Skip to content

Latest commit

 

History

History
93 lines (69 loc) · 3.31 KB

README.md

File metadata and controls

93 lines (69 loc) · 3.31 KB
Raw

SYCL Academy

Exercise 5: Device Selection

In this exercise you will learn how to create a device selector that will choose a device for you to enqueue work to.

1.) Query the device of your queue

When you default construct a queue the runtime will use the default_selector_v to choose a device.

Try querying the device associated with the queue and information about it.

Remember the device associated with a queue can be retrieved using the get_device member function and information about a device can be queried using the get_info member function template.

2.) Try other device selectors

Replace the default selector with one of the other standard device selectors that are provided by SYCL such as the cpu_selector_v, gpu_selector_v and see which device those choose.

3.) Create your own device selector

Create a device selector using the template below. Implement the function call operator, using various device and platform info queries like the one we used earlier to query the device name and then use that device selector in the queue constructor.

To construct device selector, use the following rule:

The interface for a device selector is any object that meets the C++ named requirement Callable, taking a parameter of type const device & and returning a value that is implicitly convertible to int.

  int device_selector()(const device &device) { /* scoring logic */ }

Remember the platform associated with a device can be retrieved using the get_platform member function.

Remember that the value returned from the device selector's function call operator will represent the score for each device, and a device with a negative score will never be chosen.

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_5

Alternatively from a terminal at the command line:

icpx -fsycl -o sycl-ex-5 -I../External/Catch2/single_include ../Code_Exercises/Exercise_05_Device_Selection/source.cpp
./sycl-ex-5

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/adaptivecpp -DACPP_TARGETS="<target specification>" ..
make exercise_5

alternatively, without CMake:

cd Code_Exercises/Exercise_05_Device_Selection
/path/to/adaptivecpp/bin/acpp -o sycl-ex-5 -I../../External/Catch2/single_include --acpp-targets="<target specification>" source.cpp
./sycl-ex-5