Even if the speed of one multiplication is the same, you can fit more in a vectorized operation. And the memory bandwidth is utilized better when you move around half the amount of data.
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Use case of single-precision real number | 20 | 1222 | March 17, 2024 | |
| Single, Double or Mixed Precision? | 14 | 1258 | February 7, 2024 | |
|
Question about double precision on ARM processors
|
52 | 1985 | November 19, 2023 | |
| "real" type of a calculation with mixed precisions | 10 | 487 | July 29, 2024 | |
|
Array intrinsics performances/accuracy
|
20 | 894 | May 20, 2023 |