Bug of fftw_plan_many_dft

[Qianruipku]

#include "fftw3.h"
#include <stdlib.h>
#include <cmath>

int main()
{
int nrank = 24;
int nfft = 9;
int ngrid = nrank * nfft;

fftw_complex* aux1  = (fftw_complex *) fftw_malloc(sizeof(fftw_complex) * ngrid);
int *embed = nullptr;
fftw_plan planyfor  = fftw_plan_many_dft(1, &nrank, nfft,	
                                         aux1,  embed, nfft, 1,
                                         aux1,  embed, nfft, 1,
                                         FFTW_FORWARD, FFTW_MEASURE);
for(int i = 0 ; i < ngrid; ++i)
{
    double r = std::rand() / double(RAND_MAX);
    double arg = std::rand() / double(RAND_MAX);
    aux1[i][0] = r * cos(arg);
    aux1[i][1] = r * sin(arg);
}
fftw_execute(planyfor);
double sum = 0;
for(int i = 0 ; i < ngrid; ++i)
{
    sum += aux1[i][0] + aux1[i][0];
}
printf("%.20f\n", sum);

fftw_cleanup();
fftw_destroy_plan(planyfor);
fftw_free(aux1);
}

The code produces different results every time it is executed.
When I use valgrind to test it, errors will be found:

==3216633== Invalid read of size 8
==3216633== at 0x5AD9FF: awake (dftw-direct.c:129)
==3216633== by 0x4024B5: fftw_plan_awake (plan.c:66)
==3216633== by 0x5A9DC0: awake (ct.c:64)
==3216633== by 0x4024B5: fftw_plan_awake (plan.c:66)
==3216633== by 0x40C3E4: awake (vrank-geq1.c:70)
==3216633== by 0x4024B5: fftw_plan_awake (plan.c:66)
==3216633== by 0x5A91BA: awake (buffered.c:76)
==3216633== by 0x4024B5: fftw_plan_awake (plan.c:66)
==3216633== by 0x401979: fftw_destroy_plan (apiplan.c:185)
==3216633== by 0x401511: main (test.cpp:35)
==3216633== Address 0x4dcd730 is 48 bytes inside a block of size 72 free'd
==3216633== at 0x483F0C3: free (vg_replace_malloc.c:872)
==3216633== by 0x4023CE: fftw_kernel_free (kalloc.c:143)
==3216633== by 0x4022B9: fftw_ifree (alloc.c:40)
==3216633== by 0x405391: fftw_solver_destroy (solver.c:43)
==3216633== by 0x404954: fftw_planner_destroy (planner.c:959)
==3216633== by 0x402202: fftw_cleanup (the-planner.c:39)
==3216633== by 0x401505: main (test.cpp:34)
==3216633== Block was alloc'd at
==3216633== at 0x483C855: malloc (vg_replace_malloc.c:381)
==3216633== by 0x4023A8: fftw_kernel_malloc (kalloc.c:135)
==3216633== by 0x402274: fftw_malloc_plain (alloc.c:28)
==3216633== by 0x4052F5: fftw_mksolver (solver.c:26)
==3216633== by 0x5AA5F2: fftw_mksolver_ct (ct.c:239)
==3216633== by 0x5AE317: regone (dftw-direct.c:312)
==3216633== by 0x5AE3F9: fftw_regsolver_ct_directw (dftw-direct.c:330)
==3216633== by 0x5AA6D0: fftw_kdft_dit_register (kdft-dit.c:26)
==3216633== by 0x450963: fftw_codelet_t1_6 (t1_6.c:293)
==3216633== by 0x5A5C26: fftw_solvtab_exec (solvtab.c:29)
==3216633== by 0x4077B7: fftw_dft_conf_standard (conf.c:43)
==3216633== by 0x405DBE: fftw_configure_planner (configure.c:28)

When int nrank = 24 is changed to int nrank = 20, results won't change and no errors will occur.

Enviroment:
g++ (Ubuntu 9.4.0-1ubuntu1~20.04.1) 9.4.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.