While this function can easily be written with vectors, it just fails to get any performance improvement. For reference, this is a simpler loop-free implementation that does get better performance than the current one depending on hardware, but still more or less the same metrics as the C code: func ff_sbr_neg_odd_64_rvv, zve64x li a1, 32 addi a0, a0, 7 li t0, 8 vsetvli zero, a1, e8, m2, ta, ma li t1, 0x80 vlse8.v v8, (a0), t0 vxor.vx v8, v8, t1 vsse8.v v8, (a0), t0 ret endfunc This reverts commit d06fd18f8f4c6a81ef94cbb600620d83ad51269d.
Unified · Split
| @@ -26,7 +26,6 @@ · 25 unmodified lines | |||
| 26 | 26 | ||
| 27 | void ff_sbr_sum64x5_rvv(float *z); | 27 | void ff_sbr_sum64x5_rvv(float *z); |
| 28 | float ff_sbr_sum_square_rvv(float (*x)[2], int n); | 28 | float ff_sbr_sum_square_rvv(float (*x)[2], int n); |
| 29 | void ff_sbr_neg_odd_64_rvv(float *x); | ||
| 30 | void ff_sbr_autocorrelate_rvv(const float x[40][2], float phi[3][2][2]); | 29 | void ff_sbr_autocorrelate_rvv(const float x[40][2], float phi[3][2][2]); |
| 31 | void ff_sbr_hf_gen_rvv(float (*X_high)[2], const float (*X_low)[2], | 30 | void ff_sbr_hf_gen_rvv(float (*X_high)[2], const float (*X_low)[2], |
| 32 | const float alpha0[2], const float alpha1[2], | 31 | const float alpha0[2], const float alpha1[2], |
| @@ -64,9 +63,5 @@ av_cold void ff_sbrdsp_init_riscv(SBRDSPContext *c) · 31 unmodified lines | |||
| 64 | } | 63 | } |
| 65 | c->autocorrelate = ff_sbr_autocorrelate_rvv; | 64 | c->autocorrelate = ff_sbr_autocorrelate_rvv; |
| 66 | } | 65 | } |
| 67 | #if __riscv_xlen >= 64 | ||
| 68 | if ((flags & AV_CPU_FLAG_RVV_I64) && (flags & AV_CPU_FLAG_RVB_ADDR)) | ||
| 69 | c->neg_odd_64 = ff_sbr_neg_odd_64_rvv; | ||
| 70 | #endif | ||
| 71 | #endif | 66 | #endif |
| 72 | } | 67 | } |
| @@ -68,23 +68,6 @@ NOHWF fmv.x.w a0, fa0 · 67 unmodified lines | |||
| 68 | ret | 68 | ret |
| 69 | endfunc | 69 | endfunc |
| 70 | 70 | ||
| 71 | #if __riscv_xlen >= 64 | ||
| 72 | func ff_sbr_neg_odd_64_rvv, zve64x | ||
| 73 | li a1, 32 | ||
| 74 | li t1, 1 << 63 | ||
| 75 | 1: | ||
| 76 | vsetvli t0, a1, e64, m8, ta, ma | ||
| 77 | vle64.v v8, (a0) | ||
| 78 | sub a1, a1, t0 | ||
| 79 | vxor.vx v8, v8, t1 | ||
| 80 | vse64.v v8, (a0) | ||
| 81 | sh3add a0, t0, a0 | ||
| 82 | bnez t0, 1b | ||
| 83 | |||
| 84 | ret | ||
| 85 | endfunc | ||
| 86 | #endif | ||
| 87 | |||
| 88 | func ff_sbr_autocorrelate_rvv, zve32f | 71 | func ff_sbr_autocorrelate_rvv, zve32f |
| 89 | vsetvli t0, zero, e32, m4, ta, ma | 72 | vsetvli t0, zero, e32, m4, ta, ma |
| 90 | vmv.v.x v0, zero | 73 | vmv.v.x v0, zero |