;

int main(int argc, const char* argv[]) {
  int LEN = argc > 1 ? atoi(argv[1]) : 65536;

  vector<uint8_t> a(LEN), b(LEN), c_avx2(LEN), c_avx512(LEN);
  for (int i = 0; i < LEN; ++i) {
    a[i] = i % 256;
    b[i] = (i * 2) % 256;

            

Reported by FlawFinder.

                for (int i = 0; i < N; ++i) {
    int origIdx = i * K;
    int permuteIdx = indices_data[i] * K;
    std::memcpy(Y_data + origIdx, X_data + permuteIdx, K * sizeof(T));
  }

  // Even if there is a pre-chosen quantization parameters for the output,
  // it is ignored because batch permutation output quantization should be same
  // as the input.

            

Reported by FlawFinder.