We use cookies to ensure you have the best browsing experience on our website. Please read our cookie policy for more information about how we use cookies.
  1. Prepare
  2. Algorithms
  3. Implementation
  4. Matrix Layer Rotation
  5. Discussions

Matrix Layer Rotation

  • mark_webster
     + 0 comments

    Concise C++, one-pass, O(m * n).

    Only 1 integer division per ring. Lots of branching in the inner loop but CPU branch prediction should kick in quite nicely on larger matrices.

    void matrixRotation(vector<vector<int>> matrix, int rot) {
    int hei = matrix.size(), wid = matrix[0].size(), rings = (1+min(hei, wid)) >> 1;
    int map[hei][wid];
    memset(map, 0, sizeof(map));

    for (int r=0, rh=hei-1, rw=wid-1; r<rings; r++, rh-=2, rw-=2) {
        int c2=rh+rw, c3=c2+rw, c4=c3+rh;

        for (int dof=0, sof=rot % c4, dx=0, dy=0, sx, sy; dof<c4; dof++, sof++) {
            if (sof >= c4) sof -= c4;
            if (sof < rw) sx=sof, sy=0; else
            if (sof < c2) sx=rw, sy=sof-rw; else
            if (sof < c3) sx=c3-sof, sy=rh; else
                sx=0, sy=c4-sof;

            map[r+dy][r+dx] = matrix[r+sy][r+sx];

            if (dof < rw) dx++; else if (dof < c2) dy++; else if (dof < c3) dx--; else dy--;
        }
    }

    for (int x,y=0; y<hei; y++) {
        for (x=0; x<wid; x++) cout << map[y][x] << " ";
        cout << "\n";
    }
}