void fdct16x8_1d_row(int16_t *input, int16_t *output) {
v8i16 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
v8i16 in8, in9, in10, in11, in12, in13, in14, in15;
LD_SH8(input, 16, in0, in1, in2, in3, in4, in5, in6, in7);
LD_SH8((input + 8), 16, in8, in9, in10, in11, in12, in13, in14, in15);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3,
in4, in5, in6, in7);
TRANSPOSE8x8_SH_SH(in8, in9, in10, in11, in12, in13, in14, in15, in8, in9,
in10, in11, in12, in13, in14, in15);
ADD4(in0, 1, in1, 1, in2, 1, in3, 1, in0, in1, in2, in3);
ADD4(in4, 1, in5, 1, in6, 1, in7, 1, in4, in5, in6, in7);
ADD4(in8, 1, in9, 1, in10, 1, in11, 1, in8, in9, in10, in11);
ADD4(in12, 1, in13, 1, in14, 1, in15, 1, in12, in13, in14, in15);
SRA_4V(in0, in1, in2, in3, 2);
SRA_4V(in4, in5, in6, in7, 2);
SRA_4V(in8, in9, in10, in11, 2);
SRA_4V(in12, in13, in14, in15, 2);
BUTTERFLY_16(in0, in1, in2, in3, in4, in5, in6, in7, in8, in9, in10, in11,
in12, in13, in14, in15, tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6,
tmp7, in8, in9, in10, in11, in12, in13, in14, in15);
ST_SH8(in8, in9, in10, in11, in12, in13, in14, in15, input, 16);
FDCT8x16_EVEN(tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp0, tmp1,
tmp2, tmp3, tmp4, tmp5, tmp6, tmp7);
LD_SH8(input, 16, in8, in9, in10, in11, in12, in13, in14, in15);
FDCT8x16_ODD(in8, in9, in10, in11, in12, in13, in14, in15, in0, in1, in2, in3,
in4, in5, in6, in7);
TRANSPOSE8x8_SH_SH(tmp0, in0, tmp1, in1, tmp2, in2, tmp3, in3, tmp0, in0,
tmp1, in1, tmp2, in2, tmp3, in3);
ST_SH8(tmp0, in0, tmp1, in1, tmp2, in2, tmp3, in3, output, 16);
TRANSPOSE8x8_SH_SH(tmp4, in4, tmp5, in5, tmp6, in6, tmp7, in7, tmp4, in4,
tmp5, in5, tmp6, in6, tmp7, in7);
ST_SH8(tmp4, in4, tmp5, in5, tmp6, in6, tmp7, in7, output + 8, 16);
}
static void fdct8x32_1d_row_load_butterfly(int16_t *temp_buff,
int16_t *output) {
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
v8i16 in8, in9, in10, in11, in12, in13, in14, in15;
v8i16 step0, step1, step2, step3, step4, step5, step6, step7;
LD_SH8(temp_buff, 32, in0, in1, in2, in3, in4, in5, in6, in7);
LD_SH8(temp_buff + 24, 32, in8, in9, in10, in11, in12, in13, in14, in15);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
TRANSPOSE8x8_SH_SH(in8, in9, in10, in11, in12, in13, in14, in15,
in8, in9, in10, in11, in12, in13, in14, in15);
BUTTERFLY_16(in0, in1, in2, in3, in4, in5, in6, in7,
in8, in9, in10, in11, in12, in13, in14, in15,
step0, step1, step2, step3, step4, step5, step6, step7,
in8, in9, in10, in11, in12, in13, in14, in15);
ST_SH8(step0, step1, step2, step3, step4, step5, step6, step7, output, 8);
ST_SH8(in8, in9, in10, in11, in12, in13, in14, in15, (output + 24 * 8), 8);
/* 2nd set */
LD_SH8(temp_buff + 8, 32, in0, in1, in2, in3, in4, in5, in6, in7);
LD_SH8(temp_buff + 16, 32, in8, in9, in10, in11, in12, in13, in14, in15);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
TRANSPOSE8x8_SH_SH(in8, in9, in10, in11, in12, in13, in14, in15,
in8, in9, in10, in11, in12, in13, in14, in15);
BUTTERFLY_16(in0, in1, in2, in3, in4, in5, in6, in7,
in8, in9, in10, in11, in12, in13, in14, in15,
step0, step1, step2, step3, step4, step5, step6, step7,
in8, in9, in10, in11, in12, in13, in14, in15);
ST_SH8(step0, step1, step2, step3, step4, step5, step6, step7,
(output + 8 * 8), 8);
ST_SH8(in8, in9, in10, in11, in12, in13, in14, in15, (output + 16 * 8), 8);
}
static void fadst16_transpose_msa(int16_t *input, int16_t *out) {
v8i16 r0, r1, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, r12, r13, r14, r15;
v8i16 l0, l1, l2, l3, l4, l5, l6, l7, l8, l9, l10, l11, l12, l13, l14, l15;
/* load input data */
LD_SH16(input, 8, l0, l8, l1, l9, l2, l10, l3, l11, l4, l12, l5, l13, l6, l14,
l7, l15);
TRANSPOSE8x8_SH_SH(l0, l1, l2, l3, l4, l5, l6, l7, r0, r1, r2, r3, r4, r5, r6,
r7);
TRANSPOSE8x8_SH_SH(l8, l9, l10, l11, l12, l13, l14, l15, r8, r9, r10, r11,
r12, r13, r14, r15);
ST_SH8(r0, r8, r1, r9, r2, r10, r3, r11, out, 8);
ST_SH8(r4, r12, r5, r13, r6, r14, r7, r15, (out + 64), 8);
out += 16 * 8;
/* load input data */
input += 128;
LD_SH16(input, 8, l0, l8, l1, l9, l2, l10, l3, l11, l4, l12, l5, l13, l6, l14,
l7, l15);
TRANSPOSE8x8_SH_SH(l0, l1, l2, l3, l4, l5, l6, l7, r0, r1, r2, r3, r4, r5, r6,
r7);
TRANSPOSE8x8_SH_SH(l8, l9, l10, l11, l12, l13, l14, l15, r8, r9, r10, r11,
r12, r13, r14, r15);
ST_SH8(r0, r8, r1, r9, r2, r10, r3, r11, out, 8);
ST_SH8(r4, r12, r5, r13, r6, r14, r7, r15, (out + 64), 8);
}
void vp9_fht8x8_msa(const int16_t *input, int16_t *output, int32_t stride,
int32_t tx_type) {
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
LD_SH8(input, stride, in0, in1, in2, in3, in4, in5, in6, in7);
SLLI_4V(in0, in1, in2, in3, 2);
SLLI_4V(in4, in5, in6, in7, 2);
switch (tx_type) {
case DCT_DCT:
VP9_FDCT8(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
VP9_FDCT8(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
break;
case ADST_DCT:
VP9_ADST8(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
VP9_FDCT8(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
break;
case DCT_ADST:
VP9_FDCT8(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
VP9_ADST8(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
break;
case ADST_ADST:
VP9_ADST8(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
VP9_ADST8(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
break;
default:
assert(0);
break;
}
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
VP9_SRLI_AVE_S_4V_H(in0, in1, in2, in3, in4, in5, in6, in7);
ST_SH8(in0, in1, in2, in3, in4, in5, in6, in7, output, 8);
}
static void hevc_idct_8x8_msa(int16_t *coeffs)
{
const int16_t *filter = >8x8_cnst[0];
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
LD_SH8(coeffs, 8, in0, in1, in2, in3, in4, in5, in6, in7);
HEVC_IDCT8x8_COL(in0, in1, in2, in3, in4, in5, in6, in7, 7);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
HEVC_IDCT8x8_COL(in0, in1, in2, in3, in4, in5, in6, in7, 12);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
ST_SH8(in0, in1, in2, in3, in4, in5, in6, in7, coeffs, 8);
}
static void fadst16_transpose_postproc_msa(int16_t *input, int16_t *out) {
v8i16 r0, r1, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, r12, r13, r14, r15;
v8i16 l0, l1, l2, l3, l4, l5, l6, l7, l8, l9, l10, l11, l12, l13, l14, l15;
/* load input data */
LD_SH8(input, 16, l0, l1, l2, l3, l4, l5, l6, l7);
TRANSPOSE8x8_SH_SH(l0, l1, l2, l3, l4, l5, l6, l7, r0, r1, r2, r3, r4, r5, r6,
r7);
FDCT_POSTPROC_2V_NEG_H(r0, r1);
FDCT_POSTPROC_2V_NEG_H(r2, r3);
FDCT_POSTPROC_2V_NEG_H(r4, r5);
FDCT_POSTPROC_2V_NEG_H(r6, r7);
ST_SH8(r0, r1, r2, r3, r4, r5, r6, r7, out, 8);
out += 64;
LD_SH8(input + 8, 16, l8, l9, l10, l11, l12, l13, l14, l15);
TRANSPOSE8x8_SH_SH(l8, l9, l10, l11, l12, l13, l14, l15, r8, r9, r10, r11,
r12, r13, r14, r15);
FDCT_POSTPROC_2V_NEG_H(r8, r9);
FDCT_POSTPROC_2V_NEG_H(r10, r11);
FDCT_POSTPROC_2V_NEG_H(r12, r13);
FDCT_POSTPROC_2V_NEG_H(r14, r15);
ST_SH8(r8, r9, r10, r11, r12, r13, r14, r15, out, 8);
out += 64;
/* load input data */
input += 128;
LD_SH8(input, 16, l0, l1, l2, l3, l4, l5, l6, l7);
TRANSPOSE8x8_SH_SH(l0, l1, l2, l3, l4, l5, l6, l7, r0, r1, r2, r3, r4, r5, r6,
r7);
FDCT_POSTPROC_2V_NEG_H(r0, r1);
FDCT_POSTPROC_2V_NEG_H(r2, r3);
FDCT_POSTPROC_2V_NEG_H(r4, r5);
FDCT_POSTPROC_2V_NEG_H(r6, r7);
ST_SH8(r0, r1, r2, r3, r4, r5, r6, r7, out, 8);
out += 64;
LD_SH8(input + 8, 16, l8, l9, l10, l11, l12, l13, l14, l15);
TRANSPOSE8x8_SH_SH(l8, l9, l10, l11, l12, l13, l14, l15, r8, r9, r10, r11,
r12, r13, r14, r15);
FDCT_POSTPROC_2V_NEG_H(r8, r9);
FDCT_POSTPROC_2V_NEG_H(r10, r11);
FDCT_POSTPROC_2V_NEG_H(r12, r13);
FDCT_POSTPROC_2V_NEG_H(r14, r15);
ST_SH8(r8, r9, r10, r11, r12, r13, r14, r15, out, 8);
}
void vpx_fdct8x8_msa(const int16_t *input, int16_t *output,
int32_t src_stride) {
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
LD_SH8(input, src_stride, in0, in1, in2, in3, in4, in5, in6, in7);
SLLI_4V(in0, in1, in2, in3, 2);
SLLI_4V(in4, in5, in6, in7, 2);
VP9_FDCT8(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3, in4,
in5, in6, in7);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3,
in4, in5, in6, in7);
VP9_FDCT8(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3, in4,
in5, in6, in7);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3,
in4, in5, in6, in7);
SRLI_AVE_S_4V_H(in0, in1, in2, in3, in4, in5, in6, in7);
ST_SH8(in0, in1, in2, in3, in4, in5, in6, in7, output, 8);
}
static void postproc_fdct16x8_1d_row(int16_t *intermediate, int16_t *output) {
int16_t *temp = intermediate;
int16_t *out = output;
v8i16 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
v8i16 in0, in1, in2, in3, in4, in5, in6, in7, in8, in9, in10, in11;
v8i16 in12, in13, in14, in15;
LD_SH8(temp, 16, in0, in1, in2, in3, in4, in5, in6, in7);
temp = intermediate + 8;
LD_SH8(temp, 16, in8, in9, in10, in11, in12, in13, in14, in15);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7, in0, in1, in2, in3,
in4, in5, in6, in7);
TRANSPOSE8x8_SH_SH(in8, in9, in10, in11, in12, in13, in14, in15, in8, in9,
in10, in11, in12, in13, in14, in15);
FDCT_POSTPROC_2V_NEG_H(in0, in1);
FDCT_POSTPROC_2V_NEG_H(in2, in3);
FDCT_POSTPROC_2V_NEG_H(in4, in5);
FDCT_POSTPROC_2V_NEG_H(in6, in7);
FDCT_POSTPROC_2V_NEG_H(in8, in9);
FDCT_POSTPROC_2V_NEG_H(in10, in11);
FDCT_POSTPROC_2V_NEG_H(in12, in13);
FDCT_POSTPROC_2V_NEG_H(in14, in15);
BUTTERFLY_16(in0, in1, in2, in3, in4, in5, in6, in7, in8, in9, in10, in11,
in12, in13, in14, in15, tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6,
tmp7, in8, in9, in10, in11, in12, in13, in14, in15);
temp = intermediate;
ST_SH8(in8, in9, in10, in11, in12, in13, in14, in15, temp, 16);
FDCT8x16_EVEN(tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp0, tmp1,
tmp2, tmp3, tmp4, tmp5, tmp6, tmp7);
temp = intermediate;
LD_SH8(temp, 16, in8, in9, in10, in11, in12, in13, in14, in15);
FDCT8x16_ODD(in8, in9, in10, in11, in12, in13, in14, in15, in0, in1, in2, in3,
in4, in5, in6, in7);
TRANSPOSE8x8_SH_SH(tmp0, in0, tmp1, in1, tmp2, in2, tmp3, in3, tmp0, in0,
tmp1, in1, tmp2, in2, tmp3, in3);
ST_SH8(tmp0, in0, tmp1, in1, tmp2, in2, tmp3, in3, out, 16);
TRANSPOSE8x8_SH_SH(tmp4, in4, tmp5, in5, tmp6, in6, tmp7, in7, tmp4, in4,
tmp5, in5, tmp6, in6, tmp7, in7);
out = output + 8;
ST_SH8(tmp4, in4, tmp5, in5, tmp6, in6, tmp7, in7, out, 16);
}
static void hevc_idct_transpose_8x32_to_32x8(int16_t *tmp_buf, int16_t *coeffs)
{
uint8_t i;
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
for (i = 0; i < 4; i++) {
LD_SH8(tmp_buf + i * 8 * 8, 8, in0, in1, in2, in3, in4, in5, in6, in7);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
ST_SH8(in0, in1, in2, in3, in4, in5, in6, in7, coeffs + i * 8, 32);
}
}
static void avc_idct8_addblk_msa( uint8_t *p_dst, int16_t *p_src,
int32_t i_dst_stride )
{
v8i16 src0, src1, src2, src3, src4, src5, src6, src7;
v8i16 vec0, vec1, vec2, vec3;
v8i16 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
v8i16 res0, res1, res2, res3, res4, res5, res6, res7;
v4i32 tmp0_r, tmp1_r, tmp2_r, tmp3_r, tmp4_r, tmp5_r, tmp6_r, tmp7_r;
v4i32 tmp0_l, tmp1_l, tmp2_l, tmp3_l, tmp4_l, tmp5_l, tmp6_l, tmp7_l;
v4i32 vec0_r, vec1_r, vec2_r, vec3_r, vec0_l, vec1_l, vec2_l, vec3_l;
v4i32 res0_r, res1_r, res2_r, res3_r, res4_r, res5_r, res6_r, res7_r;
v4i32 res0_l, res1_l, res2_l, res3_l, res4_l, res5_l, res6_l, res7_l;
v16i8 dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7;
v16i8 zeros = { 0 };
p_src[ 0 ] += 32;
LD_SH8( p_src, 8, src0, src1, src2, src3, src4, src5, src6, src7 );
vec0 = src0 + src4;
vec1 = src0 - src4;
vec2 = src2 >> 1;
vec2 = vec2 - src6;
vec3 = src6 >> 1;
vec3 = src2 + vec3;
BUTTERFLY_4( vec0, vec1, vec2, vec3, tmp0, tmp1, tmp2, tmp3 );
vec0 = src7 >> 1;
vec0 = src5 - vec0 - src3 - src7;
vec1 = src3 >> 1;
vec1 = src1 - vec1 + src7 - src3;
vec2 = src5 >> 1;
vec2 = vec2 - src1 + src7 + src5;
vec3 = src1 >> 1;
vec3 = vec3 + src3 + src5 + src1;
tmp4 = vec3 >> 2;
tmp4 += vec0;
tmp5 = vec2 >> 2;
tmp5 += vec1;
tmp6 = vec1 >> 2;
tmp6 -= vec2;
tmp7 = vec0 >> 2;
tmp7 = vec3 - tmp7;
BUTTERFLY_8( tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7,
res0, res1, res2, res3, res4, res5, res6, res7 );
TRANSPOSE8x8_SH_SH( res0, res1, res2, res3, res4, res5, res6, res7,
res0, res1, res2, res3, res4, res5, res6, res7 );
UNPCK_SH_SW( res0, tmp0_r, tmp0_l );
UNPCK_SH_SW( res1, tmp1_r, tmp1_l );
UNPCK_SH_SW( res2, tmp2_r, tmp2_l );
UNPCK_SH_SW( res3, tmp3_r, tmp3_l );
UNPCK_SH_SW( res4, tmp4_r, tmp4_l );
UNPCK_SH_SW( res5, tmp5_r, tmp5_l );
UNPCK_SH_SW( res6, tmp6_r, tmp6_l );
UNPCK_SH_SW( res7, tmp7_r, tmp7_l );
BUTTERFLY_4( tmp0_r, tmp0_l, tmp4_l, tmp4_r,
vec0_r, vec0_l, vec1_l, vec1_r );
vec2_r = tmp2_r >> 1;
vec2_l = tmp2_l >> 1;
vec2_r -= tmp6_r;
vec2_l -= tmp6_l;
vec3_r = tmp6_r >> 1;
vec3_l = tmp6_l >> 1;
vec3_r += tmp2_r;
vec3_l += tmp2_l;
BUTTERFLY_4( vec0_r, vec1_r, vec2_r, vec3_r,
tmp0_r, tmp2_r, tmp4_r, tmp6_r );
BUTTERFLY_4( vec0_l, vec1_l, vec2_l, vec3_l,
tmp0_l, tmp2_l, tmp4_l, tmp6_l );
vec0_r = tmp7_r >> 1;
vec0_l = tmp7_l >> 1;
vec0_r = tmp5_r - vec0_r - tmp3_r - tmp7_r;
vec0_l = tmp5_l - vec0_l - tmp3_l - tmp7_l;
vec1_r = tmp3_r >> 1;
vec1_l = tmp3_l >> 1;
vec1_r = tmp1_r - vec1_r + tmp7_r - tmp3_r;
vec1_l = tmp1_l - vec1_l + tmp7_l - tmp3_l;
vec2_r = tmp5_r >> 1;
vec2_l = tmp5_l >> 1;
vec2_r = vec2_r - tmp1_r + tmp7_r + tmp5_r;
vec2_l = vec2_l - tmp1_l + tmp7_l + tmp5_l;
vec3_r = tmp1_r >> 1;
vec3_l = tmp1_l >> 1;
vec3_r = vec3_r + tmp3_r + tmp5_r + tmp1_r;
vec3_l = vec3_l + tmp3_l + tmp5_l + tmp1_l;
tmp1_r = vec3_r >> 2;
tmp1_l = vec3_l >> 2;
tmp1_r += vec0_r;
tmp1_l += vec0_l;
tmp3_r = vec2_r >> 2;
tmp3_l = vec2_l >> 2;
tmp3_r += vec1_r;
tmp3_l += vec1_l;
tmp5_r = vec1_r >> 2;
tmp5_l = vec1_l >> 2;
tmp5_r -= vec2_r;
tmp5_l -= vec2_l;
tmp7_r = vec0_r >> 2;
tmp7_l = vec0_l >> 2;
tmp7_r = vec3_r - tmp7_r;
tmp7_l = vec3_l - tmp7_l;
BUTTERFLY_4( tmp0_r, tmp0_l, tmp7_l, tmp7_r,
res0_r, res0_l, res7_l, res7_r );
BUTTERFLY_4( tmp2_r, tmp2_l, tmp5_l, tmp5_r,
res1_r, res1_l, res6_l, res6_r );
BUTTERFLY_4( tmp4_r, tmp4_l, tmp3_l, tmp3_r,
res2_r, res2_l, res5_l, res5_r );
BUTTERFLY_4( tmp6_r, tmp6_l, tmp1_l, tmp1_r,
res3_r, res3_l, res4_l, res4_r );
SRA_4V( res0_r, res0_l, res1_r, res1_l, 6 );
SRA_4V( res2_r, res2_l, res3_r, res3_l, 6 );
SRA_4V( res4_r, res4_l, res5_r, res5_l, 6 );
SRA_4V( res6_r, res6_l, res7_r, res7_l, 6 );
PCKEV_H4_SH( res0_l, res0_r, res1_l, res1_r, res2_l, res2_r, res3_l, res3_r,
res0, res1, res2, res3 );
PCKEV_H4_SH( res4_l, res4_r, res5_l, res5_r, res6_l, res6_r, res7_l, res7_r,
res4, res5, res6, res7 );
LD_SB8( p_dst, i_dst_stride,
dst0, dst1, dst2, dst3,
dst4, dst5, dst6, dst7 );
ILVR_B4_SH( zeros, dst0, zeros, dst1, zeros, dst2, zeros, dst3,
tmp0, tmp1, tmp2, tmp3 );
ILVR_B4_SH( zeros, dst4, zeros, dst5, zeros, dst6, zeros, dst7,
tmp4, tmp5, tmp6, tmp7 );
ADD4( res0, tmp0, res1, tmp1, res2, tmp2, res3, tmp3,
res0, res1, res2, res3 );
ADD4( res4, tmp4, res5, tmp5, res6, tmp6, res7, tmp7,
res4, res5, res6, res7 );
CLIP_SH4_0_255( res0, res1, res2, res3 );
CLIP_SH4_0_255( res4, res5, res6, res7 );
PCKEV_B4_SB( res1, res0, res3, res2, res5, res4, res7, res6,
dst0, dst1, dst2, dst3 );
ST8x4_UB( dst0, dst1, p_dst, i_dst_stride );
p_dst += ( 4 * i_dst_stride );
ST8x4_UB( dst2, dst3, p_dst, i_dst_stride );
}
static void fdct8x32_1d_row_transpose_store(int16_t *temp, int16_t *output) {
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
v8i16 in0_1, in1_1, in2_1, in3_1, in4_1, in5_1, in6_1, in7_1;
/* 1st set */
in0 = LD_SH(temp);
in4 = LD_SH(temp + 32);
in2 = LD_SH(temp + 64);
in6 = LD_SH(temp + 96);
in1 = LD_SH(temp + 128);
in7 = LD_SH(temp + 152);
in3 = LD_SH(temp + 192);
in5 = LD_SH(temp + 216);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
/* 2nd set */
in0_1 = LD_SH(temp + 16);
in1_1 = LD_SH(temp + 232);
in2_1 = LD_SH(temp + 80);
in3_1 = LD_SH(temp + 168);
in4_1 = LD_SH(temp + 48);
in5_1 = LD_SH(temp + 176);
in6_1 = LD_SH(temp + 112);
in7_1 = LD_SH(temp + 240);
ST_SH8(in0, in1, in2, in3, in4, in5, in6, in7, output, 32);
TRANSPOSE8x8_SH_SH(in0_1, in1_1, in2_1, in3_1, in4_1, in5_1, in6_1, in7_1,
in0_1, in1_1, in2_1, in3_1, in4_1, in5_1, in6_1, in7_1);
/* 3rd set */
in0 = LD_SH(temp + 8);
in1 = LD_SH(temp + 136);
in2 = LD_SH(temp + 72);
in3 = LD_SH(temp + 200);
in4 = LD_SH(temp + 40);
in5 = LD_SH(temp + 208);
in6 = LD_SH(temp + 104);
in7 = LD_SH(temp + 144);
ST_SH8(in0_1, in1_1, in2_1, in3_1, in4_1, in5_1, in6_1, in7_1,
output + 8, 32);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
ST_SH8(in0, in1, in2, in3, in4, in5, in6, in7, output + 16, 32);
/* 4th set */
in0_1 = LD_SH(temp + 24);
in1_1 = LD_SH(temp + 224);
in2_1 = LD_SH(temp + 88);
in3_1 = LD_SH(temp + 160);
in4_1 = LD_SH(temp + 56);
in5_1 = LD_SH(temp + 184);
in6_1 = LD_SH(temp + 120);
in7_1 = LD_SH(temp + 248);
TRANSPOSE8x8_SH_SH(in0_1, in1_1, in2_1, in3_1, in4_1, in5_1, in6_1, in7_1,
in0_1, in1_1, in2_1, in3_1, in4_1, in5_1, in6_1, in7_1);
ST_SH8(in0_1, in1_1, in2_1, in3_1, in4_1, in5_1, in6_1, in7_1,
output + 24, 32);
}
void vp10_iht8x8_64_add_msa(const int16_t *input, uint8_t *dst,
int32_t dst_stride, int32_t tx_type) {
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
/* load vector elements of 8x8 block */
LD_SH8(input, 8, in0, in1, in2, in3, in4, in5, in6, in7);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
switch (tx_type) {
case DCT_DCT:
/* DCT in horizontal */
VP9_IDCT8x8_1D(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
/* DCT in vertical */
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
VP9_IDCT8x8_1D(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
break;
case ADST_DCT:
/* DCT in horizontal */
VP9_IDCT8x8_1D(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
/* ADST in vertical */
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
VP9_ADST8(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
break;
case DCT_ADST:
/* ADST in horizontal */
VP9_ADST8(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
/* DCT in vertical */
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
VP9_IDCT8x8_1D(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
break;
case ADST_ADST:
/* ADST in horizontal */
VP9_ADST8(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
/* ADST in vertical */
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
VP9_ADST8(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
break;
default:
assert(0);
break;
}
/* final rounding (add 2^4, divide by 2^5) and shift */
SRARI_H4_SH(in0, in1, in2, in3, 5);
SRARI_H4_SH(in4, in5, in6, in7, 5);
/* add block and store 8x8 */
VP9_ADDBLK_ST8x4_UB(dst, dst_stride, in0, in1, in2, in3);
dst += (4 * dst_stride);
VP9_ADDBLK_ST8x4_UB(dst, dst_stride, in4, in5, in6, in7);
}
static void hevc_idct_16x16_msa(int16_t *coeffs)
{
int16_t i, j, k;
int16_t buf[256];
int16_t *buf_ptr = &buf[0];
int16_t *src = coeffs;
const int16_t *filter = >16x16_cnst[0];
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
v8i16 in8, in9, in10, in11, in12, in13, in14, in15;
v8i16 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
v8i16 src0_r, src1_r, src2_r, src3_r, src4_r, src5_r, src6_r, src7_r;
v8i16 src0_l, src1_l, src2_l, src3_l, src4_l, src5_l, src6_l, src7_l;
for (i = 2; i--;) {
LD_SH16(src, 16, in0, in1, in2, in3, in4, in5, in6, in7,
in8, in9, in10, in11, in12, in13, in14, in15);
ILVR_H4_SH(in4, in0, in12, in8, in6, in2, in14, in10,
src0_r, src1_r, src2_r, src3_r);
ILVR_H4_SH(in5, in1, in13, in9, in3, in7, in11, in15,
src4_r, src5_r, src6_r, src7_r);
ILVL_H4_SH(in4, in0, in12, in8, in6, in2, in14, in10,
src0_l, src1_l, src2_l, src3_l);
ILVL_H4_SH(in5, in1, in13, in9, in3, in7, in11, in15,
src4_l, src5_l, src6_l, src7_l);
HEVC_IDCT16x16_COL(src0_r, src1_r, src2_r, src3_r, src4_r, src5_r,
src6_r, src7_r, src0_l, src1_l, src2_l, src3_l,
src4_l, src5_l, src6_l, src7_l, 7);
src += 8;
buf_ptr = (&buf[0] + 8);
filter = >16x16_cnst[0];
}
src = &buf[0];
buf_ptr = coeffs;
filter = >16x16_cnst[0];
for (i = 2; i--;) {
LD_SH16(src, 8, in0, in8, in1, in9, in2, in10, in3, in11,
in4, in12, in5, in13, in6, in14, in7, in15);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
TRANSPOSE8x8_SH_SH(in8, in9, in10, in11, in12, in13, in14, in15,
in8, in9, in10, in11, in12, in13, in14, in15);
ILVR_H4_SH(in4, in0, in12, in8, in6, in2, in14, in10,
src0_r, src1_r, src2_r, src3_r);
ILVR_H4_SH(in5, in1, in13, in9, in3, in7, in11, in15,
src4_r, src5_r, src6_r, src7_r);
ILVL_H4_SH(in4, in0, in12, in8, in6, in2, in14, in10,
src0_l, src1_l, src2_l, src3_l);
ILVL_H4_SH(in5, in1, in13, in9, in3, in7, in11, in15,
src4_l, src5_l, src6_l, src7_l);
HEVC_IDCT16x16_COL(src0_r, src1_r, src2_r, src3_r, src4_r, src5_r,
src6_r, src7_r, src0_l, src1_l, src2_l, src3_l,
src4_l, src5_l, src6_l, src7_l, 12);
src += 128;
buf_ptr = coeffs + 8;
filter = >16x16_cnst[0];
}
LD_SH8(coeffs, 16, in0, in1, in2, in3, in4, in5, in6, in7);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7);
ST_SH8(vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, coeffs, 16);
LD_SH8((coeffs + 8), 16, in0, in1, in2, in3, in4, in5, in6, in7);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7);
LD_SH8((coeffs + 128), 16, in8, in9, in10, in11, in12, in13, in14, in15);
ST_SH8(vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, (coeffs + 128), 16);
TRANSPOSE8x8_SH_SH(in8, in9, in10, in11, in12, in13, in14, in15,
vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7);
ST_SH8(vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, (coeffs + 8), 16);
LD_SH8((coeffs + 136), 16, in0, in1, in2, in3, in4, in5, in6, in7);
TRANSPOSE8x8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7,
vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7);
ST_SH8(vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, (coeffs + 136), 16);
}