[SYCL] Add exclusive scan over group test

SYCL/GroupAlgorithm/exclusive_scan_over_group.cpp

@@ -0,0 +1,58 @@
+// UNSUPPORTED: cuda || hip
+// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
+// RUN: %GPU_RUN_PLACEHOLDER %t.out
+
+#include <CL/sycl.hpp>
+#include <algorithm>
+#include <iostream>
+
+template <typename T>
+cl::sycl::event compiler_group_scan_impl(cl::sycl::queue *queue, T *in_data,
+                                         T *out_data, int num_wg,
+                                         int group_size) {
+  cl::sycl::nd_range<1> thread_range(num_wg * group_size, group_size);
+  cl::sycl::event event = queue->submit([&](cl::sycl::handler &cgh) {
+    cgh.parallel_for(thread_range, [=](cl::sycl::nd_item<1> item) {
+      auto id = item.get_global_linear_id();
+      auto group = item.get_group();
+      T data = in_data[id];
+
+      T updated_data = cl::sycl::exclusive_scan_over_group(
+          group, data, cl::sycl::multiplies<T>());
+      out_data[id] = updated_data;
+    });
+  });
+  return event;
+}
+
+template <typename T>
+void test_compiler_group_scan(cl::sycl::queue *queue, T *in_data, T *out_data,
+                              int num_wg, int group_size) {
+  compiler_group_scan_impl(queue, in_data, out_data, num_wg, group_size);
+}
+
+int main(int argc, const char **argv) {
+  int num_wg = 1;
+  int group_size = 16;
+
+  cl::sycl::queue queue{
+      cl::sycl::gpu_selector{},
+      cl::sycl::property_list{cl::sycl::property::queue::enable_profiling(),
+                              cl::sycl::property::queue::in_order()}};
+
+  typedef int T;
+  size_t nelems = num_wg * group_size;
+  T *data = cl::sycl::malloc_shared<T>(nelems, queue);
+  T *result = cl::sycl::malloc_shared<T>(nelems, queue);
+  queue.fill<T>(data, T(2), nelems).wait();
+  queue.memset(result, 0, nelems * sizeof(T)).wait();
+
+  test_compiler_group_scan(&queue, data, result, num_wg, group_size);
+  queue.wait();
+  T expected[] = {1,   2,   4,    8,    16,   32,   64,    128,
+                  256, 512, 1024, 2048, 4096, 8192, 16384, 32768};
+  for (int i = 0; i < sizeof(expected) / sizeof(T); ++i) {
+    assert(result[i] == expected[i]);
+  }
+  return 0;
+}