1 2 3 4 5
// vfmsub: vd[i] = +(vd[i] * f[rs1]) - vs2[i] VI_VFP_VF_LOOP ({ vd = f32_mulAdd(vd, rs1, f32(vs2.v ^ F32_SIGN)); })