mex run time error
in [Matlab]
|
From: skyworld on 8 Jul 2010 10:05 Hi, I wrote a mex file like this: #include <stdio.h> #include "math.h" #include "mex.h" #define DATA_NUM 512 #define COEF_NUM 7 void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[] ) { mxArray *x_data_array; double *x_data; mxArray *coef_data_array; double *coef_data; double *y_data_array; int i,j; double tmp_x[DATA_NUM]; x_data_array = prhs[0]; x_data = mxGetPr(x_data_array); coef_data_array = prhs[1]; coef_data = mxGetPr(coef_data_array); plhs[0] = mxCreateDoubleMatrix(1,DATA_NUM,mxREAL); y_data_array = mxGetPr(plhs[0]); for (i=0;i<COEF_NUM;i++) { mexPrintf("%d,data=%d,coef=%d\n",i,x_data[i],coef_data[i]); } for (i=0;i<DATA_NUM;i++) { tmp_x[i] = 12; mexPrintf("tmp_x=%d\n",tmp_x[i]); tmp_x[i] = x_data[i]; mexPrintf("i=%d, x=%d, tmp_x=%d\n",i,x_data[i],tmp_x[i]); } for (i=0;i<DATA_NUM;i++) { for (j=0;j<COEF_NUM;j++) { y_data_array[i] += x_data[i+j]*coef_data[COEF_NUM-j-1]; } } return; } my purpose is to finish a fir function. I use this command to compile it: mex fir_check.c and I infer it in matlab as this: b = firhalfband(N,fc/fs); x=cos(2*pi*t*f0)+cos(2*pi*t*f1)+cos(2*pi*t*f2)+cos(2*pi*t*f3); y_dut = fir_check(x,b); It is very interesting that I can't get right data from x_data, coef_data and tmp_x. The display shows these data are very huge numbers, which I think should be within +/- 4. And sometime when I run the m file, the matlab will report run-time error. Could anybody help me to find what is wrong with this code? thanks.
From: Jan Simon on 8 Jul 2010 11:17 Dear Yong, > double tmp_x[DATA_NUM]; > tmp_x[i] = 12; > mexPrintf("tmp_x=%d\n",tmp_x[i]); Your tmp_x is a double vector. Display doubles with the format "%f" or "%g", because "%d" formats integers. Please post just the relevant part of the code and omit all lines, which are not involved in the runtime error. Good luck, Jan
From: James Tursa on 8 Jul 2010 14:35 skyworld <chenyong20000(a)gmail.com> wrote in message <9eb6e874-c2c3-4de1-ad17-446a3c028234(a)k8g2000prh.googlegroups.com>... > > I wrote a mex file like this: > (snip) > > It is very interesting that I can't get right data from x_data, > coef_data and tmp_x. The display shows these data are very huge > numbers, which I think should be within +/- 4. And sometime when I run > the m file, the matlab will report run-time error. Could anybody help > me to find what is wrong with this code? thanks. > You are probably exceeding valid indexes for the arrays. Impossible to tell for sure because you didn't post enough information about size of the inputs. e.g., I would suspect this line: > y_data_array[i] += x_data[i+j]*coef_data[COEF_NUM-j-1]; Does x_data really contain 511+6+1 = 518 elements? Try putting in some checks to see that your sizes and indexing are all consistent. James Tursa
From: skyworld on 8 Jul 2010 21:12 On Jul 8, 11:17 pm, "Jan Simon" <matlab.THIS_Y...(a)nMINUSsimon.de> wrote: > Dear Yong, > > > double tmp_x[DATA_NUM]; > > tmp_x[i] = 12; > > mexPrintf("tmp_x=%d\n",tmp_x[i]); > > Your tmp_x is a double vector. Display doubles with the format "%f" or "%g", because "%d" formats integers. > > Please post just the relevant part of the code and omit all lines, which are not involved in the runtime error. > > Good luck, Jan Hi Jan, thanks for your kind information. You just point out one of my errors. Thanks. A very annoying problem is that when I revised %d to %f, there is no messages on command window output. When I change back to %d, also no output. totally different from yesterday. That is my life.... regards yong
From: skyworld on 8 Jul 2010 21:22
On Jul 9, 2:35 am, "James Tursa" <aclassyguy_with_a_k_not_...(a)hotmail.com> wrote: > skyworld<chenyong20...(a)gmail.com> wrote in message <9eb6e874-c2c3-4de1-ad17-446a3c028...(a)k8g2000prh.googlegroups.com>... > > > I wrote a mex file like this: > > (snip) > > > It is very interesting that I can't get right data from x_data, > > coef_data and tmp_x. The display shows these data are very huge > > numbers, which I think should be within +/- 4. And sometime when I run > > the m file, the matlab will report run-time error. Could anybody help > > me to find what is wrong with this code? thanks. > > You are probably exceeding valid indexes for the arrays. Impossible to tell for sure because you didn't post enough information about size of the inputs. e.g., I would suspect this line: > > > y_data_array[i] += x_data[i+j]*coef_data[COEF_NUM-j-1]; > > Does x_data really contain 511+6+1 = 518 elements? > > Try putting in some checks to see that your sizes and indexing are all consistent. > > James Tursa Hi James, thanks for your message. In fact I'm trying to write a code for FIR check so that it could be used in HW co-simulation. My inputs to this function are two arguments: x_data with 512 samples and coefficients with 7 taps. In order to match matlab filter outputs, I guess I need to pad 6 zeros before x samples. That is the reason I defined tmp_x. My purpose would like to do this: for (i=0 to i=5) tmp_x(i) = 0; for (i=6 to i=518) tmp_x(i) = x_data(i-6); then I will do fir multiply-addition as y_data_array[i] += x_data[i+j]*coef_data[COEF_NUM-j-1]; I met two problems when doing above: 1) I can't get right data in tmp_x from x_data. I use code like these to check if tmp_x got data from x_data for (i=0;i<DATA_NUM;i++) { tmp_x[i] = 12; mexPrintf("tmp_x=%d\n",tmp_x[i]); tmp_x[i] = x_data[i]; mexPrintf("i=%d, x=%d, tmp_x=%d\n",i,x_data[i],tmp_x[i]); } the display shows two data are different. 2) If I use only 512 x sample to do fir calculation, the result seems ok. But if I run this program several times, there maybe run-time error. I don't know what kind of hidden issues in my code. thanks. |