Add bufferguard example

[bernhardmgruber]

This example showcases a custom LLAMA mapping splitting a 2D array of 3 component vectors into 9 regions.
These 9 regions are the center part of the array, 4 guard regions at the left, right, top and bottom border and the 4 corners of the array.
The content of the blobs belonging to one guard region of a view are copied to blobs of a corresponding guard region of another view.
Such a workflow is typical for HPC applications where the guard regions are exchanged between compute nodes.

Command line output:

View 1:
[  1,  2,  3][  4,  5,  6][  7,  8,  9][ 10, 11, 12][ 13, 14, 15]
[ 16, 17, 18][ 19, 20, 21][ 22, 23, 24][ 25, 26, 27][ 28, 29, 30]
[ 31, 32, 33][ 34, 35, 36][ 37, 38, 39][ 40, 41, 42][ 43, 44, 45]
[ 46, 47, 48][ 49, 50, 51][ 52, 53, 54][ 55, 56, 57][ 58, 59, 60]
[ 61, 62, 63][ 64, 65, 66][ 67, 68, 69][ 70, 71, 72][ 73, 74, 75]
[ 76, 77, 78][ 79, 80, 81][ 82, 83, 84][ 85, 86, 87][ 88, 89, 90]
[ 91, 92, 93][ 94, 95, 96][ 97, 98, 99][100,101,102][103,104,105]

View 2:
[  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]

Copy view 1 right -> view 2 left:
View 2:
[  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[ 28, 29, 30][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[ 43, 44, 45][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[ 58, 59, 60][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[ 73, 74, 75][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[ 88, 89, 90][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]

Copy view 1 left top -> view 2 right bot:
View 2:
[  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[ 28, 29, 30][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[ 43, 44, 45][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[ 58, 59, 60][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[ 73, 74, 75][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[ 88, 89, 90][  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0]
[  0,  0,  0][  0,  0,  0][  0,  0,  0][  0,  0,  0][  1,  2,  3]

Copy view 2 center -> view 1 center:
View 1:
[  1,  2,  3][  4,  5,  6][  7,  8,  9][ 10, 11, 12][ 13, 14, 15]
[ 16, 17, 18][  0,  0,  0][  0,  0,  0][  0,  0,  0][ 28, 29, 30]
[ 31, 32, 33][  0,  0,  0][  0,  0,  0][  0,  0,  0][ 43, 44, 45]
[ 46, 47, 48][  0,  0,  0][  0,  0,  0][  0,  0,  0][ 58, 59, 60]
[ 61, 62, 63][  0,  0,  0][  0,  0,  0][  0,  0,  0][ 73, 74, 75]
[ 76, 77, 78][  0,  0,  0][  0,  0,  0][  0,  0,  0][ 88, 89, 90]
[ 91, 92, 93][ 94, 95, 96][ 97, 98, 99][100,101,102][103,104,105]

Dump of mapping:

We can beautifully see the 4 blobs for the corner regions, the 2 long blobs for left/right, the 2 shorter blobs for bot/top and the large blob for the center region.

[bernhardmgruber]

SoA:

SoA with 1 blob per element (except first 4 blobs with are mapped using llama::mapping::One):

[psychocoderHPC]

@bernhardmgruber I love llama and I am so sad that I have not enough time to integrate it directly into PIConGPU 😢

[bernhardmgruber]

@bernhardmgruber I love llama and I am so sad that I have not enough time to integrate it directly into PIConGPU 😢

Thank you for the praise! :) In fact, this example is primarily motived by some ideas you sketched in the last alpaka VC. I built it for you <3. Jokes aside, LLAMA should see at least a proof-of-concept PIConGPU integration at some point.

[bussmann]

Github love stories ❤

[bussmann]

Btw: I really like this. Excellent work!

[bernhardmgruber]

Btw: I really like this. Excellent work!

Thank you! This means something to me.