| /external/libvpx/libvpx/vpx_dsp/ppc/ |
| D | txfm_common_vsx.h | 22 static const int16x8_t cospi1_v = { 16364, 16364, 16364, 16364,
24 static const int16x8_t cospi2_v = { 16305, 16305, 16305, 16305,
26 static const int16x8_t cospi3_v = { 16207, 16207, 16207, 16207,
28 static const int16x8_t cospi4_v = { 16069, 16069, 16069, 16069,
30 static const int16x8_t cospi4m_v = { -16069, -16069, -16069, -16069,
32 static const int16x8_t cospi5_v = { 15893, 15893, 15893, 15893,
34 static const int16x8_t cospi6_v = { 15679, 15679, 15679, 15679,
36 static const int16x8_t cospi7_v = { 15426, 15426, 15426, 15426,
38 static const int16x8_t cospi8_v = { 15137, 15137, 15137, 15137,
40 static const int16x8_t cospi8m_v = { -15137, -15137, -15137, -15137,
[all …]
|
| D | hadamard_vsx.c | 16 static void vpx_hadamard_s16_8x8_one_pass(int16x8_t v[8]) { in vpx_hadamard_s16_8x8_one_pass()
17 const int16x8_t b0 = vec_add(v[0], v[1]); in vpx_hadamard_s16_8x8_one_pass()
18 const int16x8_t b1 = vec_sub(v[0], v[1]); in vpx_hadamard_s16_8x8_one_pass()
19 const int16x8_t b2 = vec_add(v[2], v[3]); in vpx_hadamard_s16_8x8_one_pass()
20 const int16x8_t b3 = vec_sub(v[2], v[3]); in vpx_hadamard_s16_8x8_one_pass()
21 const int16x8_t b4 = vec_add(v[4], v[5]); in vpx_hadamard_s16_8x8_one_pass()
22 const int16x8_t b5 = vec_sub(v[4], v[5]); in vpx_hadamard_s16_8x8_one_pass()
23 const int16x8_t b6 = vec_add(v[6], v[7]); in vpx_hadamard_s16_8x8_one_pass()
24 const int16x8_t b7 = vec_sub(v[6], v[7]); in vpx_hadamard_s16_8x8_one_pass()
26 const int16x8_t c0 = vec_add(b0, b2); in vpx_hadamard_s16_8x8_one_pass()
[all …]
|
| D | quantize_vsx.c | 18 static INLINE int16x8_t vec_sign(int16x8_t a, int16x8_t b) { in vec_sign()
19 const int16x8_t mask = vec_sra(b, vec_shift_sign_s16); in vec_sign()
32 static INLINE int16x8_t vec_mulhi(int16x8_t a, int16x8_t b) { in vec_mulhi()
39 static INLINE int16x8_t quantize_coeff(int16x8_t coeff, int16x8_t coeff_abs, in quantize_coeff()
40 int16x8_t round, int16x8_t quant, in quantize_coeff()
41 int16x8_t quant_shift, bool16x8_t mask) { in quantize_coeff()
42 const int16x8_t rounded = vec_vaddshs(coeff_abs, round); in quantize_coeff()
43 int16x8_t qcoeff = vec_mulhi(rounded, quant); in quantize_coeff()
51 static INLINE int16x8_t quantize_coeff_32(int16x8_t coeff, int16x8_t coeff_abs, in quantize_coeff_32()
52 int16x8_t round, int16x8_t quant, in quantize_coeff_32()
[all …]
|
| D | transpose_vsx.h | 17 static INLINE void vpx_transpose_s16_8x8(int16x8_t v[8]) { in vpx_transpose_s16_8x8()
40 int16x8_t b0, b1, b2, b3, b4, b5, b6, b7; in vpx_transpose_s16_8x8()
41 int16x8_t c0, c1, c2, c3, c4, c5, c6, c7; in vpx_transpose_s16_8x8()
101 static INLINE void transpose_8x8(const int16x8_t *a, int16x8_t *b) { in transpose_8x8()
103 const int16x8_t s1_0 = vec_mergeh(a[0], a[4]); in transpose_8x8()
104 const int16x8_t s1_1 = vec_mergel(a[0], a[4]); in transpose_8x8()
105 const int16x8_t s1_2 = vec_mergeh(a[1], a[5]); in transpose_8x8()
106 const int16x8_t s1_3 = vec_mergel(a[1], a[5]); in transpose_8x8()
107 const int16x8_t s1_4 = vec_mergeh(a[2], a[6]); in transpose_8x8()
108 const int16x8_t s1_5 = vec_mergel(a[2], a[6]); in transpose_8x8()
[all …]
|
| D | inv_txfm_vsx.h | 16 void vpx_round_store4x4_vsx(int16x8_t *in, int16x8_t *out, uint8_t *dest,
18 void vpx_idct4_vsx(int16x8_t *in, int16x8_t *out);
19 void vp9_iadst4_vsx(int16x8_t *in, int16x8_t *out);
21 void vpx_round_store8x8_vsx(int16x8_t *in, uint8_t *dest, int stride);
22 void vpx_idct8_vsx(int16x8_t *in, int16x8_t *out);
23 void vp9_iadst8_vsx(int16x8_t *in, int16x8_t *out);
43 void vpx_round_store16x16_vsx(int16x8_t *src0, int16x8_t *src1, uint8_t *dest,
45 void vpx_idct16_vsx(int16x8_t *src0, int16x8_t *src1);
46 void vpx_iadst16_vsx(int16x8_t *src0, int16x8_t *src1);
|
| D | fdct32x32_vsx.c | 19 static INLINE void single_butterfly(int16x8_t a, int16x8_t b, int16x8_t *add, in single_butterfly()
20 int16x8_t *sub) { in single_butterfly()
46 *add = (int16x8_t)vec_perm(ssum_e, ssum_o, vec_perm_odd_even_pack); in single_butterfly()
47 *sub = (int16x8_t)vec_perm(sdiff_e, sdiff_o, vec_perm_odd_even_pack); in single_butterfly()
51 static INLINE void double_butterfly(int16x8_t a, int16x8_t c1, int16x8_t b, in double_butterfly()
52 int16x8_t c2, int16x8_t *add, in double_butterfly()
53 int16x8_t *sub) { in double_butterfly()
81 *add = (int16x8_t)vec_perm(ssum_e, ssum_o, vec_perm_odd_even_pack); in double_butterfly()
82 *sub = (int16x8_t)vec_perm(sdiff_e, sdiff_o, vec_perm_odd_even_pack); in double_butterfly()
87 static INLINE void load(const int16_t *a, int stride, int16x8_t *b) { in load()
[all …]
|
| D | variance_vsx.c | 21 const int16x8_t a0 = unpack_to_s16_h(read4x2(src_ptr, src_stride)); in vpx_get4x4sse_cs_vsx()
22 const int16x8_t a1 = in vpx_get4x4sse_cs_vsx()
24 const int16x8_t b0 = unpack_to_s16_h(read4x2(ref_ptr, ref_stride)); in vpx_get4x4sse_cs_vsx()
25 const int16x8_t b1 = in vpx_get4x4sse_cs_vsx()
27 const int16x8_t d0 = vec_sub(a0, b0); in vpx_get4x4sse_cs_vsx()
28 const int16x8_t d1 = vec_sub(a1, b1); in vpx_get4x4sse_cs_vsx()
43 const int16x8_t v = vec_vsx_ld(0, src_ptr + i); in vpx_get_mb_ss_vsx()
112 const int16x8_t a0 = unpack_to_s16_h(va0); in variance_inner_32()
113 const int16x8_t b0 = unpack_to_s16_h(vb0); in variance_inner_32()
114 const int16x8_t a1 = unpack_to_s16_l(va0); in variance_inner_32()
[all …]
|
| D | inv_txfm_vsx.c | 22 static const int16x8_t cospi1_v = { 16364, 16364, 16364, 16364,
24 static const int16x8_t cospi1m_v = { -16364, -16364, -16364, -16364,
26 static const int16x8_t cospi2_v = { 16305, 16305, 16305, 16305,
28 static const int16x8_t cospi2m_v = { -16305, -16305, -16305, -16305,
30 static const int16x8_t cospi3_v = { 16207, 16207, 16207, 16207,
32 static const int16x8_t cospi4_v = { 16069, 16069, 16069, 16069,
34 static const int16x8_t cospi4m_v = { -16069, -16069, -16069, -16069,
36 static const int16x8_t cospi5_v = { 15893, 15893, 15893, 15893,
38 static const int16x8_t cospi5m_v = { -15893, -15893, -15893, -15893,
40 static const int16x8_t cospi6_v = { 15679, 15679, 15679, 15679,
[all …]
|
| /external/gemmlowp/fixedpoint/ |
| D | fixedpoint_neon.h | 32 struct FixedPointRawTypeTraits<int16x8_t> {
43 inline int16x8_t BitAnd(int16x8_t a, int16x8_t b) {
53 inline int16x8_t BitOr(int16x8_t a, int16x8_t b) {
63 inline int16x8_t BitXor(int16x8_t a, int16x8_t b) {
73 inline int16x8_t BitNot(int16x8_t a) {
83 inline int16x8_t Add(int16x8_t a, int16x8_t b) {
93 inline int16x8_t Sub(int16x8_t a, int16x8_t b) {
103 inline int16x8_t Neg(int16x8_t a) {
113 inline int16x8_t ShiftLeft(int16x8_t a, int offset) {
123 inline int16x8_t ShiftRight(int16x8_t a, int offset) {
[all …]
|
| /external/libvpx/libvpx/vpx_dsp/arm/ |
| D | hadamard_neon.c | 19 static void hadamard8x8_one_pass(int16x8_t *a0, int16x8_t *a1, int16x8_t *a2, in hadamard8x8_one_pass()
20 int16x8_t *a3, int16x8_t *a4, int16x8_t *a5, in hadamard8x8_one_pass()
21 int16x8_t *a6, int16x8_t *a7) { in hadamard8x8_one_pass()
22 const int16x8_t b0 = vaddq_s16(*a0, *a1); in hadamard8x8_one_pass()
23 const int16x8_t b1 = vsubq_s16(*a0, *a1); in hadamard8x8_one_pass()
24 const int16x8_t b2 = vaddq_s16(*a2, *a3); in hadamard8x8_one_pass()
25 const int16x8_t b3 = vsubq_s16(*a2, *a3); in hadamard8x8_one_pass()
26 const int16x8_t b4 = vaddq_s16(*a4, *a5); in hadamard8x8_one_pass()
27 const int16x8_t b5 = vsubq_s16(*a4, *a5); in hadamard8x8_one_pass()
28 const int16x8_t b6 = vaddq_s16(*a6, *a7); in hadamard8x8_one_pass()
[all …]
|
| D | quantize_neon.c | 18 static INLINE void calculate_dqcoeff_and_store(const int16x8_t qcoeff, in calculate_dqcoeff_and_store()
19 const int16x8_t dequant, in calculate_dqcoeff_and_store()
41 const int16x8_t one = vdupq_n_s16(1); in vpx_quantize_b_neon()
42 const int16x8_t neg_one = vdupq_n_s16(-1); in vpx_quantize_b_neon()
51 const int16x8_t zbin = vld1q_s16(zbin_ptr); in vpx_quantize_b_neon()
52 const int16x8_t round = vld1q_s16(round_ptr); in vpx_quantize_b_neon()
53 const int16x8_t quant = vld1q_s16(quant_ptr); in vpx_quantize_b_neon()
54 const int16x8_t quant_shift = vld1q_s16(quant_shift_ptr); in vpx_quantize_b_neon()
55 const int16x8_t dequant = vld1q_s16(dequant_ptr); in vpx_quantize_b_neon()
60 const int16x8_t coeff = load_tran_low_to_s16q(coeff_ptr); in vpx_quantize_b_neon()
[all …]
|
| D | highbd_idct32x32_add_neon.c | 19 const int16x8_t res, in highbd_idct32x32_1_add_pos_kernel()
20 const int16x8_t max) { in highbd_idct32x32_1_add_pos_kernel()
25 const int16x8_t b0 = vaddq_s16(res, vreinterpretq_s16_u16(a0)); in highbd_idct32x32_1_add_pos_kernel()
26 const int16x8_t b1 = vaddq_s16(res, vreinterpretq_s16_u16(a1)); in highbd_idct32x32_1_add_pos_kernel()
27 const int16x8_t b2 = vaddq_s16(res, vreinterpretq_s16_u16(a2)); in highbd_idct32x32_1_add_pos_kernel()
28 const int16x8_t b3 = vaddq_s16(res, vreinterpretq_s16_u16(a3)); in highbd_idct32x32_1_add_pos_kernel()
29 const int16x8_t c0 = vminq_s16(b0, max); in highbd_idct32x32_1_add_pos_kernel()
30 const int16x8_t c1 = vminq_s16(b1, max); in highbd_idct32x32_1_add_pos_kernel()
31 const int16x8_t c2 = vminq_s16(b2, max); in highbd_idct32x32_1_add_pos_kernel()
32 const int16x8_t c3 = vminq_s16(b3, max); in highbd_idct32x32_1_add_pos_kernel()
[all …]
|
| D | fwd_txfm_neon.c | 24 int16x8_t input_0 = vshlq_n_s16(vld1q_s16(&input[0 * stride]), 2); in vpx_fdct8x8_neon()
25 int16x8_t input_1 = vshlq_n_s16(vld1q_s16(&input[1 * stride]), 2); in vpx_fdct8x8_neon()
26 int16x8_t input_2 = vshlq_n_s16(vld1q_s16(&input[2 * stride]), 2); in vpx_fdct8x8_neon()
27 int16x8_t input_3 = vshlq_n_s16(vld1q_s16(&input[3 * stride]), 2); in vpx_fdct8x8_neon()
28 int16x8_t input_4 = vshlq_n_s16(vld1q_s16(&input[4 * stride]), 2); in vpx_fdct8x8_neon()
29 int16x8_t input_5 = vshlq_n_s16(vld1q_s16(&input[5 * stride]), 2); in vpx_fdct8x8_neon()
30 int16x8_t input_6 = vshlq_n_s16(vld1q_s16(&input[6 * stride]), 2); in vpx_fdct8x8_neon()
31 int16x8_t input_7 = vshlq_n_s16(vld1q_s16(&input[7 * stride]), 2); in vpx_fdct8x8_neon()
33 int16x8_t out_0, out_1, out_2, out_3, out_4, out_5, out_6, out_7; in vpx_fdct8x8_neon()
34 const int16x8_t v_s0 = vaddq_s16(input_0, input_7); in vpx_fdct8x8_neon()
[all …]
|
| D | fdct_partial_neon.c | 28 int16x8_t b0, b1; in vpx_fdct4x4_1_neon()
29 int16x8_t c; in vpx_fdct4x4_1_neon()
53 int16x8_t sum = vld1q_s16(&input[0]); in vpx_fdct8x8_1_neon()
56 const int16x8_t input_00 = vld1q_s16(&input[r * stride]); in vpx_fdct8x8_1_neon()
67 int16x8_t left = vld1q_s16(input); in vpx_fdct16x16_1_neon()
68 int16x8_t right = vld1q_s16(input + 8); in vpx_fdct16x16_1_neon()
73 const int16x8_t a = vld1q_s16(input); in vpx_fdct16x16_1_neon()
74 const int16x8_t b = vld1q_s16(input + 8); in vpx_fdct16x16_1_neon()
89 int16x8_t a0 = vld1q_s16(input); in vpx_fdct32x32_1_neon()
90 int16x8_t a1 = vld1q_s16(input + 8); in vpx_fdct32x32_1_neon()
[all …]
|
| D | idct_neon.h | 45 static INLINE int16x8_t final_add(const int16x8_t a, const int16x8_t b) { in final_add()
53 static INLINE int16x8_t final_sub(const int16x8_t a, const int16x8_t b) { in final_sub()
81 static INLINE int16x8_t dct_const_round_shift_low_8(const int32x4_t *const in) { in dct_const_round_shift_low_8()
87 int16x8_t *const d0, in dct_const_round_shift_low_8_dual()
88 int16x8_t *const d1) { in dct_const_round_shift_low_8_dual()
104 static INLINE int16x8_t multiply_shift_and_narrow_s16(const int16x8_t a, in multiply_shift_and_narrow_s16()
118 static INLINE int16x8_t add_multiply_shift_and_narrow_s16( in add_multiply_shift_and_narrow_s16()
119 const int16x8_t a, const int16x8_t b, const int16_t ab_const) { in add_multiply_shift_and_narrow_s16()
137 static INLINE int16x8_t sub_multiply_shift_and_narrow_s16( in sub_multiply_shift_and_narrow_s16()
138 const int16x8_t a, const int16x8_t b, const int16_t ab_const) { in sub_multiply_shift_and_narrow_s16()
[all …]
|
| /external/libvpx/libvpx/vp9/encoder/arm/neon/ |
| D | vp9_quantize_neon.c | 38 const int16x8_t v_zero = vdupq_n_s16(0); in vp9_quantize_fp_neon()
39 const int16x8_t v_one = vdupq_n_s16(1); in vp9_quantize_fp_neon()
40 int16x8_t v_eobmax_76543210 = vdupq_n_s16(-1); in vp9_quantize_fp_neon()
41 int16x8_t v_round = vmovq_n_s16(round_ptr[1]); in vp9_quantize_fp_neon()
42 int16x8_t v_quant = vmovq_n_s16(quant_ptr[1]); in vp9_quantize_fp_neon()
43 int16x8_t v_dequant = vmovq_n_s16(dequant_ptr[1]); in vp9_quantize_fp_neon()
55 const int16x8_t v_iscan = vld1q_s16(&iscan[0]); in vp9_quantize_fp_neon()
56 const int16x8_t v_coeff = load_tran_low_to_s16q(coeff_ptr); in vp9_quantize_fp_neon()
57 const int16x8_t v_coeff_sign = vshrq_n_s16(v_coeff, 15); in vp9_quantize_fp_neon()
58 const int16x8_t v_tmp = vabaq_s16(v_round, v_coeff, v_zero); in vp9_quantize_fp_neon()
[all …]
|
| /external/libhevc/encoder/arm/ |
| D | ihevce_common_utils_neon.c | 113 int16x8_t a2, a3, a4, a5, a8; in ihevce_wt_avg_2d_16x1_neon()
179 int16x8_t a0, a1, a6; in ihevce_wt_avg_2d_8x1_neon()
234 int16x8_t a8, a9, a6; in ihevce_wt_avg_2d_4xn_neon()
448 static INLINE WORD32 sad_cal(int16x8_t temp_reg) in sad_cal()
473 int16x8_t temp_reg0, temp_reg1, temp_reg2, temp_reg3, temp_reg4; in ihevce_get_luma_eo_sao_params_neon()
474 int16x8_t edgeidx_reg0, edgeidx_reg1, edgeidx_reg2, edgeidx_reg3, edgeidx_reg4; in ihevce_get_luma_eo_sao_params_neon()
475 int16x8_t edgeidx_reg5, edgeidx_reg6, edgeidx_reg7; in ihevce_get_luma_eo_sao_params_neon()
476 int16x8_t pel_error, pel_error1; in ihevce_get_luma_eo_sao_params_neon()
477 int16x8_t sign_reg0, sign_reg1, sign_reg, sign_reg2, sign_reg3; in ihevce_get_luma_eo_sao_params_neon()
478 int16x8_t edgeidx, edgeidx1; in ihevce_get_luma_eo_sao_params_neon()
[all …]
|
| D | ihevce_had_compute_neon.c | 85 hadamard4x4_2_one_pass(int16x8_t *r0, int16x8_t *r1, int16x8_t *r2, int16x8_t *r3) in hadamard4x4_2_one_pass()
87 const int16x8_t a0 = vaddq_s16(*r0, *r2); in hadamard4x4_2_one_pass()
88 const int16x8_t a1 = vaddq_s16(*r1, *r3); in hadamard4x4_2_one_pass()
89 const int16x8_t a2 = vsubq_s16(*r0, *r2); in hadamard4x4_2_one_pass()
90 const int16x8_t a3 = vsubq_s16(*r1, *r3); in hadamard4x4_2_one_pass()
103 int16x8_t *r0, in hadamard4x4_2()
104 int16x8_t *r1, in hadamard4x4_2()
105 int16x8_t *r2, in hadamard4x4_2()
106 int16x8_t *r3) in hadamard4x4_2()
129 int16x8_t *r0, in hadamard4x4_4()
[all …]
|
| /external/libaom/libaom/av1/encoder/arm/neon/ |
| D | quantize_neon.c | 42 const int16x8_t v_zero = vdupq_n_s16(0); in av1_quantize_fp_neon()
43 const int16x8_t v_one = vdupq_n_s16(1); in av1_quantize_fp_neon()
44 int16x8_t v_eobmax_76543210 = vdupq_n_s16(-1); in av1_quantize_fp_neon()
45 int16x8_t v_round = vmovq_n_s16(round_ptr[1]); in av1_quantize_fp_neon()
46 int16x8_t v_quant = vmovq_n_s16(quant_ptr[1]); in av1_quantize_fp_neon()
47 int16x8_t v_dequant = vmovq_n_s16(dequant_ptr[1]); in av1_quantize_fp_neon()
54 const int16x8_t v_iscan = vld1q_s16(&iscan[0]); in av1_quantize_fp_neon()
55 const int16x8_t v_coeff = vld1q_s16(&coeff_ptr[0]); in av1_quantize_fp_neon()
56 const int16x8_t v_coeff_sign = vshrq_n_s16(v_coeff, 15); in av1_quantize_fp_neon()
57 const int16x8_t v_tmp = vabaq_s16(v_round, v_coeff, v_zero); in av1_quantize_fp_neon()
[all …]
|
| /external/webp/src/dsp/ |
| D | yuv_neon.c | 105 const int16x8_t tmp3 = vcombine_s16(vshrn_n_s32(tmp2_lo, 16), \
122 int16x8_t U, V; in ConvertRGBA32ToUV_NEON()
142 int16x8_t U_tmp, V_tmp; in ConvertARGBToUV_NEON()
186 const int16x8_t zero = vdupq_n_s16(0); in SharpYUVUpdateY_NEON()
187 const int16x8_t max = vdupq_n_s16(MAX_Y); in SharpYUVUpdateY_NEON()
192 const int16x8_t A = vreinterpretq_s16_u16(vld1q_u16(ref + i)); in SharpYUVUpdateY_NEON()
193 const int16x8_t B = vreinterpretq_s16_u16(vld1q_u16(src + i)); in SharpYUVUpdateY_NEON()
194 const int16x8_t C = vreinterpretq_s16_u16(vld1q_u16(dst + i)); in SharpYUVUpdateY_NEON()
195 const int16x8_t D = vsubq_s16(A, B); // diff_y in SharpYUVUpdateY_NEON()
196 const int16x8_t F = vaddq_s16(C, D); // new_y in SharpYUVUpdateY_NEON()
[all …]
|
| /external/clang/test/CodeGen/ |
| D | aarch64-neon-2velem.c | 21 int16x8_t test_vmlaq_lane_s16(int16x8_t a, int16x8_t b, int16x4_t v) { in test_vmlaq_lane_s16()
48 int16x4_t test_vmla_laneq_s16(int16x4_t a, int16x4_t b, int16x8_t v) { in test_vmla_laneq_s16()
57 int16x8_t test_vmlaq_laneq_s16(int16x8_t a, int16x8_t b, int16x8_t v) { in test_vmlaq_laneq_s16()
93 int16x8_t test_vmlsq_lane_s16(int16x8_t a, int16x8_t b, int16x4_t v) { in test_vmlsq_lane_s16()
120 int16x4_t test_vmls_laneq_s16(int16x4_t a, int16x4_t b, int16x8_t v) { in test_vmls_laneq_s16()
129 int16x8_t test_vmlsq_laneq_s16(int16x8_t a, int16x8_t b, int16x8_t v) { in test_vmlsq_laneq_s16()
163 int16x8_t test_vmulq_lane_s16(int16x8_t a, int16x4_t v) { in test_vmulq_lane_s16()
219 int16x4_t test_vmul_laneq_s16(int16x4_t a, int16x8_t v) { in test_vmul_laneq_s16()
227 int16x8_t test_vmulq_laneq_s16(int16x8_t a, int16x8_t v) { in test_vmulq_laneq_s16()
531 int32x4_t test_vmlal_laneq_s16(int32x4_t a, int16x4_t b, int16x8_t v) { in test_vmlal_laneq_s16()
[all …]
|
| D | arm-v8.1a-neon-intrinsics.c | 34 int16x8_t test_vqrdmlahq_s16(int16x8_t a, int16x8_t b, int16x8_t c) { in test_vqrdmlahq_s16()
78 int16x8_t test_vqrdmlahq_lane_s16(int16x8_t a, int16x8_t b, int16x4_t c) { in test_vqrdmlahq_lane_s16()
122 int16x8_t test_vqrdmlshq_s16(int16x8_t a, int16x8_t b, int16x8_t c) { in test_vqrdmlshq_s16()
166 int16x8_t test_vqrdmlshq_lane_s16(int16x8_t a, int16x8_t b, int16x4_t c) { in test_vqrdmlshq_lane_s16()
|
| /external/libvpx/libvpx/vp9/encoder/ppc/ |
| D | vp9_quantize_vsx.c | 20 static INLINE int16x8_t vec_mulhi(int16x8_t a, int16x8_t b) { in vec_mulhi()
28 static INLINE int16x8_t vec_sign(int16x8_t a, int16x8_t b) { in vec_sign()
29 const int16x8_t mask = vec_sra(b, vec_shift_sign_s16); in vec_sign()
35 static INLINE int16x8_t vec_max_across(int16x8_t a) { in vec_max_across()
47 int16x8_t qcoeff0, qcoeff1, dqcoeff0, dqcoeff1, eob; in vp9_quantize_fp_vsx()
50 int16x8_t round = vec_vsx_ld(0, round_ptr); in vp9_quantize_fp_vsx()
51 int16x8_t quant = vec_vsx_ld(0, quant_ptr); in vp9_quantize_fp_vsx()
52 int16x8_t dequant = vec_vsx_ld(0, dequant_ptr); in vp9_quantize_fp_vsx()
53 int16x8_t coeff0 = vec_vsx_ld(0, coeff_ptr); in vp9_quantize_fp_vsx()
54 int16x8_t coeff1 = vec_vsx_ld(16, coeff_ptr); in vp9_quantize_fp_vsx()
[all …]
|
| /external/libaom/libaom/aom_dsp/arm/ |
| D | fwd_txfm_neon.c | 21 int16x8_t input_0 = vshlq_n_s16(vld1q_s16(&input[0 * stride]), 2); in aom_fdct8x8_neon()
22 int16x8_t input_1 = vshlq_n_s16(vld1q_s16(&input[1 * stride]), 2); in aom_fdct8x8_neon()
23 int16x8_t input_2 = vshlq_n_s16(vld1q_s16(&input[2 * stride]), 2); in aom_fdct8x8_neon()
24 int16x8_t input_3 = vshlq_n_s16(vld1q_s16(&input[3 * stride]), 2); in aom_fdct8x8_neon()
25 int16x8_t input_4 = vshlq_n_s16(vld1q_s16(&input[4 * stride]), 2); in aom_fdct8x8_neon()
26 int16x8_t input_5 = vshlq_n_s16(vld1q_s16(&input[5 * stride]), 2); in aom_fdct8x8_neon()
27 int16x8_t input_6 = vshlq_n_s16(vld1q_s16(&input[6 * stride]), 2); in aom_fdct8x8_neon()
28 int16x8_t input_7 = vshlq_n_s16(vld1q_s16(&input[7 * stride]), 2); in aom_fdct8x8_neon()
30 int16x8_t out_0, out_1, out_2, out_3, out_4, out_5, out_6, out_7; in aom_fdct8x8_neon()
31 const int16x8_t v_s0 = vaddq_s16(input_0, input_7); in aom_fdct8x8_neon()
[all …]
|
| /external/libaom/libaom/av1/common/arm/ |
| D | convolve_neon.h | 19 const int16x8_t s0, const int16x8_t s1, const int16x8_t s2, in wiener_convolve8_vert_4x8()
20 const int16x8_t s3, const int16x8_t s4, const int16x8_t s5, in wiener_convolve8_vert_4x8()
21 const int16x8_t s6, int16_t *filter_y, const int bd, in wiener_convolve8_vert_4x8()
23 int16x8_t ss0, ss1, ss2; in wiener_convolve8_vert_4x8()
68 const int16x8_t s0, const int16x8_t s1, const int16x8_t s2, in wiener_convolve8_horiz_8x8()
69 const int16x8_t s3, int16_t *filter_x, const int bd, in wiener_convolve8_horiz_8x8()
71 int16x8_t sum; in wiener_convolve8_horiz_8x8()
158 static INLINE int16x8_t
159 convolve8_8x8_s16(const int16x8_t s0, const int16x8_t s1, const int16x8_t s2, in convolve8_8x8_s16()
160 const int16x8_t s3, const int16x8_t s4, const int16x8_t s5, in convolve8_8x8_s16()
[all …]
|