tile_static Keyword

Article
04/28/2015

The tile_static keyword is used to declare a variable that can be accessed by all threads in a tile of threads. The lifetime of the variable starts when execution reaches the point of declaration and ends when the kernel function returns. For more information on using tiles, see Using Tiles.

The tile_static keyword has the following limitations:

It can be used only on variables that are in a function that has the restrict(amp) modifier.
It cannot be used on variables that are pointer or reference types.
A tile_static variable cannot have an initializer. Default constructors and destructors are not invoked automatically.
The value of an uninitialized tile_static variable is undefined.
If a tile_static variable is declared in a call graph that is rooted by a non-tiled call to parallel_for_each, a warning is generated and the behavior of the variable is undefined.

Example

The following example shows how a tile_static variable can be used to accumulate data across several threads in a tile.

// Sample data:
int sampledata[] = {
    2, 2, 9, 7, 1, 4,
    4, 4, 8, 8, 3, 4,
    1, 5, 1, 2, 5, 2,
    6, 8, 3, 2, 7, 2};

// The tiles:
// 2 2    9 7    1 4
// 4 4    8 8    3 4
//
// 1 5    1 2    5 2
// 6 8    3 2    7 2

// Averages:
int averagedata[] = { 
    0, 0, 0, 0, 0, 0, 
    0, 0, 0, 0, 0, 0, 
    0, 0, 0, 0, 0, 0, 
    0, 0, 0, 0, 0, 0, 
};

array_view<int, 2> sample(4, 6, sampledata);
array_view<int, 2> average(4, 6, averagedata);

parallel_for_each(
    // Create threads for sample.extent and divide the extent into 2 x 2 tiles.
    sample.extent.tile<2,2>(),
    [=](tiled_index<2,2> idx) restrict(amp)
    {
        // Create a 2 x 2 array to hold the values in this tile.
        tile_static int nums[2][2];
        // Copy the values for the tile into the 2 x 2 array.
        nums[idx.local[1]][idx.local[0]] = sample[idx.global];
        // When all the threads have executed and the 2 x 2 array is complete, find the average.
        idx.barrier.wait();
        int sum = nums[0][0] + nums[0][1] + nums[1][0] + nums[1][1];
        // Copy the average into the array_view.
        average[idx.global] = sum / 4;
      }
);

for (int i = 0; i < 4; i++) {
    for (int j = 0; j < 6; j++) {
        std::cout << average(i,j) << " ";
    }
    std::cout << "\n";
}

// Output:
// 3 3 8 8 3 3
// 3 3 8 8 3 3
// 5 5 2 2 4 4
// 5 5 2 2 4 4
// Sample data.
int sampledata[] = {
    2, 2, 9, 7, 1, 4,
    4, 4, 8, 8, 3, 4,
    1, 5, 1, 2, 5, 2,
    6, 8, 3, 2, 7, 2};

// The tiles are:
// 2 2    9 7    1 4
// 4 4    8 8    3 4
//
// 1 5    1 2    5 2
// 6 8    3 2    7 2

// Averages.
int averagedata[] = { 
    0, 0, 0, 0, 0, 0, 
    0, 0, 0, 0, 0, 0, 
    0, 0, 0, 0, 0, 0, 
    0, 0, 0, 0, 0, 0, 
};

array_view<int, 2> sample(4, 6, sampledata);
array_view<int, 2> average(4, 6, averagedata);

parallel_for_each(
    // Create threads for sample.grid and divide the grid into 2 x 2 tiles.
    sample.extent.tile<2,2>(),
    [=](tiled_index<2,2> idx) restrict(amp)
    {
        // Create a 2 x 2 array to hold the values in this tile.
        tile_static int nums[2][2];
        // Copy the values for the tile into the 2 x 2 array.
        nums[idx.local[1]][idx.local[0]] = sample[idx.global];
        // When all the threads have executed and the 2 x 2 array is complete, find the average.
        idx.barrier.wait();
        int sum = nums[0][0] + nums[0][1] + nums[1][0] + nums[1][1];
        // Copy the average into the array_view.
        average[idx.global] = sum / 4;
      }
);

for (int i = 0; i < 4; i++) {
    for (int j = 0; j < 6; j++) {
        std::cout << average(i,j) << " ";
    }
    std::cout << "\n";
}

// Output.
// 3 3 8 8 3 3
// 3 3 8 8 3 3
// 5 5 2 2 4 4
// 5 5 2 2 4 4

Partager via

tile_static Keyword

Example

See Also

Tasks

Reference

Concepts

Ressources supplémentaires