[SYCL] Improve Graphs testing

* Extend testing

* Fix reduction test

* Add test to verify node ordering

* Update sycl include

* Switch to assertions in graph tests

* Formatting

Author: julianmi

sycl/test/graph/graph-explicit-dotp.cpp

@@ -1,6 +1,6 @@
 // RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
-#include <CL/sycl.hpp>
 #include <iostream>
+#include <sycl/sycl.hpp>
 
 #include <sycl/ext/oneapi/experimental/graph.hpp>
 
@@ -83,16 +83,12 @@ int main() {
   // Using shortcut for executing a graph of commands
   q.ext_oneapi_graph(executable_graph).wait();
 
-  if (*dotp != host_gold_result()) {
-    std::cout << "Error unexpected result!\n";
-  }
+  assert(*dotp == host_gold_result());
 
   sycl::free(dotp, q);
   sycl::free(x, q);
   sycl::free(y, q);
   sycl::free(z, q);
 
-  std::cout << "done.\n";
-
   return 0;
 }

sycl/test/graph/graph-explicit-node-ordering.cpp

@@ -0,0 +1,56 @@
+// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
+#include <iostream>
+#include <sycl/sycl.hpp>
+
+#include <sycl/ext/oneapi/experimental/graph.hpp>
+
+int main() {
+
+  sycl::property_list properties{
+      sycl::property::queue::in_order{},
+      sycl::ext::oneapi::property::queue::lazy_execution{}};
+
+  sycl::queue q{sycl::gpu_selector_v, properties};
+
+  sycl::ext::oneapi::experimental::command_graph g;
+
+  const size_t n = 10;
+  float *x = sycl::malloc_shared<float>(n, q);
+
+  auto init = g.add([&](sycl::handler &h) {
+    h.parallel_for(sycl::range<1>{n}, [=](sycl::id<1> idx) {
+      size_t i = idx;
+      x[i] = 2.0f;
+    });
+  });
+
+  auto add = g.add([&](sycl::handler &h) {
+    h.parallel_for(sycl::range<1>{n}, [=](sycl::id<1> idx) {
+      size_t i = idx;
+      x[i] += 2.0f;
+    });
+  });
+
+  auto mult = g.add([&](sycl::handler &h) {
+    h.parallel_for(sycl::range<1>{n}, [=](sycl::id<1> idx) {
+      size_t i = idx;
+      x[i] *= 3.0f;
+    });
+  });
+
+  g.make_edge(init, mult);
+  g.make_edge(mult, add);
+
+  auto executable_graph = g.finalize(q.get_context());
+
+  q.submit([&](sycl::handler &h) {
+     h.ext_oneapi_graph(executable_graph);
+   }).wait();
+
+  for (int i = 0; i < n; i++)
+    assert(x[i] == 8.0f);
+
+  sycl::free(x, q);
+
+  return 0;
+}

sycl/test/graph/graph-explicit-queue-shortcuts.cpp

@@ -1,6 +1,6 @@
 // RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
-#include <CL/sycl.hpp>
 #include <iostream>
+#include <sycl/sycl.hpp>
 
 #include <sycl/ext/oneapi/experimental/graph.hpp>
 
@@ -33,9 +33,10 @@ int main() {
   auto e3 = q.ext_oneapi_graph(executable_graph, e1);
   q.ext_oneapi_graph(executable_graph, {e2, e3}).wait();
 
-  sycl::free(arr, q);
+  for (int i = 0; i < n; i++)
+    assert(arr[i] == 1);
 
-  std::cout << "done " << arr[0] << std::endl;
+  sycl::free(arr, q);
 
   return 0;
 }

sycl/test/graph/graph-explicit-reduction.cpp

@@ -1,6 +1,6 @@
 // RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
-#include <CL/sycl.hpp>
 #include <iostream>
+#include <sycl/sycl.hpp>
 
 #include <sycl/ext/oneapi/experimental/graph.hpp>
 
@@ -26,12 +26,13 @@ int main() {
                    [=](sycl::id<1> idx, auto &sum) { sum += input[idx]; });
   });
 
-  e.wait();
+  auto executable_graph = g.finalize(q.get_context());
+  q.ext_oneapi_graph(executable_graph).wait();
+
+  assert(*output == 45);
 
   sycl::free(input, q);
   sycl::free(output, q);
 
-  std::cout << "done\n";
-
   return 0;
 }

sycl/test/graph/graph-explicit-repeated-exec.cpp

@@ -0,0 +1,58 @@
+// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
+#include <iostream>
+#include <sycl/sycl.hpp>
+
+#include <sycl/ext/oneapi/experimental/graph.hpp>
+
+int main() {
+
+  sycl::property_list properties{
+      sycl::property::queue::in_order{},
+      sycl::ext::oneapi::property::queue::lazy_execution{}};
+
+  sycl::queue q{sycl::gpu_selector_v, properties};
+
+  sycl::ext::oneapi::experimental::command_graph g;
+
+  const size_t n = 10;
+  float *arr = sycl::malloc_shared<float>(n, q);
+  for (int i = 0; i < n; i++) {
+    arr[i] = 0;
+  }
+
+  g.add([&](sycl::handler &h) {
+    h.parallel_for(sycl::range<1>{n}, [=](sycl::id<1> idx) {
+      size_t i = idx;
+      arr[i] += 1;
+    });
+  });
+
+  bool check = true;
+  for (int i = 0; i < n; i++) {
+    if (arr[i] != 0)
+      check = false;
+  }
+
+  auto executable_graph = g.finalize(q.get_context());
+
+  for (int i = 0; i < n; i++) {
+    if (arr[i] != 0)
+      check = false;
+  }
+
+  q.submit([&](sycl::handler &h) { h.ext_oneapi_graph(executable_graph); });
+
+  for (int i = 0; i < n; i++) {
+    if (arr[i] != 1)
+      check = false;
+  }
+
+  q.submit([&](sycl::handler &h) { h.ext_oneapi_graph(executable_graph); });
+
+  for (int i = 0; i < n; i++)
+    assert(arr[i] == 2);
+
+  sycl::free(arr, q);
+
+  return 0;
+}

sycl/test/graph/graph-explicit-saxpy.cpp

@@ -1,6 +1,6 @@
 // RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
-#include <sycl/sycl.hpp>
 #include <iostream>
+#include <sycl/sycl.hpp>
 
 #include <sycl/ext/oneapi/experimental/graph.hpp>
 
@@ -38,12 +38,15 @@ int main() {
 
   auto executable_graph = g.finalize(q.get_context());
 
-  q.submit([&](sycl::handler &h) { h.ext_oneapi_graph(executable_graph); }).wait();
+  q.submit([&](sycl::handler &h) {
+     h.ext_oneapi_graph(executable_graph);
+   }).wait();
+
+  for (int i = 0; i < n; i++)
+    assert(y[i] == 5.0f);
 
   sycl::free(x, q);
   sycl::free(y, q);
 
-  std::cout << "done.\n";
-
   return 0;
 }

sycl/test/graph/graph-explicit-single-node.cpp

@@ -1,6 +1,6 @@
 // RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
-#include <CL/sycl.hpp>
 #include <iostream>
+#include <sycl/sycl.hpp>
 
 #include <sycl/ext/oneapi/experimental/graph.hpp>
 
@@ -16,6 +16,9 @@ int main() {
 
   const size_t n = 10;
   float *arr = sycl::malloc_shared<float>(n, q);
+  for (int i = 0; i < n; i++) {
+    arr[i] = 0;
+  }
 
   g.add([&](sycl::handler &h) {
     h.parallel_for(sycl::range<1>{n}, [=](sycl::id<1> idx) {
@@ -24,19 +27,25 @@ int main() {
     });
   });
 
-  auto result_before_exec1 = arr[0];
+  bool check = true;
+  for (int i = 0; i < n; i++) {
+    if (arr[i] != 0)
+      check = false;
+  }
 
   auto executable_graph = g.finalize(q.get_context());
 
-  auto result_before_exec2 = arr[0];
+  for (int i = 0; i < n; i++) {
+    if (arr[i] != 0)
+      check = false;
+  }
 
   q.submit([&](sycl::handler &h) { h.ext_oneapi_graph(executable_graph); });
 
-  auto result = arr[0];
+  for (int i = 0; i < n; i++)
+    assert(arr[i] == 1);
 
   sycl::free(arr, q);
 
-  std::cout << "done.\n";
-
   return 0;
 }

sycl/test/graph/graph-explicit-subgraph.cpp

@@ -2,8 +2,8 @@
 
 // Modified version of the dotp example which submits part of the graph as a
 // sub-graph
-#include <CL/sycl.hpp>
 #include <iostream>
+#include <sycl/sycl.hpp>
 
 #include <sycl/ext/oneapi/experimental/graph.hpp>
 
@@ -70,7 +70,7 @@ int main() {
   auto subGraphExec = subGraph.finalize(q.get_context());
 
   auto node_sub =
-      g.add([&](sycl::handler &h) { h.exec_graph(subGraphExec); }, {n_i});
+      g.add([&](sycl::handler &h) { h.ext_oneapi_graph(subGraphExec); }, {n_i});
 
   auto node_c = g.add(
       [&](sycl::handler &h) {
@@ -86,18 +86,14 @@ int main() {
   auto executable_graph = g.finalize(q.get_context());
 
   // Using shortcut for executing a graph of commands
-  q.exec_graph(executable_graph).wait();
+  q.ext_oneapi_graph(executable_graph).wait();
 
-  if (*dotp != host_gold_result()) {
-    std::cout << "Error unexpected result!\n";
-  }
+  assert(*dotp == host_gold_result());
 
   sycl::free(dotp, q);
   sycl::free(x, q);
   sycl::free(y, q);
   sycl::free(z, q);
 
-  std::cout << "done.\n";
-
   return 0;
 }