Added use-onchip-memory sample

samples/use-onchip-memory/CMakeLists.txt

@@ -0,0 +1,21 @@
+set(SOURCE_NAME "use-onchip-memory")
+
+add_executable(
+ ${SOURCE_NAME}
+ ${CMAKE_CURRENT_SOURCE_DIR}/${SOURCE_NAME}.cpp
+)
+include_directories(
+ ${PROJECT_SOURCE_DIR}/include
+)
+add_sycl_to_target(
+ TARGET ${SOURCE_NAME}
+ SOURCES ${CMAKE_CURRENT_SOURCE_DIR}/${SOURCE_NAME}.cpp
+)
+add_test(
+ NAME ${SOURCE_NAME}
+ COMMAND ${SOURCE_NAME}
+)
+install(
+ TARGETS ${SOURCE_NAME}
+ RUNTIME DESTINATION bin
+)

samples/use-onchip-memory/use-onchip-memory.cpp

@@ -0,0 +1,142 @@
+/***************************************************************************
+ *
+ * Copyright Codeplay Software Limited
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * For your convenience, a copy of the License has been included in this
+ * repository.
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * Codeplay's ComputeCpp SDK
+ *
+ * example_vptr.cpp
+ *
+ * Description:
+ * Sample code that demonstrates the use of the use_onchip_memory extension
+ * to SYCL provided by ComputeCpp.
+ *
+ **************************************************************************/
+
+#include <CL/sycl.hpp>
+#include <SYCL/codeplay.hpp>
+#include <iostream>
+
+namespace sycl = cl::sycl;
+namespace access = sycl::access;
+
+// Sample kernel
+// -------------
+// Since kernels are no longer allowed to be declared inline, I dislike
+// declaring a random name in the global namespace that can potentially be
+// repurposed.
+//
+// This kernel performs the same operation as std::iota, but also scales
+// the result by two.
+//
+class scaled_iota {
+ public:
+ using accessor = sycl::accessor<int, 1, access::mode::discard_write,
+ access::target::global_buffer>;
+
+ explicit scaled_iota(accessor deviceAccess) noexcept
+ : m_deviceAccess{std::move(deviceAccess)} {}
+
+ void operator()(sycl::nd_item<2> itemId) const noexcept {
+ const auto linearId = itemId.get_global_linear_id();
+ m_deviceAccess[linearId] = linearId * 2;
+ }
+
+ private:
+ accessor m_deviceAccess;
+};
+
+namespace codeplay = sycl::codeplay;
+
+template <class Policy>
+void use_with_policy(Policy policy, sycl::queue& queue) {
+ auto hostData = sycl::vector_class<int>(1024);
+ {
+ auto taskContext = queue.get_context();
+
+ // Notice that the Codeplay property takes a policy argument: this is used
+ // to indicate whether the property is advantageous or genuinely necessary.
+ //
+ auto deviceData = sycl::buffer<int, 1>{
+ hostData.data(), //
+ sycl::range<1>(hostData.size()), //
+ sycl::property_list{
+ sycl::property::buffer::context_bound(taskContext),
+ codeplay::property::buffer::use_onchip_memory(policy) // <--------
+ } // ^---------------------
+ };
+
+ deviceData.set_final_data(hostData.data());
+
+ queue.submit([&](sycl::handler & cgh) {
+ constexpr auto dimension_size = 2;
+ auto r = sycl::nd_range<dimension_size>{
+ sycl::range<dimension_size>{
+ hostData.size() / dimension_size, //
+ dimension_size //
+ },
+ sycl::range<dimension_size>{dimension_size, 1} //
+ };
+ cgh.parallel_for(
+ r,
+ scaled_iota(deviceData.get_access<access::mode::discard_write>(cgh)));
+ });
+
+ queue.wait_and_throw();
+ }
+}
+
+// Codeplay policy extensions have two different enabling mechanisms: the first
+// is to indicate that a policy is preferred. Using this policy means that if
+// the system supports the feature, then the feature will be enabled. If the
+// feature is not present on the system, then it will not be enabled.
+//
+// Puns aside, this is the preferred default.
+//
+void how_to_use_with_prefer() {
+ auto queue = sycl::queue{};
+ ::use_with_policy(codeplay::property::prefer, queue);
+}
+
+// Alternatively, if you can guarantee that your system will support this
+// policy, or if it is expected any system using your software must support the
+// policy, then you can use the require tag to indicate that the feature is
+// required by your software.
+//
+// In the event that the property isn't supported, an exception will be thrown.
+//
+void how_to_use_with_require() {
+ try {
+ auto queue = sycl::queue{};
+ ::use_with_policy(codeplay::property::require, queue);
+ } catch (const sycl::exception& e) {
+ std::cerr << "An error occurred: " << e.what()
+ << "\n"
+ "\n"
+ "This particular error has occurred because you are requiring "
+ "the policy use_onchip_memory be available, and your hardware "
+ "doesn't support the use_onchip_memory, so the SYCL ecosystem "
+ "will raise an error.\n";
+ }
+}
+
+int main() {
+ // Using the on-chip memory policy with the require tag.
+ how_to_use_with_require();
+
+ // Using the on-chip memory policy with the prefer tag.
+ how_to_use_with_prefer();
+}