C908: sse_0_c: 614.7 sse_0_rvv_i32: 138.2 sse_1_c: 302.7 sse_1_rvv_i32: 107.2 sse_2_c: 175.7 sse_2_rvv_i32: 104.2
Unified · Split
| @@ -39,6 +39,13 @@ int ff_pix_abs16_y2_rvv(MpegEncContext *v, const uint8_t *pix1, const uint8_t *p · 38 unmodified lines | ||
| 39 | 39 | int ff_pix_abs8_y2_rvv(MpegEncContext *v, const uint8_t *pix1, const uint8_t *pix2, |
| 40 | 40 | ptrdiff_t stride, int h); |
| 41 | 41 | |
| 42 | +int ff_sse16_rvv(MpegEncContext *v, const uint8_t *pix1, const uint8_t *pix2, | |
| 43 | + ptrdiff_t stride, int h); | |
| 44 | +int ff_sse8_rvv(MpegEncContext *v, const uint8_t *pix1, const uint8_t *pix2, | |
| 45 | + ptrdiff_t stride, int h); | |
| 46 | +int ff_sse4_rvv(MpegEncContext *v, const uint8_t *pix1, const uint8_t *pix2, | |
| 47 | + ptrdiff_t stride, int h); | |
| 48 | + | |
| 42 | 49 | av_cold void ff_me_cmp_init_riscv(MECmpContext *c, AVCodecContext *avctx) |
| 43 | 50 | { |
| 44 | 51 | #if HAVE_RVV |
| @@ -53,6 +60,10 @@ av_cold void ff_me_cmp_init_riscv(MECmpContext *c, AVCodecContext *avctx) · 8 unmodified lines | ||
| 53 | 60 | c->pix_abs[1][1] = ff_pix_abs8_x2_rvv; |
| 54 | 61 | c->pix_abs[0][2] = ff_pix_abs16_y2_rvv; |
| 55 | 62 | c->pix_abs[1][2] = ff_pix_abs8_y2_rvv; |
| 63 | + | |
| 64 | + c->sse[0] = ff_sse16_rvv; | |
| 65 | + c->sse[1] = ff_sse8_rvv; | |
| 66 | + c->sse[2] = ff_sse4_rvv; | |
| 56 | 67 | } |
| 57 | 68 | #endif |
| 58 | 69 | } |
| @@ -165,3 +165,69 @@ func ff_pix_abs8_y2_rvv, zve32x · 164 unmodified lines | ||
| 165 | 165 | |
| 166 | 166 | pix_abs_ret |
| 167 | 167 | endfunc |
| 168 | + | |
| 169 | +func ff_sse16_rvv, zve32x | |
| 170 | + vsetivli t0, 16, e32, m4, ta, ma | |
| 171 | + vmv.v.x v24, zero | |
| 172 | + vmv.s.x v0, zero | |
| 173 | +1: | |
| 174 | + vsetvli zero, zero, e8, m1, tu, ma | |
| 175 | + vle8.v v4, (a1) | |
| 176 | + vle8.v v12, (a2) | |
| 177 | + addi a4, a4, -1 | |
| 178 | + vwsubu.vv v16, v4, v12 | |
| 179 | + vsetvli zero, zero, e16, m2, tu, ma | |
| 180 | + vwmacc.vv v24, v16, v16 | |
| 181 | + add a1, a1, a3 | |
| 182 | + add a2, a2, a3 | |
| 183 | + bnez a4, 1b | |
| 184 | + | |
| 185 | + vsetvli zero, zero, e32, m4, tu, ma | |
| 186 | + vredsum.vs v0, v24, v0 | |
| 187 | + vmv.x.s a0, v0 | |
| 188 | + ret | |
| 189 | +endfunc | |
| 190 | + | |
| 191 | +func ff_sse8_rvv, zve32x | |
| 192 | + vsetivli t0, 8, e32, m2, ta, ma | |
| 193 | + vmv.v.x v24, zero | |
| 194 | + vmv.s.x v0, zero | |
| 195 | +1: | |
| 196 | + vsetvli zero, zero, e8, mf2, tu, ma | |
| 197 | + vle8.v v4, (a1) | |
| 198 | + vle8.v v12, (a2) | |
| 199 | + addi a4, a4, -1 | |
| 200 | + vwsubu.vv v16, v4, v12 | |
| 201 | + vsetvli zero, zero, e16, m1, tu, ma | |
| 202 | + vwmacc.vv v24, v16, v16 | |
| 203 | + add a1, a1, a3 | |
| 204 | + add a2, a2, a3 | |
| 205 | + bnez a4, 1b | |
| 206 | + | |
| 207 | + vsetvli zero, zero, e32, m2, tu, ma | |
| 208 | + vredsum.vs v0, v24, v0 | |
| 209 | + vmv.x.s a0, v0 | |
| 210 | + ret | |
| 211 | +endfunc | |
| 212 | + | |
| 213 | +func ff_sse4_rvv, zve32x | |
| 214 | + vsetivli t0, 4, e32, m1, ta, ma | |
| 215 | + vmv.v.x v24, zero | |
| 216 | + vmv.s.x v0, zero | |
| 217 | +1: | |
| 218 | + vsetvli zero, zero, e8, mf4, tu, ma | |
| 219 | + vle8.v v4, (a1) | |
| 220 | + vle8.v v12, (a2) | |
| 221 | + addi a4, a4, -1 | |
| 222 | + vwsubu.vv v16, v4, v12 | |
| 223 | + vsetvli zero, zero, e16, mf2, tu, ma | |
| 224 | + vwmacc.vv v24, v16, v16 | |
| 225 | + add a1, a1, a3 | |
| 226 | + add a2, a2, a3 | |
| 227 | + bnez a4, 1b | |
| 228 | + | |
| 229 | + vsetvli zero, zero, e32, m1, tu, ma | |
| 230 | + vredsum.vs v0, v24, v0 | |
| 231 | + vmv.x.s a0, v0 | |
| 232 | + ret | |
| 233 | +endfunc | |