hotfix for 4 layer mapping on aarch64
This commit is contained in:
@@ -454,9 +454,9 @@ void nr_layer_mapping(int nbCodes,
|
|||||||
tx0++;
|
tx0++;
|
||||||
*(uint32_t *)tx1 = vgetq_lane_u32(d4, 1);
|
*(uint32_t *)tx1 = vgetq_lane_u32(d4, 1);
|
||||||
tx1++;
|
tx1++;
|
||||||
*(uint32_t *)tx2 = vgetq_lane_u32(d4, 0);
|
*(uint32_t *)tx2 = vgetq_lane_u32(d4, 2);
|
||||||
tx2++;
|
tx2++;
|
||||||
*(uint32_t *)tx3 = vgetq_lane_u32(d4, 1);
|
*(uint32_t *)tx3 = vgetq_lane_u32(d4, 3);
|
||||||
tx3++;
|
tx3++;
|
||||||
}
|
}
|
||||||
#endif
|
#endif
|
||||||
|
|||||||
Reference in New Issue
Block a user