To explain this slide, in the 1D blocked assignment:
For every block of $N^2/P$ elements computed, there is one row sent and one row received, giving us $ 2N $ elements communicated per computation block.
In the 1D interleaved assignment:
For every row of $N$ elements computed, there is one row sent and one row received, giving us $2N$ elements per row.
pdp
@machine6, I think its computed as in 1D blocked, 2 rows sent (one to the top and one to the bottom) and in 1D interleaved, its 1 row sent twice which accounts for 2 rows worth of elements. Please correct me if I am wrong.
To explain this slide, in the 1D blocked assignment:
In the 1D interleaved assignment:
@machine6, I think its computed as in 1D blocked, 2 rows sent (one to the top and one to the bottom) and in 1D interleaved, its 1 row sent twice which accounts for 2 rows worth of elements. Please correct me if I am wrong.