Skip to content

T2: minor fix for 4x4 processing

Jaroslava Fiedlerova requested to merge t2-fix-4x4 into develop
  • correct max number of LDPC segments for T2 processing
  • use pointer arithmetics in retrieve_ldpc_enc_op() in T2 lib to optimize loop processing

Comparison of processing time measured in nr_dlsim for LDPC encoding performed by CPU, T2 and T2 with this MR.

testcase develop (8 cores) develop with T2 this MR with T2
SNR = 30, MCS = 5, 106 PRBs, 1 layer 184 us 37 us 31 us
SNR = 30, MCS = 15, 106 PRBs, 1 layer 243 us 61 us 48 us
SNR = 30, MCS = 25, 106 PRBs, 1 layer 248 us 89 us 69 us
SNR = 30, MCS = 5, 273 PRBs, 1 layer 256 us 68 us 53 us
SNR = 30, MCS = 15, 273 PRBs, 1 layer 308 us 135 us 101 us
SNR = 30, MCS = 25, 273 PRBs, 1 layer 337 us 202 us 153 us
SNR = 30, MCS = 5, 273 PRBs, 2 layers 334 us 126 us 125 us
SNR = 30, MCS = 15, 273 PRBs, 2 layers 365 us 246 us 181 us
SNR = 30, MCS = 25, 273 PRBs, 2 layers 394 us 386 us 273 us
SNR = 30, MCS = 5, 273 PRBs, 2 layers, 4x4 349 us 131 us 98 us
SNR = 30, MCS = 15, 273 PRBs, 2 layers, 4x4 355 us 250 us 186 us
SNR = 30, MCS = 25, 273 PRBs, 2 layers, 4x4 388 us 377 us 279 us
Edited by Jaroslava Fiedlerova

Merge request reports