Implement cudaq::run API

[annagrin]

Required prerequisites

• Search the issue tracker to check if your feature has already been mentioned or rejected in other issues.

Describe the feature

Implement cudaq::run API

template<typename T, typename … ARGS> 
std::vector<T> cudaq::run( 
     size_t shots, 
     std::function<T(ARGS…)>&& kernel, ARGS…); 

Examples (c++)

__qpu__ bool kernel(int n) {
  cudaq::qvector q(n);
  x(q[0]);
  return mz(q[0]);
}

int main() {
  auto results = cudaq::run<bool>(/*shots=*/1000, kernel, 2);
  std::cout << results << std::endl;
}

Output:
[1,1, ... 1] // 1000 results

Currently we would like to support returning the following types:

bool
int
float
double
std::vector<bool> // For measurements of qvectors (const size after opts)
structs of elements above (next version)

Examples (python)

@cudaq.kernel()
def kernel(n: int): bool
  q = cudaq.qvector(n)
  x(q[0])
  return mz(q[0])

results = cudaq.run(shots=1000, kernel, 2)
print(results)

Output:
[1,1, ... 1] // 1000 results

Currently we would like to support returning the following types:

bool
int
np.float32
np.float64
list[bool] // For measurements of qvectors (const size after opts)
structs of elements above (next version)

Tasks

• Define log parsing API in c++ (no templates)

Idea:

struct OutputRecord {
 void* ptr;
 std::size_t byteSize;
};

class OutputParser {
 virtual std::vector<OutputRecord> parse(std::istream& s);
};

• Implement log parsing API for backends that will use logs for output (quantum devices and remote simulator)
• Implement log parsing API for backends that use memory for output (local simulators)
  • will need runtime QIR functions implementation
  • make sure we handle multiple and single shot scenarios
• Implement cudaq::run API in c++
• Implement cudaq.run API in Python
• Implement retrieving logs from quantum backends (start with quantinuum)
• Implement creating and retrieving logs from remote simulator server
• Make sure quantum emulation works with new output log functionality
  • run LLVM generated from QIR on the simulator (needs runtime functions implementations)
• Convert returns to log function calls in QIR