topic Re:Crash for a simple 2D MKL FFT example when matrix size is not too larger in Intel® oneAPI Math Kernel Library

Crash for a simple 2D MKL FFT example when matrix size is not too larger

strngtn — Mon, 19 Oct 2020 17:57:41 GMT

Environment: MKL Version: 2020.2.254

Toolchain: Visual Studio Community 16.7.6, Host x64, Target x64

CMakeLists.txt:
add_executable(mkl_test_s main.cpp)
target_link_libraries(mkl_test_s mkl_intel_lp64.lib mkl_sequential.lib mkl_core.lib)

#include <chrono>
#include <iostream>
#include <random>

#include "mkl.h"
#include "mkl_dfti.h"

void init_array_random(int size, double *a) {
  std::uniform_real_distribution<double> unif(0, 1);
  std::random_device rd;
  std::default_random_engine re(rd());
  for (int i = 0; i < size; ++i) a[i] = unif(re);
}

void fft(double *datain, double *dataout, int sz) {
  DFTI_DESCRIPTOR_HANDLE handle;
  MKL_LONG status;
  MKL_LONG sizes[2] = {sz, sz};
  status = DftiCreateDescriptor(&handle, DFTI_DOUBLE, DFTI_REAL, 2, sizes);
  status = DftiSetValue(handle, DFTI_PLACEMENT, DFTI_NOT_INPLACE);
  status = DftiCommitDescriptor(handle);
  auto start = std::chrono::high_resolution_clock::now();
  status = DftiComputeForward(handle, datain, dataout);
  auto dur = std::chrono::high_resolution_clock::now() - start;
  std::cout
      << "FFT Done. Elapse time = "
      << std::chrono::duration_cast<std::chrono::microseconds>(dur).count()
      << "us" << std::endl;
  status = DftiFreeDescriptor(&handle);
}

int main(int argc, char *argv[]) {
  int size = 299; // Crash when size  >= 300
  double *in = (double *)mkl_calloc(size * size, sizeof(double), 64);
  double *out = (double *)mkl_calloc(size * size, sizeof(double), 64);
  init_array_random(size * size, in);
  fft(in, out, size);
  mkl_free(in);
  mkl_free(out);
  return 0;
}

Re:Crash for a simple 2D MKL FFT example when matrix size is not too larger

Gennady_F_Intel — Tue, 20 Oct 2020 13:46:38 GMT

building this example as is by the trivial way: icl /Qmkl test.cpp and enabling the verbose mode, I see no problems with my run:

>set MKL_VERBOSE=1

>test_2dfft.exe

MKL_VERBOSE Intel(R) MKL 2020.0 Update 2 Product build 20200624 for Intel(R) 64 architecture Intel(R) Advanced Vector Extensions 2 (Intel(R) AVX2) enabled processors, Win 2.60GHz intel_thread

MKL_VERBOSE FFT(drfo299:299:299x299:1:1,pack:ccs,tLim:2,desc:0000020CD5E3C700) 1.53ms CNR:OFF Dyn:1 FastMM:1 TID:0 NThr:2

FFT Done. Elapse time = 111128us

> building like icl /Qmkl=seqential ( exactly like you did) run well also with the outputs:

test_2dfft.exe

MKL_VERBOSE Intel(R) MKL 2020.0 Update 2 Product build 20200624 for Intel(R) 64 architecture Intel(R) Advanced Vector Extensions 2 (Intel(R) AVX2) enabled processors, Win 2.60GHz sequential

MKL_VERBOSE FFT(drfo299:299:299x299:1:1,pack:ccs,tLim:1,desc:0000022C2470BAC0) 1.73ms CNR:OFF Dyn:1 FastMM:1 TID:0 NThr:1

FFT Done. Elapse time = 110499us

Re:Crash for a simple 2D MKL FFT example when matrix size is not too larger

Gennady_F_Intel — Mon, 16 Nov 2020 01:44:35 GMT

This issue has been resolved and we will no longer respond to this thread. If you require additional assistance from Intel, please start a new thread. Any further interaction in this thread will be considered community only.