static WEBP_INLINE void ITransformOne(const uint8_t* ref, const int16_t* in,
uint8_t* dst) {
v8i16 input0, input1;
v4i32 in0, in1, in2, in3, hz0, hz1, hz2, hz3, vt0, vt1, vt2, vt3;
v4i32 res0, res1, res2, res3;
v16i8 dest0, dest1, dest2, dest3;
const v16i8 zero = { 0 };
LD_SH2(in, 8, input0, input1);
UNPCK_SH_SW(input0, in0, in1);
UNPCK_SH_SW(input1, in2, in3);
IDCT_1D_W(in0, in1, in2, in3, hz0, hz1, hz2, hz3);
TRANSPOSE4x4_SW_SW(hz0, hz1, hz2, hz3, hz0, hz1, hz2, hz3);
IDCT_1D_W(hz0, hz1, hz2, hz3, vt0, vt1, vt2, vt3);
SRARI_W4_SW(vt0, vt1, vt2, vt3, 3);
TRANSPOSE4x4_SW_SW(vt0, vt1, vt2, vt3, vt0, vt1, vt2, vt3);
LD_SB4(ref, BPS, dest0, dest1, dest2, dest3);
ILVR_B4_SW(zero, dest0, zero, dest1, zero, dest2, zero, dest3,
res0, res1, res2, res3);
ILVR_H4_SW(zero, res0, zero, res1, zero, res2, zero, res3,
res0, res1, res2, res3);
ADD4(res0, vt0, res1, vt1, res2, vt2, res3, vt3, res0, res1, res2, res3);
CLIP_SW4_0_255(res0, res1, res2, res3);
PCKEV_B2_SW(res0, res1, res2, res3, vt0, vt1);
res0 = (v4i32)__msa_pckev_b((v16i8)vt0, (v16i8)vt1);
ST4x4_UB(res0, res0, 3, 2, 1, 0, dst, BPS);
}
void vp8_short_walsh4x4_msa(int16_t *input, int16_t *output, int32_t pitch) {
v8i16 in0_h, in1_h, in2_h, in3_h;
v4i32 in0_w, in1_w, in2_w, in3_w, temp0, temp1, temp2, temp3;
LD_SH4(input, pitch / 2, in0_h, in1_h, in2_h, in3_h);
TRANSPOSE4x4_SH_SH(in0_h, in1_h, in2_h, in3_h, in0_h, in1_h, in2_h, in3_h);
UNPCK_R_SH_SW(in0_h, in0_w);
UNPCK_R_SH_SW(in1_h, in1_w);
UNPCK_R_SH_SW(in2_h, in2_w);
UNPCK_R_SH_SW(in3_h, in3_w);
BUTTERFLY_4(in0_w, in1_w, in3_w, in2_w, temp0, temp3, temp2, temp1);
SLLI_4V(temp0, temp1, temp2, temp3, 2);
BUTTERFLY_4(temp0, temp1, temp2, temp3, in0_w, in1_w, in2_w, in3_w);
temp0 = RET_1_IF_NZERO_W(temp0);
in0_w += temp0;
TRANSPOSE4x4_SW_SW(in0_w, in1_w, in2_w, in3_w, in0_w, in1_w, in2_w, in3_w);
BUTTERFLY_4(in0_w, in1_w, in3_w, in2_w, temp0, temp3, temp2, temp1);
BUTTERFLY_4(temp0, temp1, temp2, temp3, in0_w, in1_w, in2_w, in3_w);
in0_w += RET_1_IF_NEG_W(in0_w);
in1_w += RET_1_IF_NEG_W(in1_w);
in2_w += RET_1_IF_NEG_W(in2_w);
in3_w += RET_1_IF_NEG_W(in3_w);
ADD4(in0_w, 3, in1_w, 3, in2_w, 3, in3_w, 3, in0_w, in1_w, in2_w, in3_w);
SRA_4V(in0_w, in1_w, in2_w, in3_w, 3);
PCKEV_H2_SH(in1_w, in0_w, in3_w, in2_w, in0_h, in1_h);
ST_SH2(in0_h, in1_h, output, 8);
}
static void avc_dct4x4dc_msa( int16_t *p_src, int16_t *p_dst,
int32_t i_src_stride )
{
v8i16 src0, src1, src2, src3, ver_res0, ver_res1, ver_res2, ver_res3;
v4i32 src0_r, src1_r, src2_r, src3_r, tmp0, tmp1, tmp2, tmp3;
v4i32 hor_res0, hor_res1, hor_res2, hor_res3;
v4i32 ver_res0_r, ver_res1_r, ver_res2_r, ver_res3_r;
LD_SH4( p_src, i_src_stride, src0, src1, src2, src3 );
UNPCK_R_SH_SW( src0, src0_r );
UNPCK_R_SH_SW( src1, src1_r );
UNPCK_R_SH_SW( src2, src2_r );
UNPCK_R_SH_SW( src3, src3_r );
BUTTERFLY_4( src0_r, src2_r, src3_r, src1_r,
tmp0, tmp3, tmp2, tmp1 );
BUTTERFLY_4( tmp0, tmp1, tmp2, tmp3,
hor_res0, hor_res3, hor_res2, hor_res1 );
TRANSPOSE4x4_SW_SW( hor_res0, hor_res1, hor_res2, hor_res3,
hor_res0, hor_res1, hor_res2, hor_res3 );
BUTTERFLY_4( hor_res0, hor_res2, hor_res3, hor_res1,
tmp0, tmp3, tmp2, tmp1 );
BUTTERFLY_4( tmp0, tmp1, tmp2, tmp3,
ver_res0_r, ver_res3_r, ver_res2_r, ver_res1_r );
SRARI_W4_SW( ver_res0_r, ver_res1_r, ver_res2_r, ver_res3_r, 1 );
PCKEV_H4_SH( ver_res0_r, ver_res0_r, ver_res1_r, ver_res1_r,
ver_res2_r, ver_res2_r, ver_res3_r, ver_res3_r,
ver_res0, ver_res1, ver_res2, ver_res3 );
PCKOD_D2_SH( ver_res1, ver_res0, ver_res3, ver_res2, ver_res0, ver_res2 );
ST_SH2( ver_res0, ver_res2, p_dst, 8 );
}
static void avc_idct4x4dc_msa( int16_t *p_src, int32_t i_src_stride,
int16_t *p_dst, int32_t i_dst_stride )
{
v8i16 src0, src1, src2, src3;
v4i32 src0_r, src1_r, src2_r, src3_r;
v4i32 hres0, hres1, hres2, hres3;
v8i16 vres0, vres1, vres2, vres3;
v4i32 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
v2i64 res0, res1;
LD_SH4( p_src, i_src_stride, src0, src1, src2, src3 );
UNPCK_R_SH_SW( src0, src0_r );
UNPCK_R_SH_SW( src1, src1_r );
UNPCK_R_SH_SW( src2, src2_r );
UNPCK_R_SH_SW( src3, src3_r );
BUTTERFLY_4( src0_r, src2_r, src3_r, src1_r, vec0, vec3, vec2, vec1 );
BUTTERFLY_4( vec0, vec1, vec2, vec3, hres0, hres3, hres2, hres1 );
TRANSPOSE4x4_SW_SW( hres0, hres1, hres2, hres3,
hres0, hres1, hres2, hres3 );
BUTTERFLY_4( hres0, hres2, hres3, hres1, vec0, vec3, vec2, vec1 );
BUTTERFLY_4( vec0, vec1, vec2, vec3, vec4, vec7, vec6, vec5 );
PCKEV_H4_SH( vec4, vec4, vec5, vec5, vec6, vec6, vec7, vec7,
vres0, vres1, vres2, vres3 );
PCKOD_D2_SD( vres1, vres0, vres3, vres2, res0, res1 );
ST8x4_UB( res0, res1, p_dst, i_dst_stride * 2 );
}
static void hevc_idct_luma_4x4_msa(int16_t *coeffs)
{
v8i16 in0, in1, dst0, dst1;
v4i32 in_r0, in_l0, in_r1, in_l1, res0, res1, res2, res3;
LD_SH2(coeffs, 8, in0, in1);
UNPCK_SH_SW(in0, in_r0, in_l0);
UNPCK_SH_SW(in1, in_r1, in_l1);
HEVC_IDCT_LUMA4x4_COL(in_r0, in_l0, in_r1, in_l1, res0, res1, res2, res3,
7);
TRANSPOSE4x4_SW_SW(res0, res1, res2, res3, in_r0, in_l0, in_r1, in_l1);
HEVC_IDCT_LUMA4x4_COL(in_r0, in_l0, in_r1, in_l1, res0, res1, res2, res3,
12);
/* Pack and transpose */
PCKEV_H2_SH(res2, res0, res3, res1, dst0, dst1);
ILVRL_H2_SW(dst1, dst0, res0, res1);
ILVRL_W2_SH(res1, res0, dst0, dst1);
ST_SH2(dst0, dst1, coeffs, 8);
}
static void hevc_idct_4x4_msa(int16_t *coeffs)
{
v8i16 in0, in1;
v4i32 in_r0, in_l0, in_r1, in_l1;
v4i32 sum0, sum1, sum2, sum3;
v8i16 zeros = { 0 };
LD_SH2(coeffs, 8, in0, in1);
ILVRL_H2_SW(zeros, in0, in_r0, in_l0);
ILVRL_H2_SW(zeros, in1, in_r1, in_l1);
HEVC_IDCT4x4_COL(in_r0, in_l0, in_r1, in_l1, sum0, sum1, sum2, sum3, 7);
TRANSPOSE4x4_SW_SW(sum0, sum1, sum2, sum3, in_r0, in_l0, in_r1, in_l1);
HEVC_IDCT4x4_COL(in_r0, in_l0, in_r1, in_l1, sum0, sum1, sum2, sum3, 12);
/* Pack and transpose */
PCKEV_H2_SH(sum2, sum0, sum3, sum1, in0, in1);
ILVRL_H2_SW(in1, in0, sum0, sum1);
ILVRL_W2_SH(sum1, sum0, in0, in1);
ST_SH2(in0, in1, coeffs, 8);
}
void aom_iwht4x4_16_add_msa(const int16_t *input, uint8_t *dst,
int32_t dst_stride) {
v8i16 in0, in1, in2, in3;
v4i32 in0_r, in1_r, in2_r, in3_r, in4_r;
/* load vector elements of 4x4 block */
LD4x4_SH(input, in0, in2, in3, in1);
TRANSPOSE4x4_SH_SH(in0, in2, in3, in1, in0, in2, in3, in1);
UNPCK_R_SH_SW(in0, in0_r);
UNPCK_R_SH_SW(in2, in2_r);
UNPCK_R_SH_SW(in3, in3_r);
UNPCK_R_SH_SW(in1, in1_r);
SRA_4V(in0_r, in1_r, in2_r, in3_r, UNIT_QUANT_SHIFT);
in0_r += in2_r;
in3_r -= in1_r;
in4_r = (in0_r - in3_r) >> 1;
in1_r = in4_r - in1_r;
in2_r = in4_r - in2_r;
in0_r -= in1_r;
in3_r += in2_r;
TRANSPOSE4x4_SW_SW(in0_r, in1_r, in2_r, in3_r, in0_r, in1_r, in2_r, in3_r);
in0_r += in1_r;
in2_r -= in3_r;
in4_r = (in0_r - in2_r) >> 1;
in3_r = in4_r - in3_r;
in1_r = in4_r - in1_r;
in0_r -= in3_r;
in2_r += in1_r;
PCKEV_H4_SH(in0_r, in0_r, in1_r, in1_r, in2_r, in2_r, in3_r, in3_r, in0, in1,
in2, in3);
ADDBLK_ST4x4_UB(in0, in3, in1, in2, dst, dst_stride);
}