dense_matvec.hpp

Go to the documentation of this file.

#pragma once
 
namespace monolish {
 
namespace {
// double ///////////////////
template <typename VEC1, typename VEC2>
void Dmatvec_core(const matrix::Dense<double> &A, const VEC1 &x, VEC2 &y) {
  Logger &logger = Logger::get_instance();
  logger.func_in(monolish_func);
 
  // err, M = MN * N
  assert(A.get_row() == y.size());
  assert(A.get_col() == x.size());
  assert(util::is_same_device_mem_stat(A, x, y));
 
  const double *xd = x.data();
  double *yd = y.data();
  const double *vald = A.val.data();
  const size_t m = A.get_row();
  const size_t n = A.get_col();
  const double alpha = 1.0;
  const double beta = 0.0;
  const size_t xoffset = x.get_offset();
  const size_t yoffset = y.get_offset();
 
  if (A.get_device_mem_stat() == true) {
#if MONOLISH_USE_GPU
    cublasHandle_t h;
    internal::check_CUDA(cublasCreate(&h));
#pragma omp target data use_device_ptr(xd, yd, vald)
    {
      // cublas is col major
      internal::check_CUDA(cublasDgemv(h, CUBLAS_OP_T, n, m, &alpha, vald, n,
                                       xd + xoffset, 1, &beta, yd + yoffset,
                                       1));
    }
    cublasDestroy(h);
#else
    throw std::runtime_error("error USE_GPU is false, but gpu_status == true");
#endif
  } else {
    cblas_dgemv(CblasRowMajor, CblasNoTrans, m, n, alpha, vald, n, xd + xoffset,
                1, beta, yd + yoffset, 1);
  }
 
  logger.func_out();
}
 
// float ///////////////////
template <typename VEC1, typename VEC2>
void Smatvec_core(const matrix::Dense<float> &A, const VEC1 &x, VEC2 &y) {
  Logger &logger = Logger::get_instance();
  logger.func_in(monolish_func);
 
  // err, M = MN * N
  assert(A.get_row() == y.size());
  assert(A.get_col() == x.size());
  assert(util::is_same_device_mem_stat(A, x, y));
 
  const float *xd = x.data();
  float *yd = y.data();
  const float *vald = A.val.data();
  const size_t n = A.get_row();
  const size_t m = A.get_col();
  const float alpha = 1.0;
  const float beta = 0.0;
  const size_t xoffset = x.get_offset();
  const size_t yoffset = y.get_offset();
 
  if (A.get_device_mem_stat() == true) {
#if MONOLISH_USE_GPU
    cublasHandle_t h;
    internal::check_CUDA(cublasCreate(&h));
#pragma omp target data use_device_ptr(xd, yd, vald)
    {
      // cublas is col major
      internal::check_CUDA(cublasSgemv(h, CUBLAS_OP_T, m, n, &alpha, vald, m,
                                       xd + xoffset, 1, &beta, yd + yoffset,
                                       1));
    }
    cublasDestroy(h);
#else
    throw std::runtime_error("error USE_GPU is false, but gpu_status == true");
#endif
  } else {
    cblas_sgemv(CblasRowMajor, CblasNoTrans, n, m, alpha, vald, m, xd + xoffset,
                1, beta, yd + yoffset, 1);
  }
 
  logger.func_out();
}
} // namespace
 
} // namespace monolish