// vfnmsac.vv vd, vs1, vs2, vm   # vd[i] = -(vs2[i] * vs1[i]) + vd[i]
VI_VFP_VV_LOOP
({
  vd = f32_mulAdd(f32(vs1.v ^ F32_SIGN), vs2, vd);
})