`cuda-quake` mishandles measure results, causing compiler crash

[IlanIwumbwe]

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

cudaq-quake fails to emit a quake.discriminate to map measure result type to bool when performing range based for loop.

Steps to reproduce the bug

Build from source and run the circuit below

struct main_circuit { 
uint64_t operator()() __qpu__ {
	cudaq::qarray<2>reg_1190;

        int64_t packed_result = 0; int bit_offset = 0;	
	
	for (bool b : mz(reg_1190)){
               if (b) packed_result |= (1ULL << bit_offset);
               bit_offset++;
        }

	return packed_result;
}
};

and get a compiler crash, stemming from cudaq-quake.
loc("test.cpp":10:2):error: 'cc.store' op failed to verify that type of value matches element type of pointer

Line 10 being for (bool b : mz(reg_1190)){

The MLIR that contains that cc.store is

%18 = "cc.alloca"() <{elementType = i1}> : () -> !cc.ptr<i1>
%19 = "cc.load"(%17) : (!cc.ptr<!cc.measure_handle>) -> !cc.measure_handle
"cc.store"(%19, %18) : (!cc.measure_handle, !cc.ptr<i1>) -> ()

The cc.store tries map something of type !cc.measure_handle to !cc.ptr<i1> which is ptr to a bool.

Running the code but with auto instead of bool works. The MLIR at the relevant section is

cc.store %9, %8 : !cc.ptr<!cc.measure_handle> loc(#loc6)
%10 = cc.load %8 : !cc.ptr<!cc.measure_handle> loc(#loc8)
%11 = quake.discriminate %10 : (!cc.measure_handle) -> i1 loc(#loc8)

quake.discriminate added here maps the measure result to the correct type.

Expected behavior

Compiler should not crash

Is this a regression? If it is, put the last known working version (or commit) here.

Not a regression

Environment

• CUDA-Q version: b04dee5
• Python version: 3.12.3
• C++ compiler: clang++-22
• Operating system: Ubuntu 24.04

Suggestions

No response

[atgeller]

This is a result of breaking changes from #4409, where the result of mz on a vector of qubits is a std::vector<measure_result>, which is not implicitly convertible to std::vector<bool>. When you use auto the type of b is actually measure_result, which is implicitly converted when used as the conditional in the if statement. If you wish to keep the type of b as bool, you can explicitly convert the type using to_bools:

struct main_circuit { 
uint64_t operator()() __qpu__ {
  cudaq::qarray<2> reg_1190;

  int64_t packed_result = 0;
  int bit_offset = 0;

  for (bool b : cudaq::to_bools(mz(reg_1190))) {
    if (b) packed_result |= (1ULL << bit_offset);
      bit_offset++;
  }

  return packed_result;
}
};

[IlanIwumbwe]

I see, thanks for the response and workaround