Merge branch 'main' into use-zstd-layers

Author: mitchdz

lib/Frontend/nvqpp/ConvertDecl.cpp

@@ -536,8 +536,28 @@ bool QuakeBridgeVisitor::VisitFunctionDecl(clang::FunctionDecl *x) {
   auto typeFromStack = peelPointerFromFunction(popType());
   if (auto f = module.lookupSymbol<func::FuncOp>(kernSym)) {
     auto fTy = f.getFunctionType();
-    assert(typeFromStack == fTy);
     auto fSym = f.getSymNameAttr();
+    if (typeFromStack != fTy) {
+      // This may be a call to an entry-point kernel. Determine if that is the
+      // case, and convert this to a direct call. Otherwise, this an calling
+      // convention violation.
+      bool found = false;
+      for (auto pair : namesMap)
+        if (pair.second == kernName) {
+          if (auto f = module.lookupSymbol<func::FuncOp>(pair.first)) {
+            fTy = f.getFunctionType();
+            fSym = f.getSymNameAttr();
+            found = true;
+          }
+          break;
+        }
+      if (!found) {
+        reportClangError(
+            x, mangler,
+            "invalid call from kernel: calling convention violation");
+        return false;
+      }
+    }
     return pushValue(builder.create<func::ConstantOp>(loc, fTy, fSym));
   }
   auto [funcOp, alreadyAdded] = getOrAddFunc(loc, kernName, typeFromStack);

test/AST-Quake/separate_compilation.cpp

@@ -0,0 +1,23 @@
+/*******************************************************************************
+ * Copyright (c) 2025 NVIDIA Corporation & Affiliates.                         *
+ * All rights reserved.                                                        *
+ *                                                                             *
+ * This source code and the accompanying materials are made available under    *
+ * the terms of the Apache License 2.0 which accompanies this distribution.    *
+ ******************************************************************************/
+
+// RUN: cudaq-quake %cpp_std %s | cudaq-opt | FileCheck %s
+
+#include "cudaq.h"
+
+__qpu__ uint64_t otherKernel(std::vector<cudaq::measure_result> &x);
+
+__qpu__ uint64_t test_entry_point() {
+  cudaq::qvector q(5);
+  auto results = cudaq::mz(q);
+  return otherKernel(results);
+}
+
+// CHECK-LABEL:   func.func @__nvqpp__mlirgen__function_test_entry_point.
+// CHECK-SAME:      () -> i64 attributes {"cudaq-entrypoint", "cudaq-kernel", no_this} {
+// CHECK:           %[[VAL_3:.*]] = call @__nvqpp__mlirgen__function_otherKernel.{{.*}}(%{{.*}}) : (!cc.stdvec<i1>) -> i64