Downlink performance degradation in nrUE
Here at Northeastern we observed a degradation of the downlink performances using the latest commit of develop w.r.t a 'stable' version we were using up to now (~september 21), since we are running OAI on colosseum, we weren't sure whether it was an issue with the channel emulation.
At the weekly soft-ue meeting we exposed the problem and today @luis_pereira87 was able to reproduce it on its own testbed. Later, I also reproduced it using two B210 and antennas.
Using 106PRB I was able to measure ~9Mbps of uplink TCP troughput, but when testing downlink TCP troughput the connection was failing:
[NR_PHY] ============================================
[NR_PHY] Harq round stats for Downlink: 20668/1755/117/8/0
[NR_PHY] ============================================
[NR_MAC] [8.0] Received TA_COMMAND 31 TAGID 0 CC_id 0
[NR_MAC] [18.0] Received TA_COMMAND 31 TAGID 0 CC_id 0
[NR_MAC] [28.0] Received TA_COMMAND 31 TAGID 0 CC_id 0
LLLLLLLLULLLULU[PDCP] discard NR PDU rcvd_count=18002, entity->rx_deliv 18007,sdu_in_list 0
[PDCP] discard NR PDU rcvd_count=18003, entity->rx_deliv 18007,sdu_in_list 0
LLLLLLLLLLLULLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLULLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLULLLLLLLLLLLLLLLLLLLLL
LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL
LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL[NR_PHY] ============================================
[NR_PHY] Harq round stats for Downlink: 20712/1771/123/11/0
[NR_PHY] ============================================
[NR_MAC] [UE 0] Release all SRs
[NR_MAC] [UE 0] Release all SRs
[NR_PHY] ============================================
[NR_PHY] Harq round stats for Downlink: 20712/1771/123/11/0
[NR_PHY] ============================================
Someone was mentioning that a commit, in sight of properly reimplementing a multi-threaded UE, dropped the parallel dlsch decoding.
I managed to test the previous commit and I measured a stable symmetrical throughput of around 10Mbps.