Prevent buffer under/overflow in ULSCH decoding through alignment and sufficient array size
This commit is contained in:
committed by
Sakthivel Velumani
parent
3cfb169b37
commit
66cbe490e1
@@ -308,8 +308,8 @@ void nr_processULSegment(void* arg) {
|
||||
int max_ldpc_iterations = p_decoderParms->numMaxIter;
|
||||
int8_t llrProcBuf[OAI_UL_LDPC_MAX_NUM_LLR] __attribute__ ((aligned(32)));
|
||||
|
||||
int16_t z [68*384];
|
||||
int8_t l [68*384];
|
||||
int16_t z [68*384 + 16] __attribute__ ((aligned(16)));
|
||||
int8_t l [68*384 + 16] __attribute__ ((aligned(16)));
|
||||
|
||||
__m128i *pv = (__m128i*)&z;
|
||||
__m128i *pl = (__m128i*)&l;
|
||||
|
||||
Reference in New Issue
Block a user