Fix QPUD Library Mode Installer Build

docker/build/assets.Dockerfile

@@ -169,7 +169,10 @@ RUN python3 -m ensurepip --upgrade && python3 -m pip install lit && \
     dnf install -y --nobest --setopt=install_weak_deps=False file which
 RUN cd /cuda-quantum && source scripts/configure_build.sh && \
     "$LLVM_INSTALL_PREFIX/bin/llvm-lit" -v build/test \
-        --param nvqpp_site_config=build/test/lit.site.cfg.py
+        --param nvqpp_site_config=build/test/lit.site.cfg.py && \
+    "$LLVM_INSTALL_PREFIX/bin/llvm-lit" -v build/targettests \
+        --param nvqpp_site_config=build/targettests/lit.site.cfg.py
+
 
 # Tests for the Python wheel are run post-installation.
 COPY --from=python_build /wheelhouse /cuda_quantum/wheelhouse

runtime/common/JIT.cpp

@@ -28,6 +28,8 @@
 #include "mlir/ExecutionEngine/ExecutionEngine.h"
 #include <cxxabi.h>
 
+#define DEBUG_TYPE "cudaq-qpud"
+
 namespace cudaq {
 
 void invokeWrappedKernel(std::string_view irString,
@@ -79,14 +81,14 @@ void invokeWrappedKernel(std::string_view irString,
         std::string demangledName(demangledPtr);
         if (demangledName.rfind(wrappedKernelSymbol, 0) == 0 &&
             demangledName.find(templatedTypeName) != std::string::npos) {
-          llvm::dbgs() << "Found symbol " << func.getName() << " for "
-                       << wrappedKernelSymbol;
+          LLVM_DEBUG(llvm::dbgs() << "Found symbol " << func.getName()
+                                  << " for " << wrappedKernelSymbol);
           mangledWrapper = func.getName().str();
           fixUpLinkage(func);
         }
         if (demangledName.rfind(funcName, 0) == 0) {
-          llvm::dbgs() << "Found symbol " << func.getName() << " for "
-                       << funcName;
+          LLVM_DEBUG(llvm::dbgs() << "Found symbol " << func.getName()
+                                  << " for " << funcName);
           mangledKernel = func.getName().str();
           fixUpLinkage(func);
         }
@@ -109,10 +111,6 @@ void invokeWrappedKernel(std::string_view irString,
           return std::make_unique<llvm::SectionMemoryManager>();
         });
     llvm::Triple targetTriple(llvm::Twine(llvmModule->getTargetTriple()));
-    // IMPORTANT: need to setAutoClaimResponsibilityForObjectSymbols to true to
-    // prevent debug asserts about symbol responsibility.
-    objectLayer->setAutoClaimResponsibilityForObjectSymbols(true);
-
     return objectLayer;
   };
 

runtime/cudaq/algorithms/observe.h

@@ -502,10 +502,16 @@ auto observe_async(const std::size_t qpu_id, QuantumKernel &&kernel, spin_op &H,
       H, platform, shots, kernelName, qpu_id);
 #else
   return details::runObservationAsync(
-      [&kernel,
-       args = std::forward_as_tuple(std::forward<Args>(args)...)]() mutable {
-        std::apply(std::move(kernel), std::move(args));
-      },
+      detail::make_copyable_function([&kernel,
+                                      args = std::make_tuple(std::forward<Args>(
+                                          args)...)]() mutable {
+        std::apply(
+            [&kernel](Args &&...args) {
+              return cudaq::invokeKernel(std::forward<QuantumKernel>(kernel),
+                                         std::forward<Args>(args)...);
+            },
+            std::move(args));
+      }),
       H, platform, shots, kernelName, qpu_id);
 #endif
 }
@@ -535,10 +541,16 @@ auto observe_async(std::size_t shots, std::size_t qpu_id,
       H, platform, shots, kernelName, qpu_id);
 #else
   return details::runObservationAsync(
-      [&kernel,
-       args = std::forward_as_tuple(std::forward<Args>(args)...)]() mutable {
-        std::apply(std::move(kernel), std::move(args));
-      },
+      detail::make_copyable_function([&kernel,
+                                      args = std::make_tuple(std::forward<Args>(
+                                          args)...)]() mutable {
+        std::apply(
+            [&kernel](Args &&...args) {
+              return cudaq::invokeKernel(std::forward<QuantumKernel>(kernel),
+                                         std::forward<Args>(args)...);
+            },
+            std::move(args));
+      }),
       H, platform, shots, kernelName, qpu_id);
 #endif
 }

runtime/cudaq/algorithms/sample.h

@@ -385,10 +385,16 @@ async_sample_result sample_async(const std::size_t qpu_id,
       platform, kernelName, shots, qpu_id);
 #else
   return details::runSamplingAsync(
-      [&kernel,
-       args = std::forward_as_tuple(std::forward<Args>(args)...)]() mutable {
-        std::apply(std::move(kernel), std::move(args));
-      },
+      detail::make_copyable_function([&kernel,
+                                      args = std::make_tuple(std::forward<Args>(
+                                          args)...)]() mutable {
+        std::apply(
+            [&kernel](Args &&...args) {
+              return cudaq::invokeKernel(std::forward<QuantumKernel>(kernel),
+                                         std::forward<Args>(args)...);
+            },
+            std::move(args));
+      }),
       platform, kernelName, shots, qpu_id);
 #endif
 }
@@ -437,10 +443,16 @@ async_sample_result sample_async(std::size_t shots, std::size_t qpu_id,
       platform, kernelName, shots, qpu_id);
 #else
   return details::runSamplingAsync(
-      [&kernel,
-       args = std::forward_as_tuple(std::forward<Args>(args)...)]() mutable {
-        std::apply(std::move(kernel), std::move(args));
-      },
+      detail::make_copyable_function([&kernel,
+                                      args = std::make_tuple(std::forward<Args>(
+                                          args)...)]() mutable {
+        std::apply(
+            [&kernel](Args &&...args) {
+              return cudaq::invokeKernel(std::forward<QuantumKernel>(kernel),
+                                         std::forward<Args>(args)...);
+            },
+            std::move(args));
+      }),
       platform, kernelName, shots, qpu_id);
 #endif
 }

targettests/Remote-Sim/args_parsing.cpp

@@ -7,12 +7,9 @@
  ******************************************************************************/
 
 // REQUIRES: remote-sim
-// REQUIRES: c++20
-// FIXME: https://github.com/NVIDIA/cuda-quantum/issues/1111
-
 // clang-format off
-// RUN: nvq++ --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t 
-// RUN: nvq++ --enable-mlir --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t
+// RUN: nvq++ %cpp_std --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t 
+// RUN: nvq++ %cpp_std --enable-mlir --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t
 // clang-format on
 
 #include <cudaq.h>

targettests/Remote-Sim/args_synthesis.cpp

@@ -8,8 +8,8 @@
 
 // REQUIRES: remote-sim
 // clang-format off
-// RUN: nvq++ --enable-mlir --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t
-// RUN: nvq++ --enable-mlir --no-aggressive-early-inline --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t
+// RUN: nvq++ %cpp_std --enable-mlir --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t
+// RUN: nvq++ %cpp_std --enable-mlir --no-aggressive-early-inline --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t
 // clang-format on
 
 // This is a comprehensive set of tests for kernel argument synthesis for remote

targettests/Remote-Sim/free_func.cpp

@@ -7,12 +7,10 @@
  ******************************************************************************/
 
 // REQUIRES: remote-sim
-// REQUIRES: c++20
-// FIXME: https://github.com/NVIDIA/cuda-quantum/issues/1111
 
 // clang-format off
-// RUN: nvq++ --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t 
-// RUN: nvq++ --enable-mlir --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t
+// RUN: nvq++ %cpp_std --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t 
+// RUN: nvq++ %cpp_std --enable-mlir --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t
 // clang-format on
 
 #include <cudaq.h>
@@ -21,7 +19,11 @@ void ghz(std::size_t N) __qpu__ {
   cudaq::qvector q(N);
   h(q[0]);
   for (int i = 0; i < N - 1; i++) {
+#if CUDAQ_USE_STD20
     x<cudaq::ctrl>(q[i], q[i + 1]);
+#else
+    cx(q[i], q[i + 1]);
+#endif
   }
   mz(q);
 }
@@ -30,7 +32,11 @@ void ansatz(double theta) __qpu__ {
   cudaq::qvector q(2);
   x(q[0]);
   ry(theta, q[1]);
+#if CUDAQ_USE_STD20
   x<cudaq::ctrl>(q[1], q[0]);
+#else
+  cx(q[1], q[0]);
+#endif
 }
 
 int main() {

targettests/Remote-Sim/observe.cpp

@@ -6,13 +6,11 @@
  * the terms of the Apache License 2.0 which accompanies this distribution.    *
  ******************************************************************************/
 
-// REQUIRES: c++20
 // REQUIRES: remote-sim
-// FIXME: https://github.com/NVIDIA/cuda-quantum/issues/1111
 
 // clang-format off
-// RUN: nvq++ --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t 
-// RUN: nvq++ --enable-mlir --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t
+// RUN: nvq++ %cpp_std --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t 
+// RUN: nvq++ %cpp_std --enable-mlir --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t
 // clang-format on
 
 #include <cudaq.h>
@@ -26,7 +24,11 @@ struct ansatz {
     cudaq::qvector q(2);
     x(q[0]);
     ry(theta, q[1]);
+#if CUDAQ_USE_STD20
     x<cudaq::ctrl>(q[1], q[0]);
+#else
+    cx(q[1], q[0]);
+#endif
   }
 };
 

targettests/Remote-Sim/observe_async.cpp

@@ -7,12 +7,10 @@
  ******************************************************************************/
 
 // REQUIRES: remote-sim
-// REQUIRES: c++20
-// FIXME: https://github.com/NVIDIA/cuda-quantum/issues/1111
 
 // clang-format off
-// RUN: nvq++ --target remote-mqpu --remote-mqpu-auto-launch 3 %s -o %t && %t 
-// RUN: nvq++ --enable-mlir --target remote-mqpu --remote-mqpu-auto-launch 3 %s -o %t && %t
+// RUN: nvq++ %cpp_std --target remote-mqpu --remote-mqpu-auto-launch 3 %s -o %t && %t 
+// RUN: nvq++ %cpp_std --enable-mlir --target remote-mqpu --remote-mqpu-auto-launch 3 %s -o %t && %t
 // clang-format on
 
 #include <cudaq.h>
@@ -22,7 +20,7 @@ struct ansatz {
     cudaq::qvector q(2);
     x(q[0]);
     ry(theta, q[1]);
-    x<cudaq::ctrl>(q[1], q[0]);
+    cx(q[1], q[0]);
   }
 };
 
@@ -45,5 +43,12 @@ int main() {
   printf("Energy is %lf\n", energy);
   printf("Gradient is %lf\n", gradient);
   assert(std::abs(energy + 1.748794) < 1e-3);
+  // Shots-based observe async. API
+  cudaq::set_random_seed(13);
+  auto energyFutureShots =
+      cudaq::observe_async(/*shots=*/8192, /*qpu_id=*/0, ansatz{}, h, .59);
+  const auto energyShots = energyFutureShots.get().expectation();
+  printf("Energy (shots) is %lf\n", energyShots);
+  assert(std::abs(energyShots + 1.748794) < 0.1);
   return 0;
 }

targettests/Remote-Sim/sample.cpp

@@ -6,13 +6,11 @@
  * the terms of the Apache License 2.0 which accompanies this distribution.    *
  ******************************************************************************/
 
-// REQUIRES: c++20
 // REQUIRES: remote-sim
-// FIXME: https://github.com/NVIDIA/cuda-quantum/issues/1111
 
 // clang-format off
-// RUN: nvq++ --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t 
-// RUN: nvq++ --enable-mlir --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t
+// RUN: nvq++ %cpp_std --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t 
+// RUN: nvq++ %cpp_std --enable-mlir --target remote-mqpu --remote-mqpu-auto-launch 1 %s -o %t && %t
 // clang-format on
 
 #include <cudaq.h>
@@ -23,7 +21,11 @@ struct ghz {
     cudaq::qarray<N> q;
     h(q[0]);
     for (int i = 0; i < N - 1; i++) {
+#if CUDAQ_USE_STD20
       x<cudaq::ctrl>(q[i], q[i + 1]);
+#else
+      cx(q[i], q[i + 1]);
+#endif
     }
     mz(q);
   }
@@ -42,7 +44,11 @@ int main() {
       cudaq::qvector q(N);
       h(q[0]);
       for (int i = 0; i < N - 1; i++) {
+#if CUDAQ_USE_STD20
         x<cudaq::ctrl>(q[i], q[i + 1]);
+#else
+        cx(q[i], q[i + 1]);
+#endif
       }
       mz(q);
     };

targettests/Remote-Sim/sample_async.cpp

@@ -7,12 +7,10 @@
  ******************************************************************************/
 
 // REQUIRES: remote-sim
-// REQUIRES: c++20
-// FIXME: https://github.com/NVIDIA/cuda-quantum/issues/1111
 
 // clang-format off
-// RUN: nvq++ --target remote-mqpu --remote-mqpu-auto-launch 4 %s -o %t && %t 
-// RUN: nvq++ --enable-mlir --target remote-mqpu --remote-mqpu-auto-launch 4 %s -o %t && %t
+// RUN: nvq++ %cpp_std --target remote-mqpu --remote-mqpu-auto-launch 4 %s -o %t && %t 
+// RUN: nvq++ %cpp_std --enable-mlir --target remote-mqpu --remote-mqpu-auto-launch 4 %s -o %t && %t
 // clang-format on
 
 #include <cudaq.h>
@@ -31,16 +29,25 @@ int main() {
   printf("Number of QPUs: %zu\n", num_qpus);
   assert(num_qpus == 4);
   std::vector<cudaq::async_sample_result> countFutures;
+  // sample_async API with default shots
   for (std::size_t i = 0; i < num_qpus; i++) {
     countFutures.emplace_back(cudaq::sample_async(i, simpleX{}, i + 1));
   }
+  // Shots-based sample_async API
+  std::vector<cudaq::async_sample_result> countFuturesWithShots;
+  for (std::size_t i = 0; i < num_qpus; i++) {
+    countFuturesWithShots.emplace_back(
+        cudaq::sample_async(/*shots=*/1024, /*qpu_id=*/i, simpleX{}, i + 1));
+  }
 
   for (std::size_t i = 0; i < num_qpus; i++) {
-    auto counts = countFutures[i].get();
-    counts.dump();
-    const std::string expectedBitStr(i + 1, '1');
-    assert(counts.size() == 1);
-    assert(counts.begin()->first == expectedBitStr);
+    for (auto counts :
+         {countFutures[i].get(), countFuturesWithShots[i].get()}) {
+      counts.dump();
+      const std::string expectedBitStr(i + 1, '1');
+      assert(counts.size() == 1);
+      assert(counts.begin()->first == expectedBitStr);
+    }
   }
   return 0;
 }

targettests/Target/IQM/for_loop.cpp

@@ -18,7 +18,11 @@ struct ghz {
     cudaq::qarray<N> q;
     h(q[0]);
     for (int i = 0; i < N - 1; i++) {
+#if CUDAQ_USE_STD20
       x<cudaq::ctrl>(q[i], q[i + 1]);
+#else
+      cx(q[i], q[i + 1]);
+#endif
     }
     auto result = mz(q[0]);
   }

targettests/Target/IQM/z_gate.cpp

@@ -16,7 +16,11 @@ template <std::size_t N>
 struct kernel_with_z {
   auto operator()() __qpu__ {
     cudaq::qarray<N> q;
+#if CUDAQ_USE_STD20
     z<cudaq::ctrl>(q[0], q[1]);
+#else
+    cz(q[0], q[1]);
+#endif
     auto result = mz(q[0]);
   }
 };

targettests/execution/auto_kernel.cpp

@@ -20,8 +20,13 @@ struct ak2 {
     cudaq::qarray<3> q;
     x(q[0]);
     h(q[1]);
+#if CUDAQ_USE_STD20
     x<cudaq::ctrl>(q[1], q[2]);
     x<cudaq::ctrl>(q[0], q[1]);
+#else
+    cx(q[1], q[2]);
+    cx(q[0], q[1]);
+#endif
     h(q[0]);
     x(q[1]);
     y(q[2]);