Invalid QASM generated from Quake

[srulre]

Required prerequisites

• 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.

Describe the bug

This is a continuation of #603, which has not been resolved entirely.
Some kernels don't translate to a valid QASM - specifically, gate definitions appear after their calls.
In this case, it has something to do with empty kernels (identity operators).

Steps to reproduce the bug

I ran the code below:

#include <cudaq.h>

#include <iostream>

int main() {
        auto [kernel1, qubit1] = cudaq::make_kernel<cudaq::qubit>();

        auto [kernel2, qubit2] = cudaq::make_kernel<cudaq::qubit>();
        kernel2.x(qubit2);
        kernel2.call(kernel1, qubit2);

        auto [kernel3, qubit3] = cudaq::make_kernel<cudaq::qubit>();
        kernel3.h(qubit3);
        kernel3.call(kernel2, qubit3);
        kernel3.h(qubit3);
        kernel3.mz(qubit3);

        auto kernel4 = cudaq::make_kernel();
        auto qubit4 = kernel4.qalloc();
        kernel4.call(kernel1, qubit4);
        kernel4.call(kernel3, qubit4);

        std::cout << kernel4.to_quake() << std::endl;
        return 0;
}

And it generated the following Quake code:

module {
  func.func @__nvqpp__mlirgen____nvqppBuilderKernel_361293194784() attributes {"cudaq-entrypoint"} {
    %0 = quake.alloca !quake.ref
    call @__nvqpp__mlirgen____nvqppBuilderKernel_367535629127(%0) : (!quake.ref) -> ()
    call @__nvqpp__mlirgen____nvqppBuilderKernel_202375922897(%0) : (!quake.ref) -> ()
    return
  }
  func.func @__nvqpp__mlirgen____nvqppBuilderKernel_367535629127(%arg0: !quake.ref) {
    return
  }
  func.func @__nvqpp__mlirgen____nvqppBuilderKernel_202375922897(%arg0: !quake.ref) {
    quake.h %arg0 : (!quake.ref) -> ()
    call @__nvqpp__mlirgen____nvqppBuilderKernel_093606261879(%arg0) : (!quake.ref) -> ()
    quake.h %arg0 : (!quake.ref) -> ()
    %bits = quake.mz %arg0 name "" : (!quake.ref) -> i1
    return
  }
  func.func @__nvqpp__mlirgen____nvqppBuilderKernel_093606261879(%arg0: !quake.ref) {
    quake.x %arg0 : (!quake.ref) -> ()
    call @__nvqpp__mlirgen____nvqppBuilderKernel_367535629127(%arg0) : (!quake.ref) -> ()
    return
  }
}

Then, I ran cudaq-translate --convert-to=openqasm example.qke. The following QASM code has been generated, with the gate nvqpp__mlirgen____nvqppBuilderKernel_367535629127 illegally being called before being defined.

// Code generated by NVIDIA's nvq++ compiler
OPENQASM 2.0;

include "qelib1.inc";

gate nvqpp__mlirgen____nvqppBuilderKernel_093606261879 q0 {
  x q0;
  nvqpp__mlirgen____nvqppBuilderKernel_367535629127 q0;
}

gate nvqpp__mlirgen____nvqppBuilderKernel_202375922897 q0 {
  h q0;
  nvqpp__mlirgen____nvqppBuilderKernel_093606261879 q0;
  h q0;
  creg var1[1];
  measure q0 -> var1[0];
}

gate nvqpp__mlirgen____nvqppBuilderKernel_367535629127 q0 {
}

qreg var0[1];
nvqpp__mlirgen____nvqppBuilderKernel_367535629127 var0[0];
nvqpp__mlirgen____nvqppBuilderKernel_202375922897 var0[0];

Expected behavior

The code below or an equivalent:

// Code generated by NVIDIA's nvq++ compiler
OPENQASM 2.0;

include "qelib1.inc";

gate nvqpp__mlirgen____nvqppBuilderKernel_367535629127 q0 {
}

gate nvqpp__mlirgen____nvqppBuilderKernel_093606261879 q0 {
  x q0;
  nvqpp__mlirgen____nvqppBuilderKernel_367535629127 q0;
}

gate nvqpp__mlirgen____nvqppBuilderKernel_202375922897 q0 {
  h q0;
  nvqpp__mlirgen____nvqppBuilderKernel_093606261879 q0;
  h q0;
  creg var1[1];
  measure q0 -> var1[0];
}

qreg var0[1];
nvqpp__mlirgen____nvqppBuilderKernel_367535629127 var0[0];
nvqpp__mlirgen____nvqppBuilderKernel_202375922897 var0[0];

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

Not a regression

Environment

Cudaq Docker image ID: sha256:672fdd4393aab3b27162a5ede45bc5f651801442e372441bb1e3e516b643e0ea
Additional data:

"org.opencontainers.image.created": "2023-09-09T02:16:38.587Z",
"org.opencontainers.image.description": "CUDA Quantum toolkit for heterogeneous quantum-classical workflows",
"org.opencontainers.image.licenses": "NOASSERTION",
"org.opencontainers.image.ref.name": "ubuntu",
"org.opencontainers.image.revision": "77cbcd1ab00ef597727858abb9e248602c53c27d",
"org.opencontainers.image.source": "https://github.com/NVIDIA/cuda-quantum",
"org.opencontainers.image.title": "cuda-quantum",
"org.opencontainers.image.url": "https://github.com/NVIDIA/cuda-quantum",
"org.opencontainers.image.version": "latest"

Suggestions

No response

[schweitzpgi]

Appears to be fixed. Please verify and close.

[srulre]

Thanks, I'll test and let you know after the new docker image is available.

[srulre]

Works now, thanks!