pointers and vectors

I cant manage to sum a few vectors. Values are gigantic and only iterates about 800 times until theres no more memory. I guess I am not handling pointers correctly, but I cant work it out. anybody see anything strange?

Code:

---

float *amp, *sum_amp;

SOME SKIPPED CODE

// memory allocations
        sum_amp = (float *)calloc(nx*ny, sizeof(float));
        if (!sum_amp) perror("Insufficient memory!");

        amp = (float *)calloc(nx*ny, sizeof(float));
        if (!amp) perror("Insufficient memory!");

       
        // for each line
        printf("Creating mean amplitude from files...\n"); //info
        for (ii=0;ii<kk;ii++)
        {
       
                fgets(line,sizeof(line),fp);
                // remove newline
                if ((p = strchr(line, '\n')) != NULL) *p = '\0';

                //read headerfile
                read_header(line);

                // create amplitude filename to open
                strcpy(amp_filename, path);
                strcat(amp_filename, "amp_");
                strcat(amp_filename, filename);
                strcat(amp_filename, ".bin");


                //try fopen amp_in
                if (!(amp_in = fopen(amp_filename, "rb")))
                {
                        printf("\nAmplitude file not found!\n");
                        exit(0);
                }
               
                // read data
                fread( (float *)amp, sizeof(float), nx*ny, amp_in);
               
                // accumulate amplitudes
                for (i=0;i<nx*ny;i++)
                {
                        sum_amp[i] += amp[i];
                        printf("i = %d, sum_amp = %f\n",i, sum_amp[i]); //trace
                }

                fclose(amp_in);

        }//for
        fclose(fp);

---

If I put the calloc allocations inside the for loop, it seems to work (numbers dont get gigantic at least) but it doesnt solve my problem since I want sum up multiple vectors, so I dont want to reset sum_amp to 0 next iteration.

OK, managed to get it to work but I am puzzled. If I put my calloc memory allocations inside the for loop instead of before the loop it works. Like this:

Code:

---

                amp = (float *)calloc(nx*ny, sizeof(float));
                if (!amp) perror("Insufficient memory!");
               
                if (ii==0)
                {
                        sum_amp = (float *)calloc(nx*ny, sizeof(float));
                        if (!sum_amp) perror("Insufficient memory!");
                }

---

Why is this different?

A difference is that you allocate with each iteration memory again, which isn't of course good for performance

Code:

---

fread( (float *)amp, sizeof(float), nx*ny, amp_in);

---

why the casting? amp is already float *

That doesn't make sense, and my best guess would be that you are somehow doing something else wrong in your code. I can't spot anything, but I'm sure that's what happens.

Code:

---

fread( (float *)amp, sizeof(float), nx*ny, amp_in);

---

Why the cast here? It is already a float *, so the cast does absolutely nothing.


One possibility, if you moved the calloc down to the lower part of the loop is that something in the upper part of the loop causes your amp/sum_amp variables to get clobbered, and that's what is causing the failure.

--
Mats

Thanks for your replies! I realize now that my values nx and ny are read from the header which hasnt been read before the loop. I am tired. Thanks for the casting info aswell.

Quote:

---

Originally Posted by larne

Thanks for your replies! I realize now that my values nx and ny are read from the header which hasnt been read before the loop. I am tired. Thanks for the casting info aswell.

---

Ah, yes, that would also make that difference ;-)

If you are doing this many times, you may want to free the memory in the loop as well.

--
Mats

As a final note, you might want to consider realloc(), since you want to change the memory you have according to nx, ny.

BUT, I understand that you want one sum_amp that adds the other amps. So what values of nx, ny are you going to use for that? You have to use the maximum values of nx, ny in order to have what you want, correct? I
f that is true than you should allocate only ONCE amp with also the maximum values. Allocating takes time, don't do it if not necessary.

Do you initialize sum_amp? If it is the sum then you should initialize it to 0, or random numbers automatically initialized will be given giving you gigantic results (it might be initialized at 2143244 for you).
You can use a loop for that or memset() is even better.