Calling cudaq.control or cudaq.adjoint on a kernel passed as argument to a device kernel fails during synthesis #3499
Description
Required prerequisites
- Consult the security policy. If reporting a security vulnerability, do not report the bug using this form. Use the process described in the policy to report the issue.
- Make sure you've read the documentation. Your issue may be addressed there.
- Search the issue tracker to verify that this hasn't already been reported. +1 or comment there if it has.
- If possible, make a PR with a failing test to give us a starting point to work on!
Describe the bug
The PySynthCallableBlockArgs pass (and subsequent support in the runtime) requires that all quake.ApplyOp instructions are eliminated during compilation. However, the substitution fails when the indirectly invoked kernel is an argument to a device kernel.
Steps to reproduce the bug
A set of failing tests are added in #3489. After that merge, search for the link to this issue in the code base to find them.
Expected behavior
It should be possible to compile and run the examples in the tests.
Is this a regression? If it is, put the last known working version (or commit) here.
Not a regression
Environment
- CUDA-Q version: latest (status merge PR Various Python bridge fixes and improvements #3489)
- Python version: N/A
- C++ compiler: N/A
- Operating system: N/A
Suggestions
No response