<table border="1" cellspacing="0" cellpadding="8">
    <tr>
        <th>Issue</th>
        <td>
            <a href=http://email.email.llvm.org/c/eJylVEuv2ywQ_TXOBsUC_Ei88KJPKYtuPqnq8grbY5sWQ8Tg5Lv99R1wch-97aoSSgAz55yZOdC54bH9NmsDzDj3Q9uJaRscUwyNnubAzuBH5xdle2AeJg-I2lk6xLKa90bZaS9KmrIT69UCTPXeITJ0C4Q5wnkVZvAMg6ezEANpnYKVMZPXdojRbg3nNeQZ_5jxd9vvV4zhv7HE2JNnJIkk0sbDwwJLP_sHRJDxQEQlkt6DQhhY98gE30ueyYqN3i1MWabM4jCw3lkMyoYIQwciuEvzhEMEcR5JX6k6MUUgBh1DgCjwOVPUizbKs6sO86aNkrb9co7iyoiaP0u-7T6ISKbs8BRgwD4l47ZiuVTBOZ5SHph1galxhD5QgormxryS-C3y34omWRJ80xrBMHbposwKyBRuYm-Hi9eZBqbx9_aFmQgpXaTyrYbqC89SqNaRoHfLmfxEMTqahrYQkm6jO6_8Y2oR5uzDDP2Pu6rTfwxnd0VKjg2aED1QMN7tcuvqxfUqkP1e6fyiHjvYNNH4vlJrVQTzgZREfDe-sVu3UnKBDE2bUbWP1G4cE9nZu1GbGHmdwbJxtX0kRbYG2v65SXa4qcKUGq7DAJag1s20xSeyXfTcyb91z3o39tdutWHdqs8u4NPNEjLnOd8Xa_oossMnSesyj07M5Ie_xxUUJ_aZfE9Dkou5oCF4Qx6W7w8VFMV4EL0c1D6Cwv_nDXs7KAvBucir4ws7_pmmTPJe0lBoJY9VQRtc1E1Vd_WRH8VbmkKIg2wqkYvybu8bx5bibmiLoSkatVMrVdi3o_ZwVRfYrd60cwhnzIp3mfxMYyLfrl1OLaaFMZf735669538SEuNSCanSVVUx8NubmXD60MNqj5UBS_LsSuE7IcDr5uhFMem3hnVgcE2qyg5aeHKEgTNs-rj7t8V6DbVQfJSHEpBdWi6oqLiwbGohviTlRwWpU0ecXLnp51vE2S3TkgfjcaAzx8VvcWTBUiCSWHQwUAbH6R41_qX_jTuSjeSbtP9MaWHund-SGZ27Oly7JLkNun9BctT8Vw>53587</a>
        </td>
    </tr>

    <tr>
        <th>Summary</th>
        <td>
            some libc calls are slower in clang-14 according to callgrind
        </td>
    </tr>

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

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

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

<pre>
    While looking into a slight performance regression in `clang-14` I came across something rather strange in the `callgrind` output.

Using `clang-14` the Ir for a `__memchr_sse2` call increased by 10-20% from an almost constant `20` to `22` or `24`.

I am also seeing something similar with `__strncmp_sse42`. `__memcmp_sse4_1` and `__strlen_sse2` on the other hand are not affected at all.

With clang-12 I am seeing the same values as with clang-13.

It is rather strange that this could be affected by the compiler since these are library calls. Checking the IR shows no differences in the call invocation.

Maybe this is just a shortcoming of `callgrind` but it really throws off the profiling when functions utilizing those calls are suddenly using > 10% Ir.

I am using `Ubuntu clang version 12.0.0-3ubuntu1~20.04.4`, `Ubuntu clang version 13.0.1-++20220120110924+75e33f71c2da-1~exp1~20220120231001.58` and `Ubuntu clang version 14.0.0-++20220131052853+016956b68081-1~exp1~20220131172951.142` on Ubuntu 20.04.
</pre>
<img width="1px" height="1px" alt="" src="http://email.email.llvm.org/o/eJx1VMuuozgQ_RqysYKwCSEsWPRrpCxmM1Krl1fGFOBuY0cuO5k7Xz9lk9xH3xnJSmxTdc6pqgODG5_7H4s2wIxzv7SdmbbBMcnQ6HkJ7AJ-cn6VVgHzMHtA1M5SECuOlTLSznt-oC07MyVXYFJ5h8jQrRCWBOdlWMAzDJ5iISXSOSdLY2av7ZiyXQyXGMqi-lpUn7bf75jSf2NJuWfPSBJJpIunpxVWtfgnRBApIKESifIgEUY2PDNe7UVViIZN3q1MWibN6jAw5SwGaUOCoYAE7vI-4xBB2ifSd6rOTBKIQccQIAl8rRT1qo307KbDsmmjoq1aL0ncIaGWr5Lvt088kUk7viQYsC_FuK1ZLndwSVHSA7MuMDlNoAIVKGlvzDuJPxL_vWmCZcF3rQkM05Su0kRAJnETew-u31camMbfxxcWIqRykdoXDfUXXqVQrxOBcuuF_EQ5OpmGrhCybqMHL_1zHhGW7MsC6tdD1fkvhou7IRXHRk2IHigZH3a5T_XqlAxkv3c6_5TPA2yaaP2MNFqZwHwgJQnfTR_sNkQqLpCh6TKp9onaTVMmu3g3aZMybwtYNkWrEimyGOj6n02yw00V5tIwjiNYgoqbaetvZLvkubP_6J74MPb3IdoQt-6zK_j8ZnFRVmW1r2N-yIv2m6DzoUxOLMSX_8-rKY_vC_GZliAXV5wWrzrysPjcNlDXU8uVGOU-gcLflw17CxQ1rypeNqc3dvxvmkOW95aGUhtxamq6qPixa47D8VSd-EeamvNWdA0v-eFh7zvHVuJu7Ouxqzu5CzoY6NOrlVyj3nbauBt5i3zx-CzQJ0c5P-axOPYy5l30pl9CuGBRfyrEH7Rm8nocSrIFHYy5Pv72NPGf5GE6akR6MWjT1M2p3S29qGQnFUmvu7YVcGwHWTUdH6dTw48AamfkAAb7oqGGCAs3liFoXzRfd7rPlYvqwNsDp8q7oW6oXXCqmzH9FIcKVqlNmXSUzs8732dJQ5yRHhqNAV8fSvr6zhYg0xG-jORE30_aw01eYZe5-6z9Xztr3JU">