<table border="1" cellspacing="0" cellpadding="8">
    <tr>
        <th>Issue</th>
        <td>
            <a href=https://github.com/llvm/llvm-project/issues/107219>107219</a>
        </td>
    </tr>

    <tr>
        <th>Summary</th>
        <td>
            [NVPTX] Clang errors with non-trivial scalar-to-vector conversion when using opencl vector of two 8-bit integer as input
        </td>
    </tr>

    <tr>
      <th>Labels</th>
      <td>
            clang
      </td>
    </tr>

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

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

<pre>
    This issue started occurring with LLVM 18. Below is the reproduction code and the command used to reproduce the issue.

```cuda
typedef char cchar __attribute__((ext_vector_type(2)));

__global__ void kernel(cchar* out, cchar in) {
    int tid = blockDim.x * blockIdx.x + threadIdx.x;
 out[tid] = in;
}
```
```bash
$ clang --cuda-path=<cuda_root> --cuda-gpu-arch=sm_60 --cuda-device-only -O3 -S repro.cu -o repro.ptx
error: non-trivial scalar-to-vector conversion
1 error generated when compiling for sm_60.
```
Here is a link of compiler explorer where issue is reproduced - https://cuda.godbolt.org/z/c8jscoach

More Observations:
1. The issue only occurs when using a scalar input of the 8-bit integer vector of size 2. For other sizes and when a pointer is used, PTX is generated.
2. The issue doesn't occur when using other lower bit integers such as short for 16-bit integers
</pre>
<img width="1px" height="1px" alt="" src="http://email.email.llvm.org/o/eJyMVFFv2zgM_jXKC2HDlp3GefBD21xwB2y3AVcMezNkibG1KpIh0Wm6X3-Q7HZZcQ8HGE5Ii-T3kR8lQtCDRWzZ9oFtDxsx0-h860UYUZ1fadM79do-jTqADmFGCCQ8oQIn5ey9tgO8aBrh06dvn6FscnhA415AB6ARwePknZolaWdBOoUgrEpfpDuf4_85oAJy7ycxfU2lclYcWHG_vu-K5ZGzEouLXidUeAI5Cg8yvbtOEHndz4Rdx3jDeINX6i4oyfkuBjDecMb361M93NbousG4Xpiug4vTCp7RWzSMNyk54_fgZmL8cS2mLeN7YLs1BwCAtgSkFbDqAL1x8vmgz_kVYmgy_1LXZD4AjR6FSvY7ipR--0Base0h5dD2F8Td4UMnPpi9COPq4jVII-wAWRbblU2CRlYdWPUYzc47R6z64-3rMM2Z8DKeCOfurnjzK7xoiZmz5hWyLxVk_yxTyuUM2TqxfKLrUhS9d55V92CdzcjrixYGghRG-IxctowApLMX9EE7u0SVkOJgQIteRF29jBilcp60ieI6OQ8JVf6f7P9EH9UCAoy2z-BOayh6wOtknEcfM_pVUvHou9IUZDASTYFV94wfGT9G2vngVO8M5c4PjB9_RnfzI0gn5Hgrls_OI3zpA_qLiPJOWRZOOTy9aRhS99KuhIXaHCItsbYGtJ1mirij7Jus1xRFhAN6WFvmThD0TwSewzGaNKJPnpCWKSUVMLkY5iPBuFJRpV-fvkfzvbVrB_ktPOUwWMZ3tGC8hbgUMu4FPdzAChBmOYIIEEbnKQ2ovLsFHjaqrdS-2osNtuWOb-t6d7dvNmNblifeb2XdbAvkzV4KrE-7XYWyVlW1L8uNbnnB62Jf1GVZlFWTi1NTlupUV6dSlL2SrC7wLLTJjbmc44g2iUdbFjte7jdG9GhCuss4TyvAOI_Xmm9jQNbPQ2B1YXSg8CsFaTLpAvz729en73H3HtP2JGmG5X77n6r-rYETWmluxkgv7sOIRVgEsJm9aX_X4qBpnPtcujPjxwh1_ckm736gJMaPiXpg_Liyv7T83wAAAP__J63klA">