static WEBP_INLINE void VE4(uint8_t* dst, const uint8_t* top) { // vertical
const uint64_t val_m = LD(top - 1);
const v16u8 A = (v16u8)__msa_insert_d((v2i64)A, 0, val_m);
const v16u8 B = SLDI_UB(A, A, 1);
const v16u8 C = SLDI_UB(A, A, 2);
const v16u8 AC = __msa_ave_u_b(A, C);
const v16u8 B2 = __msa_ave_u_b(B, B);
const v16u8 R = __msa_aver_u_b(AC, B2);
const uint32_t out = __msa_copy_s_w((v4i32)R, 0);
SW4(out, out, out, out, dst, BPS);
}
static WEBP_INLINE void LD4(uint8_t* dst, const uint8_t* top) {
const uint64_t val_m = LD(top);
const v16u8 A = (v16u8)__msa_insert_d((v2i64)A, 0, val_m);
const v16u8 B = SLDI_UB(A, A, 1);
const v16u8 C1 = SLDI_UB(A, A, 2);
const v16u8 C = (v16u8)__msa_insert_b((v16i8)C1, 6, top[7]);
const v16u8 AC = __msa_ave_u_b(A, C);
const v16u8 B2 = __msa_ave_u_b(B, B);
const v16u8 R0 = __msa_aver_u_b(AC, B2);
const v16u8 R1 = SLDI_UB(R0, R0, 1);
const v16u8 R2 = SLDI_UB(R1, R1, 1);
const v16u8 R3 = SLDI_UB(R2, R2, 1);
const uint32_t val0 = __msa_copy_s_w((v4i32)R0, 0);
const uint32_t val1 = __msa_copy_s_w((v4i32)R1, 0);
const uint32_t val2 = __msa_copy_s_w((v4i32)R2, 0);
const uint32_t val3 = __msa_copy_s_w((v4i32)R3, 0);
SW4(val0, val1, val2, val3, dst, BPS);
}
LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
ILVR_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_r,
src87_r, src98_r, src109_r);
ILVR_D2_SB(src87_r, src76_r, src109_r, src98_r, src8776, src10998);
XORI_B2_128_SB(src8776, src10998);
out10 = FILT_8TAP_DPADD_S_H(src2110, src4332, src6554, src8776, filt0,
filt1, filt2, filt3);
out32 = FILT_8TAP_DPADD_S_H(src4332, src6554, src8776, src10998, filt0,
filt1, filt2, filt3);
SRARI_H2_SH(out10, out32, FILTER_BITS);
SAT_SH2_SH(out10, out32, 7);
out = PCKEV_XORI128_UB(out10, out32);
ILVR_W2_UB(dst1, dst0, dst3, dst2, dst0, dst2);
dst0 = (v16u8)__msa_ilvr_d((v2i64)dst2, (v2i64)dst0);
out = __msa_aver_u_b(out, dst0);
ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
dst += (4 * dst_stride);
src2110 = src6554;
src4332 = src8776;
src6554 = src10998;
src6 = src10;
}
}
static void common_vt_8t_and_aver_dst_8w_msa(const uint8_t *src,
int32_t src_stride, uint8_t *dst,
int32_t dst_stride, int8_t *filter,
int32_t height) {
