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);
}
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 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 fdct8x32_1d_row_even_rd(int16_t *temp, int16_t *out) {
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, temp0, temp1;
/* fdct32 even */
/* stage 2 */
LD_SH8(temp, 8, in0, in1, in2, in3, in4, in5, in6, in7);
LD_SH8(temp + 64, 8, 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,
vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7,
in8, in9, in10, in11, in12, in13, in14, in15);
FDCT_POSTPROC_2V_NEG_H(vec0, vec1);
FDCT_POSTPROC_2V_NEG_H(vec2, vec3);
FDCT_POSTPROC_2V_NEG_H(vec4, vec5);
FDCT_POSTPROC_2V_NEG_H(vec6, vec7);
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);
/* Stage 3 */
ADD4(vec0, vec7, vec1, vec6, vec2, vec5, vec3, vec4, in0, in1, in2, in3);
temp0 = in0 + in3;
in0 = in0 - in3;
in3 = in1 + in2;
in1 = in1 - in2;
DOTP_CONST_PAIR(temp0, in3, cospi_16_64, cospi_16_64, temp1, temp0);
ST_SH(temp0, out);
ST_SH(temp1, out + 8);
DOTP_CONST_PAIR(in0, in1, cospi_24_64, cospi_8_64, temp1, temp0);
ST_SH(temp0, out + 16);
ST_SH(temp1, out + 24);
SUB4(vec3, vec4, vec2, vec5, vec1, vec6, vec0, vec7, vec4, vec5, vec6, vec7);
DOTP_CONST_PAIR(vec6, vec5, cospi_16_64, cospi_16_64, vec5, vec6);
ADD2(vec4, vec5, vec7, vec6, vec0, vec1);
DOTP_CONST_PAIR(vec1, vec0, cospi_28_64, cospi_4_64, temp1, temp0);
ST_SH(temp0, out + 32);
ST_SH(temp1, out + 56);
SUB2(vec4, vec5, vec7, vec6, vec4, vec7);
DOTP_CONST_PAIR(vec7, vec4, cospi_12_64, cospi_20_64, temp1, temp0);
ST_SH(temp0, out + 40);
ST_SH(temp1, out + 48);
DOTP_CONST_PAIR(in13, in10, cospi_16_64, cospi_16_64, vec2, vec5);
DOTP_CONST_PAIR(in12, in11, cospi_16_64, cospi_16_64, vec3, vec4);
ADD4(in8, vec3, in9, vec2, in14, vec5, in15, vec4, in0, vec1, vec6, in2);
DOTP_CONST_PAIR(vec6, vec1, cospi_24_64, cospi_8_64, in1, in3);
ADD2(in0, in1, in2, in3, vec0, vec7);
DOTP_CONST_PAIR(vec7, vec0, cospi_30_64, cospi_2_64, temp1, temp0);
ST_SH(temp0, out + 64);
ST_SH(temp1, out + 120);
SUB2(in0, in1, in2, in3, in0, in2);
DOTP_CONST_PAIR(in2, in0, cospi_14_64, cospi_18_64, temp1, temp0);
ST_SH(temp0, out + 72);
ST_SH(temp1, out + 112);
SUB2(in9, vec2, in14, vec5, vec2, vec5);
DOTP_CONST_PAIR((-vec2), vec5, cospi_24_64, cospi_8_64, in2, in1);
SUB4(in8, vec3, in15, vec4, in3, in2, in0, in1, in3, in0, vec2, vec5);
DOTP_CONST_PAIR(vec5, vec2, cospi_22_64, cospi_10_64, temp1, temp0);
ST_SH(temp0, out + 80);
ST_SH(temp1, out + 104);
ADD2(in3, in2, in0, in1, vec3, vec4);
DOTP_CONST_PAIR(vec4, vec3, cospi_6_64, cospi_26_64, temp0, temp1);
ST_SH(temp0, out + 96);
ST_SH(temp1, out + 88);
}
static void fdct8x32_1d_row_even_4x(int16_t *input, int16_t *interm_ptr,
int16_t *out) {
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;
v4i32 vec0_l, vec1_l, vec2_l, vec3_l, vec4_l, vec5_l, vec6_l, vec7_l;
v4i32 vec0_r, vec1_r, vec2_r, vec3_r, vec4_r, vec5_r, vec6_r, vec7_r;
v4i32 tmp0_w, tmp1_w, tmp2_w, tmp3_w;
/* fdct32 even */
/* stage 2 */
LD_SH8(input, 8, in0, in1, in2, in3, in4, in5, in6, in7);
LD_SH8(input + 64, 8, 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,
vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7,
in8, in9, in10, in11, in12, in13, in14, in15);
ST_SH8(vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, interm_ptr, 8);
ST_SH8(in8, in9, in10, in11, in12, in13, in14, in15, interm_ptr + 64, 8);
/* Stage 3 */
UNPCK_SH_SW(vec0, vec0_l, vec0_r);
UNPCK_SH_SW(vec1, vec1_l, vec1_r);
UNPCK_SH_SW(vec2, vec2_l, vec2_r);
UNPCK_SH_SW(vec3, vec3_l, vec3_r);
UNPCK_SH_SW(vec4, vec4_l, vec4_r);
UNPCK_SH_SW(vec5, vec5_l, vec5_r);
UNPCK_SH_SW(vec6, vec6_l, vec6_r);
UNPCK_SH_SW(vec7, vec7_l, vec7_r);
ADD4(vec0_r, vec7_r, vec1_r, vec6_r, vec2_r, vec5_r, vec3_r, vec4_r,
tmp0_w, tmp1_w, tmp2_w, tmp3_w);
BUTTERFLY_4(tmp0_w, tmp1_w, tmp2_w, tmp3_w, vec4_r, vec6_r, vec7_r, vec5_r);
ADD4(vec0_l, vec7_l, vec1_l, vec6_l, vec2_l, vec5_l, vec3_l, vec4_l,
vec0_r, vec1_r, vec2_r, vec3_r);
tmp3_w = vec0_r + vec3_r;
vec0_r = vec0_r - vec3_r;
vec3_r = vec1_r + vec2_r;
vec1_r = vec1_r - vec2_r;
DOTP_CONST_PAIR_W(vec4_r, vec6_r, tmp3_w, vec3_r, cospi_16_64,
cospi_16_64, vec4_r, tmp3_w, vec6_r, vec3_r);
FDCT32_POSTPROC_NEG_W(vec4_r);
FDCT32_POSTPROC_NEG_W(tmp3_w);
FDCT32_POSTPROC_NEG_W(vec6_r);
FDCT32_POSTPROC_NEG_W(vec3_r);
PCKEV_H2_SH(vec4_r, tmp3_w, vec6_r, vec3_r, vec4, vec5);
ST_SH2(vec5, vec4, out, 8);
DOTP_CONST_PAIR_W(vec5_r, vec7_r, vec0_r, vec1_r, cospi_24_64,
cospi_8_64, vec4_r, tmp3_w, vec6_r, vec3_r);
FDCT32_POSTPROC_NEG_W(vec4_r);
FDCT32_POSTPROC_NEG_W(tmp3_w);
FDCT32_POSTPROC_NEG_W(vec6_r);
FDCT32_POSTPROC_NEG_W(vec3_r);
PCKEV_H2_SH(vec4_r, tmp3_w, vec6_r, vec3_r, vec4, vec5);
ST_SH2(vec5, vec4, out + 16, 8);
LD_SH8(interm_ptr, 8, vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7);
SUB4(vec3, vec4, vec2, vec5, vec1, vec6, vec0, vec7, vec4, vec5, vec6, vec7);
DOTP_CONST_PAIR(vec6, vec5, cospi_16_64, cospi_16_64, vec5, vec6);
ADD2(vec4, vec5, vec7, vec6, vec0, vec1);
DOTP_CONST_PAIR(vec1, vec0, cospi_28_64, cospi_4_64, in5, in4);
FDCT_POSTPROC_2V_NEG_H(in4, in5);
ST_SH(in4, out + 32);
ST_SH(in5, out + 56);
SUB2(vec4, vec5, vec7, vec6, vec4, vec7);
DOTP_CONST_PAIR(vec7, vec4, cospi_12_64, cospi_20_64, in5, in4);
FDCT_POSTPROC_2V_NEG_H(in4, in5);
ST_SH(in4, out + 40);
ST_SH(in5, out + 48);
LD_SH8(interm_ptr + 64, 8, in8, in9, in10, in11, in12, in13, in14, in15);
DOTP_CONST_PAIR(in13, in10, cospi_16_64, cospi_16_64, vec2, vec5);
DOTP_CONST_PAIR(in12, in11, cospi_16_64, cospi_16_64, vec3, vec4);
ADD4(in8, vec3, in9, vec2, in14, vec5, in15, vec4, in0, vec1, vec6, in2);
DOTP_CONST_PAIR(vec6, vec1, cospi_24_64, cospi_8_64, in1, in3);
ADD2(in0, in1, in2, in3, vec0, vec7);
DOTP_CONST_PAIR(vec7, vec0, cospi_30_64, cospi_2_64, in5, in4);
FDCT_POSTPROC_2V_NEG_H(in4, in5);
ST_SH(in4, out + 64);
ST_SH(in5, out + 120);
SUB2(in0, in1, in2, in3, in0, in2);
DOTP_CONST_PAIR(in2, in0, cospi_14_64, cospi_18_64, in5, in4);
FDCT_POSTPROC_2V_NEG_H(in4, in5);
ST_SH(in4, out + 72);
ST_SH(in5, out + 112);
SUB2(in9, vec2, in14, vec5, vec2, vec5);
DOTP_CONST_PAIR((-vec2), vec5, cospi_24_64, cospi_8_64, in2, in1);
SUB4(in8, vec3, in15, vec4, in3, in2, in0, in1, in3, in0, vec2, vec5);
DOTP_CONST_PAIR(vec5, vec2, cospi_22_64, cospi_10_64, in5, in4);
FDCT_POSTPROC_2V_NEG_H(in4, in5);
ST_SH(in4, out + 80);
ST_SH(in5, out + 104);
ADD2(in3, in2, in0, in1, vec3, vec4);
DOTP_CONST_PAIR(vec4, vec3, cospi_6_64, cospi_26_64, in4, in5);
FDCT_POSTPROC_2V_NEG_H(in4, in5);
ST_SH(in4, out + 96);
ST_SH(in5, out + 88);
}