ms-20240504.cu
Differences
This shows you the differences between two versions of the page.
| Both sides previous revisionPrevious revisionNext revision | Previous revision | ||
| ms-20240504.cu [2024/07/03 19:42] – mino | ms-20240504.cu [2024/10/16 15:05] (current) – mino | ||
|---|---|---|---|
| Line 33: | Line 33: | ||
| because the upper 32 bits of v2 should be ignored when compared with the 32bit unsigned rev_max32_c. | because the upper 32 bits of v2 should be ignored when compared with the 32bit unsigned rev_max32_c. | ||
| - | However, this flaw is irrelevant to the validity of the result because this logical operation is totally unnecessary and has no effect for even N (including 6). And for N < 6, the upper 32 bits of v2 are always 0s and the cast has no effect in any way. | + | However, this flaw is irrelevant to the validity of the result because this logical operation is totally unnecessary and has no effect for even N (including 6) because complements of any rows never appear as columns in even order magic squares. And for N < 6, the upper 32 bits of v2 are always 0s and the cast has no effect in any way. |
| - | This part of the code and other parts related to the self complemenarity check should be removed from functions MakeCols() and LastCols() in order to optimize the performance for N=6. Complements | + | This part of the code and other parts related to the self complemenarity check should be removed from functions MakeCols() and LastCols() in order to optimize the performance for N=6. |
| + | We will gain 3% improvement | ||
| <file cpp ms-20240504.cu> | <file cpp ms-20240504.cu> | ||
ms-20240504.cu.1720003326.txt.gz · Last modified: 2024/07/03 19:42 by mino