<table border="1" cellspacing="0" cellpadding="8">
    <tr>
        <th>Issue</th>
        <td>
            <a href=https://github.com/llvm/llvm-project/issues/63971>63971</a>
        </td>
    </tr>

    <tr>
        <th>Summary</th>
        <td>
            [SIMD] auto-vectorization using instruction usdot
        </td>
    </tr>

    <tr>
      <th>Labels</th>
      <td>
            new issue
      </td>
    </tr>

    <tr>
      <th>Assignees</th>
      <td>
      </td>
    </tr>

    <tr>
      <th>Reporter</th>
      <td>
          vfdff
      </td>
    </tr>
</table>

<pre>
    test: https://gcc.godbolt.org/z/f86hxd8cT
```
#define N 480

unsigned int
f (unsigned int res, signed char *restrict a,
   unsigned char *restrict b)
{
  for (__INTPTR_TYPE__ i = 0; i < N; ++i)
    {
 int av = a[i];
      int bv = b[i];
      signed short mult = av * bv;
      res += mult;
    }
  return res;
}
```

According [gcc-12](https://community.arm.com/arm-community-blogs/b/tools-software-ides-blog/posts/gcc-12), **Armv8.6-A** introduced a new dot-product instruction for when the sign of the operands differ called [usdot](https://developer.arm.com/documentation/ddi0596/2021-12/SIMD-FP-Instructions/USDOT--by-element---Dot-Product-with-unsigned-and-signed-integers--vector--by-element--). This instruction is introduced behind the **+i8mm** compiler flag.

Starting with GCC 12 the auto-vectorizer can now automatically recognize and use this instruction, while llvm can't.
</pre>
<img width="1px" height="1px" alt="" src="http://email.email.llvm.org/o/eJxsVM9v-6gT_WvGlxGWA0lsH3xI6q-_6mG71TZ72FOFAdusbKgAJ9v-9StI-nM_UuXC8Jh58-YF7r0ejVIN7I6wazO-hsm65jzIYch6K1-boHwAdsAphBcP7AC0A9qNQuSjlb2dQ27dCLR7A9oN1X76R1biBEULxQH2xe3vuqVMqkEbhQ-4rd6D6buaREOiNuEaGRBo9TWMTnmgd3iLiIk7BHpwygenRUAO9O56FRE_Lv4H1gOtb5XL4zt-sBFUPT_fP5weT388n_56_N_zM2oE1mIB7JiWd_gQl0CPQI_6Iw8i4meuyJSf00UOu6OGXQvs-InEhOiviP7XiBt3P1kXcFnncE13jo1gf_6BdsonTqxN2G-nULbvG6fC6kxS8R3xcfpzUOl7EMI6qc2IsDuOQpANjVRp9d0Jwi7LanR4zblbcmEXoB13C_mIk362owfa9UC7YO3sibdDuHCniJbKp3Og3Yv1wV-tFUvROk4b6AHo4eCWc5XvyeG6jRI6K1ehJHI06oLSBvKSQgG18cGtImhr0mAvkzIYJpVkRTuktX1RjhvpUephUA4Fn2clY6erlzb8qlGpzmqO9740Kq1YF2UCj9XiXupiV--BdrSgm9RG93T_W0u6R3L_ySu2-edT-_uJkP6VqFnFHISQ1gbyeG2DXHSYyLuPCTeS3JbaBDUq5wk5KxGs-54DaJ3jadL-mw5p-6FZryZtZNLhKmj0c7UsN3WFXV70rBwOMx_zr5Z4CtyF6IhIDv9_d4cbmtLwNdgbHf2W5DRo7CXFFx50lPcVnRJ2NPpNITcSV68w_CAaR36Z9Kxwns9LTAO0DHkmGyZrVvNMNZt9VVesKnZVNjWcUVaxertlfMO3jJV7IeRQVKISJRW9yHRDC8qKkhZFSUtW5ayWTLC6UAPd7cvNFraFWrie81gwPmWZ9n5VzZ7V5Sabea9mnx5HSqPR0iHQ-EvIXBPvkH4dPWyLWfvgP7MEHeb0qsbxw679LlHyC64-avl1Tsl82erm5sdzq8O09jfTxRq3f9H0fysRgHaJWTRWYv5vAAAA__-o3s9F">