This slightly improves performance with the Device Under Test.
Unified · Split
| @@ -22,13 +22,13 @@ · 21 unmodified lines | |||
| 22 | 22 | ||
| 23 | func ff_ps_add_squares_rvv, zve32f | 23 | func ff_ps_add_squares_rvv, zve32f |
| 24 | 1: | 24 | 1: |
| 25 | vsetvli t0, a2, e32, m1, ta, ma | 25 | vsetvli t0, a2, e32, m4, ta, ma |
| 26 | vlseg2e32.v v24, (a1) | 26 | vlseg2e32.v v24, (a1) |
| 27 | sub a2, a2, t0 | 27 | sub a2, a2, t0 |
| 28 | vle32.v v16, (a0) | 28 | vle32.v v16, (a0) |
| 29 | sh3add a1, t0, a1 | 29 | sh3add a1, t0, a1 |
| 30 | vfmacc.vv v16, v24, v24 | 30 | vfmacc.vv v16, v24, v24 |
| 31 | vfmacc.vv v16, v25, v25 | 31 | vfmacc.vv v16, v28, v28 |
| 32 | vse32.v v16, (a0) | 32 | vse32.v v16, (a0) |
| 33 | sh2add a0, t0, a0 | 33 | sh2add a0, t0, a0 |
| 34 | bnez a2, 1b | 34 | bnez a2, 1b |