|
|
|
|
|
|
by layer8
1481 days ago
|
|
|
TLDR: Due to loop-carried dependencies preventing parallelized execution: https://en.wikipedia.org/wiki/Loop_dependence_analysis#Loop-... In the 2 additions version, computation of the next iteration depends on the results of the preceding iteration. In the multiplication version, the computations are independent for each iteration, enabling parallel execution (by SIMD and/or pipelined/superscalar execution). |
|
|