generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Merge "Take out VP9_ prefix from mips/msa macros"

Browse Source
This commit is contained in:
Jingning Han 2015-07-23 01:15:52 +00:00 committed by Gerrit Code Review
commit 2b4f49cb1a
4 changed files with 343 additions and 343 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

156
vp9/encoder/mips/msa/vp9_fdct16x16_msa.c
View File

@ -37,8 +37,8 @@ static void fdct8x16_1d_column(const int16_t *input, int16_t *tmp_ptr,
SLLI_4V(in12, in13, in14, in15, 2);
ADD4(in0, in15, in1, in14, in2, in13, in3, in12, tmp0, tmp1, tmp2, tmp3);
ADD4(in4, in11, in5, in10, in6, in9, in7, in8, tmp4, tmp5, tmp6, tmp7);
VP9_FDCT8x16_EVEN(tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7,
tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7);
FDCT8x16_EVEN(tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7,
tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7);
ST_SH8(tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp_ptr, 32);
SUB4(in0, in15, in1, in14, in2, in13, in3, in12, in15, in14, in13, in12);
SUB4(in4, in11, in5, in10, in6, in9, in7, in8, in11, in10, in9, in8);
@ -50,13 +50,13 @@ static void fdct8x16_1d_column(const int16_t *input, int16_t *tmp_ptr,
ILVR_H2_SH(in10, in13, in11, in12, vec3, vec5);
cnst4 = __msa_splati_h(coeff, 0);
stp25 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec2, vec3, cnst4);
stp25 = DOT_SHIFT_RIGHT_PCK_H(vec2, vec3, cnst4);
cnst5 = __msa_splati_h(coeff, 1);
cnst5 = __msa_ilvev_h(cnst5, cnst4);
stp22 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec2, vec3, cnst5);
stp24 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec4, vec5, cnst4);
stp23 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec4, vec5, cnst5);
stp22 = DOT_SHIFT_RIGHT_PCK_H(vec2, vec3, cnst5);
stp24 = DOT_SHIFT_RIGHT_PCK_H(vec4, vec5, cnst4);
stp23 = DOT_SHIFT_RIGHT_PCK_H(vec4, vec5, cnst5);
/* stp2 */
BUTTERFLY_4(in8, in9, stp22, stp23, stp30, stp31, stp32, stp33);
@ -65,45 +65,45 @@ static void fdct8x16_1d_column(const int16_t *input, int16_t *tmp_ptr,
ILVR_H2_SH(stp36, stp31, stp35, stp32, vec3, vec5);
SPLATI_H2_SH(coeff, 2, 3, cnst0, cnst1);
cnst0 = __msa_ilvev_h(cnst0, cnst1);
stp26 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec2, vec3, cnst0);
stp26 = DOT_SHIFT_RIGHT_PCK_H(vec2, vec3, cnst0);
cnst0 = __msa_splati_h(coeff, 4);
cnst1 = __msa_ilvev_h(cnst1, cnst0);
stp21 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec2, vec3, cnst1);
stp21 = DOT_SHIFT_RIGHT_PCK_H(vec2, vec3, cnst1);
BUTTERFLY_4(stp30, stp37, stp26, stp21, in8, in15, in14, in9);
ILVRL_H2_SH(in15, in8, vec1, vec0);
SPLATI_H2_SH(coeff1, 0, 1, cnst0, cnst1);
cnst0 = __msa_ilvev_h(cnst0, cnst1);
in8 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
in8 = DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
ST_SH(in8, tmp_ptr);
cnst0 = __msa_splati_h(coeff2, 0);
cnst0 = __msa_ilvev_h(cnst1, cnst0);
in8 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
in8 = DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
ST_SH(in8, tmp_ptr + 224);
ILVRL_H2_SH(in14, in9, vec1, vec0);
SPLATI_H2_SH(coeff1, 2, 3, cnst0, cnst1);
cnst1 = __msa_ilvev_h(cnst1, cnst0);
in8 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst1);
in8 = DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst1);
ST_SH(in8, tmp_ptr + 128);
cnst1 = __msa_splati_h(coeff2, 2);
cnst0 = __msa_ilvev_h(cnst0, cnst1);
in8 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
in8 = DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
ST_SH(in8, tmp_ptr + 96);
SPLATI_H2_SH(coeff, 2, 5, cnst0, cnst1);
cnst1 = __msa_ilvev_h(cnst1, cnst0);
stp25 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec4, vec5, cnst1);
stp25 = DOT_SHIFT_RIGHT_PCK_H(vec4, vec5, cnst1);
cnst1 = __msa_splati_h(coeff, 3);
cnst1 = __msa_ilvev_h(cnst0, cnst1);
stp22 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec4, vec5, cnst1);
stp22 = DOT_SHIFT_RIGHT_PCK_H(vec4, vec5, cnst1);
/* stp4 */
ADD2(stp34, stp25, stp33, stp22, in13, in10);
@ -111,12 +111,12 @@ static void fdct8x16_1d_column(const int16_t *input, int16_t *tmp_ptr,
ILVRL_H2_SH(in13, in10, vec1, vec0);
SPLATI_H2_SH(coeff1, 4, 5, cnst0, cnst1);
cnst0 = __msa_ilvev_h(cnst0, cnst1);
in8 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
in8 = DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
ST_SH(in8, tmp_ptr + 64);
cnst0 = __msa_splati_h(coeff2, 1);
cnst0 = __msa_ilvev_h(cnst1, cnst0);
in8 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
in8 = DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
ST_SH(in8, tmp_ptr + 160);
SUB2(stp34, stp25, stp33, stp22, in12, in11);
@ -124,12 +124,12 @@ static void fdct8x16_1d_column(const int16_t *input, int16_t *tmp_ptr,
SPLATI_H2_SH(coeff1, 6, 7, cnst0, cnst1);
cnst1 = __msa_ilvev_h(cnst1, cnst0);
in8 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst1);
in8 = DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst1);
ST_SH(in8, tmp_ptr + 192);
cnst1 = __msa_splati_h(coeff2, 3);
cnst0 = __msa_ilvev_h(cnst0, cnst1);
in8 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
in8 = DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
ST_SH(in8, tmp_ptr + 32);
}
@ -156,11 +156,11 @@ static void fdct16x8_1d_row(int16_t *input, int16_t *output) {
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);
VP9_FDCT8x16_EVEN(tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7,
tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7);
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);
VP9_FDCT8x16_ODD(in8, in9, in10, in11, in12, in13, in14, in15,
in0, in1, in2, in3, in4, in5, in6, in7);
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);
@ -188,10 +188,10 @@ void vp9_fdct16x16_msa(const int16_t *input, int16_t *output,
void vp9_fdct16x16_1_msa(const int16_t *input, int16_t *out, int32_t stride) {
out[1] = 0;
out[0] = VP9_LD_HADD(input, stride);
out[0] += VP9_LD_HADD(input + 8, stride);
out[0] += VP9_LD_HADD(input + 16 * 8, stride);
out[0] += VP9_LD_HADD(input + 16 * 8 + 8, stride);
out[0] = LD_HADD(input, stride);
out[0] += LD_HADD(input + 8, stride);
out[0] += LD_HADD(input + 16 * 8, stride);
out[0] += LD_HADD(input + 16 * 8 + 8, stride);
out[0] >>= 1;
}
@ -211,7 +211,7 @@ static void fadst16_cols_step1_msa(const int16_t *input, int32_t stride,
/* stage 1 */
LD_SW2(const0, 4, k0, k1);
LD_SW2(const0 + 8, 4, k2, k3);
VP9_MADD_BF(r15, r0, r7, r8, k0, k1, k2, k3, g0, g1, g2, g3);
MADD_BF(r15, r0, r7, r8, k0, k1, k2, k3, g0, g1, g2, g3);
r3 = LD_SH(input + 3 * stride);
r4 = LD_SH(input + 4 * stride);
@ -221,7 +221,7 @@ static void fadst16_cols_step1_msa(const int16_t *input, int32_t stride,
LD_SW2(const0 + 4 * 4, 4, k0, k1);
LD_SW2(const0 + 4 * 6, 4, k2, k3);
VP9_MADD_BF(r11, r4, r3, r12, k0, k1, k2, k3, g8, g9, g10, g11);
MADD_BF(r11, r4, r3, r12, k0, k1, k2, k3, g8, g9, g10, g11);
/* stage 2 */
BUTTERFLY_4(g0, g2, g10, g8, tp0, tp2, tp3, tp1);
@ -230,7 +230,7 @@ static void fadst16_cols_step1_msa(const int16_t *input, int32_t stride,
LD_SW2(const0 + 4 * 8, 4, k0, k1);
k2 = LD_SW(const0 + 4 * 10);
VP9_MADD_BF(g1, g3, g9, g11, k0, k1, k2, k0, h0, h1, h2, h3);
MADD_BF(g1, g3, g9, g11, k0, k1, k2, k0, h0, h1, h2, h3);
ST_SH2(h0, h1, int_buf + 8 * 8, 8);
ST_SH2(h3, h2, int_buf + 12 * 8, 8);
@ -243,7 +243,7 @@ static void fadst16_cols_step1_msa(const int16_t *input, int32_t stride,
LD_SW2(const0 + 4 * 11, 4, k0, k1);
LD_SW2(const0 + 4 * 13, 4, k2, k3);
VP9_MADD_BF(r9, r6, r1, r14, k0, k1, k2, k3, g0, g1, g2, g3);
MADD_BF(r9, r6, r1, r14, k0, k1, k2, k3, g0, g1, g2, g3);
ST_SH2(g1, g3, int_buf + 3 * 8, 4 * 8);
@ -255,7 +255,7 @@ static void fadst16_cols_step1_msa(const int16_t *input, int32_t stride,
LD_SW2(const0 + 4 * 15, 4, k0, k1);
LD_SW2(const0 + 4 * 17, 4, k2, k3);
VP9_MADD_BF(r13, r2, r5, r10, k0, k1, k2, k3, h0, h1, h2, h3);
MADD_BF(r13, r2, r5, r10, k0, k1, k2, k3, h0, h1, h2, h3);
ST_SH2(h1, h3, int_buf + 11 * 8, 4 * 8);
@ -276,7 +276,7 @@ static void fadst16_cols_step2_msa(int16_t *int_buf, const int32_t *const0,
LD_SH2(int_buf + 11 * 8, 4 * 8, g5, g7);
LD_SW2(const0 + 4 * 19, 4, k0, k1);
k2 = LD_SW(const0 + 4 * 21);
VP9_MADD_BF(g7, g5, g15, g13, k0, k1, k2, k0, h4, h5, h6, h7);
MADD_BF(g7, g5, g15, g13, k0, k1, k2, k0, h4, h5, h6, h7);
tp0 = LD_SH(int_buf + 4 * 8);
tp1 = LD_SH(int_buf + 5 * 8);
@ -284,14 +284,14 @@ static void fadst16_cols_step2_msa(int16_t *int_buf, const int32_t *const0,
tp2 = LD_SH(int_buf + 14 * 8);
LD_SW2(const0 + 4 * 22, 4, k0, k1);
k2 = LD_SW(const0 + 4 * 24);
VP9_MADD_BF(tp0, tp1, tp2, tp3, k0, k1, k2, k0, out4, out6, out5, out7);
MADD_BF(tp0, tp1, tp2, tp3, k0, k1, k2, k0, out4, out6, out5, out7);
out4 = -out4;
ST_SH(out4, (out + 3 * 16));
ST_SH(out5, (out_ptr + 4 * 16));
h1 = LD_SH(int_buf + 9 * 8);
h3 = LD_SH(int_buf + 12 * 8);
VP9_MADD_BF(h1, h3, h5, h7, k0, k1, k2, k0, out12, out14, out13, out15);
MADD_BF(h1, h3, h5, h7, k0, k1, k2, k0, out12, out14, out13, out15);
out13 = -out13;
ST_SH(out12, (out + 2 * 16));
ST_SH(out13, (out_ptr + 5 * 16));
@ -317,19 +317,19 @@ static void fadst16_cols_step2_msa(int16_t *int_buf, const int32_t *const0,
/* stage 4 */
LD_SW2(const0 + 4 * 25, 4, k0, k1);
LD_SW2(const0 + 4 * 27, 4, k2, k3);
VP9_MADD_SHORT(h10, h11, k1, k2, out2, out3);
MADD_SHORT(h10, h11, k1, k2, out2, out3);
ST_SH(out2, (out + 7 * 16));
ST_SH(out3, (out_ptr));
VP9_MADD_SHORT(out6, out7, k0, k3, out6, out7);
MADD_SHORT(out6, out7, k0, k3, out6, out7);
ST_SH(out6, (out + 4 * 16));
ST_SH(out7, (out_ptr + 3 * 16));
VP9_MADD_SHORT(out10, out11, k0, k3, out10, out11);
MADD_SHORT(out10, out11, k0, k3, out10, out11);
ST_SH(out10, (out + 6 * 16));
ST_SH(out11, (out_ptr + 16));
VP9_MADD_SHORT(out14, out15, k1, k2, out14, out15);
MADD_SHORT(out14, out15, k1, k2, out14, out15);
ST_SH(out14, (out + 5 * 16));
ST_SH(out15, (out_ptr + 2 * 16));
}
@ -342,20 +342,20 @@ static void fadst16_transpose_postproc_msa(int16_t *input, int16_t *out) {
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);
VP9_FDCT_POSTPROC_2V_NEG_H(r0, r1);
VP9_FDCT_POSTPROC_2V_NEG_H(r2, r3);
VP9_FDCT_POSTPROC_2V_NEG_H(r4, r5);
VP9_FDCT_POSTPROC_2V_NEG_H(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);
VP9_FDCT_POSTPROC_2V_NEG_H(r8, r9);
VP9_FDCT_POSTPROC_2V_NEG_H(r10, r11);
VP9_FDCT_POSTPROC_2V_NEG_H(r12, r13);
VP9_FDCT_POSTPROC_2V_NEG_H(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;
@ -364,20 +364,20 @@ static void fadst16_transpose_postproc_msa(int16_t *input, int16_t *out) {
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);
VP9_FDCT_POSTPROC_2V_NEG_H(r0, r1);
VP9_FDCT_POSTPROC_2V_NEG_H(r2, r3);
VP9_FDCT_POSTPROC_2V_NEG_H(r4, r5);
VP9_FDCT_POSTPROC_2V_NEG_H(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);
VP9_FDCT_POSTPROC_2V_NEG_H(r8, r9);
VP9_FDCT_POSTPROC_2V_NEG_H(r10, r11);
VP9_FDCT_POSTPROC_2V_NEG_H(r12, r13);
VP9_FDCT_POSTPROC_2V_NEG_H(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);
}
@ -396,7 +396,7 @@ static void fadst16_rows_step1_msa(int16_t *input, const int32_t *const0,
/* stage 1 */
LD_SW2(const0, 4, k0, k1);
LD_SW2(const0 + 4 * 2, 4, k2, k3);
VP9_MADD_BF(r15, r0, r7, r8, k0, k1, k2, k3, g0, g1, g2, g3);
MADD_BF(r15, r0, r7, r8, k0, k1, k2, k3, g0, g1, g2, g3);
r3 = LD_SH(input + 3 * 8);
r4 = LD_SH(input + 4 * 8);
@ -405,7 +405,7 @@ static void fadst16_rows_step1_msa(int16_t *input, const int32_t *const0,
LD_SW2(const0 + 4 * 4, 4, k0, k1);
LD_SW2(const0 + 4 * 6, 4, k2, k3);
VP9_MADD_BF(r11, r4, r3, r12, k0, k1, k2, k3, g8, g9, g10, g11);
MADD_BF(r11, r4, r3, r12, k0, k1, k2, k3, g8, g9, g10, g11);
/* stage 2 */
BUTTERFLY_4(g0, g2, g10, g8, tp0, tp2, tp3, tp1);
@ -414,7 +414,7 @@ static void fadst16_rows_step1_msa(int16_t *input, const int32_t *const0,
LD_SW2(const0 + 4 * 8, 4, k0, k1);
k2 = LD_SW(const0 + 4 * 10);
VP9_MADD_BF(g1, g3, g9, g11, k0, k1, k2, k0, h0, h1, h2, h3);
MADD_BF(g1, g3, g9, g11, k0, k1, k2, k0, h0, h1, h2, h3);
ST_SH2(h0, h3, int_buf + 8 * 8, 4 * 8);
ST_SH2(h1, h2, int_buf + 9 * 8, 4 * 8);
@ -425,7 +425,7 @@ static void fadst16_rows_step1_msa(int16_t *input, const int32_t *const0,
LD_SW2(const0 + 4 * 11, 4, k0, k1);
LD_SW2(const0 + 4 * 13, 4, k2, k3);
VP9_MADD_BF(r9, r6, r1, r14, k0, k1, k2, k3, g0, g1, g2, g3);
MADD_BF(r9, r6, r1, r14, k0, k1, k2, k3, g0, g1, g2, g3);
ST_SH2(g1, g3, int_buf + 3 * 8, 4 * 8);
r2 = LD_SH(input + 2 * 8);
@ -435,7 +435,7 @@ static void fadst16_rows_step1_msa(int16_t *input, const int32_t *const0,
LD_SW2(const0 + 4 * 15, 4, k0, k1);
LD_SW2(const0 + 4 * 17, 4, k2, k3);
VP9_MADD_BF(r13, r2, r5, r10, k0, k1, k2, k3, h0, h1, h2, h3);
MADD_BF(r13, r2, r5, r10, k0, k1, k2, k3, h0, h1, h2, h3);
ST_SH2(h1, h3, int_buf + 11 * 8, 4 * 8);
BUTTERFLY_4(h0, h2, g2, g0, tp0, tp1, tp2, tp3);
ST_SH4(tp0, tp1, tp2, tp3, int_buf + 2 * 8, 4 * 8);
@ -457,7 +457,7 @@ static void fadst16_rows_step2_msa(int16_t *int_buf, const int32_t *const0,
LD_SW2(const0 + 4 * 19, 4, k0, k1);
k2 = LD_SW(const0 + 4 * 21);
VP9_MADD_BF(g7, g5, g15, g13, k0, k1, k2, k0, h4, h5, h6, h7);
MADD_BF(g7, g5, g15, g13, k0, k1, k2, k0, h4, h5, h6, h7);
tp0 = LD_SH(int_buf + 4 * 8);
tp1 = LD_SH(int_buf + 5 * 8);
@ -466,14 +466,14 @@ static void fadst16_rows_step2_msa(int16_t *int_buf, const int32_t *const0,
LD_SW2(const0 + 4 * 22, 4, k0, k1);
k2 = LD_SW(const0 + 4 * 24);
VP9_MADD_BF(tp0, tp1, tp2, tp3, k0, k1, k2, k0, out4, out6, out5, out7);
MADD_BF(tp0, tp1, tp2, tp3, k0, k1, k2, k0, out4, out6, out5, out7);
out4 = -out4;
ST_SH(out4, (out + 3 * 16));
ST_SH(out5, (out_ptr + 4 * 16));
h1 = LD_SH(int_buf + 9 * 8);
h3 = LD_SH(int_buf + 12 * 8);
VP9_MADD_BF(h1, h3, h5, h7, k0, k1, k2, k0, out12, out14, out13, out15);
MADD_BF(h1, h3, h5, h7, k0, k1, k2, k0, out12, out14, out13, out15);
out13 = -out13;
ST_SH(out12, (out + 2 * 16));
ST_SH(out13, (out_ptr + 5 * 16));
@ -498,19 +498,19 @@ static void fadst16_rows_step2_msa(int16_t *int_buf, const int32_t *const0,
/* stage 4 */
LD_SW2(const0 + 4 * 25, 4, k0, k1);
LD_SW2(const0 + 4 * 27, 4, k2, k3);
VP9_MADD_SHORT(h10, h11, k1, k2, out2, out3);
MADD_SHORT(h10, h11, k1, k2, out2, out3);
ST_SH(out2, (out + 7 * 16));
ST_SH(out3, (out_ptr));
VP9_MADD_SHORT(out6, out7, k0, k3, out6, out7);
MADD_SHORT(out6, out7, k0, k3, out6, out7);
ST_SH(out6, (out + 4 * 16));
ST_SH(out7, (out_ptr + 3 * 16));
VP9_MADD_SHORT(out10, out11, k0, k3, out10, out11);
MADD_SHORT(out10, out11, k0, k3, out10, out11);
ST_SH(out10, (out + 6 * 16));
ST_SH(out11, (out_ptr + 16));
VP9_MADD_SHORT(out14, out15, k1, k2, out14, out15);
MADD_SHORT(out14, out15, k1, k2, out14, out15);
ST_SH(out14, (out + 5 * 16));
ST_SH(out15, (out_ptr + 2 * 16));
}
@ -556,26 +556,26 @@ static void postproc_fdct16x8_1d_row(int16_t *intermediate, int16_t *output) {
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);
VP9_FDCT_POSTPROC_2V_NEG_H(in0, in1);
VP9_FDCT_POSTPROC_2V_NEG_H(in2, in3);
VP9_FDCT_POSTPROC_2V_NEG_H(in4, in5);
VP9_FDCT_POSTPROC_2V_NEG_H(in6, in7);
VP9_FDCT_POSTPROC_2V_NEG_H(in8, in9);
VP9_FDCT_POSTPROC_2V_NEG_H(in10, in11);
VP9_FDCT_POSTPROC_2V_NEG_H(in12, in13);
VP9_FDCT_POSTPROC_2V_NEG_H(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);
VP9_FDCT8x16_EVEN(tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7,
tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7);
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);
VP9_FDCT8x16_ODD(in8, in9, in10, in11, in12, in13, in14, in15,
in0, in1, in2, in3, in4, in5, in6, in7);
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);

384
vp9/encoder/mips/msa/vp9_fdct32x32_msa.c
View File

@ -75,57 +75,57 @@ static void fdct8x32_1d_column_even_store(int16_t *input, int16_t *temp) {
/* Stage 3 */
ADD4(vec0, vec7, vec1, vec6, vec2, vec5, vec3, vec4, in0, in1, in2, in3);
BUTTERFLY_4(in0, in1, in2, in3, temp0, in4, in1, in0);
VP9_DOTP_CONST_PAIR(temp0, in4, cospi_16_64, cospi_16_64, temp1, temp0);
VP9_FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
DOTP_CONST_PAIR(temp0, in4, cospi_16_64, cospi_16_64, temp1, temp0);
FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
ST_SH(temp0, temp);
ST_SH(temp1, temp + 512);
VP9_DOTP_CONST_PAIR(in0, in1, cospi_24_64, cospi_8_64, temp1, temp0);
VP9_FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
DOTP_CONST_PAIR(in0, in1, cospi_24_64, cospi_8_64, temp1, temp0);
FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
ST_SH(temp0, temp + 256);
ST_SH(temp1, temp + 768);
SUB4(vec0, vec7, vec1, vec6, vec2, vec5, vec3, vec4, vec7, vec6, vec5, vec4);
VP9_DOTP_CONST_PAIR(vec6, vec5, cospi_16_64, cospi_16_64, vec5, vec6);
DOTP_CONST_PAIR(vec6, vec5, cospi_16_64, cospi_16_64, vec5, vec6);
ADD2(vec4, vec5, vec7, vec6, vec0, vec1);
VP9_DOTP_CONST_PAIR(vec1, vec0, cospi_28_64, cospi_4_64, temp1, temp0);
VP9_FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
DOTP_CONST_PAIR(vec1, vec0, cospi_28_64, cospi_4_64, temp1, temp0);
FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
ST_SH(temp0, temp + 128);
ST_SH(temp1, temp + 896);
SUB2(vec4, vec5, vec7, vec6, vec4, vec7);
VP9_DOTP_CONST_PAIR(vec7, vec4, cospi_12_64, cospi_20_64, temp1, temp0);
VP9_FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
DOTP_CONST_PAIR(vec7, vec4, cospi_12_64, cospi_20_64, temp1, temp0);
FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
ST_SH(temp0, temp + 640);
ST_SH(temp1, temp + 384);
VP9_DOTP_CONST_PAIR(in13, in10, cospi_16_64, cospi_16_64, vec2, vec5);
VP9_DOTP_CONST_PAIR(in12, in11, cospi_16_64, cospi_16_64, vec3, vec4);
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);
VP9_DOTP_CONST_PAIR(vec6, vec1, cospi_24_64, cospi_8_64, in1, in3);
DOTP_CONST_PAIR(vec6, vec1, cospi_24_64, cospi_8_64, in1, in3);
ADD2(in0, in1, in2, in3, vec0, vec7);
VP9_DOTP_CONST_PAIR(vec7, vec0, cospi_30_64, cospi_2_64, temp1, temp0);
VP9_FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
DOTP_CONST_PAIR(vec7, vec0, cospi_30_64, cospi_2_64, temp1, temp0);
FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
ST_SH(temp0, temp + 64);
ST_SH(temp1, temp + 960);
SUB2(in0, in1, in2, in3, in0, in2);
VP9_DOTP_CONST_PAIR(in2, in0, cospi_14_64, cospi_18_64, temp1, temp0);
VP9_FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
DOTP_CONST_PAIR(in2, in0, cospi_14_64, cospi_18_64, temp1, temp0);
FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
ST_SH(temp0, temp + 576);
ST_SH(temp1, temp + 448);
SUB2(in9, vec2, in14, vec5, vec2, vec5);
VP9_DOTP_CONST_PAIR((-vec2), vec5, cospi_24_64, cospi_8_64, in2, in1);
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);
VP9_DOTP_CONST_PAIR(vec5, vec2, cospi_22_64, cospi_10_64, temp1, temp0);
VP9_FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
DOTP_CONST_PAIR(vec5, vec2, cospi_22_64, cospi_10_64, temp1, temp0);
FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
ST_SH(temp0, temp + 320);
ST_SH(temp1, temp + 704);
ADD2(in3, in2, in0, in1, vec3, vec4);
VP9_DOTP_CONST_PAIR(vec4, vec3, cospi_6_64, cospi_26_64, temp0, temp1);
VP9_FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
DOTP_CONST_PAIR(vec4, vec3, cospi_6_64, cospi_26_64, temp0, temp1);
FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
ST_SH(temp0, temp + 192);
ST_SH(temp1, temp + 832);
}
@ -139,8 +139,8 @@ static void fdct8x32_1d_column_odd_store(int16_t *input, int16_t *temp_ptr) {
in26 = LD_SH(input + 80);
in27 = LD_SH(input + 88);
VP9_DOTP_CONST_PAIR(in27, in20, cospi_16_64, cospi_16_64, in20, in27);
VP9_DOTP_CONST_PAIR(in26, in21, cospi_16_64, cospi_16_64, in21, in26);
DOTP_CONST_PAIR(in27, in20, cospi_16_64, cospi_16_64, in20, in27);
DOTP_CONST_PAIR(in26, in21, cospi_16_64, cospi_16_64, in21, in26);
in18 = LD_SH(input + 16);
in19 = LD_SH(input + 24);
@ -162,8 +162,8 @@ static void fdct8x32_1d_column_odd_store(int16_t *input, int16_t *temp_ptr) {
in26 = in29 + in26;
LD_SH4(input + 48, 8, in22, in23, in24, in25);
VP9_DOTP_CONST_PAIR(in25, in22, cospi_16_64, cospi_16_64, in22, in25);
VP9_DOTP_CONST_PAIR(in24, in23, cospi_16_64, cospi_16_64, in23, in24);
DOTP_CONST_PAIR(in25, in22, cospi_16_64, cospi_16_64, in22, in25);
DOTP_CONST_PAIR(in24, in23, cospi_16_64, cospi_16_64, in23, in24);
in16 = LD_SH(input);
in17 = LD_SH(input + 8);
@ -180,33 +180,33 @@ static void fdct8x32_1d_column_odd_store(int16_t *input, int16_t *temp_ptr) {
ST_SH(vec4, input + 104);
ADD4(in16, in23, in17, in22, in30, in25, in31, in24, in16, in17, in30, in31);
VP9_DOTP_CONST_PAIR(in26, in21, cospi_24_64, cospi_8_64, in18, in29);
VP9_DOTP_CONST_PAIR(in27, in20, cospi_24_64, cospi_8_64, in19, in28);
DOTP_CONST_PAIR(in26, in21, cospi_24_64, cospi_8_64, in18, in29);
DOTP_CONST_PAIR(in27, in20, cospi_24_64, cospi_8_64, in19, in28);
ADD4(in16, in19, in17, in18, in30, in29, in31, in28, in27, in22, in21, in25);
VP9_DOTP_CONST_PAIR(in21, in22, cospi_28_64, cospi_4_64, in26, in24);
DOTP_CONST_PAIR(in21, in22, cospi_28_64, cospi_4_64, in26, in24);
ADD2(in27, in26, in25, in24, in23, in20);
VP9_DOTP_CONST_PAIR(in20, in23, cospi_31_64, cospi_1_64, vec4, vec5);
VP9_FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
DOTP_CONST_PAIR(in20, in23, cospi_31_64, cospi_1_64, vec4, vec5);
FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
ST_SH(vec5, temp_ptr);
ST_SH(vec4, temp_ptr + 960);
SUB2(in27, in26, in25, in24, in22, in21);
VP9_DOTP_CONST_PAIR(in21, in22, cospi_15_64, cospi_17_64, vec5, vec4);
VP9_FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
DOTP_CONST_PAIR(in21, in22, cospi_15_64, cospi_17_64, vec5, vec4);
FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
ST_SH(vec5, temp_ptr + 448);
ST_SH(vec4, temp_ptr + 512);
SUB4(in17, in18, in16, in19, in31, in28, in30, in29, in23, in26, in24, in20);
VP9_DOTP_CONST_PAIR((-in23), in20, cospi_28_64, cospi_4_64, in27, in25);
DOTP_CONST_PAIR((-in23), in20, cospi_28_64, cospi_4_64, in27, in25);
SUB2(in26, in27, in24, in25, in23, in20);
VP9_DOTP_CONST_PAIR(in20, in23, cospi_23_64, cospi_9_64, vec4, vec5);
VP9_FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
DOTP_CONST_PAIR(in20, in23, cospi_23_64, cospi_9_64, vec4, vec5);
FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
ST_SH(vec4, temp_ptr + 704);
ST_SH(vec5, temp_ptr + 256);
ADD2(in26, in27, in24, in25, in22, in21);
VP9_DOTP_CONST_PAIR(in21, in22, cospi_7_64, cospi_25_64, vec4, vec5);
VP9_FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
DOTP_CONST_PAIR(in21, in22, cospi_7_64, cospi_25_64, vec4, vec5);
FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
ST_SH(vec4, temp_ptr + 192);
ST_SH(vec5, temp_ptr + 768);
@ -214,32 +214,32 @@ static void fdct8x32_1d_column_odd_store(int16_t *input, int16_t *temp_ptr) {
LD_SH4(input + 80, 8, in26, in27, in24, in25);
in16 = in20;
in17 = in21;
VP9_DOTP_CONST_PAIR(-in16, in27, cospi_24_64, cospi_8_64, in20, in27);
VP9_DOTP_CONST_PAIR(-in17, in26, cospi_24_64, cospi_8_64, in21, in26);
DOTP_CONST_PAIR(-in16, in27, cospi_24_64, cospi_8_64, in20, in27);
DOTP_CONST_PAIR(-in17, in26, cospi_24_64, cospi_8_64, in21, in26);
SUB4(in23, in20, in22, in21, in25, in26, in24, in27, in28, in17, in18, in31);
VP9_DOTP_CONST_PAIR(in18, in17, cospi_12_64, cospi_20_64, in29, in30);
DOTP_CONST_PAIR(in18, in17, cospi_12_64, cospi_20_64, in29, in30);
ADD2(in28, in29, in31, in30, in16, in19);
VP9_DOTP_CONST_PAIR(in19, in16, cospi_27_64, cospi_5_64, vec5, vec4);
VP9_FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
DOTP_CONST_PAIR(in19, in16, cospi_27_64, cospi_5_64, vec5, vec4);
FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
ST_SH(vec5, temp_ptr + 832);
ST_SH(vec4, temp_ptr + 128);
SUB2(in28, in29, in31, in30, in17, in18);
VP9_DOTP_CONST_PAIR(in18, in17, cospi_11_64, cospi_21_64, vec5, vec4);
VP9_FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
DOTP_CONST_PAIR(in18, in17, cospi_11_64, cospi_21_64, vec5, vec4);
FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
ST_SH(vec5, temp_ptr + 320);
ST_SH(vec4, temp_ptr + 640);
ADD4(in22, in21, in23, in20, in24, in27, in25, in26, in16, in29, in30, in19);
VP9_DOTP_CONST_PAIR(-in16, in19, cospi_12_64, cospi_20_64, in28, in31);
DOTP_CONST_PAIR(-in16, in19, cospi_12_64, cospi_20_64, in28, in31);
SUB2(in29, in28, in30, in31, in16, in19);
VP9_DOTP_CONST_PAIR(in19, in16, cospi_19_64, cospi_13_64, vec5, vec4);
VP9_FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
DOTP_CONST_PAIR(in19, in16, cospi_19_64, cospi_13_64, vec5, vec4);
FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
ST_SH(vec5, temp_ptr + 576);
ST_SH(vec4, temp_ptr + 384);
ADD2(in29, in28, in30, in31, in17, in18);
VP9_DOTP_CONST_PAIR(in18, in17, cospi_3_64, cospi_29_64, vec5, vec4);
VP9_FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
DOTP_CONST_PAIR(in18, in17, cospi_3_64, cospi_29_64, vec5, vec4);
FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
ST_SH(vec5, temp_ptr + 64);
ST_SH(vec4, temp_ptr + 896);
}
@ -327,67 +327,67 @@ static void fdct8x32_1d_row_even_4x(int16_t *input, int16_t *interm_ptr,
vec3_r = vec1_r + vec2_r;
vec1_r = vec1_r - vec2_r;
VP9_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);
VP9_FDCT32_POSTPROC_NEG_W(vec4_r);
VP9_FDCT32_POSTPROC_NEG_W(tmp3_w);
VP9_FDCT32_POSTPROC_NEG_W(vec6_r);
VP9_FDCT32_POSTPROC_NEG_W(vec3_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);
VP9_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);
VP9_FDCT32_POSTPROC_NEG_W(vec4_r);
VP9_FDCT32_POSTPROC_NEG_W(tmp3_w);
VP9_FDCT32_POSTPROC_NEG_W(vec6_r);
VP9_FDCT32_POSTPROC_NEG_W(vec3_r);
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);
VP9_DOTP_CONST_PAIR(vec6, vec5, cospi_16_64, cospi_16_64, vec5, vec6);
DOTP_CONST_PAIR(vec6, vec5, cospi_16_64, cospi_16_64, vec5, vec6);
ADD2(vec4, vec5, vec7, vec6, vec0, vec1);
VP9_DOTP_CONST_PAIR(vec1, vec0, cospi_28_64, cospi_4_64, in5, in4);
VP9_FDCT_POSTPROC_2V_NEG_H(in4, in5);
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);
VP9_DOTP_CONST_PAIR(vec7, vec4, cospi_12_64, cospi_20_64, in5, in4);
VP9_FDCT_POSTPROC_2V_NEG_H(in4, in5);
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);
VP9_DOTP_CONST_PAIR(in13, in10, cospi_16_64, cospi_16_64, vec2, vec5);
VP9_DOTP_CONST_PAIR(in12, in11, cospi_16_64, cospi_16_64, vec3, vec4);
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);
VP9_DOTP_CONST_PAIR(vec6, vec1, cospi_24_64, cospi_8_64, in1, in3);
DOTP_CONST_PAIR(vec6, vec1, cospi_24_64, cospi_8_64, in1, in3);
ADD2(in0, in1, in2, in3, vec0, vec7);
VP9_DOTP_CONST_PAIR(vec7, vec0, cospi_30_64, cospi_2_64, in5, in4);
VP9_FDCT_POSTPROC_2V_NEG_H(in4, in5);
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);
VP9_DOTP_CONST_PAIR(in2, in0, cospi_14_64, cospi_18_64, in5, in4);
VP9_FDCT_POSTPROC_2V_NEG_H(in4, in5);
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);
VP9_DOTP_CONST_PAIR((-vec2), vec5, cospi_24_64, cospi_8_64, in2, in1);
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);
VP9_DOTP_CONST_PAIR(vec5, vec2, cospi_22_64, cospi_10_64, in5, in4);
VP9_FDCT_POSTPROC_2V_NEG_H(in4, in5);
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);
VP9_DOTP_CONST_PAIR(vec4, vec3, cospi_6_64, cospi_26_64, in4, in5);
VP9_FDCT_POSTPROC_2V_NEG_H(in4, in5);
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);
}
@ -410,57 +410,57 @@ static void fdct8x32_1d_row_even(int16_t *temp, int16_t *out) {
/* Stage 3 */
ADD4(vec0, vec7, vec1, vec6, vec2, vec5, vec3, vec4, in0, in1, in2, in3);
BUTTERFLY_4(in0, in1, in2, in3, temp0, in4, in1, in0);
VP9_DOTP_CONST_PAIR(temp0, in4, cospi_16_64, cospi_16_64, temp1, temp0);
VP9_FDCT_POSTPROC_2V_NEG_H(temp0, temp1);
DOTP_CONST_PAIR(temp0, in4, cospi_16_64, cospi_16_64, temp1, temp0);
FDCT_POSTPROC_2V_NEG_H(temp0, temp1);
ST_SH(temp0, out);
ST_SH(temp1, out + 8);
VP9_DOTP_CONST_PAIR(in0, in1, cospi_24_64, cospi_8_64, temp1, temp0);
VP9_FDCT_POSTPROC_2V_NEG_H(temp0, temp1);
DOTP_CONST_PAIR(in0, in1, cospi_24_64, cospi_8_64, temp1, temp0);
FDCT_POSTPROC_2V_NEG_H(temp0, temp1);
ST_SH(temp0, out + 16);
ST_SH(temp1, out + 24);
SUB4(vec3, vec4, vec2, vec5, vec1, vec6, vec0, vec7, vec4, vec5, vec6, vec7);
VP9_DOTP_CONST_PAIR(vec6, vec5, cospi_16_64, cospi_16_64, vec5, vec6);
DOTP_CONST_PAIR(vec6, vec5, cospi_16_64, cospi_16_64, vec5, vec6);
ADD2(vec4, vec5, vec7, vec6, vec0, vec1);
VP9_DOTP_CONST_PAIR(vec1, vec0, cospi_28_64, cospi_4_64, temp1, temp0);
VP9_FDCT_POSTPROC_2V_NEG_H(temp0, temp1);
DOTP_CONST_PAIR(vec1, vec0, cospi_28_64, cospi_4_64, temp1, temp0);
FDCT_POSTPROC_2V_NEG_H(temp0, temp1);
ST_SH(temp0, out + 32);
ST_SH(temp1, out + 56);
SUB2(vec4, vec5, vec7, vec6, vec4, vec7);
VP9_DOTP_CONST_PAIR(vec7, vec4, cospi_12_64, cospi_20_64, temp1, temp0);
VP9_FDCT_POSTPROC_2V_NEG_H(temp0, temp1);
DOTP_CONST_PAIR(vec7, vec4, cospi_12_64, cospi_20_64, temp1, temp0);
FDCT_POSTPROC_2V_NEG_H(temp0, temp1);
ST_SH(temp0, out + 40);
ST_SH(temp1, out + 48);
VP9_DOTP_CONST_PAIR(in13, in10, cospi_16_64, cospi_16_64, vec2, vec5);
VP9_DOTP_CONST_PAIR(in12, in11, cospi_16_64, cospi_16_64, vec3, vec4);
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);
VP9_DOTP_CONST_PAIR(vec6, vec1, cospi_24_64, cospi_8_64, in1, in3);
DOTP_CONST_PAIR(vec6, vec1, cospi_24_64, cospi_8_64, in1, in3);
ADD2(in0, in1, in2, in3, vec0, vec7);
VP9_DOTP_CONST_PAIR(vec7, vec0, cospi_30_64, cospi_2_64, temp1, temp0);
VP9_FDCT_POSTPROC_2V_NEG_H(temp0, temp1);
DOTP_CONST_PAIR(vec7, vec0, cospi_30_64, cospi_2_64, temp1, temp0);
FDCT_POSTPROC_2V_NEG_H(temp0, temp1);
ST_SH(temp0, out + 64);
ST_SH(temp1, out + 120);
SUB2(in0, in1, in2, in3, in0, in2);
VP9_DOTP_CONST_PAIR(in2, in0, cospi_14_64, cospi_18_64, temp1, temp0);
VP9_FDCT_POSTPROC_2V_NEG_H(temp0, temp1);
DOTP_CONST_PAIR(in2, in0, cospi_14_64, cospi_18_64, temp1, temp0);
FDCT_POSTPROC_2V_NEG_H(temp0, temp1);
ST_SH(temp0, out + 72);
ST_SH(temp1, out + 112);
SUB2(in9, vec2, in14, vec5, vec2, vec5);
VP9_DOTP_CONST_PAIR((-vec2), vec5, cospi_24_64, cospi_8_64, in2, in1);
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)
VP9_DOTP_CONST_PAIR(vec5, vec2, cospi_22_64, cospi_10_64, temp1, temp0);
VP9_FDCT_POSTPROC_2V_NEG_H(temp0, temp1);
DOTP_CONST_PAIR(vec5, vec2, cospi_22_64, cospi_10_64, temp1, temp0);
FDCT_POSTPROC_2V_NEG_H(temp0, temp1);
ST_SH(temp0, out + 80);
ST_SH(temp1, out + 104);
ADD2(in3, in2, in0, in1, vec3, vec4);
VP9_DOTP_CONST_PAIR(vec4, vec3, cospi_6_64, cospi_26_64, temp0, temp1);
VP9_FDCT_POSTPROC_2V_NEG_H(temp0, temp1);
DOTP_CONST_PAIR(vec4, vec3, cospi_6_64, cospi_26_64, temp0, temp1);
FDCT_POSTPROC_2V_NEG_H(temp0, temp1);
ST_SH(temp0, out + 96);
ST_SH(temp1, out + 88);
}
@ -475,8 +475,8 @@ static void fdct8x32_1d_row_odd(int16_t *temp, int16_t *interm_ptr,
in26 = LD_SH(temp + 80);
in27 = LD_SH(temp + 88);
VP9_DOTP_CONST_PAIR(in27, in20, cospi_16_64, cospi_16_64, in20, in27);
VP9_DOTP_CONST_PAIR(in26, in21, cospi_16_64, cospi_16_64, in21, in26);
DOTP_CONST_PAIR(in27, in20, cospi_16_64, cospi_16_64, in20, in27);
DOTP_CONST_PAIR(in26, in21, cospi_16_64, cospi_16_64, in21, in26);
in18 = LD_SH(temp + 16);
in19 = LD_SH(temp + 24);
@ -499,8 +499,8 @@ static void fdct8x32_1d_row_odd(int16_t *temp, int16_t *interm_ptr,
in24 = LD_SH(temp + 64);
in25 = LD_SH(temp + 72);
VP9_DOTP_CONST_PAIR(in25, in22, cospi_16_64, cospi_16_64, in22, in25);
VP9_DOTP_CONST_PAIR(in24, in23, cospi_16_64, cospi_16_64, in23, in24);
DOTP_CONST_PAIR(in25, in22, cospi_16_64, cospi_16_64, in22, in25);
DOTP_CONST_PAIR(in24, in23, cospi_16_64, cospi_16_64, in23, in24);
in16 = LD_SH(temp);
in17 = LD_SH(temp + 8);
@ -517,37 +517,37 @@ static void fdct8x32_1d_row_odd(int16_t *temp, int16_t *interm_ptr,
ST_SH(vec4, interm_ptr + 80);
ADD4(in16, in23, in17, in22, in30, in25, in31, in24, in16, in17, in30, in31);
VP9_DOTP_CONST_PAIR(in26, in21, cospi_24_64, cospi_8_64, in18, in29);
VP9_DOTP_CONST_PAIR(in27, in20, cospi_24_64, cospi_8_64, in19, in28);
DOTP_CONST_PAIR(in26, in21, cospi_24_64, cospi_8_64, in18, in29);
DOTP_CONST_PAIR(in27, in20, cospi_24_64, cospi_8_64, in19, in28);
ADD4(in16, in19, in17, in18, in30, in29, in31, in28, in27, in22, in21, in25);
VP9_DOTP_CONST_PAIR(in21, in22, cospi_28_64, cospi_4_64, in26, in24);
DOTP_CONST_PAIR(in21, in22, cospi_28_64, cospi_4_64, in26, in24);
ADD2(in27, in26, in25, in24, in23, in20);
VP9_DOTP_CONST_PAIR(in20, in23, cospi_31_64, cospi_1_64, vec4, vec5);
VP9_FDCT_POSTPROC_2V_NEG_H(vec5, vec4);
DOTP_CONST_PAIR(in20, in23, cospi_31_64, cospi_1_64, vec4, vec5);
FDCT_POSTPROC_2V_NEG_H(vec5, vec4);
ST_SH(vec5, out);
ST_SH(vec4, out + 120);
SUB2(in27, in26, in25, in24, in22, in21);
VP9_DOTP_CONST_PAIR(in21, in22, cospi_15_64, cospi_17_64, vec5, vec4);
VP9_FDCT_POSTPROC_2V_NEG_H(vec5, vec4);
DOTP_CONST_PAIR(in21, in22, cospi_15_64, cospi_17_64, vec5, vec4);
FDCT_POSTPROC_2V_NEG_H(vec5, vec4);
ST_SH(vec5, out + 112);
ST_SH(vec4, out + 8);
SUB4(in17, in18, in16, in19, in31, in28, in30, in29, in23, in26, in24, in20);
VP9_DOTP_CONST_PAIR((-in23), in20, cospi_28_64, cospi_4_64, in27, in25);
DOTP_CONST_PAIR((-in23), in20, cospi_28_64, cospi_4_64, in27, in25);
SUB2(in26, in27, in24, in25, in23, in20);
VP9_DOTP_CONST_PAIR(in20, in23, cospi_23_64, cospi_9_64, vec4, vec5);
VP9_FDCT_POSTPROC_2V_NEG_H(vec5, vec4);
DOTP_CONST_PAIR(in20, in23, cospi_23_64, cospi_9_64, vec4, vec5);
FDCT_POSTPROC_2V_NEG_H(vec5, vec4);
ST_SH(vec4, out + 16);
ST_SH(vec5, out + 104);
ADD2(in26, in27, in24, in25, in22, in21);
VP9_DOTP_CONST_PAIR(in21, in22, cospi_7_64, cospi_25_64, vec4, vec5);
VP9_FDCT_POSTPROC_2V_NEG_H(vec5, vec4);
DOTP_CONST_PAIR(in21, in22, cospi_7_64, cospi_25_64, vec4, vec5);
FDCT_POSTPROC_2V_NEG_H(vec5, vec4);
ST_SH(vec4, out + 24);
ST_SH(vec5, out + 96);
@ -558,8 +558,8 @@ static void fdct8x32_1d_row_odd(int16_t *temp, int16_t *interm_ptr,
in16 = in20;
in17 = in21;
VP9_DOTP_CONST_PAIR(-in16, in27, cospi_24_64, cospi_8_64, in20, in27);
VP9_DOTP_CONST_PAIR(-in17, in26, cospi_24_64, cospi_8_64, in21, in26);
DOTP_CONST_PAIR(-in16, in27, cospi_24_64, cospi_8_64, in20, in27);
DOTP_CONST_PAIR(-in17, in26, cospi_24_64, cospi_8_64, in21, in26);
in22 = LD_SH(interm_ptr + 40);
in25 = LD_SH(interm_ptr + 48);
@ -567,32 +567,32 @@ static void fdct8x32_1d_row_odd(int16_t *temp, int16_t *interm_ptr,
in23 = LD_SH(interm_ptr + 80);
SUB4(in23, in20, in22, in21, in25, in26, in24, in27, in28, in17, in18, in31);
VP9_DOTP_CONST_PAIR(in18, in17, cospi_12_64, cospi_20_64, in29, in30);
DOTP_CONST_PAIR(in18, in17, cospi_12_64, cospi_20_64, in29, in30);
ADD2(in28, in29, in31, in30, in16, in19);
VP9_DOTP_CONST_PAIR(in19, in16, cospi_27_64, cospi_5_64, vec5, vec4);
VP9_FDCT_POSTPROC_2V_NEG_H(vec5, vec4);
DOTP_CONST_PAIR(in19, in16, cospi_27_64, cospi_5_64, vec5, vec4);
FDCT_POSTPROC_2V_NEG_H(vec5, vec4);
ST_SH(vec5, out + 32);
ST_SH(vec4, out + 88);
SUB2(in28, in29, in31, in30, in17, in18);
VP9_DOTP_CONST_PAIR(in18, in17, cospi_11_64, cospi_21_64, vec5, vec4);
VP9_FDCT_POSTPROC_2V_NEG_H(vec5, vec4);
DOTP_CONST_PAIR(in18, in17, cospi_11_64, cospi_21_64, vec5, vec4);
FDCT_POSTPROC_2V_NEG_H(vec5, vec4);
ST_SH(vec5, out + 40);
ST_SH(vec4, out + 80);
ADD4(in22, in21, in23, in20, in24, in27, in25, in26, in16, in29, in30, in19);
VP9_DOTP_CONST_PAIR(-in16, in19, cospi_12_64, cospi_20_64, in28, in31);
DOTP_CONST_PAIR(-in16, in19, cospi_12_64, cospi_20_64, in28, in31);
SUB2(in29, in28, in30, in31, in16, in19);
VP9_DOTP_CONST_PAIR(in19, in16, cospi_19_64, cospi_13_64, vec5, vec4);
VP9_FDCT_POSTPROC_2V_NEG_H(vec5, vec4);
DOTP_CONST_PAIR(in19, in16, cospi_19_64, cospi_13_64, vec5, vec4);
FDCT_POSTPROC_2V_NEG_H(vec5, vec4);
ST_SH(vec5, out + 72);
ST_SH(vec4, out + 48);
ADD2(in29, in28, in30, in31, in17, in18);
VP9_DOTP_CONST_PAIR(in18, in17, cospi_3_64, cospi_29_64, vec5, vec4);
VP9_FDCT_POSTPROC_2V_NEG_H(vec5, vec4);
DOTP_CONST_PAIR(in18, in17, cospi_3_64, cospi_29_64, vec5, vec4);
FDCT_POSTPROC_2V_NEG_H(vec5, vec4);
ST_SH(vec4, out + 56);
ST_SH(vec5, out + 64);
}
@ -700,22 +700,22 @@ void vp9_fdct32x32_msa(const int16_t *input, int16_t *output,
void vp9_fdct32x32_1_msa(const int16_t *input, int16_t *out, int32_t stride) {
out[1] = 0;
out[0] = VP9_LD_HADD(input, stride);
out[0] += VP9_LD_HADD(input + 8, stride);
out[0] += VP9_LD_HADD(input + 16, stride);
out[0] += VP9_LD_HADD(input + 24, stride);
out[0] += VP9_LD_HADD(input + 32 * 8, stride);
out[0] += VP9_LD_HADD(input + 32 * 8 + 8, stride);
out[0] += VP9_LD_HADD(input + 32 * 8 + 16, stride);
out[0] += VP9_LD_HADD(input + 32 * 8 + 24, stride);
out[0] += VP9_LD_HADD(input + 32 * 16, stride);
out[0] += VP9_LD_HADD(input + 32 * 16 + 8, stride);
out[0] += VP9_LD_HADD(input + 32 * 16 + 16, stride);
out[0] += VP9_LD_HADD(input + 32 * 16 + 24, stride);
out[0] += VP9_LD_HADD(input + 32 * 24, stride);
out[0] += VP9_LD_HADD(input + 32 * 24 + 8, stride);
out[0] += VP9_LD_HADD(input + 32 * 24 + 16, stride);
out[0] += VP9_LD_HADD(input + 32 * 24 + 24, stride);
out[0] = LD_HADD(input, stride);
out[0] += LD_HADD(input + 8, stride);
out[0] += LD_HADD(input + 16, stride);
out[0] += LD_HADD(input + 24, stride);
out[0] += LD_HADD(input + 32 * 8, stride);
out[0] += LD_HADD(input + 32 * 8 + 8, stride);
out[0] += LD_HADD(input + 32 * 8 + 16, stride);
out[0] += LD_HADD(input + 32 * 8 + 24, stride);
out[0] += LD_HADD(input + 32 * 16, stride);
out[0] += LD_HADD(input + 32 * 16 + 8, stride);
out[0] += LD_HADD(input + 32 * 16 + 16, stride);
out[0] += LD_HADD(input + 32 * 16 + 24, stride);
out[0] += LD_HADD(input + 32 * 24, stride);
out[0] += LD_HADD(input + 32 * 24 + 8, stride);
out[0] += LD_HADD(input + 32 * 24 + 16, stride);
out[0] += LD_HADD(input + 32 * 24 + 24, stride);
out[0] >>= 3;
}
@ -733,14 +733,14 @@ static void fdct8x32_1d_row_even_rd(int16_t *temp, int16_t *out) {
in8, in9, in10, in11, in12, in13, in14, in15,
vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7,
in8, in9, in10, in11, in12, in13, in14, in15);
VP9_FDCT_POSTPROC_2V_NEG_H(vec0, vec1);
VP9_FDCT_POSTPROC_2V_NEG_H(vec2, vec3);
VP9_FDCT_POSTPROC_2V_NEG_H(vec4, vec5);
VP9_FDCT_POSTPROC_2V_NEG_H(vec6, vec7);
VP9_FDCT_POSTPROC_2V_NEG_H(in8, in9);
VP9_FDCT_POSTPROC_2V_NEG_H(in10, in11);
VP9_FDCT_POSTPROC_2V_NEG_H(in12, in13);
VP9_FDCT_POSTPROC_2V_NEG_H(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);
@ -750,49 +750,49 @@ static void fdct8x32_1d_row_even_rd(int16_t *temp, int16_t *out) {
in3 = in1 + in2;
in1 = in1 - in2;
VP9_DOTP_CONST_PAIR(temp0, in3, cospi_16_64, cospi_16_64, temp1, temp0);
DOTP_CONST_PAIR(temp0, in3, cospi_16_64, cospi_16_64, temp1, temp0);
ST_SH(temp0, out);
ST_SH(temp1, out + 8);
VP9_DOTP_CONST_PAIR(in0, in1, cospi_24_64, cospi_8_64, temp1, temp0);
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);
VP9_DOTP_CONST_PAIR(vec6, vec5, cospi_16_64, cospi_16_64, vec5, vec6);
DOTP_CONST_PAIR(vec6, vec5, cospi_16_64, cospi_16_64, vec5, vec6);
ADD2(vec4, vec5, vec7, vec6, vec0, vec1);
VP9_DOTP_CONST_PAIR(vec1, vec0, cospi_28_64, cospi_4_64, temp1, temp0);
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);
VP9_DOTP_CONST_PAIR(vec7, vec4, cospi_12_64, cospi_20_64, temp1, temp0);
DOTP_CONST_PAIR(vec7, vec4, cospi_12_64, cospi_20_64, temp1, temp0);
ST_SH(temp0, out + 40);
ST_SH(temp1, out + 48);
VP9_DOTP_CONST_PAIR(in13, in10, cospi_16_64, cospi_16_64, vec2, vec5);
VP9_DOTP_CONST_PAIR(in12, in11, cospi_16_64, cospi_16_64, vec3, vec4);
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);
VP9_DOTP_CONST_PAIR(vec6, vec1, cospi_24_64, cospi_8_64, in1, in3);
DOTP_CONST_PAIR(vec6, vec1, cospi_24_64, cospi_8_64, in1, in3);
ADD2(in0, in1, in2, in3, vec0, vec7);
VP9_DOTP_CONST_PAIR(vec7, vec0, cospi_30_64, cospi_2_64, temp1, temp0);
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);
VP9_DOTP_CONST_PAIR(in2, in0, cospi_14_64, cospi_18_64, temp1, temp0);
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);
VP9_DOTP_CONST_PAIR((-vec2), vec5, cospi_24_64, cospi_8_64, in2, in1);
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);
VP9_DOTP_CONST_PAIR(vec5, vec2, cospi_22_64, cospi_10_64, temp1, temp0);
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);
VP9_DOTP_CONST_PAIR(vec4, vec3, cospi_6_64, cospi_26_64, temp0, temp1);
DOTP_CONST_PAIR(vec4, vec3, cospi_6_64, cospi_26_64, temp0, temp1);
ST_SH(temp0, out + 96);
ST_SH(temp1, out + 88);
}
@ -808,19 +808,19 @@ static void fdct8x32_1d_row_odd_rd(int16_t *temp, int16_t *interm_ptr,
in26 = LD_SH(temp + 80);
in27 = LD_SH(temp + 88);
VP9_DOTP_CONST_PAIR(in27, in20, cospi_16_64, cospi_16_64, in20, in27);
VP9_DOTP_CONST_PAIR(in26, in21, cospi_16_64, cospi_16_64, in21, in26);
DOTP_CONST_PAIR(in27, in20, cospi_16_64, cospi_16_64, in20, in27);
DOTP_CONST_PAIR(in26, in21, cospi_16_64, cospi_16_64, in21, in26);
VP9_FDCT_POSTPROC_2V_NEG_H(in20, in21);
VP9_FDCT_POSTPROC_2V_NEG_H(in26, in27);
FDCT_POSTPROC_2V_NEG_H(in20, in21);
FDCT_POSTPROC_2V_NEG_H(in26, in27);
in18 = LD_SH(temp + 16);
in19 = LD_SH(temp + 24);
in28 = LD_SH(temp + 96);
in29 = LD_SH(temp + 104);
VP9_FDCT_POSTPROC_2V_NEG_H(in18, in19);
VP9_FDCT_POSTPROC_2V_NEG_H(in28, in29);
FDCT_POSTPROC_2V_NEG_H(in18, in19);
FDCT_POSTPROC_2V_NEG_H(in28, in29);
vec4 = in19 - in20;
ST_SH(vec4, interm_ptr + 32);
@ -838,18 +838,18 @@ static void fdct8x32_1d_row_odd_rd(int16_t *temp, int16_t *interm_ptr,
in24 = LD_SH(temp + 64);
in25 = LD_SH(temp + 72);
VP9_DOTP_CONST_PAIR(in25, in22, cospi_16_64, cospi_16_64, in22, in25);
VP9_DOTP_CONST_PAIR(in24, in23, cospi_16_64, cospi_16_64, in23, in24);
VP9_FDCT_POSTPROC_2V_NEG_H(in22, in23);
VP9_FDCT_POSTPROC_2V_NEG_H(in24, in25);
DOTP_CONST_PAIR(in25, in22, cospi_16_64, cospi_16_64, in22, in25);
DOTP_CONST_PAIR(in24, in23, cospi_16_64, cospi_16_64, in23, in24);
FDCT_POSTPROC_2V_NEG_H(in22, in23);
FDCT_POSTPROC_2V_NEG_H(in24, in25);
in16 = LD_SH(temp);
in17 = LD_SH(temp + 8);
in30 = LD_SH(temp + 112);
in31 = LD_SH(temp + 120);
VP9_FDCT_POSTPROC_2V_NEG_H(in16, in17);
VP9_FDCT_POSTPROC_2V_NEG_H(in30, in31);
FDCT_POSTPROC_2V_NEG_H(in16, in17);
FDCT_POSTPROC_2V_NEG_H(in30, in31);
vec4 = in17 - in22;
ST_SH(vec4, interm_ptr + 40);
@ -861,29 +861,29 @@ static void fdct8x32_1d_row_odd_rd(int16_t *temp, int16_t *interm_ptr,
ST_SH(vec4, interm_ptr + 80);
ADD4(in16, in23, in17, in22, in30, in25, in31, in24, in16, in17, in30, in31);
VP9_DOTP_CONST_PAIR(in26, in21, cospi_24_64, cospi_8_64, in18, in29);
VP9_DOTP_CONST_PAIR(in27, in20, cospi_24_64, cospi_8_64, in19, in28);
DOTP_CONST_PAIR(in26, in21, cospi_24_64, cospi_8_64, in18, in29);
DOTP_CONST_PAIR(in27, in20, cospi_24_64, cospi_8_64, in19, in28);
ADD4(in16, in19, in17, in18, in30, in29, in31, in28, in27, in22, in21, in25);
VP9_DOTP_CONST_PAIR(in21, in22, cospi_28_64, cospi_4_64, in26, in24);
DOTP_CONST_PAIR(in21, in22, cospi_28_64, cospi_4_64, in26, in24);
ADD2(in27, in26, in25, in24, in23, in20);
VP9_DOTP_CONST_PAIR(in20, in23, cospi_31_64, cospi_1_64, vec4, vec5);
DOTP_CONST_PAIR(in20, in23, cospi_31_64, cospi_1_64, vec4, vec5);
ST_SH(vec5, out);
ST_SH(vec4, out + 120);
SUB2(in27, in26, in25, in24, in22, in21);
VP9_DOTP_CONST_PAIR(in21, in22, cospi_15_64, cospi_17_64, vec5, vec4);
DOTP_CONST_PAIR(in21, in22, cospi_15_64, cospi_17_64, vec5, vec4);
ST_SH(vec5, out + 112);
ST_SH(vec4, out + 8);
SUB4(in17, in18, in16, in19, in31, in28, in30, in29, in23, in26, in24, in20);
VP9_DOTP_CONST_PAIR((-in23), in20, cospi_28_64, cospi_4_64, in27, in25);
DOTP_CONST_PAIR((-in23), in20, cospi_28_64, cospi_4_64, in27, in25);
SUB2(in26, in27, in24, in25, in23, in20);
VP9_DOTP_CONST_PAIR(in20, in23, cospi_23_64, cospi_9_64, vec4, vec5);
DOTP_CONST_PAIR(in20, in23, cospi_23_64, cospi_9_64, vec4, vec5);
ST_SH(vec4, out + 16);
ST_SH(vec5, out + 104);
ADD2(in26, in27, in24, in25, in22, in21);
VP9_DOTP_CONST_PAIR(in21, in22, cospi_7_64, cospi_25_64, vec4, vec5);
DOTP_CONST_PAIR(in21, in22, cospi_7_64, cospi_25_64, vec4, vec5);
ST_SH(vec4, out + 24);
ST_SH(vec5, out + 96);
@ -894,8 +894,8 @@ static void fdct8x32_1d_row_odd_rd(int16_t *temp, int16_t *interm_ptr,
in16 = in20;
in17 = in21;
VP9_DOTP_CONST_PAIR(-in16, in27, cospi_24_64, cospi_8_64, in20, in27);
VP9_DOTP_CONST_PAIR(-in17, in26, cospi_24_64, cospi_8_64, in21, in26);
DOTP_CONST_PAIR(-in16, in27, cospi_24_64, cospi_8_64, in20, in27);
DOTP_CONST_PAIR(-in17, in26, cospi_24_64, cospi_8_64, in21, in26);
in22 = LD_SH(interm_ptr + 40);
in25 = LD_SH(interm_ptr + 48);
@ -903,27 +903,27 @@ static void fdct8x32_1d_row_odd_rd(int16_t *temp, int16_t *interm_ptr,
in23 = LD_SH(interm_ptr + 80);
SUB4(in23, in20, in22, in21, in25, in26, in24, in27, in28, in17, in18, in31);
VP9_DOTP_CONST_PAIR(in18, in17, cospi_12_64, cospi_20_64, in29, in30);
DOTP_CONST_PAIR(in18, in17, cospi_12_64, cospi_20_64, in29, in30);
in16 = in28 + in29;
in19 = in31 + in30;
VP9_DOTP_CONST_PAIR(in19, in16, cospi_27_64, cospi_5_64, vec5, vec4);
DOTP_CONST_PAIR(in19, in16, cospi_27_64, cospi_5_64, vec5, vec4);
ST_SH(vec5, out + 32);
ST_SH(vec4, out + 88);
SUB2(in28, in29, in31, in30, in17, in18);
VP9_DOTP_CONST_PAIR(in18, in17, cospi_11_64, cospi_21_64, vec5, vec4);
DOTP_CONST_PAIR(in18, in17, cospi_11_64, cospi_21_64, vec5, vec4);
ST_SH(vec5, out + 40);
ST_SH(vec4, out + 80);
ADD4(in22, in21, in23, in20, in24, in27, in25, in26, in16, in29, in30, in19);
VP9_DOTP_CONST_PAIR(-in16, in19, cospi_12_64, cospi_20_64, in28, in31);
DOTP_CONST_PAIR(-in16, in19, cospi_12_64, cospi_20_64, in28, in31);
SUB2(in29, in28, in30, in31, in16, in19);
VP9_DOTP_CONST_PAIR(in19, in16, cospi_19_64, cospi_13_64, vec5, vec4);
DOTP_CONST_PAIR(in19, in16, cospi_19_64, cospi_13_64, vec5, vec4);
ST_SH(vec5, out + 72);
ST_SH(vec4, out + 48);
ADD2(in29, in28, in30, in31, in17, in18);
VP9_DOTP_CONST_PAIR(in18, in17, cospi_3_64, cospi_29_64, vec5, vec4);
DOTP_CONST_PAIR(in18, in17, cospi_3_64, cospi_29_64, vec5, vec4);
ST_SH(vec4, out + 56);
ST_SH(vec5, out + 64);
}

6
vp9/encoder/mips/msa/vp9_fdct8x8_msa.c
View File

@ -28,12 +28,12 @@ void vp9_fdct8x8_msa(const int16_t *input, int16_t *output,
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_SRLI_AVE_S_4V_H(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);
}
void vp9_fdct8x8_1_msa(const int16_t *input, int16_t *out, int32_t stride) {
out[0] = VP9_LD_HADD(input, stride);
out[0] = LD_HADD(input, stride);
out[1] = 0;
}
@ -85,6 +85,6 @@ void vp9_fht8x8_msa(const int16_t *input, int16_t *output, int32_t stride,
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);
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);
}

140
vp9/encoder/mips/msa/vp9_fdct_msa.h
View File

@ -15,7 +15,7 @@
#include "vp9/common/vp9_idct.h"
#include "vpx_dsp/mips/macros_msa.h"
#define VP9_DOTP_CONST_PAIR(reg0, reg1, cnst0, cnst1, out0, out1) { \
#define DOTP_CONST_PAIR(reg0, reg1, cnst0, cnst1, out0, out1) { \
v8i16 k0_m = __msa_fill_h(cnst0); \
v4i32 s0_m, s1_m, s2_m, s3_m; \
\
@ -33,8 +33,8 @@
out1 = __msa_pckev_h((v8i16)s0_m, (v8i16)s1_m); \
}
#define VP9_DOT_ADD_SUB_SRARI_PCK(in0, in1, in2, in3, in4, in5, in6, in7, \
dst0, dst1, dst2, dst3) { \
#define DOT_ADD_SUB_SRARI_PCK(in0, in1, in2, in3, in4, in5, in6, in7, \
dst0, dst1, dst2, dst3) { \
v4i32 tp0_m, tp1_m, tp2_m, tp3_m, tp4_m; \
v4i32 tp5_m, tp6_m, tp7_m, tp8_m, tp9_m; \
\
@ -50,7 +50,7 @@
dst0, dst1, dst2, dst3); \
}
#define VP9_DOT_SHIFT_RIGHT_PCK_H(in0, in1, in2) ({ \
#define DOT_SHIFT_RIGHT_PCK_H(in0, in1, in2) ({ \
v8i16 dst_m; \
v4i32 tp0_m, tp1_m; \
\
@ -79,9 +79,9 @@
\
ILVRL_H2_SH(in0, in7, vec1_m, vec0_m); \
ILVRL_H2_SH(in4, in3, vec3_m, vec2_m); \
VP9_DOT_ADD_SUB_SRARI_PCK(vec0_m, vec1_m, vec2_m, vec3_m, cnst0_m, \
cnst1_m, cnst2_m, cnst3_m, in7, in0, \
in4, in3); \
DOT_ADD_SUB_SRARI_PCK(vec0_m, vec1_m, vec2_m, vec3_m, cnst0_m, \
cnst1_m, cnst2_m, cnst3_m, in7, in0, \
in4, in3); \
\
SPLATI_H2_SH(coeff0_m, 2, 5, cnst0_m, cnst1_m); \
cnst2_m = -cnst0_m; \
@ -93,9 +93,9 @@
ILVRL_H2_SH(in2, in5, vec1_m, vec0_m); \
ILVRL_H2_SH(in6, in1, vec3_m, vec2_m); \
\
VP9_DOT_ADD_SUB_SRARI_PCK(vec0_m, vec1_m, vec2_m, vec3_m, cnst0_m, \
cnst1_m, cnst2_m, cnst3_m, in5, in2, \
in6, in1); \
DOT_ADD_SUB_SRARI_PCK(vec0_m, vec1_m, vec2_m, vec3_m, cnst0_m, \
cnst1_m, cnst2_m, cnst3_m, in5, in2, \
in6, in1); \
BUTTERFLY_4(in7, in0, in2, in5, s1_m, s0_m, in2, in5); \
out7 = -s0_m; \
out0 = s1_m; \
@ -108,26 +108,26 @@
\
ILVRL_H2_SH(in4, in3, vec1_m, vec0_m); \
ILVRL_H2_SH(in6, in1, vec3_m, vec2_m); \
VP9_DOT_ADD_SUB_SRARI_PCK(vec0_m, vec1_m, vec2_m, vec3_m, cnst0_m, \
cnst2_m, cnst3_m, cnst1_m, out1, out6, \
s0_m, s1_m); \
DOT_ADD_SUB_SRARI_PCK(vec0_m, vec1_m, vec2_m, vec3_m, cnst0_m, \
cnst2_m, cnst3_m, cnst1_m, out1, out6, \
s0_m, s1_m); \
\
SPLATI_H2_SH(coeff1_m, 2, 3, cnst0_m, cnst1_m); \
cnst1_m = __msa_ilvev_h(cnst1_m, cnst0_m); \
\
ILVRL_H2_SH(in2, in5, vec1_m, vec0_m); \
ILVRL_H2_SH(s0_m, s1_m, vec3_m, vec2_m); \
out3 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
out4 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst1_m); \
out2 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst0_m); \
out5 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst1_m); \
out3 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
out4 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst1_m); \
out2 = DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst0_m); \
out5 = DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst1_m); \
\
out1 = -out1; \
out3 = -out3; \
out5 = -out5; \
}
#define VP9_MADD_SHORT(m0, m1, c0, c1, res0, res1) { \
#define MADD_SHORT(m0, m1, c0, c1, res0, res1) { \
v4i32 madd0_m, madd1_m, madd2_m, madd3_m; \
v8i16 madd_s0_m, madd_s1_m; \
\
@ -138,8 +138,8 @@
PCKEV_H2_SH(madd1_m, madd0_m, madd3_m, madd2_m, res0, res1); \
}
#define VP9_MADD_BF(inp0, inp1, inp2, inp3, cst0, cst1, cst2, cst3, \
out0, out1, out2, out3) { \
#define MADD_BF(inp0, inp1, inp2, inp3, cst0, cst1, cst2, cst3, \
out0, out1, out2, out3) { \
v8i16 madd_s0_m, madd_s1_m, madd_s2_m, madd_s3_m; \
v4i32 tmp0_m, tmp1_m, tmp2_m, tmp3_m, m4_m, m5_m; \
\
@ -159,7 +159,7 @@
PCKEV_H2_SH(m5_m, m4_m, tmp3_m, tmp2_m, out2, out3); \
}
#define VP9_LD_HADD(psrc, stride) ({ \
#define LD_HADD(psrc, stride) ({ \
v8i16 in0_m, in1_m, in2_m, in3_m, in4_m, in5_m, in6_m, in7_m; \
v4i32 vec_w_m; \
\
@ -174,7 +174,7 @@
HADD_SW_S32(vec_w_m); \
})
#define VP9_FDCT_POSTPROC_2V_NEG_H(vec0, vec1) { \
#define FDCT_POSTPROC_2V_NEG_H(vec0, vec1) { \
v8i16 tp0_m, tp1_m; \
v8i16 one_m = __msa_ldi_h(1); \
\
@ -267,23 +267,23 @@
ILVR_H2_SH(x1_m, x0_m, x3_m, x2_m, s1_m, s3_m); \
SPLATI_H2_SH(coeff_m, 0, 1, x0_m, x1_m); \
x1_m = __msa_ilvev_h(x1_m, x0_m); \
out4 = VP9_DOT_SHIFT_RIGHT_PCK_H(s0_m, s1_m, x1_m); \
out4 = DOT_SHIFT_RIGHT_PCK_H(s0_m, s1_m, x1_m); \
\
SPLATI_H2_SH(coeff_m, 2, 3, x2_m, x3_m); \
x2_m = -x2_m; \
x2_m = __msa_ilvev_h(x3_m, x2_m); \
out6 = VP9_DOT_SHIFT_RIGHT_PCK_H(s2_m, s3_m, x2_m); \
out6 = DOT_SHIFT_RIGHT_PCK_H(s2_m, s3_m, x2_m); \
\
out0 = VP9_DOT_SHIFT_RIGHT_PCK_H(s0_m, s1_m, x0_m); \
out0 = DOT_SHIFT_RIGHT_PCK_H(s0_m, s1_m, x0_m); \
x2_m = __msa_splati_h(coeff_m, 2); \
x2_m = __msa_ilvev_h(x2_m, x3_m); \
out2 = VP9_DOT_SHIFT_RIGHT_PCK_H(s2_m, s3_m, x2_m); \
out2 = DOT_SHIFT_RIGHT_PCK_H(s2_m, s3_m, x2_m); \
\
/* stage2 */ \
ILVRL_H2_SH(s5_m, s6_m, s1_m, s0_m); \
\
s6_m = VP9_DOT_SHIFT_RIGHT_PCK_H(s0_m, s1_m, x0_m); \
s5_m = VP9_DOT_SHIFT_RIGHT_PCK_H(s0_m, s1_m, x1_m); \
s6_m = DOT_SHIFT_RIGHT_PCK_H(s0_m, s1_m, x0_m); \
s5_m = DOT_SHIFT_RIGHT_PCK_H(s0_m, s1_m, x1_m); \
\
/* stage3 */ \
BUTTERFLY_4(s4_m, s7_m, s6_m, s5_m, x0_m, x3_m, x2_m, x1_m); \
@ -294,24 +294,24 @@
\
SPLATI_H2_SH(coeff_m, 4, 5, x0_m, x1_m); \
x1_m = __msa_ilvev_h(x0_m, x1_m); \
out1 = VP9_DOT_SHIFT_RIGHT_PCK_H(s4_m, s5_m, x1_m); \
out1 = DOT_SHIFT_RIGHT_PCK_H(s4_m, s5_m, x1_m); \
\
SPLATI_H2_SH(coeff_m, 6, 7, x2_m, x3_m); \
x2_m = __msa_ilvev_h(x3_m, x2_m); \
out5 = VP9_DOT_SHIFT_RIGHT_PCK_H(s6_m, s7_m, x2_m); \
out5 = DOT_SHIFT_RIGHT_PCK_H(s6_m, s7_m, x2_m); \
\
x1_m = __msa_splati_h(coeff_m, 5); \
x0_m = -x0_m; \
x0_m = __msa_ilvev_h(x1_m, x0_m); \
out7 = VP9_DOT_SHIFT_RIGHT_PCK_H(s4_m, s5_m, x0_m); \
out7 = DOT_SHIFT_RIGHT_PCK_H(s4_m, s5_m, x0_m); \
\
x2_m = __msa_splati_h(coeff_m, 6); \
x3_m = -x3_m; \
x2_m = __msa_ilvev_h(x2_m, x3_m); \
out3 = VP9_DOT_SHIFT_RIGHT_PCK_H(s6_m, s7_m, x2_m); \
out3 = DOT_SHIFT_RIGHT_PCK_H(s6_m, s7_m, x2_m); \
}
#define VP9_SRLI_AVE_S_4V_H(in0, in1, in2, in3, in4, in5, in6, in7) { \
#define SRLI_AVE_S_4V_H(in0, in1, in2, in3, in4, in5, in6, in7) { \
v8i16 vec0_m, vec1_m, vec2_m, vec3_m, vec4_m, vec5_m, vec6_m, vec7_m; \
\
SRLI_H4_SH(in0, in1, in2, in3, vec0_m, vec1_m, vec2_m, vec3_m, 15); \
@ -322,8 +322,8 @@
in4, in5, in6, in7); \
}
#define VP9_FDCT8x16_EVEN(in0, in1, in2, in3, in4, in5, in6, in7, \
out0, out1, out2, out3, out4, out5, out6, out7) { \
#define FDCT8x16_EVEN(in0, in1, in2, in3, in4, in5, in6, in7, \
out0, out1, out2, out3, out4, out5, out6, out7) { \
v8i16 s0_m, s1_m, s2_m, s3_m, s4_m, s5_m, s6_m, s7_m; \
v8i16 x0_m, x1_m, x2_m, x3_m; \
v8i16 coeff_m = { cospi_16_64, -cospi_16_64, cospi_8_64, cospi_24_64, \
@ -337,23 +337,23 @@
ILVR_H2_SH(x1_m, x0_m, x3_m, x2_m, s1_m, s3_m); \
SPLATI_H2_SH(coeff_m, 0, 1, x0_m, x1_m); \
x1_m = __msa_ilvev_h(x1_m, x0_m); \
out4 = VP9_DOT_SHIFT_RIGHT_PCK_H(s0_m, s1_m, x1_m); \
out4 = DOT_SHIFT_RIGHT_PCK_H(s0_m, s1_m, x1_m); \
\
SPLATI_H2_SH(coeff_m, 2, 3, x2_m, x3_m); \
x2_m = -x2_m; \
x2_m = __msa_ilvev_h(x3_m, x2_m); \
out6 = VP9_DOT_SHIFT_RIGHT_PCK_H(s2_m, s3_m, x2_m); \
out6 = DOT_SHIFT_RIGHT_PCK_H(s2_m, s3_m, x2_m); \
\
out0 = VP9_DOT_SHIFT_RIGHT_PCK_H(s0_m, s1_m, x0_m); \
out0 = DOT_SHIFT_RIGHT_PCK_H(s0_m, s1_m, x0_m); \
x2_m = __msa_splati_h(coeff_m, 2); \
x2_m = __msa_ilvev_h(x2_m, x3_m); \
out2 = VP9_DOT_SHIFT_RIGHT_PCK_H(s2_m, s3_m, x2_m); \
out2 = DOT_SHIFT_RIGHT_PCK_H(s2_m, s3_m, x2_m); \
\
/* stage2 */ \
ILVRL_H2_SH(s5_m, s6_m, s1_m, s0_m); \
\
s6_m = VP9_DOT_SHIFT_RIGHT_PCK_H(s0_m, s1_m, x0_m); \
s5_m = VP9_DOT_SHIFT_RIGHT_PCK_H(s0_m, s1_m, x1_m); \
s6_m = DOT_SHIFT_RIGHT_PCK_H(s0_m, s1_m, x0_m); \
s5_m = DOT_SHIFT_RIGHT_PCK_H(s0_m, s1_m, x1_m); \
\
/* stage3 */ \
BUTTERFLY_4(s4_m, s7_m, s6_m, s5_m, x0_m, x3_m, x2_m, x1_m); \
@ -364,27 +364,27 @@
\
SPLATI_H2_SH(coeff_m, 4, 5, x0_m, x1_m); \
x1_m = __msa_ilvev_h(x0_m, x1_m); \
out1 = VP9_DOT_SHIFT_RIGHT_PCK_H(s4_m, s5_m, x1_m); \
out1 = DOT_SHIFT_RIGHT_PCK_H(s4_m, s5_m, x1_m); \
\
SPLATI_H2_SH(coeff_m, 6, 7, x2_m, x3_m); \
x2_m = __msa_ilvev_h(x3_m, x2_m); \
out5 = VP9_DOT_SHIFT_RIGHT_PCK_H(s6_m, s7_m, x2_m); \
out5 = DOT_SHIFT_RIGHT_PCK_H(s6_m, s7_m, x2_m); \
\
x1_m = __msa_splati_h(coeff_m, 5); \
x0_m = -x0_m; \
x0_m = __msa_ilvev_h(x1_m, x0_m); \
out7 = VP9_DOT_SHIFT_RIGHT_PCK_H(s4_m, s5_m, x0_m); \
out7 = DOT_SHIFT_RIGHT_PCK_H(s4_m, s5_m, x0_m); \
\
x2_m = __msa_splati_h(coeff_m, 6); \
x3_m = -x3_m; \
x2_m = __msa_ilvev_h(x2_m, x3_m); \
out3 = VP9_DOT_SHIFT_RIGHT_PCK_H(s6_m, s7_m, x2_m); \
out3 = DOT_SHIFT_RIGHT_PCK_H(s6_m, s7_m, x2_m); \
}
#define VP9_FDCT8x16_ODD(input0, input1, input2, input3, \
input4, input5, input6, input7, \
out1, out3, out5, out7, \
out9, out11, out13, out15) { \
#define FDCT8x16_ODD(input0, input1, input2, input3, \
input4, input5, input6, input7, \
out1, out3, out5, out7, \
out9, out11, out13, out15) { \
v8i16 stp21_m, stp22_m, stp23_m, stp24_m, stp25_m, stp26_m; \
v8i16 stp30_m, stp31_m, stp32_m, stp33_m, stp34_m, stp35_m; \
v8i16 stp36_m, stp37_m, vec0_m, vec1_m; \
@ -404,13 +404,13 @@
ILVR_H2_SH(input2, input5, input3, input4, vec3_m, vec5_m); \
\
cnst4_m = __msa_splati_h(coeff_m, 0); \
stp25_m = VP9_DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst4_m); \
stp25_m = DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst4_m); \
\
cnst5_m = __msa_splati_h(coeff_m, 1); \
cnst5_m = __msa_ilvev_h(cnst5_m, cnst4_m); \
stp22_m = VP9_DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst5_m); \
stp24_m = VP9_DOT_SHIFT_RIGHT_PCK_H(vec4_m, vec5_m, cnst4_m); \
stp23_m = VP9_DOT_SHIFT_RIGHT_PCK_H(vec4_m, vec5_m, cnst5_m); \
stp22_m = DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst5_m); \
stp24_m = DOT_SHIFT_RIGHT_PCK_H(vec4_m, vec5_m, cnst4_m); \
stp23_m = DOT_SHIFT_RIGHT_PCK_H(vec4_m, vec5_m, cnst5_m); \
\
/* stp2 */ \
BUTTERFLY_4(input0, input1, stp22_m, stp23_m, \
@ -423,19 +423,19 @@
\
SPLATI_H2_SH(coeff_m, 2, 3, cnst0_m, cnst1_m); \
cnst0_m = __msa_ilvev_h(cnst0_m, cnst1_m); \
stp26_m = VP9_DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst0_m); \
stp26_m = DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst0_m); \
\
cnst0_m = __msa_splati_h(coeff_m, 4); \
cnst1_m = __msa_ilvev_h(cnst1_m, cnst0_m); \
stp21_m = VP9_DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst1_m); \
stp21_m = DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst1_m); \
\
SPLATI_H2_SH(coeff_m, 5, 2, cnst0_m, cnst1_m); \
cnst1_m = __msa_ilvev_h(cnst0_m, cnst1_m); \
stp25_m = VP9_DOT_SHIFT_RIGHT_PCK_H(vec4_m, vec5_m, cnst1_m); \
stp25_m = DOT_SHIFT_RIGHT_PCK_H(vec4_m, vec5_m, cnst1_m); \
\
cnst0_m = __msa_splati_h(coeff_m, 3); \
cnst1_m = __msa_ilvev_h(cnst1_m, cnst0_m); \
stp22_m = VP9_DOT_SHIFT_RIGHT_PCK_H(vec4_m, vec5_m, cnst1_m); \
stp22_m = DOT_SHIFT_RIGHT_PCK_H(vec4_m, vec5_m, cnst1_m); \
\
/* stp4 */ \
BUTTERFLY_4(stp30_m, stp37_m, stp26_m, stp21_m, \
@ -447,43 +447,43 @@
SPLATI_H2_SH(coeff1_m, 0, 1, cnst0_m, cnst1_m); \
cnst0_m = __msa_ilvev_h(cnst0_m, cnst1_m); \
\
out1 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
out1 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
\
cnst0_m = __msa_splati_h(coeff2_m, 0); \
cnst0_m = __msa_ilvev_h(cnst1_m, cnst0_m); \
out15 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
out15 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
\
ILVRL_H2_SH(vec4_m, vec5_m, vec1_m, vec0_m); \
SPLATI_H2_SH(coeff1_m, 2, 3, cnst0_m, cnst1_m); \
cnst1_m = __msa_ilvev_h(cnst1_m, cnst0_m); \
\
out9 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst1_m); \
out9 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst1_m); \
\
cnst1_m = __msa_splati_h(coeff2_m, 2); \
cnst0_m = __msa_ilvev_h(cnst0_m, cnst1_m); \
out7 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
out7 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
\
ILVRL_H2_SH(stp23_m, stp21_m, vec1_m, vec0_m); \
SPLATI_H2_SH(coeff1_m, 4, 5, cnst0_m, cnst1_m); \
cnst0_m = __msa_ilvev_h(cnst0_m, cnst1_m); \
out5 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
out5 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
\
cnst0_m = __msa_splati_h(coeff2_m, 1); \
cnst0_m = __msa_ilvev_h(cnst1_m, cnst0_m); \
out11 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
out11 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
\
ILVRL_H2_SH(stp24_m, stp31_m, vec1_m, vec0_m); \
SPLATI_H2_SH(coeff1_m, 6, 7, cnst0_m, cnst1_m); \
cnst1_m = __msa_ilvev_h(cnst1_m, cnst0_m); \
\
out13 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst1_m); \
out13 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst1_m); \
\
cnst1_m = __msa_splati_h(coeff2_m, 3); \
cnst0_m = __msa_ilvev_h(cnst0_m, cnst1_m); \
out3 = VP9_DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
out3 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
}
#define VP9_FDCT32_POSTPROC_NEG_W(vec) { \
#define FDCT32_POSTPROC_NEG_W(vec) { \
v4i32 temp_m; \
v4i32 one_m = __msa_ldi_w(1); \
\
@ -494,7 +494,7 @@
vec >>= 2; \
}
#define VP9_FDCT32_POSTPROC_2V_POS_H(vec0, vec1) { \
#define FDCT32_POSTPROC_2V_POS_H(vec0, vec1) { \
v8i16 tp0_m, tp1_m; \
v8i16 one = __msa_ldi_h(1); \
\
@ -512,9 +512,9 @@
vec1 >>= 2; \
}
#define VP9_DOTP_CONST_PAIR_W(reg0_left, reg1_left, reg0_right, \
reg1_right, const0, const1, \
out0, out1, out2, out3) { \
#define DOTP_CONST_PAIR_W(reg0_left, reg1_left, reg0_right, \
reg1_right, const0, const1, \
out0, out1, out2, out3) { \
v4i32 s0_m, s1_m, s2_m, s3_m, s4_m, s5_m, s6_m, s7_m; \
v2i64 tp0_m, tp1_m, tp2_m, tp3_m; \
v4i32 k0_m = __msa_fill_w((int32_t) const0); \