kernels_parameters.asciidoc

Test plan for parameter passing to kernels

This is a test plan for rules for parameter passing to kernels defined in 4.12.4. Rules for parameter passing to kernels.

Other legal types of parameters covered in other groups of tests.

sampled_image_accessor and unsampled_image_accessor are expected to be tested in future for passing to kernels in tests for these accessors.

1. Testing scope

This plan does not include negative tests.

1.1. Device coverage

All the tests described below are performed only on the default device that is selected on the CTS command line.

1.2. Type coverage

Tests are performed for types:

• marray<T, 5>

• vec<T, 4>

• T[5]

• class S with a non-static member variable of type T

• class K non-virtual base class S

• range<N> where N = 1, 2, 3

• id<N> where N = 1, 2, 3

For following T:

• int

• float

• bool

2. Tests

2.1. Kernel is a named function object

• Define kernel as a named function object with member variable a of tested type A and constructor that takes sycl::accessor<A, 1> as a parameter.

• In kernel operator() write value of member variable a to accessor.

• In command group create buffer accessor to a variable b of type A.

• Use single_task with kernel<A>(accessor) as a parameter.

• Check that after command is executed variable b is equal to kernel member variable a.

2.2. Kernel is a lambda function

• Create 2 variables of tested type a and b with different values.

• Define kernel as a lambda function with by-copy capture default.

• In this lambda function copy value of a to b via buffer accessor.

• Check that variables are equal after command with this lambda is executed and buffer destructor is called.