README.md

SYCL Academy

Exercise 3: Scalar Add

---

In this exercise you will learn how to create buffers to manage data and accessors to access the data within a kernel function.

---

1.) Allocate your input and output

Allocate memory on the host for your input and output data variables and assign values to the inputs.

2.) Construct buffers

In SYCL buffers are used to manage data across the host and device(s).

Construct a buffer to manage your input and output data. The parameters to construct a buffer are a pointer to the host data and a 1 dimensional range of 1 to represent a single value. The element type and dimensionality can be inferred from the pointer and the range.

3.) Construct accessors

In SYCL accessors are used to declare data dependencies to a SYCL kernel function as well as to access the data within a SYCL kernel function.

Construct an accessor for each of the buffers to access the data of each within the kernel function. The parameters to construct an accessor are the buffer and the handler.

4.) Declare your kernel

Declare a SYCL kernel function using the single_task command provide a lambda as the kernel function. The kernel function should use the operator[] of the accessor objects to read from the inputs and write the sum to the output. As each accessor is only accessing a single element you can simply specify 0.

5.) Try a temporary buffer

You can construct a temporary buffer that doesn't copy back on destruction by initializing it with just a range and no host pointer.

Build And Execution Hints

For DPC++: Using CMake to configure then build the exercise:

mkdir build
cd build
cmake .. "-GUnix Makefiles" -DSYCL_ACADEMY_USE_DPCPP=ON -DSYCL_ACADEMY_ENABLE_SOLUTIONS=OFF -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx
make exercise_3

Alternatively from a terminal at the command line:

icpx -fsycl -o sycl-ex-3 -I../External/Catch2/single_include ../Code_Exercises/Exercise_03_Scalar_Add/source.cpp
./sycl-ex-3

In Intel DevCloud, to run computational applications, you will submit jobs to a queue for execution on compute nodes, especially some features like longer walltime and multi-node computation is only available through the job queue. Please refer to the guide.

So wrap the binary into a script job_submission and run:

qsub job_submission

For AdaptiveCpp:

# <target specification> is a list of backends and devices to target, for example
# "omp;generic" compiles for CPUs with the OpenMP backend and GPUs using the generic single-pass compiler.
# The simplest target specification is "omp" which compiles for CPUs using the OpenMP backend.
cmake -DSYCL_ACADEMY_USE_ADAPTIVECPP=ON -DSYCL_ACADEMY_INSTALL_ROOT=/insert/path/to/adapivecpp -DACPP_TARGETS="<target specification>" ..
make exercise_3

alternatively, without CMake:

cd Code_Exercises/Exercise_03_Scalar_Add
/path/to/AdaptiveCpp/bin/acpp -o sycl-ex-3 -I../../External/Catch2/single_include --acpp-targets="<target specification>" source.cpp
./sycl-ex-3