CUFFT_INVALID_PLAN error Error on using CUFFT

Accelerated Computing CUDA CUDA Programming and Performance
#1

I am executing the following program and the program compiles and links correctly. But on execution I get the following error. I believe I have used the signal size as the argument in cufftExecC2C and not the data size. Any help is much appreciated.

#include <stdio.h>

#include <assert.h>

#include <cuda.h>

#include <cufft.h>

#define NX 256

#define BATCH 10

int main(void)

{

cufftComplex *h_data, *data;

int i;

cufftHandle plan;

h_data=(cufftComplex ) malloc(NXBATCH*sizeof(cufftComplex));

cudaMalloc((void**)&data, sizeof(cufftComplex)NXBATCH);

cufftPlan1d (&plan, NX, CUFFT_C2C, BATCH);

//data=(cufftComplex *) data;

for(i=0;i<NX*BATCH;i++)

(h_data+i)->x=1.0f;

cudaMemcpy(data, h_data, NXBATCHsizeof(cufftComplex),cudaMemcpyHostToDevice);

cufftExecC2C(plan, data, data, CUFFT_FORWARD);

cudaMemcpy(h_data, data, NXBATCHsizeof(cufftComplex), cudaMemcpyDeviceToHost);

printf(“After Transform:\n”);

for (i=0;i<NX*BATCH;i+=NX)

printf("Real=%lf Complex=%lf\n", (h_data+i)->x, (h_data+i+1)->y);

cufftExecC2C(plan, data, data, CUFFT_INVERSE);

printf(“After Inverse Transform:\n”);

for (i=0;i<NX*BATCH;i+=NX)

printf("Real=%lf Complex=%lf\n", (h_data+i)->x, (h_data+i+1)->y);

}

I am sure I am missing out something but after trying for a very long time I am not able to spot my mistake.

Thanks,

Guru

#2

Leaving away the printed input and output I get the following error:
cufft: ERROR: /root/cuda-stuff/sw/rel/gpgpu/toolkit/r2.2/cufft/src/cufft.cu, line 143
cufft: ERROR: CUFFT_INVALID_PLAN

P. S : I also dint do the last cudaMemcpy() before printing out the Inverse transformed data and I believe this is irrelevant to the error I get.

#3

Eagerly awaiting some help!!

Thanks,
Guru

#4

Your code compiles and runs without problems on my system.
Are you sure you’re using the right SDK version/driver version combo?

N.

#5

Hi Nico,

      Thanks for your reply! It compiles and runs perfectly in the emulation mode, but only when i do it directly on the device I get the error.  I can get other programs run correctly on the device without any problems.  But I get the error if I run any program related to FFT.  The simpleCUFFT example given gives the following error.

cufftSafeCall() CUFFT error in file <simpleCUFFT.cu>, line 124.

Thanks,

Guru

#6

Hence my question: Are you sure you’re using the right SDK version/driver version combo? :)

N.

#7

Nico,
I am using the CUDA 2.2 SDK toolkit and the 180.11 Nvidia Driver. I can get other examples working in the Release mode. Only the FFT examples are not working. I am using the GTX 275 card for which there is no supported driver for 64 bit linux by NVIDIA.

Thanks,
Guru

#8

Huh? I’m using the 185.18.14 driver in 64-bit ubuntu. Which linux distribution do you have?

N.

#9

I am using Ubuntu 8.10. I got the driver information from the nvidia-settings command. If I should be doing something else to get the driver version please let me know.

Thanks,

Guru

#10

You can download the driver here.

N.

#11

Thanks Nico. Will try changing my driver. Initially I had problems with the driver. But after some changes everything seemed to work correctly. Will check with this driver.

Thanks a lot for your help!
Guru