Does NM16 work anymore? From src/tools dir

[cartrite]

I can't find anything in the code from the latest cvs or official source code of nm16 that writes any data. When I compiled it and ran it all it did was print the header in the ppm file and quit.

[John Van Vliet]

hi what did you use to build it

for me visual stuido builds it and so dose MinGW
on my xp box, and gcc4.1 in fedora

you may need to run ti in mingw and output it to a file with a >

[t00fri]

I can't find anything in the code from the latest cvs or official source code of nm16 that writes any data. When I compiled it and ran it all it did was print the header in the ppm file and quit.

Cartrite,

the input data are read from the STDIN file handle and the output goes to STDOUT. These filehandles are called "standard input (STDIN)" and "standard output(STDOUT)". This is a very handy setup, since one may use redirection to (>) and from (<) files.

eg.

nm16 <options> < elevations16bit.bin > normalmap.ppm

So the program reads from elevations16bit.bin and writes to normalmap.ppm in this example.

The best with programs using STDIN and STDOUT instead of filenames is that they may be piped (|) together. Suppose you have a tool called 'halfsize' that reduces the size of your texture by a factor of two, then you may write e.g.

halfsize < elevations16bit.bin | halfsize | halfsize > out16bit.bin

got it?

The file out16bit.bin then only has 1/8 size of the input texture elevations16bit.bin

Unfortunately under Win32 this all works not well for binary files. But for Linux it's wonderful!
There are hacks for Win32, however.

Bye Fridger

[cartrite]

First, This is an old thread that was written before I became aware of this normalmap problem and the new tools that will become available soon to rebuild them.

Second, just to bring me up to date if I'm missing something here, can someone show me were in the following code are the statements that writes actual data to the outputfile and highlight same in red?

Code: Select all

#include <iostream>
#include <cstdio>

using namespace std;

int kernelSize = 2;
int halfKernelSize = 1;


float readS16(istream& in)
{
    unsigned char c[2];
    cin.read((char*) c, 2);

    return (((short) c[0] << 8) | c[1]) * (1.0f / 65535.0f);
}


float* readRowS16(istream& in, int width)
{
    float* row = new float[width];
    for (int i = 0; i < width; i++)
        row[i] = readS16(in);

    return row;
}


int main(int argc, char* argv[])
{
    if (argc < 5)
    {
        cerr << "Usage: nm16 <width> <height> <bumpheight> <filter method>\n";
        return 1;
    }

    int width = 0;
    int height = 0;
    float bumpheight = 1.0f;
    int method = 0;

    if (sscanf(argv[1], " %d", &width) != 1 ||
        sscanf(argv[2], " %d", &height) != 1)
    {
        cerr << "Bad image dimensions.\n";
        return 1;
    }

    if (sscanf(argv[3], " %f", &bumpheight) != 1)
    {
        cerr << "Invalid bump height.\n";
        return 1;
    }

    if (sscanf(argv[4], " %d", &method) != 1)
    {
        cerr << "Bad filter method.\n";
        return 1;
    }

    // Binary 8-bit grayscale header
    // cout << "P5\n";
    // Binary 8-bit/channel RGB header
    cout << "P6\n";
    cout << width << ' ' << height << '\n' << "255\n";

    float** samples = new float*[height];

    for (int i = 0; i < kernelSize - 1; i++)
        samples[i] = readRowS16(cin, width);

    for (int y = 0; y < height; y++)
    {
        // Out with the old . . .
        if (y > halfKernelSize)
            delete[] samples[y - halfKernelSize - 1];

        // . . . and in with the new.
        if (y < height - halfKernelSize)
            samples[y + halfKernelSize] = readRowS16(cin, width);

        for (int x = 0; x < width; x++)
        {
#if 0
            unsigned char v = (unsigned char) (samples[y][x]  * 255.99f);
            cout.write((char*) &v, 1);
#endif
            float dx;
            if (x == width - 1)
                dx = samples[y][0] - samples[y][x];
            else
                dx = samples[y][x + 1] - samples[y][x];

            float dy;
            if (y == height - 1)
                dy = samples[y][x] - samples[y - 1][x];
            else
                dy = samples[y + 1][x] - samples[y][x];
        }
    }

    return 0;
}


This is what the file looks like back to Celestia version 1.4.0

I compiled the above code ( from celestia/src/tools/nm16 - cvs version 05/2006 ) with

Code: Select all

g++ -O3 nm16.cpp -o nm16

and then ran it with

Code: Select all

nm16 10800 4800 150.0 0 < a10g.bin > /srv/disk2/home/cartrite/BMNG/topo/out.ppm

with Suse 10 x86_64 using gcc 4.0.2 and Suse 9.2 using gcc 3.3.4 and here are the contents of out.ppm from both

Code: Select all

P6
10800 4800
255

Anyhow I ended up using the old code of nm16 which looks this at the end of the file

Code: Select all

            float dx;
            if (x == width - 1)
      dx = samples[y][x] - samples[y][0];
            else
      dx = samples[y][x] - samples[y][x + 1];

            float dy;
            if (y == height - 1)
      dy = samples[y - 1][x] - samples[y][x];
            else
      dy = samples[y][x] - samples[y + 1][x];

       dx *= bumpheight;
       dy *= bumpheight;
       float mag = (float) sqrt(dx * dx + dy * dy + 1.0f);
       float rmag = 1.0f / mag;
#if 0
       if (y % 100 == 0 && x == 1000)
      cerr << "dx: " << dx << '\n';
#endif

       rgb[x * 3 + 0] = (unsigned char) (128 + 127 * dx * rmag);
       rgb[x * 3 + 1] = (unsigned char) (128 + 127 * dy * rmag);
       rgb[x * 3 + 2] = (unsigned char) (128 + 127 * rmag);
        }

   fwrite(rgb, width * 3, 1, stdout);
    }

    return 0;
}


and the output file was normal.

cartrite

[t00fri]

Cartrite,

you are correct, even the present CVS code is grossly incomplete. The whole rgb[] definition and the final fwrite() is missing. So this code cannot work!

I have never checked Chris' code that he had committed to CVS.

Bye Fridger

[cartrite]

Many Thanks,