simple_v_extension/daxpy_example.mdwn

   1     # c code
   2     void daxpy(size_t n, double a, const double x[], double y[])
   3     {
   4      for (size_t i = 0; i < n; i++) {
   5        y[i] = a*x[i] + y[i];
   6      }
   7     }
   8
   9     # a0 is n, a1 is ptr to x[0], a2 is ptr to y[0], fa0 is a (scalar)
  10      VBLK.REG[0] = {type: F, isvec: 1, regkey: a3, regidx: a3, elwidth: dflt}
  11      VBLK.REG[1] = {type: F, isvec: 1, regkey: a7, regidx: a7, elwidth: dflt}
  12     loop:
  13      VBLK.SETVL  t0, a0, #4   # MVL=4, vl = t0 = min(a0, MVL))
  14      ld     a3, a1            # load 4 registers a3-6 from x
  15      c.slli   t1, t0, 3       # t1 = vl * 8 (in bytes)
  16      c.ld     a7, a2          # load 4 registers a7-10 from y
  17      c.add    a1, a1, t1      # increment pointer to x by vl*8
  18      fmadd  a7, a3, fa0, a7   # v1 += v0 * fa0 (y = a * x + y)
  19      c.sub    a0, a0, t0      # n -= vl (t0)
  20      c.st     a7, a2          # store 4 registers a7-10 to y
  21      c.add    a2, a2, t1      # increment pointer to y by vl*8
  22      c.bnez   a0, loop        # repeat if n != 0