Test releasing a command-buffer after submission but before execution has finished (#2414)

Add cl_khr_command_buffer test that is it valid to release a
command-buffer after it has been enqueued but before execution is
finished.

This stresses the semantics from
[clReleaseCommandBufferKHR](https://registry.khronos.org/OpenCL/sdk/3.0/docs/man/html/clReleaseCommandBufferKHR.html#_description)
that: "After the command_buffer reference count becomes zero **and has
finished execution**, the command-buffer is deleted"

test_conformance/extensions/cl_khr_command_buffer/basic_command_buffer.cpp

@@ -435,3 +435,40 @@ bool InterleavedEnqueueTest::Skip()
 {
     return BasicCommandBufferTest::Skip() || !simultaneous_use_support;
 }
+
+cl_int EnqueueAndReleaseTest::Run()
+{
+    cl_int error = clCommandNDRangeKernelKHR(
+        command_buffer, nullptr, nullptr, kernel, 1, nullptr, &num_elements,
+        nullptr, 0, nullptr, nullptr, nullptr);
+    test_error(error, "clCommandNDRangeKernelKHR failed");
+
+    error = clFinalizeCommandBufferKHR(command_buffer);
+    test_error(error, "clFinalizeCommandBufferKHR failed");
+
+    cl_int pattern = 42;
+    error = clEnqueueFillBuffer(queue, in_mem, &pattern, sizeof(cl_int), 0,
+                                data_size(), 0, nullptr, nullptr);
+    test_error(error, "clEnqueueFillBuffer failed");
+
+    error = clEnqueueCommandBufferKHR(0, nullptr, command_buffer, 0, nullptr,
+                                      nullptr);
+    test_error(error, "clEnqueueCommandBufferKHR failed");
+
+    // Calls release on cl_command_buffer_khr handle inside wrapper class, and
+    // sets the handle to nullptr, so that release doesn't get called again at
+    // end of test when wrapper object is destroyed.
+    command_buffer.reset();
+
+    std::vector<cl_int> output_data(num_elements);
+    error = clEnqueueReadBuffer(queue, out_mem, CL_TRUE, 0, data_size(),
+                                output_data.data(), 0, nullptr, nullptr);
+    test_error(error, "clEnqueueReadBuffer failed");
+
+    for (size_t i = 0; i < num_elements; i++)
+    {
+        CHECK_VERIFICATION_ERROR(pattern, output_data[i], i);
+    }
+
+    return CL_SUCCESS;
+}

test_conformance/extensions/cl_khr_command_buffer/basic_command_buffer.h

@@ -128,6 +128,15 @@ struct InterleavedEnqueueTest : public BasicCommandBufferTest
     bool Skip() override;
 };
 
+// Test releasing a command-buffer after it has been submitted for execution,
+// but before the user has waited on completion of the enqueue.
+struct EnqueueAndReleaseTest : public BasicCommandBufferTest
+{
+    using BasicCommandBufferTest::BasicCommandBufferTest;
+
+    cl_int Run() override;
+};
+
 template <class T>
 int MakeAndRunTest(cl_device_id device, cl_context context,
                    cl_command_queue queue, int num_elements)

test_conformance/extensions/cl_khr_command_buffer/basic_command_buffer_tests.cpp

@@ -44,3 +44,9 @@ REGISTER_TEST(explicit_flush)
     return MakeAndRunTest<ExplicitFlushTest>(device, context, queue,
                                              num_elements);
 }
+
+REGISTER_TEST(enqueue_and_release)
+{
+    return MakeAndRunTest<EnqueueAndReleaseTest>(device, context, queue,
+                                                 num_elements);
+}