112     int32x4_t reg0[4], reg1[4];  in ihevce_wt_avg_2d_16x1_neon()  local
129     reg0[0] = vmovl_s16(vget_low_s16(a2));  in ihevce_wt_avg_2d_16x1_neon()
130     reg0[1] = vmovl_s16(vget_high_s16(a2));  in ihevce_wt_avg_2d_16x1_neon()
131     reg0[2] = vmovl_s16(vget_low_s16(a3));  in ihevce_wt_avg_2d_16x1_neon()
132     reg0[3] = vmovl_s16(vget_high_s16(a3));  in ihevce_wt_avg_2d_16x1_neon()
139     reg0[0] = vmulq_s32(reg0[0], a6);  in ihevce_wt_avg_2d_16x1_neon()
140     reg0[1] = vmulq_s32(reg0[1], a6);  in ihevce_wt_avg_2d_16x1_neon()
141     reg0[2] = vmulq_s32(reg0[2], a6);  in ihevce_wt_avg_2d_16x1_neon()
142     reg0[3] = vmulq_s32(reg0[3], a6);  in ihevce_wt_avg_2d_16x1_neon()
149     reg0[0] = vaddq_s32(reg0[0], reg1[0]);  in ihevce_wt_avg_2d_16x1_neon()
150     reg0[1] = vaddq_s32(reg0[1], reg1[1]);  in ihevce_wt_avg_2d_16x1_neon()
151     reg0[2] = vaddq_s32(reg0[2], reg1[2]);  in ihevce_wt_avg_2d_16x1_neon()
152     reg0[3] = vaddq_s32(reg0[3], reg1[3]);  in ihevce_wt_avg_2d_16x1_neon()
154     reg0[0] = vshlq_s32(reg0[0], a9);  in ihevce_wt_avg_2d_16x1_neon()
155     reg0[1] = vshlq_s32(reg0[1], a9);  in ihevce_wt_avg_2d_16x1_neon()
156     reg0[2] = vshlq_s32(reg0[2], a9);  in ihevce_wt_avg_2d_16x1_neon()
157     reg0[3] = vshlq_s32(reg0[3], a9);  // (p0*w0 + p1*w1) >> shift  in ihevce_wt_avg_2d_16x1_neon()
159     a2 = vcombine_s16(vmovn_s32(reg0[0]), vmovn_s32(reg0[1]));  in ihevce_wt_avg_2d_16x1_neon()
160     a3 = vcombine_s16(vmovn_s32(reg0[2]), vmovn_s32(reg0[3]));  in ihevce_wt_avg_2d_16x1_neon()
232     int32x4_t reg0[4], reg1[4];  in ihevce_wt_avg_2d_4xn_neon()  local
255             reg0[0] = vmovl_s16(vreinterpret_s16_u16(vget_low_u16(a0)));  in ihevce_wt_avg_2d_4xn_neon()
256             reg0[1] = vmovl_s16(vreinterpret_s16_u16(vget_high_u16(a0)));  in ihevce_wt_avg_2d_4xn_neon()
257             reg0[2] = vmovl_s16(vreinterpret_s16_u16(vget_low_u16(a1)));  in ihevce_wt_avg_2d_4xn_neon()
258             reg0[3] = vmovl_s16(vreinterpret_s16_u16(vget_high_u16(a1)));  in ihevce_wt_avg_2d_4xn_neon()
265             reg0[0] = vmulq_s32(reg0[0], a4);  in ihevce_wt_avg_2d_4xn_neon()
266             reg0[1] = vmulq_s32(reg0[1], a4);  in ihevce_wt_avg_2d_4xn_neon()
267             reg0[2] = vmulq_s32(reg0[2], a4);  in ihevce_wt_avg_2d_4xn_neon()
268             reg0[3] = vmulq_s32(reg0[3], a4);  in ihevce_wt_avg_2d_4xn_neon()
275             reg0[0] = vaddq_s32(reg0[0], reg1[0]);  in ihevce_wt_avg_2d_4xn_neon()
276             reg0[1] = vaddq_s32(reg0[1], reg1[1]);  in ihevce_wt_avg_2d_4xn_neon()
277             reg0[2] = vaddq_s32(reg0[2], reg1[2]);  in ihevce_wt_avg_2d_4xn_neon()
278             reg0[3] = vaddq_s32(reg0[3], reg1[3]);  in ihevce_wt_avg_2d_4xn_neon()
280             reg0[0] = vshlq_s32(reg0[0], a7);  in ihevce_wt_avg_2d_4xn_neon()
281             reg0[1] = vshlq_s32(reg0[1], a7);  in ihevce_wt_avg_2d_4xn_neon()
282             reg0[2] = vshlq_s32(reg0[2], a7);  in ihevce_wt_avg_2d_4xn_neon()
283             reg0[3] = vshlq_s32(reg0[3], a7);  in ihevce_wt_avg_2d_4xn_neon()
285             a8 = vcombine_s16(vmovn_s32(reg0[0]), vmovn_s32(reg0[1]));  in ihevce_wt_avg_2d_4xn_neon()
286             a9 = vcombine_s16(vmovn_s32(reg0[2]), vmovn_s32(reg0[3]));  in ihevce_wt_avg_2d_4xn_neon()