Integration 2020 wk40 comprises the following udpates:
MR893 ULSCH decode mthread:
- Enabled parallelization of ULSCH segments decoding, deinterleaving and rx rate matching.
MR896 Nsa remove band hardcodings:
- A missing commit from wk39, just a log change
MR901 NR UE DCI improvements:
- Handling of DCI formats 0_0 and 0_1, including removing hard-coding and scheduling PUSCH with DCI parameters
- Blind decoding of DCIs based on size and format indicator
- Fix for segmentation fault during ULSCH encoding
MR895 support of RA procedure by implementing Msg3 on UE
MR908 NR DL scheduler improvements:
Preliminary work on the DL scheduler
- cleanup of DL scheduler, add interface preprocessor-postprocessor (`NR_UE_sched_ctrl_t`)
- postprocessor can handle multiple UEs. Preprocessor will follow soon (proportional fair scheduler)
- No MAC UE Context creation before end of RA (Msg 3)
These updates do not change anything performance wise (no additional scheduled slots, etc).
This MR also reduces gNB log.
-Since the aim of itti RRC_SUBFRAME_PROCESS message (MAC->RRC) is to trigger an itti X2AP_SUBFRAME_PROCESS message (RRC->X2AP), avoid creating
RRC_SUBFRAME_PROCESS message if x2ap task is not enabled. Not sure if this is the best solution. We could also avoid only sending the second itti X2AP_SUBFRAME_PROCESS message.
-Since the aim of itti RRC_SUBFRAME_PROCESS message (MAC->RRC) is to trigger an itti X2AP_SUBFRAME_PROCESS message (RRC->X2AP), avoid creating
RRC_SUBFRAME_PROCESS message if x2ap task is not enabled. Not sure if this is the best solution. We could also avoid only sending the second itti X2AP_SUBFRAME_PROCESS message.
A lock help during executing a timer event prevented such event from
creating/removing new timers due to a dead lock. This commit changes the
logic of the timing loop and adding/removing code such that this dead
lock does not happen.
Remi Hardy