Commit 5b16542
committed
SWITCH_RATIO for Arm(R) Neoverse(TM) architecture
This seems like a good balance of values for reasonably sized matrices. With `SWITCH_RATIO=16` the DGEMM scales better to bigger sizes but the better solution would be some kind of
thread throttling so I've gone with `SWITCH_RATIO=8`.1 parent a5e1fdd commit 5b16542
1 file changed
Lines changed: 18 additions & 2 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
1 | 1 | | |
2 | | - | |
| 2 | + | |
3 | 3 | | |
4 | 4 | | |
5 | 5 | | |
| |||
3338 | 3338 | | |
3339 | 3339 | | |
3340 | 3340 | | |
| 3341 | + | |
| 3342 | + | |
| 3343 | + | |
| 3344 | + | |
| 3345 | + | |
| 3346 | + | |
3341 | 3347 | | |
3342 | 3348 | | |
3343 | 3349 | | |
| |||
3367 | 3373 | | |
3368 | 3374 | | |
3369 | 3375 | | |
3370 | | - | |
| 3376 | + | |
| 3377 | + | |
| 3378 | + | |
| 3379 | + | |
| 3380 | + | |
3371 | 3381 | | |
3372 | 3382 | | |
3373 | 3383 | | |
| |||
3398 | 3408 | | |
3399 | 3409 | | |
3400 | 3410 | | |
| 3411 | + | |
| 3412 | + | |
| 3413 | + | |
| 3414 | + | |
| 3415 | + | |
| 3416 | + | |
3401 | 3417 | | |
3402 | 3418 | | |
3403 | 3419 | | |
| |||
0 commit comments