[llvm-branch-commits] [llvm] [libsycl] Add parallel_for feature (PR #189068)

[Kseniya Tikhomirova via llvm-branch-commits]

================
@@ -199,18 +178,169 @@ class _LIBSYCL_EXPORT queue {
         "group. ");
 
     setKernelParameters(depEvents);
-    submitSingleTask<KernelName, KernelType>(kernelFunc);
+    using NameT =
+        typename detail::get_kernel_name_t<KernelName, KernelType>::name;
+    submitSingleTask<NameT, KernelType>(kernelFunc);
     return getLastEvent();
   }
 
+  /// Defines and invokes a SYCL kernel function as a lambda expression or a
+  /// named function object type, for the specified range.
+  ///
+  /// \param numWorkItems specifies the global work space of the kernel.
+  /// \param rest acts as-if: const KernelType &KernelFunc".
+  // TODO: Rest will represent reduction types once it is supported.
+  template <typename KernelName = detail::AutoName, typename... Rest>
+  event parallel_for(range<1> numWorkItems, Rest &&...rest) {
----------------
KseniyaTikhomirova wrote:

no, it will cause issues with integral type conversion to range. I implemented it initially exactly this way and broke this functionality.  

https://github.com/llvm/llvm-project/pull/189068