arm: Use vreinterpret instead of a plain cast for converting between neon vector types
This fixes building with MSVC for arm. Change-Id: Iffae0408e0c68760e87e96b9e17d9df8e8cadb1a
This commit is contained in:
parent
6e706ed783
commit
e5647d6826
@ -119,8 +119,10 @@ int vp8_denoiser_filter_neon(YV12_BUFFER_CONFIG *mc_running_avg,
|
||||
v_abs_adjustment);
|
||||
v_running_avg_y = vqaddq_u8(v_sig, v_pos_adjustment);
|
||||
v_running_avg_y = vqsubq_u8(v_running_avg_y, v_neg_adjustment);
|
||||
v_sum_diff = vqaddq_s8(v_sum_diff, (int8x16_t)v_pos_adjustment);
|
||||
v_sum_diff = vqsubq_s8(v_sum_diff, (int8x16_t)v_neg_adjustment);
|
||||
v_sum_diff = vqaddq_s8(v_sum_diff,
|
||||
vreinterpretq_s8_u8(v_pos_adjustment));
|
||||
v_sum_diff = vqsubq_s8(v_sum_diff,
|
||||
vreinterpretq_s8_u8(v_neg_adjustment));
|
||||
|
||||
/* Store results. */
|
||||
vst1q_u8(running_avg_y, v_running_avg_y);
|
||||
|
Loading…
x
Reference in New Issue
Block a user