Visible to Intel only — GUID: GUID-51580037-C890-4D82-B157-70622BCD3217
Introduction to SYCL Essentials
Module 1: oneAPI Intro
Module 2: DPCPP Program Structure
Module 3: DPCPP Unified Shared Memory
Module 4: DPCPP Sub-Groups
Module 5: Intel® Advisor
Module 6: VTune™ Profiler
Module 7: DPCPP Library
Module 8: DPCPP Reduction
Module 9: DPCPP Buffers And Accessors In Depth
Module 10: DPCPP Graphs Scheduling Data Management
Module 11: Intel® Distribution for GDB
Module 12: DPCPP Local Memory And Atomics
Visible to Intel only — GUID: GUID-51580037-C890-4D82-B157-70622BCD3217
USM vector_add with SYCL
In this discussion, we implement vector addition using Unified Shared Memory (USM), highlighting how the SYCL programming model simplifies the utilization of heterogeneous computing resources. The main takeaway is that with USM there is no requirement to define buffers, handlers, and accessors. In this case, the USM malloc_shared function is sufficient to allocate and manage memory access:
//==============================================================
// Copyright © Intel Corporation
//
// SPDX-License-Identifier: MIT
// =============================================================
#include <CL/sycl.hpp>
using namespace sycl;
static const int N = 256;
int main() {
queue q;
// STEP 1 : Get device information
std::cout << "Device : " << q.get_device().get_info<info::device::name>() << "\n";
// STEP 2 : Initialize vectors in shared memory
int *data1 = malloc_shared<int>(N, q);
int *data2 = malloc_shared<int>(N, q);
int *dataR = malloc_shared<int>(N, q);
// STEP 3 : Asign values to vectors
for (int i = 0; i < N; i++) {
data1[i] = 10;
data2[i] = 20;
dataR[i] = 0;
}
// STEP 4 : Proceed with the calculation
q.parallel_for(range<1>(N), [=](id<1> i) {
dataR[i] = data1[i] + data2[i];
}).wait();
// STEP 5 : Print output values
for (int i = 0; i < N; i++) std::cout << dataR[i] << " ";
std::cout << "\n";
// STEP 6 : Release memory
free(data1, q);
free(data2, q);
free(dataR, q);
return 0;
}This part of the code:
std::cout << "Device : " << q.get_device().get_info<info::device::name>() << "\n";gets device information from the q, and prints information about the selected device.
USM gets established here:
int *data1 = malloc_shared<int>(N, q);
int *data2 = malloc_shared<int>(N, q);
int *dataR = malloc_shared<int>(N, q);where three shared memory arrays data1, data2, and dataR are allocated using malloc_shared<int>(N, q). They are accessible during computation by both the host and device. USM also offers malloc_device and malloc_host to explicitly allocate memory on the host or device, but they will not be covered in this course.
So, when should you use USM versus the conventional SYCL? In summary, using USM is a good approach to consider when:
The code has many pointers that are susceptible to future changes.
When it is expected that the application will gain complexity over time.