/* calculate the 32-bit product of unsigned 16-bit op16 and 32-bit op32 */
static inline uint32_t mul_u16_u32(uint32_t op16, uint32_t op32)
{
    unsigned t1, r;
    asm (
        "swap.w %[b], %[t1]\n\t"
        "mulu   %[a], %[t1]\n\t"
        "sts    macl, %[t1]\n\t"
        "mulu   %[a], %[b]\n\t"
        "sts    macl, %[r]\n\t"
        "shll16 %[t1]\n\t"
        "add    %[t1], %[r]\n\t"
        : [r]  "=r" (r),
          [t1] "=&r" (t1)
        : [a]  "r"   (op16),
          [b]  "r"   (op32)
    );
    return r;
}

/* calculate the 32-bit product of signed 32-bit op16 and unsigned 32-bit op32,
 * where op16 is known to have only 16 significant bits */
static inline uint32_t mul_s16_u32(int32_t op16, int32_t op32)
{
    unsigned t1, t2, r;
    asm (
        "swap.w %[b], %[t1]\n\t"
        "mulu   %[a], %[b]\n\t"
        "sts    macl, %[r]\n\t"
        "mulu   %[a], %[t1]\n\t"
        "sts    macl, %[t1]\n\t"
        "cmp/pz %[a]\n\t"
        "bt 1f\n\t"
        "extu.w %[b], %[t2]\n\t"
        "sub    %[t2], %[t1]\n"
    "1:\n\t"
        "shll16 %[t1]\n\t"
        "add    %[t1], %[r]\n\t"
        : [r]  "=&r" (r),
          [t1] "=&r" (t1),
          [t2] "=r" (t2)
        : [a]  "r"   (op16),
          [b]  "r"   (op32)
    );
    return r;
}