static void intra_predict_tm_32x32_msa(const uint8_t *src_top,
const uint8_t *src_left,
uint8_t *dst, int32_t dst_stride) {
uint8_t top_left = src_top[-1];
uint32_t loop_cnt;
v16i8 src_top0, src_top1, src_left0, src_left1, src_left2, src_left3;
v8u16 src_top_left, res_r0, res_r1, res_l0, res_l1;
LD_SB2(src_top, 16, src_top0, src_top1);
src_top_left = (v8u16)__msa_fill_h(top_left);
for (loop_cnt = 8; loop_cnt--;) {
src_left0 = __msa_fill_b(src_left[0]);
src_left1 = __msa_fill_b(src_left[1]);
src_left2 = __msa_fill_b(src_left[2]);
src_left3 = __msa_fill_b(src_left[3]);
src_left += 4;
ILVR_B2_UH(src_left0, src_top0, src_left0, src_top1, res_r0, res_r1);
ILVL_B2_UH(src_left0, src_top0, src_left0, src_top1, res_l0, res_l1);
HADD_UB4_UH(res_r0, res_l0, res_r1, res_l1, res_r0, res_l0, res_r1, res_l1);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r0, res_l0);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r1, res_l1);
SAT_UH4_UH(res_r0, res_l0, res_r1, res_l1, 7);
PCKEV_ST_SB(res_r0, res_l0, dst);
PCKEV_ST_SB(res_r1, res_l1, dst + 16);
dst += dst_stride;
ILVR_B2_UH(src_left1, src_top0, src_left1, src_top1, res_r0, res_r1);
ILVL_B2_UH(src_left1, src_top0, src_left1, src_top1, res_l0, res_l1);
HADD_UB4_UH(res_r0, res_l0, res_r1, res_l1, res_r0, res_l0, res_r1, res_l1);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r0, res_l0);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r1, res_l1);
SAT_UH4_UH(res_r0, res_l0, res_r1, res_l1, 7);
PCKEV_ST_SB(res_r0, res_l0, dst);
PCKEV_ST_SB(res_r1, res_l1, dst + 16);
dst += dst_stride;
ILVR_B2_UH(src_left2, src_top0, src_left2, src_top1, res_r0, res_r1);
ILVL_B2_UH(src_left2, src_top0, src_left2, src_top1, res_l0, res_l1);
HADD_UB4_UH(res_r0, res_l0, res_r1, res_l1, res_r0, res_l0, res_r1, res_l1);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r0, res_l0);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r1, res_l1);
SAT_UH4_UH(res_r0, res_l0, res_r1, res_l1, 7);
PCKEV_ST_SB(res_r0, res_l0, dst);
PCKEV_ST_SB(res_r1, res_l1, dst + 16);
dst += dst_stride;
ILVR_B2_UH(src_left3, src_top0, src_left3, src_top1, res_r0, res_r1);
ILVL_B2_UH(src_left3, src_top0, src_left3, src_top1, res_l0, res_l1);
HADD_UB4_UH(res_r0, res_l0, res_r1, res_l1, res_r0, res_l0, res_r1, res_l1);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r0, res_l0);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r1, res_l1);
SAT_UH4_UH(res_r0, res_l0, res_r1, res_l1, 7);
PCKEV_ST_SB(res_r0, res_l0, dst);
PCKEV_ST_SB(res_r1, res_l1, dst + 16);
dst += dst_stride;
}
}
static void intra_predict_tm_4x4_msa(const uint8_t *src_top_ptr,
const uint8_t *src_left,
uint8_t *dst, int32_t dst_stride) {
uint32_t val;
uint8_t top_left = src_top_ptr[-1];
v16i8 src_left0, src_left1, src_left2, src_left3, tmp0, tmp1, src_top = { 0 };
v16u8 src0, src1, src2, src3;
v8u16 src_top_left, vec0, vec1, vec2, vec3;
src_top_left = (v8u16)__msa_fill_h(top_left);
val = LW(src_top_ptr);
src_top = (v16i8)__msa_insert_w((v4i32)src_top, 0, val);
src_left0 = __msa_fill_b(src_left[0]);
src_left1 = __msa_fill_b(src_left[1]);
src_left2 = __msa_fill_b(src_left[2]);
src_left3 = __msa_fill_b(src_left[3]);
ILVR_B4_UB(src_left0, src_top, src_left1, src_top, src_left2, src_top,
src_left3, src_top, src0, src1, src2, src3);
HADD_UB4_UH(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, vec0, vec1);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, vec2, vec3);
SAT_UH4_UH(vec0, vec1, vec2, vec3, 7);
PCKEV_B2_SB(vec1, vec0, vec3, vec2, tmp0, tmp1);
ST4x4_UB(tmp0, tmp1, 0, 2, 0, 2, dst, dst_stride);
}
static void intra_predict_tm_8x8_msa(const uint8_t *src_top_ptr,
const uint8_t *src_left,
uint8_t *dst, int32_t dst_stride) {
uint64_t val;
uint8_t top_left = src_top_ptr[-1];
uint32_t loop_cnt;
v16i8 src_left0, src_left1, src_left2, src_left3, tmp0, tmp1, src_top = { 0 };
v8u16 src_top_left, vec0, vec1, vec2, vec3;
v16u8 src0, src1, src2, src3;
val = LD(src_top_ptr);
src_top = (v16i8)__msa_insert_d((v2i64)src_top, 0, val);
src_top_left = (v8u16)__msa_fill_h(top_left);
for (loop_cnt = 2; loop_cnt--;) {
src_left0 = __msa_fill_b(src_left[0]);
src_left1 = __msa_fill_b(src_left[1]);
src_left2 = __msa_fill_b(src_left[2]);
src_left3 = __msa_fill_b(src_left[3]);
src_left += 4;
ILVR_B4_UB(src_left0, src_top, src_left1, src_top, src_left2, src_top,
src_left3, src_top, src0, src1, src2, src3);
HADD_UB4_UH(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, vec0, vec1);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, vec2, vec3);
SAT_UH4_UH(vec0, vec1, vec2, vec3, 7);
PCKEV_B2_SB(vec1, vec0, vec3, vec2, tmp0, tmp1);
ST8x4_UB(tmp0, tmp1, dst, dst_stride);
dst += (4 * dst_stride);
}
}
uint32_t vp10_avg_8x8_msa(const uint8_t *src, int32_t src_stride) {
uint32_t sum_out;
v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
v8u16 sum0, sum1, sum2, sum3, sum4, sum5, sum6, sum7;
v4u32 sum = { 0 };
LD_UB8(src, src_stride, src0, src1, src2, src3, src4, src5, src6, src7);
HADD_UB4_UH(src0, src1, src2, src3, sum0, sum1, sum2, sum3);
HADD_UB4_UH(src4, src5, src6, src7, sum4, sum5, sum6, sum7);
ADD4(sum0, sum1, sum2, sum3, sum4, sum5, sum6, sum7, sum0, sum2, sum4, sum6);
ADD2(sum0, sum2, sum4, sum6, sum0, sum4);
sum0 += sum4;
sum = __msa_hadd_u_w(sum0, sum0);
sum0 = (v8u16)__msa_pckev_h((v8i16)sum, (v8i16)sum);
sum = __msa_hadd_u_w(sum0, sum0);
sum = (v4u32)__msa_srari_w((v4i32)sum, 6);
sum_out = __msa_copy_u_w((v4i32)sum, 0);
return sum_out;
}
static int32_t avc_coeff_last64_msa( int16_t *p_src )
{
uint32_t u_res;
v8i16 src0, src1, src2, src3, src4, src5, src6, src7;
v8i16 tmp_h0, tmp_h1, tmp_h2, tmp_h3, tmp_h4, tmp_h5, tmp_h6, tmp_h7;
v16u8 tmp0, tmp1, tmp2, tmp3;
v8u16 vec0, vec1, vec2, vec3;
v4i32 out0;
v16u8 mask = { 1, 2, 4, 8, 16, 32, 64, 128, 1, 2, 4, 8, 16, 32, 64, 128 };
LD_SH8( p_src, 8, src0, src1, src2, src3, src4, src5, src6, src7 );
tmp_h0 = __msa_ceqi_h( src0, 0 );
tmp_h1 = __msa_ceqi_h( src1, 0 );
tmp_h2 = __msa_ceqi_h( src2, 0 );
tmp_h3 = __msa_ceqi_h( src3, 0 );
tmp_h4 = __msa_ceqi_h( src4, 0 );
tmp_h5 = __msa_ceqi_h( src5, 0 );
tmp_h6 = __msa_ceqi_h( src6, 0 );
tmp_h7 = __msa_ceqi_h( src7, 0 );
PCKEV_B4_UB( tmp_h1, tmp_h0, tmp_h3, tmp_h2, tmp_h5, tmp_h4, tmp_h7, tmp_h6,
tmp0, tmp1, tmp2, tmp3 );
tmp0 = tmp0 & mask;
tmp1 = tmp1 & mask;
tmp2 = tmp2 & mask;
tmp3 = tmp3 & mask;
HADD_UB4_UH( tmp0, tmp1, tmp2, tmp3, vec0, vec1, vec2, vec3 );
PCKEV_B2_UB( vec1, vec0, vec3, vec2, tmp0, tmp1 );
HADD_UB2_UH( tmp0, tmp1, vec0, vec1 );
tmp0 = ( v16u8 ) __msa_pckev_b( ( v16i8 ) vec1, ( v16i8 ) vec0 );
vec0 = __msa_hadd_u_h( tmp0, tmp0 );
tmp0 = ( v16u8 ) __msa_pckev_b( ( v16i8 ) vec0, ( v16i8 ) vec0 );
out0 = ( v4i32 ) __msa_nloc_d( ( v2i64 ) tmp0 );
u_res = __msa_copy_u_w( out0, 0 );
return ( 63 - u_res );
}
