[SYCL] Make cl::reqd_work_group_size applicable to lambdas

Signed-off-by: Dmitry Sidorov <dmitry.sidorov@intel.com>

Author: MrSidims

clang/include/clang/Basic/AttributeCommonInfo.h

@@ -152,7 +152,14 @@ class AttributeCommonInfo {
     // FIXME: Eventually we want to do a list here populated via tablegen.  But
     // we want C++ attributes to be permissible on Lambdas, and get propagated
     // to the call operator declaration.
-    return getParsedKind() == AT_SYCLIntelKernelArgsRestrict;
+    // FIXME: Attribute lambda capturing is not a part of SYCL 1.2.1. We shall
+    // check if a program is being compiled with -sycl-ext option.
+    auto ParsedAttr = getParsedKind();
+    if (ParsedAttr == AT_SYCLIntelKernelArgsRestrict ||
+        (ParsedAttr == AT_ReqdWorkGroupSize && isCXX11Attribute()))
+      return true;
+
+    return false;
   }
 
   bool isC2xAttribute() const { return SyntaxUsed == AS_C2x; }

clang/test/CodeGenSYCL/reqd-work-group-size.cpp

@@ -25,10 +25,15 @@ void bar() {
 
   Functor f;
   kernel<class kernel_name2>(f);
+
+  kernel<class kernel_name3>(
+  []() [[cl::reqd_work_group_size(8, 8, 8)]] {});
 }
 
 // CHECK: define spir_kernel void @{{.*}}kernel_name1() {{.*}} !reqd_work_group_size ![[WGSIZE32:[0-9]+]]
 // CHECK: define spir_kernel void @{{.*}}kernel_name2() {{.*}} !reqd_work_group_size ![[WGSIZE8:[0-9]+]]
+// CHECK: define spir_kernel void @{{.*}}kernel_name3() {{.*}} !reqd_work_group_size ![[WGSIZE88:[0-9]+]]
 // CHECK: ![[WGSIZE32]] = !{i32 32, i32 16, i32 16}
 // CHECK: ![[WGSIZE8]] = !{i32 8, i32 1, i32 1}
+// CHECK: ![[WGSIZE88]] = !{i32 8, i32 8, i32 8}
 

clang/test/SemaSYCL/reqd-work-group-size.cpp

@@ -96,17 +96,8 @@ void bar() {
     f32x32x1();
   });
 
-  // Support for reqd_work_group_size (and other SYCL attributes) that apply to
-  // lambda expressions is not implemented in clang yet.
-  // When it lands, the following code is expected to compile successfully.
-  //
-  // expected-error@+1 {{'reqd_work_group_size' attribute cannot be applied to types}}
-  kernel<class kernel_name9>([]() [[cl::reqd_work_group_size(32, 32, 32)]] {
-    f32x32x32();
-  });
-  // While this case is not going to work (wrong syntax):
   // expected-error@+1 {{expected variable name or 'this' in lambda capture list}}
-  kernel<class kernel_name10>([[cl::reqd_work_group_size(32, 32, 32)]] []() {
+  kernel<class kernel_name9>([[cl::reqd_work_group_size(32, 32, 32)]] []() {
     f32x32x32();
   });