[SYCL] Fix a bug when using no device split and reqd_work_group_size

llvm/include/llvm/Support/PropertySetIO.h

@@ -230,6 +230,17 @@ class PropertySetRegistry {
     PropSet.insert({PropName, PropertyValue(PropVal)});
   }
 
+  void remove(StringRef Category, StringRef PropName) {
+    auto PropertySetIt = PropSetMap.find(Category);
+    if (PropertySetIt == PropSetMap.end())
+      return;
+    auto &PropertySet = PropertySetIt->second;
+    auto PropIt = PropertySet.find(PropName);
+    if (PropIt == PropertySet.end())
+      return;
+    PropertySet.erase(PropIt);
+  }
+
   /// Parses from the given \p Buf a property set registry.
   static Expected<std::unique_ptr<PropertySetRegistry>>
   read(const MemoryBuffer *Buf);

llvm/lib/SYCLLowerIR/SYCLDeviceRequirements.cpp

@@ -40,7 +40,6 @@ SYCLDeviceRequirements
 llvm::computeDeviceRequirements(const Module &M,
                                 const SetVector<Function *> &EntryPoints) {
   SYCLDeviceRequirements Reqs;
-  bool MultipleReqdWGSize = false;
   // Process all functions in the module
   for (const Function &F : M) {
     if (auto *MDN = F.getMetadata("sycl_used_aspects")) {
@@ -81,8 +80,6 @@ llvm::computeDeviceRequirements(const Module &M,
             ExtractUnsignedIntegerFromMDNodeOperand(MDN, I));
       if (!Reqs.ReqdWorkGroupSize.has_value())
         Reqs.ReqdWorkGroupSize = NewReqdWorkGroupSize;
-      if (Reqs.ReqdWorkGroupSize != NewReqdWorkGroupSize)
-        MultipleReqdWGSize = true;
     }
 
     if (auto *MDN = F.getMetadata("sycl_joint_matrix")) {
@@ -119,13 +116,6 @@ llvm::computeDeviceRequirements(const Module &M,
     }
   }
 
-  // Usually, we would only expect one ReqdWGSize, as the module passed to
-  // this function would be split according to that. However, when splitting
-  // is disabled, this cannot be guaranteed. In this case, we reset the value,
-  // which makes so that no value is reqd_work_group_size data is attached in
-  // in the device image.
-  if (MultipleReqdWGSize)
-    Reqs.ReqdWorkGroupSize.reset();
   return Reqs;
 }
 

llvm/tools/sycl-post-link/sycl-post-link.cpp

@@ -310,6 +310,15 @@ std::string saveModuleProperties(module_split::ModuleDesc &MD,
   auto PropSet =
       computeModuleProperties(MD.getModule(), MD.entries(), GlobProps);
 
+  // When the split mode is none, the required work group size will be added
+  // to the whole module, which will make the runtime unable to
+  // launch the other kernels in the module that have different
+  // required work group sizes or no requried work group sizes. So we need to
+  // remove the required work group size metadata in this case.
+  if (SplitMode == module_split::SPLIT_NONE)
+    PropSet.remove(PropSetRegTy::SYCL_DEVICE_REQUIREMENTS,
+                   "reqd_work_group_size_uint64_t");
+
   std::string NewSuff = Suff.str();
   if (!Target.empty()) {
     PropSet.add(PropSetRegTy::SYCL_DEVICE_REQUIREMENTS, "compile_target",

sycl/test-e2e/Regression/no-split-reqd-wg-size-2.cpp

@@ -0,0 +1,19 @@
+// This test checks that with -fsycl-device-code-split=off, kernels
+// with different reqd_work_group_size dimensions can be launched.
+
+// RUN: %{build} -fsycl -fsycl-device-code-split=off -o %t.out
+// RUN: %{run} %t.out
+
+// UNSUPPORTED: hip
+
+#include <sycl/detail/core.hpp>
+
+using namespace sycl;
+
+int main(int argc, char **argv) {
+  queue q;
+  q.single_task([] {});
+  q.parallel_for(range<2>(24, 1),
+                 [=](auto) [[sycl::reqd_work_group_size(24, 1)]] {});
+  return 0;
+}