This is pretty much the same as for lpc16, though it only improves half as large prediction orders. With 128-bit vectors, this gives: C V old V new 1 69.2 181.5 95.5 2 107.7 180.7 95.2 3 145.5 180.0 103.5 4 183.0 179.2 102.7 5 220.7 178.5 128.0 6 257.7 194.0 127.5 7 294.5 193.7 126.7 8 331.0 193.0 126.5 Larger prediction orders see no significant changes at that size.
Unified · Split
| @@ -71,17 +71,18 @@ av_cold void ff_flacdsp_init_riscv(FLACDSPContext *c, enum AVSampleFormat fmt, · 70 unmodified lines | ||
| 71 | 71 | if ((flags & AV_CPU_FLAG_RVV_I32) && (flags & AV_CPU_FLAG_RVB_ADDR)) { |
| 72 | 72 | int vlenb = ff_get_rv_vlenb(); |
| 73 | 73 | |
| 74 | − if ((flags & AV_CPU_FLAG_RVB_BASIC) && vlenb >= 16) | |
| 74 | + if ((flags & AV_CPU_FLAG_RVB_BASIC) && vlenb >= 16) { | |
| 75 | 75 | c->lpc16 = ff_flac_lpc16_rvv; |
| 76 | 76 | |
| 77 | 77 | # if (__riscv_xlen >= 64) |
| 78 | − if (flags & AV_CPU_FLAG_RVV_I64) { | |
| 79 | − if (vlenb > 16) | |
| 80 | − c->lpc32 = ff_flac_lpc32_rvv_simple; | |
| 81 | − else | |
| 82 | − c->lpc32 = ff_flac_lpc32_rvv; | |
| 83 | − } | |
| 78 | + if (flags & AV_CPU_FLAG_RVV_I64) { | |
| 79 | + if (vlenb > 16) | |
| 80 | + c->lpc32 = ff_flac_lpc32_rvv_simple; | |
| 81 | + else | |
| 82 | + c->lpc32 = ff_flac_lpc32_rvv; | |
| 83 | + } | |
| 84 | 84 | # endif |
| 85 | + } | |
| 85 | 86 | |
| 86 | 87 | c->wasted32 = ff_flac_wasted32_rvv; |
| 87 | 88 | |
| @@ -76,22 +76,24 @@ func ff_flac_lpc32_rvv, zve64x · 75 unmodified lines | ||
| 76 | 76 | ret |
| 77 | 77 | endfunc |
| 78 | 78 | |
| 79 | −func ff_flac_lpc32_rvv_simple, zve64x | |
| 80 | − vsetivli zero, 1, e64, m1, ta, ma | |
| 79 | +func ff_flac_lpc32_rvv_simple, zve64x, zbb | |
| 80 | + vtype_vli t3, a2, t1, e64, ta, ma | |
| 81 | + vntypei t2, t3 | |
| 82 | + vsetvl zero, a2, t3 // e64 | |
| 81 | 83 | vmv.s.x v0, zero |
| 82 | − vsetvli zero, a2, e32, m4, ta, ma | |
| 84 | + vsetvl zero, zero, t2 // e32 | |
| 83 | 85 | vle32.v v8, (a1) |
| 84 | 86 | sub a4, a4, a2 |
| 85 | 87 | vle32.v v16, (a0) |
| 86 | 88 | sh2add a0, a2, a0 |
| 87 | 89 | 1: |
| 88 | 90 | vwmul.vv v24, v8, v16 |
| 89 | − vsetvli zero, zero, e64, m8, ta, ma | |
| 91 | + vsetvl zero, zero, t3 // e64 | |
| 90 | 92 | vredsum.vs v24, v24, v0 |
| 91 | 93 | lw t0, (a0) |
| 92 | 94 | addi a4, a4, -1 |
| 93 | 95 | vmv.x.s t1, v24 |
| 94 | − vsetvli zero, zero, e32, m4, ta, ma | |
| 96 | + vsetvl zero, zero, t2 // e32 | |
| 95 | 97 | sra t1, t1, a3 |
| 96 | 98 | add t0, t0, t1 |
| 97 | 99 | vslide1down.vx v16, v16, t0 |