| /external/libvpx/libvpx/vpx_dsp/arm/ |
| D | vpx_convolve8_neon.c | 87 tt0 = vreinterpretq_s16_u16(vmovl_u8(t0)); in vpx_convolve8_horiz_neon()
88 tt1 = vreinterpretq_s16_u16(vmovl_u8(t1)); in vpx_convolve8_horiz_neon()
89 tt2 = vreinterpretq_s16_u16(vmovl_u8(t2)); in vpx_convolve8_horiz_neon()
90 tt3 = vreinterpretq_s16_u16(vmovl_u8(t3)); in vpx_convolve8_horiz_neon()
107 tt0 = vreinterpretq_s16_u16(vmovl_u8(t0)); in vpx_convolve8_horiz_neon()
108 tt1 = vreinterpretq_s16_u16(vmovl_u8(t1)); in vpx_convolve8_horiz_neon()
109 tt2 = vreinterpretq_s16_u16(vmovl_u8(t2)); in vpx_convolve8_horiz_neon()
110 tt3 = vreinterpretq_s16_u16(vmovl_u8(t3)); in vpx_convolve8_horiz_neon()
161 s0 = vreinterpretq_s16_u16(vmovl_u8(t0)); in vpx_convolve8_horiz_neon()
162 s1 = vreinterpretq_s16_u16(vmovl_u8(t1)); in vpx_convolve8_horiz_neon()
[all …]
|
| D | vpx_convolve8_neon.h | 125 ss[0] = vreinterpretq_s16_u16(vmovl_u8(s[0])); in scale_filter_8()
126 ss[1] = vreinterpretq_s16_u16(vmovl_u8(s[1])); in scale_filter_8()
127 ss[2] = vreinterpretq_s16_u16(vmovl_u8(s[2])); in scale_filter_8()
128 ss[3] = vreinterpretq_s16_u16(vmovl_u8(s[3])); in scale_filter_8()
129 ss[4] = vreinterpretq_s16_u16(vmovl_u8(s[4])); in scale_filter_8()
130 ss[5] = vreinterpretq_s16_u16(vmovl_u8(s[5])); in scale_filter_8()
131 ss[6] = vreinterpretq_s16_u16(vmovl_u8(s[6])); in scale_filter_8()
132 ss[7] = vreinterpretq_s16_u16(vmovl_u8(s[7])); in scale_filter_8()
|
| D | vpx_scaled_convolve8_neon.c | 50 ss[0] = vreinterpretq_s16_u16(vmovl_u8(s[0])); in scaledconvolve_horiz_w4()
51 ss[1] = vreinterpretq_s16_u16(vmovl_u8(s[1])); in scaledconvolve_horiz_w4()
52 ss[2] = vreinterpretq_s16_u16(vmovl_u8(s[2])); in scaledconvolve_horiz_w4()
53 ss[3] = vreinterpretq_s16_u16(vmovl_u8(s[3])); in scaledconvolve_horiz_w4()
177 t[0] = vget_low_s16(vreinterpretq_s16_u16(vmovl_u8(s[0]))); in scaledconvolve_vert_w4()
178 t[1] = vget_low_s16(vreinterpretq_s16_u16(vmovl_u8(s[1]))); in scaledconvolve_vert_w4()
179 t[2] = vget_low_s16(vreinterpretq_s16_u16(vmovl_u8(s[2]))); in scaledconvolve_vert_w4()
180 t[3] = vget_low_s16(vreinterpretq_s16_u16(vmovl_u8(s[3]))); in scaledconvolve_vert_w4()
181 t[4] = vget_low_s16(vreinterpretq_s16_u16(vmovl_u8(s[4]))); in scaledconvolve_vert_w4()
182 t[5] = vget_low_s16(vreinterpretq_s16_u16(vmovl_u8(s[5]))); in scaledconvolve_vert_w4()
[all …]
|
| D | idct_neon.h | 250 c[0] = vrsraq_n_s16(vreinterpretq_s16_u16(vmovl_u8(b[0])), a[0], 6); in add_and_store_u8_s16()
251 c[1] = vrsraq_n_s16(vreinterpretq_s16_u16(vmovl_u8(b[1])), a[1], 6); in add_and_store_u8_s16()
252 c[2] = vrsraq_n_s16(vreinterpretq_s16_u16(vmovl_u8(b[2])), a[2], 6); in add_and_store_u8_s16()
253 c[3] = vrsraq_n_s16(vreinterpretq_s16_u16(vmovl_u8(b[3])), a[3], 6); in add_and_store_u8_s16()
254 c[4] = vrsraq_n_s16(vreinterpretq_s16_u16(vmovl_u8(b[4])), a[4], 6); in add_and_store_u8_s16()
255 c[5] = vrsraq_n_s16(vreinterpretq_s16_u16(vmovl_u8(b[5])), a[5], 6); in add_and_store_u8_s16()
256 c[6] = vrsraq_n_s16(vreinterpretq_s16_u16(vmovl_u8(b[6])), a[6], 6); in add_and_store_u8_s16()
257 c[7] = vrsraq_n_s16(vreinterpretq_s16_u16(vmovl_u8(b[7])), a[7], 6); in add_and_store_u8_s16()
852 const uint16x8_t d = vmovl_u8(vqmovun_s16(res)); in highbd_idct16x16_add8x1_bd8()
|
| /external/libaom/libaom/av1/common/arm/ |
| D | convolve_neon.c | 237 s0 = vget_low_s16(vreinterpretq_s16_u16(vmovl_u8(t0))); in av1_convolve_x_sr_neon()
238 s1 = vget_low_s16(vreinterpretq_s16_u16(vmovl_u8(t1))); in av1_convolve_x_sr_neon()
239 s2 = vget_low_s16(vreinterpretq_s16_u16(vmovl_u8(t2))); in av1_convolve_x_sr_neon()
240 s3 = vget_low_s16(vreinterpretq_s16_u16(vmovl_u8(t3))); in av1_convolve_x_sr_neon()
241 s4 = vget_high_s16(vreinterpretq_s16_u16(vmovl_u8(t0))); in av1_convolve_x_sr_neon()
242 s5 = vget_high_s16(vreinterpretq_s16_u16(vmovl_u8(t1))); in av1_convolve_x_sr_neon()
243 s6 = vget_high_s16(vreinterpretq_s16_u16(vmovl_u8(t2))); in av1_convolve_x_sr_neon()
254 s7 = vget_low_s16(vreinterpretq_s16_u16(vmovl_u8(t0))); in av1_convolve_x_sr_neon()
255 s8 = vget_low_s16(vreinterpretq_s16_u16(vmovl_u8(t1))); in av1_convolve_x_sr_neon()
256 s9 = vget_low_s16(vreinterpretq_s16_u16(vmovl_u8(t2))); in av1_convolve_x_sr_neon()
[all …]
|
| D | jnt_convolve_neon.c | 355 tt0 = vreinterpretq_s16_u16(vmovl_u8(t0)); in dist_wtd_convolve_2d_horiz_neon()
356 tt1 = vreinterpretq_s16_u16(vmovl_u8(t1)); in dist_wtd_convolve_2d_horiz_neon()
357 tt2 = vreinterpretq_s16_u16(vmovl_u8(t2)); in dist_wtd_convolve_2d_horiz_neon()
358 tt3 = vreinterpretq_s16_u16(vmovl_u8(t3)); in dist_wtd_convolve_2d_horiz_neon()
374 tt0 = vreinterpretq_s16_u16(vmovl_u8(t0)); in dist_wtd_convolve_2d_horiz_neon()
375 tt1 = vreinterpretq_s16_u16(vmovl_u8(t1)); in dist_wtd_convolve_2d_horiz_neon()
376 tt2 = vreinterpretq_s16_u16(vmovl_u8(t2)); in dist_wtd_convolve_2d_horiz_neon()
377 tt3 = vreinterpretq_s16_u16(vmovl_u8(t3)); in dist_wtd_convolve_2d_horiz_neon()
404 tt0 = vreinterpretq_s16_u16(vmovl_u8(t0)); // a0 a1 a2 a3 a4 a5 a6 a7 in dist_wtd_convolve_2d_horiz_neon()
412 s7 = vget_low_s16(vreinterpretq_s16_u16(vmovl_u8(t0))); in dist_wtd_convolve_2d_horiz_neon()
[all …]
|
| D | wiener_convolve_neon.c | 121 res3 = vreinterpretq_s16_u16(vmovl_u8(t3)); in av1_wiener_convolve_add_src_neon()
128 res3 = vreinterpretq_s16_u16(vmovl_u8(t4)); in av1_wiener_convolve_add_src_neon()
135 res3 = vreinterpretq_s16_u16(vmovl_u8(t5)); in av1_wiener_convolve_add_src_neon()
142 res3 = vreinterpretq_s16_u16(vmovl_u8(t6)); in av1_wiener_convolve_add_src_neon()
149 res3 = vreinterpretq_s16_u16(vmovl_u8(t7)); in av1_wiener_convolve_add_src_neon()
156 res3 = vreinterpretq_s16_u16(vmovl_u8(t8)); in av1_wiener_convolve_add_src_neon()
163 res3 = vreinterpretq_s16_u16(vmovl_u8(t9)); in av1_wiener_convolve_add_src_neon()
170 res3 = vreinterpretq_s16_u16(vmovl_u8(t10)); in av1_wiener_convolve_add_src_neon()
225 res3 = vreinterpretq_s16_u16(vmovl_u8(t3)); in av1_wiener_convolve_add_src_neon()
262 tt0 = vreinterpretq_s16_u16(vmovl_u8(t0)); in av1_wiener_convolve_add_src_neon()
[all …]
|
| /external/libhevc/common/arm/ |
| D | ihevc_inter_pred_chroma_copy_w16out.s | 141 vmovl.u8 q0,d0 @vmovl_u8(vld1_u8(pu1_src_tmp)
149 vmovl.u8 q11,d22 @vmovl_u8(vld1_u8(pu1_src_tmp)
152 vmovl.u8 q12,d24 @vmovl_u8(vld1_u8(pu1_src_tmp)
157 vmovl.u8 q13,d26 @vmovl_u8(vld1_u8(pu1_src_tmp)
183 vmovl.u8 q0,d0 @vmovl_u8(vld1_u8(pu1_src_tmp)
191 vmovl.u8 q11,d22 @vmovl_u8(vld1_u8(pu1_src_tmp)
194 vmovl.u8 q12,d24 @vmovl_u8(vld1_u8(pu1_src_tmp)
220 vmovl.u8 q8,d8 @vmovl_u8(vld1_u8(pu1_src_tmp))
221 vmovl.u8 q9,d10 @vmovl_u8(vld1_u8(pu1_src_tmp)
222 vmovl.u8 q10,d12 @vmovl_u8(vld1_u8(pu1_src_tmp)
[all …]
|
| D | ihevc_inter_pred_luma_copy_w16out.s | 109 vmovl.u8 q0,d0 @vmovl_u8(vld1_u8(pu1_src_tmp)
117 vmovl.u8 q11,d22 @vmovl_u8(vld1_u8(pu1_src_tmp)
120 vmovl.u8 q12,d24 @vmovl_u8(vld1_u8(pu1_src_tmp)
125 vmovl.u8 q13,d26 @vmovl_u8(vld1_u8(pu1_src_tmp)
158 vmovl.u8 q8,d8 @vmovl_u8(vld1_u8(pu1_src_tmp))
159 vmovl.u8 q9,d10 @vmovl_u8(vld1_u8(pu1_src_tmp)
160 vmovl.u8 q10,d12 @vmovl_u8(vld1_u8(pu1_src_tmp)
161 vmovl.u8 q11,d14 @vmovl_u8(vld1_u8(pu1_src_tmp)
188 vmovl.u8 q8,d8 @vmovl_u8(vld1_u8(pu1_src_tmp))
191 vmovl.u8 q9,d10 @vmovl_u8(vld1_u8(pu1_src_tmp)
[all …]
|
| D | ihevc_sao_edge_offset_class1_chroma.s | 170 … @II pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_c…
173 … @II pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_…
192 …D10 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_c…
204 …D11 @pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_…
223 … @II pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_…
264 …D10 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_c…
270 …D11 @pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_…
325 … @II pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_c…
349 …D10 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_c…
400 …D10 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_c…
|
| D | ihevc_sao_edge_offset_class1.s | 166 … @II pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_c…
169 … @II pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_…
188 …D10 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_c…
197 …11 @pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_…
247 …D10 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_c…
253 …D11 @pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_…
308 … @II pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_c…
325 …D10 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_c…
365 …D10 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_c…
|
| /external/XNNPACK/src/q8-dwconv/ |
| D | up8x9-neon.c | 73 const int16x8_t vxi00 = vreinterpretq_s16_u16(vmovl_u8(vi00)); in xnn_q8_dwconv_ukernel_up8x9__neon()
74 const int16x8_t vxi01 = vreinterpretq_s16_u16(vmovl_u8(vi01)); in xnn_q8_dwconv_ukernel_up8x9__neon()
75 const int16x8_t vxi02 = vreinterpretq_s16_u16(vmovl_u8(vi02)); in xnn_q8_dwconv_ukernel_up8x9__neon()
88 const int16x8_t vxi10 = vreinterpretq_s16_u16(vmovl_u8(vi10)); in xnn_q8_dwconv_ukernel_up8x9__neon()
89 const int16x8_t vxi11 = vreinterpretq_s16_u16(vmovl_u8(vi11)); in xnn_q8_dwconv_ukernel_up8x9__neon()
90 const int16x8_t vxi12 = vreinterpretq_s16_u16(vmovl_u8(vi12)); in xnn_q8_dwconv_ukernel_up8x9__neon()
103 const int16x8_t vxi20 = vreinterpretq_s16_u16(vmovl_u8(vi20)); in xnn_q8_dwconv_ukernel_up8x9__neon()
104 const int16x8_t vxi21 = vreinterpretq_s16_u16(vmovl_u8(vi21)); in xnn_q8_dwconv_ukernel_up8x9__neon()
105 const int16x8_t vxi22 = vreinterpretq_s16_u16(vmovl_u8(vi22)); in xnn_q8_dwconv_ukernel_up8x9__neon()
116 const int16x8_t vxi03 = vreinterpretq_s16_u16(vmovl_u8(vi03)); in xnn_q8_dwconv_ukernel_up8x9__neon()
[all …]
|
| /external/tensorflow/tensorflow/lite/kernels/internal/optimized/ |
| D | depthwiseconv_uint8.h | 48 filter[i] = vaddq_s16(vreinterpretq_s16_u16(vmovl_u8(filter_u8.val[i])),
62 const int16x8_t input_s16 = vreinterpretq_s16_u16(vmovl_u8(input_u8));
91 const int16x8_t filter_s16 = vreinterpretq_s16_u16(vmovl_u8(filter_u8));
110 input[i] = vreinterpretq_s16_u16(vmovl_u8(input_u8[i]));
138 const int16x8_t input_s16 = vreinterpretq_s16_u16(vmovl_u8(input_u8));
159 const int16x8_t filter_s16 = vreinterpretq_s16_u16(vmovl_u8(filter_u8));
173 const int16x8_t input_s16 = vreinterpretq_s16_u16(vmovl_u8(input_u8));
205 vreinterpret_s16_u16(vget_low_u16(vmovl_u8(input_u8)));
231 const int16x8_t filter_s16 = vreinterpretq_s16_u16(vmovl_u8(filter_u8));
250 vreinterpret_s16_u16(vget_low_u16(vmovl_u8(input_u8)));
[all …]
|
| /external/skia/src/core/ |
| D | SkBlitRow_D32.cpp | 165 vsrc_wide = vmovl_u8(vsrc); in blit_row_s32_blend()
187 vsrc_wide = vmovl_u8(vsrc); in blit_row_s32_blend()
213 vsrc_wide = vmovl_u8(vsrc); in blit_row_s32a_blend()
216 vdst_wide = vmovl_u8(vdst); in blit_row_s32a_blend()
246 vdst_scale = vmovl_u8(vsrc_alphas); in blit_row_s32a_blend()
256 vsrc_wide = vmovl_u8(vsrc); in blit_row_s32a_blend()
259 vdst_wide = vmovl_u8(vdst); in blit_row_s32a_blend()
|
| /external/skqp/src/core/ |
| D | SkBlitRow_D32.cpp | 166 vsrc_wide = vmovl_u8(vsrc); in blit_row_s32_blend()
188 vsrc_wide = vmovl_u8(vsrc); in blit_row_s32_blend()
214 vsrc_wide = vmovl_u8(vsrc); in blit_row_s32a_blend()
217 vdst_wide = vmovl_u8(vdst); in blit_row_s32a_blend()
247 vdst_scale = vmovl_u8(vsrc_alphas); in blit_row_s32a_blend()
257 vsrc_wide = vmovl_u8(vsrc); in blit_row_s32a_blend()
260 vdst_wide = vmovl_u8(vdst); in blit_row_s32a_blend()
|
| /external/libhevc/encoder/arm/ |
| D | ihevce_hme_utils_neon.c | 137 src0_8x16b = vreinterpretq_s16_u16(vmovl_u8(src0_8x8b)); in ihevce_get_wt_inp_4x8_neon()
138 src1_8x16b = vreinterpretq_s16_u16(vmovl_u8(src1_8x8b)); in ihevce_get_wt_inp_4x8_neon()
139 src2_8x16b = vreinterpretq_s16_u16(vmovl_u8(src2_8x8b)); in ihevce_get_wt_inp_4x8_neon()
140 src3_8x16b = vreinterpretq_s16_u16(vmovl_u8(src3_8x8b)); in ihevce_get_wt_inp_4x8_neon()
567 src0_8x16b = vreinterpretq_s16_u16(vmovl_u8(src0_8x8b)); in hme_get_wt_inp_ctb_neon()
568 src1_8x16b = vreinterpretq_s16_u16(vmovl_u8(src1_8x8b)); in hme_get_wt_inp_ctb_neon()
569 src2_8x16b = vreinterpretq_s16_u16(vmovl_u8(src2_8x8b)); in hme_get_wt_inp_ctb_neon()
570 src3_8x16b = vreinterpretq_s16_u16(vmovl_u8(src3_8x8b)); in hme_get_wt_inp_ctb_neon()
700 src0_8x16b = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(src0_16x8b))); in hme_get_wt_inp_ctb_neon()
701 src1_8x16b = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(src0_16x8b))); in hme_get_wt_inp_ctb_neon()
|
| D | ihevce_itrans_recon_neon.c | 67 a1 = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(src_u8))); in ihevce_itrans_recon_dc_4x4_luma_neon()
68 a2 = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(src_u8))); in ihevce_itrans_recon_dc_4x4_luma_neon()
98 a1 = vreinterpretq_s16_u16(vmovl_u8(vld1_u8(pu1_pred + i * pred_strd))); in ihevce_itrans_recon_dc_4x4_chroma_neon()
|
| /external/skqp/src/opts/ |
| D | Sk4px_NEON.h | 11 return Sk16h(vmovl_u8(vget_low_u8 (this->fVec)), in widen()
12 vmovl_u8(vget_high_u8(this->fVec))); in widen()
|
| /external/skia/src/opts/ |
| D | Sk4px_NEON.h | 11 return Sk16h(vmovl_u8(vget_low_u8 (this->fVec)), in widen()
12 vmovl_u8(vget_high_u8(this->fVec))); in widen()
|
| /external/XNNPACK/src/q8-igemm/ |
| D | 8x8-neon.c | 134 const int16x8_t vxa0 = vreinterpretq_s16_u16(vmovl_u8(va0)); in xnn_q8_igemm_ukernel_8x8__neon()
135 const int16x8_t vxa1 = vreinterpretq_s16_u16(vmovl_u8(va1)); in xnn_q8_igemm_ukernel_8x8__neon()
136 const int16x8_t vxa2 = vreinterpretq_s16_u16(vmovl_u8(va2)); in xnn_q8_igemm_ukernel_8x8__neon()
137 const int16x8_t vxa3 = vreinterpretq_s16_u16(vmovl_u8(va3)); in xnn_q8_igemm_ukernel_8x8__neon()
138 const int16x8_t vxa4 = vreinterpretq_s16_u16(vmovl_u8(va4)); in xnn_q8_igemm_ukernel_8x8__neon()
139 const int16x8_t vxa5 = vreinterpretq_s16_u16(vmovl_u8(va5)); in xnn_q8_igemm_ukernel_8x8__neon()
140 const int16x8_t vxa6 = vreinterpretq_s16_u16(vmovl_u8(va6)); in xnn_q8_igemm_ukernel_8x8__neon()
141 const int16x8_t vxa7 = vreinterpretq_s16_u16(vmovl_u8(va7)); in xnn_q8_igemm_ukernel_8x8__neon()
330 const int16x8_t vxa0 = vreinterpretq_s16_u16(vmovl_u8(va0)); in xnn_q8_igemm_ukernel_8x8__neon()
331 const int16x8_t vxa1 = vreinterpretq_s16_u16(vmovl_u8(va1)); in xnn_q8_igemm_ukernel_8x8__neon()
[all …]
|
| D | 4x8-neon.c | 90 const int16x8_t vxa0 = vreinterpretq_s16_u16(vmovl_u8(va0)); in xnn_q8_igemm_ukernel_4x8__neon()
91 const int16x8_t vxa1 = vreinterpretq_s16_u16(vmovl_u8(va1)); in xnn_q8_igemm_ukernel_4x8__neon()
92 const int16x8_t vxa2 = vreinterpretq_s16_u16(vmovl_u8(va2)); in xnn_q8_igemm_ukernel_4x8__neon()
93 const int16x8_t vxa3 = vreinterpretq_s16_u16(vmovl_u8(va3)); in xnn_q8_igemm_ukernel_4x8__neon()
214 const int16x8_t vxa0 = vreinterpretq_s16_u16(vmovl_u8(va0)); in xnn_q8_igemm_ukernel_4x8__neon()
215 const int16x8_t vxa1 = vreinterpretq_s16_u16(vmovl_u8(va1)); in xnn_q8_igemm_ukernel_4x8__neon()
216 const int16x8_t vxa2 = vreinterpretq_s16_u16(vmovl_u8(va2)); in xnn_q8_igemm_ukernel_4x8__neon()
217 const int16x8_t vxa3 = vreinterpretq_s16_u16(vmovl_u8(va3)); in xnn_q8_igemm_ukernel_4x8__neon()
|
| /external/XNNPACK/src/q8-gemm/ |
| D | 8x8-neon.c | 102 const int16x8_t vxa0 = vreinterpretq_s16_u16(vmovl_u8(va0)); a0 += 8; in xnn_q8_gemm_ukernel_8x8__neon()
104 const int16x8_t vxa1 = vreinterpretq_s16_u16(vmovl_u8(va1)); a1 += 8; in xnn_q8_gemm_ukernel_8x8__neon()
106 const int16x8_t vxa2 = vreinterpretq_s16_u16(vmovl_u8(va2)); a2 += 8; in xnn_q8_gemm_ukernel_8x8__neon()
108 const int16x8_t vxa3 = vreinterpretq_s16_u16(vmovl_u8(va3)); a3 += 8; in xnn_q8_gemm_ukernel_8x8__neon()
110 const int16x8_t vxa4 = vreinterpretq_s16_u16(vmovl_u8(va4)); a4 += 8; in xnn_q8_gemm_ukernel_8x8__neon()
112 const int16x8_t vxa5 = vreinterpretq_s16_u16(vmovl_u8(va5)); a5 += 8; in xnn_q8_gemm_ukernel_8x8__neon()
114 const int16x8_t vxa6 = vreinterpretq_s16_u16(vmovl_u8(va6)); a6 += 8; in xnn_q8_gemm_ukernel_8x8__neon()
116 const int16x8_t vxa7 = vreinterpretq_s16_u16(vmovl_u8(va7)); a7 += 8; in xnn_q8_gemm_ukernel_8x8__neon()
282 const int16x8_t vxa0 = vreinterpretq_s16_u16(vmovl_u8(va0)); in xnn_q8_gemm_ukernel_8x8__neon()
284 const int16x8_t vxa1 = vreinterpretq_s16_u16(vmovl_u8(va1)); in xnn_q8_gemm_ukernel_8x8__neon()
[all …]
|
| D | 4x8-neon.c | 70 const int16x8_t vxa0 = vreinterpretq_s16_u16(vmovl_u8(va0)); in xnn_q8_gemm_ukernel_4x8__neon()
72 const int16x8_t vxa1 = vreinterpretq_s16_u16(vmovl_u8(va1)); in xnn_q8_gemm_ukernel_4x8__neon()
74 const int16x8_t vxa2 = vreinterpretq_s16_u16(vmovl_u8(va2)); in xnn_q8_gemm_ukernel_4x8__neon()
76 const int16x8_t vxa3 = vreinterpretq_s16_u16(vmovl_u8(va3)); in xnn_q8_gemm_ukernel_4x8__neon()
178 const int16x8_t vxa0 = vreinterpretq_s16_u16(vmovl_u8(va0)); in xnn_q8_gemm_ukernel_4x8__neon()
180 const int16x8_t vxa1 = vreinterpretq_s16_u16(vmovl_u8(va1)); in xnn_q8_gemm_ukernel_4x8__neon()
182 const int16x8_t vxa2 = vreinterpretq_s16_u16(vmovl_u8(va2)); in xnn_q8_gemm_ukernel_4x8__neon()
184 const int16x8_t vxa3 = vreinterpretq_s16_u16(vmovl_u8(va3)); in xnn_q8_gemm_ukernel_4x8__neon()
|
| /external/tensorflow/tensorflow/core/kernels/ |
| D | quantized_instance_norm.cc | 71 const uint16x8_t v_high_u16 = vmovl_u8(v_high); in ColMeanAndVariance()
72 const uint16x8_t v_low_u16 = vmovl_u8(v_low); in ColMeanAndVariance()
165 const uint16x8_t v_high = vmovl_u8(vget_high_u8(v)); in MinAndMax()
166 const uint16x8_t v_low = vmovl_u8(vget_low_u8(v)); in MinAndMax()
221 const uint16x8_t v_high = vmovl_u8(vget_high_u8(v)); in InstanceNorm()
222 const uint16x8_t v_low = vmovl_u8(vget_low_u8(v)); in InstanceNorm()
|
| /external/webp/src/dsp/ |
| D | dec_neon.c | 380 return vreinterpretq_s16_u16(vmovl_u8(v)); in ConvertU8ToS16_NEON()
1286 const uint16x8_t L0 = vmovl_u8(vld1_u8(dst + 0 * BPS - 1)); in DC4_NEON()
1287 const uint16x8_t L1 = vmovl_u8(vld1_u8(dst + 1 * BPS - 1)); in DC4_NEON()
1288 const uint16x8_t L2 = vmovl_u8(vld1_u8(dst + 2 * BPS - 1)); in DC4_NEON()
1289 const uint16x8_t L3 = vmovl_u8(vld1_u8(dst + 3 * BPS - 1)); in DC4_NEON()
1432 const uint16x8_t B = vmovl_u8(A); in DC8_NEON()
1444 const uint16x8_t L0 = vmovl_u8(vld1_u8(dst + 0 * BPS - 1)); in DC8_NEON()
1445 const uint16x8_t L1 = vmovl_u8(vld1_u8(dst + 1 * BPS - 1)); in DC8_NEON()
1446 const uint16x8_t L2 = vmovl_u8(vld1_u8(dst + 2 * BPS - 1)); in DC8_NEON()
1447 const uint16x8_t L3 = vmovl_u8(vld1_u8(dst + 3 * BPS - 1)); in DC8_NEON()
[all …]
|