donut.c without any libraries

donut.c

// Original source: https://www.a1k0n.net/2021/01/13/optimizing-donut.html

#define UART_ADDR 0x20000000

char b[1760], z[1760];

int mult(int a, int b);
int div(int a, int b);
int mod(int a, int b);

void * memset(void * p, int c, unsigned int n);

int print_str(char * c);
int putchar(int c);

#define R(mul,shift,x,y) \
  _=x; \
  x -= mult(mul, y)>>shift; \
  y += mult(mul, _)>>shift; \
  _ = 3145728-mult(x, x)-mult(y, y)>>11; \
  x = mult(x, _)>>10; \
  y = mult(y, _)>>10;

void main() {
  int sA=1024,cA=0,sB=1024,cB=0,_;
  for (;;) {
    memset(b, 32, 1760);  // text buffer
    memset(z, 127, 1760);   // z buffer
    int sj=0, cj=1024;
    for (int j = 0; j < 90; j++) {
      int si = 0, ci = 1024;  // sine and cosine of angle i
      for (int i = 0; i < 324; i++) {
        int R1 = 1, R2 = 2048, K2 = 5120*1024;

        int x0 = mult(R1, cj) + R2,
            x1 = mult(ci, x0) >> 10,
            x2 = mult(cA, sj) >> 10,
            x3 = mult(si, x0) >> 10,
            x4 = mult(R1, x2) - (mult(sA, x3) >> 10),
            x5 = mult(sA, sj) >> 10,
            x6 = K2 + mult(mult(R1, 1024), x5) + mult(cA, x3),
            x7 = mult(cj, si) >> 10,
            x = 40 + div(mult(30, (mult(cB, x1) - mult(sB, x4))), x6),
            y = 12 + div(mult(15, (mult(cB, x4) + mult(sB, x1))), x6),
            N = (-mult(cA, x7) - mult(cB, ((-mult(sA, x7)>>10) + x2)) - mult(ci, (mult(cj, sB) >> 10)) >> 10) - x5 >> 7;

        int o = x + mult(80, y);
        char zz = (x6-K2)>>15;
        if (22 > y && y > 0 && x > 0 && 80 > x && zz < z[o]) {
          z[o] = zz;
          b[o] = ".,-~:;=!*#$@"[N > 0 ? N : 0];
        }
        R(5, 8, ci, si)  // rotate i
      }
      R(9, 7, cj, sj)  // rotate j
    }
    for (int k = 0; 1761 > k; k++)
      putchar(mod(k, 80) ? b[k] : 10);
    R(5, 7, cA, sA);
    R(5, 8, cB, sB);
    print_str("\x1b[23A");
  }
}

int mult(int a, int b)
{
  int acc = 0;

  for (int i = 0; i < 32; i++)
  {
    if (a & (1 << i))
      acc += b << i;
  }

  return acc;
}

int div(int a, int b)
{
  int ret = 0;

  int res_neg = (a < 0) ^ (b < 0);
  int a_tmp = a < 0 ? -a : a;
  int b_tmp = b < 0 ? -b : b;

  while (a_tmp >= b_tmp)
  {
    a_tmp -= b_tmp;
    ret++;
  }

  return res_neg ? -ret : ret;
}

int mod(int a, int b)
{
  while (a >= b)
  {
    a -= b;
  }

  return a;
}

void * memset(void * p, int c, unsigned int n)
{
  for(int i = 0; i < n; i++)
  {
    *((char *)p) = c;
    p++;
  }
}

int print_str(char * s)
{
  char * p = (char *)UART_ADDR;

  while (*s != '\0')
  {
    *p = *s;
    s++;
  }
}

int putchar(int c)
{
  char * p = (char *)UART_ADDR;
  *p = c;
  return 0;
}