<table border="1" cellspacing="0" cellpadding="8">
    <tr>
        <th>Issue</th>
        <td>
            <a href=https://github.com/llvm/llvm-project/issues/130172>130172</a>
        </td>
    </tr>

    <tr>
        <th>Summary</th>
        <td>
            Optimzations make assumption about function names
        </td>
    </tr>

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

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

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

<pre>
    Given the following module as an input
```llvm ir
define double @anon_expr_1() {
entry:
  %calltmp = call double @cos(double 1.234000e+00)
  ret double %calltmp
}

declare double @cos(double)
```

optimization with a minimal set of passes "simplifycfg,instcombine,reassociate,gvn" will produce the following module:
```llvm ir
define double @anon_expr_1() {
entry:
  %calltmp = call double @cos(double 1.234000e+00)
  ret double 0x3FD526571FE8C7A5
}

declare double @cos(double)
```
NOTE: The return of this function is the IEEE Hex form of 0.33046510807172985 which is the mathematical Cosine of 1.234. Thus, the optimizations have ***Assumed*** that the function named `cos` is in fact the mathematical function of the same name common in C runtime libraries - but it doesn't ***KNOW*** that. It could well be that this is just a function that happens to have the name `cos` (and has NO reliance on the C runtime). The expectation is that this does something completely different that might even have side effects so ignoring the return value and substituting a const is just ***WRONG*** behavior. (One would hope that subsequent optimization stages would remove the call due to the possibility of side effects, but ignoring the return and substituting a const value is still wrong)
</pre>
<img width="1" height="1" alt="" src="http://email.email.llvm.org/o/eJzMVU2P2zYQ_TX0ZRCDomTLPujg7K7ToMAuUATIMaCokcWUH6pmaO_21xeUdu0smpx6KSDYIjTz-N6b4VAT2VNAbMTmo9jcr3TiIU4NeW2dQ161sXtpPtkzBuABoY_OxYsNJ_CxSw5BE-gANoyJhTyIrVwe584e7CTkocPeBoQuptYhiErqEMM3fB6nb4VQO6H2IOqPQh4w8PQiyoOQBwChNkY7x34EUd5Dfv8BwkQSave6LtaqrKSUKNRHKYXazwAT8jXhipUZ1vf5N_MyTk_4C9QF5ipnSYkjW2__1mxjgIvlATR4G6zXDggZYg-jJkICoRRZPzrbv5j-JNSdDcQm-tYGFOpuQk0UjdWcV6dzEErBxToH4xS7ZPCnXi_e_C8tls_l8X6jtpu6OD7s7urD5r9Y_fj05UGUB_gyYN4kTSFby4Ml6FMws_2WZo8-Pzw8wG_4DH2cfI6S67KU1XZTyJ2si1rtdxu4DNYMbxle84BeszXawV2k7FzsF4lr-DIkEupujvyx3ASDPudWOizPgSh57K5r4EHzUrU3hkF77EBsZ7Fbmfe3AXpt-N88rkmzTgTSHmcAMNH7LDfAHUwpsPUIzraTniwSfIA2MdhcB6QgVM03ir8_Pn19z28NnxlMTK6DCzoHLb7RztwIvidi0Dcy88dBjyMGAo6LBZnezOymTKidDh0MmuDxCSZ0VgeDEJeRceUt1H491xSfRzSsb3V845BVAEWPPOS2N9GPDhndC3S273HCwEu0t6eBAfNUmkmR7RCw79FwBgB7CnHKEHxrobN2CSETpdQSW06cIzSYGIiv-q-Wff3j6fHTzcAWB322cVpnuU8B4TIbOcTx1cWMin-lzPHdoCDWJ6TX8Al9fDVxOXAJs7N5PUYi21pn-SW3wY-SckfOhf6JrF8KWvRaAuI8WS5TDCeh9quuKbt9udcrbIq6KtRe7bbVamh2RWtUpbbSlLpvq05jX5tyKwtUVV-09co2SqqNLOVWSVVX1borNsa0RbXd7bZmazaikpjvjXUeTes4nVaWKGFTlLKo1crpFh3NF41SAS8wfxVK5XtnanLShzadSFTSWWK6wbBlh81TdvXtOHr9Z758KPlxdlm3MfH7w0erNLlmYB4pzzx1FOp4sjykdm2iF-qY8V__PoxT_I6GhTrOrEio4yvtc6P-CQAA___oAFkk">