Avoid classifying methods with `[numthreads]` as entry points for CUDA-related targets (#4063)

4 files changed, 23 insertions, 5 deletions

diff --git a/source/slang/slang-check-shader.cpp b/source/slang/slang-check-shader.cpp
index d4e8bda6f..2c1f8651c 100644
--- a/source/slang/slang-check-shader.cpp
+++ b/source/slang/slang-check-shader.cpp
@@ -918,7 +918,7 @@ namespace Slang
             for (Index tt = 0; tt < translationUnitCount; ++tt)
             {
                 auto translationUnit = translationUnits[tt];
-                translationUnit->getModule()->_discoverEntryPoints(sink);
+                translationUnit->getModule()->_discoverEntryPoints(sink, this->getLinkage()->targets);
             }
         }
     }
diff --git a/source/slang/slang-compiler.cpp b/source/slang/slang-compiler.cpp
index 851e3115b..1f1bed902 100644
--- a/source/slang/slang-compiler.cpp
+++ b/source/slang/slang-compiler.cpp
@@ -2431,7 +2431,7 @@ namespace Slang
         EntryPoint* entryPoint,
         DiagnosticSink* sink);
 
-    void Module::_discoverEntryPoints(DiagnosticSink* sink)
+    void Module::_discoverEntryPoints(DiagnosticSink* sink, const List<RefPtr<TargetRequest>>& targets)
     {
         for (auto globalDecl : m_moduleDecl->members)
         {
@@ -2466,12 +2466,30 @@ namespace Slang
             else
             {
                 // If there isn't a [shader] attribute, look for a [numthreads] attribute
-                // since that implicitly means a compute shader.
+                // since that implicitly means a compute shader. We'll not do this when compiling for
+                // CUDA/Torch since [numthreads] attributes are utilized differently for those targets.
+                // 
+
+                bool allTargetsCUDARelated = true;
+                for (auto target : targets)
+                {
+                    if (!isCUDATarget(target) && 
+                        target->getTarget() != CodeGenTarget::PyTorchCppBinding)
+                    {
+                        allTargetsCUDARelated = false;
+                        break;
+                    }
+                }
+
+                if (allTargetsCUDARelated && targets.getCount() > 0)
+                    continue;
+
                 auto numThreadsAttr = funcDecl->findModifier<NumThreadsAttribute>();
                 if (numThreadsAttr)
                     profile.setStage(Stage::Compute);
                 else
                     continue;
+                
             }
 
             RefPtr<EntryPoint> entryPoint = EntryPoint::create(
diff --git a/source/slang/slang-compiler.h b/source/slang/slang-compiler.h
index f9fa24622..cd4e7fdd5 100755
--- a/source/slang/slang-compiler.h
+++ b/source/slang/slang-compiler.h
@@ -1483,7 +1483,7 @@ namespace Slang
             ///
         void _collectShaderParams();
 
-        void _discoverEntryPoints(DiagnosticSink* sink);
+        void _discoverEntryPoints(DiagnosticSink* sink, const List<RefPtr<TargetRequest>>& targets);
 
         class ModuleSpecializationInfo : public SpecializationInfo
         {
diff --git a/source/slang/slang.cpp b/source/slang/slang.cpp
index 4a446a351..654295211 100644
--- a/source/slang/slang.cpp
+++ b/source/slang/slang.cpp
@@ -5335,7 +5335,7 @@ void Linkage::prepareDeserializedModule(SerialContainerData::Module& moduleEntry
     module->setPathInfo(filePathInfo);
     module->setDigest(moduleEntry.digest);
     module->_collectShaderParams();
-    module->_discoverEntryPoints(sink);
+    module->_discoverEntryPoints(sink, targets);
 
     // Hook up fileDecl's scope to module's scope.
     auto moduleDecl = module->getModuleDecl();