this fixes a potential access of invalid PHY or MAC metrics by checking
the user entry actually exists.
the RFCI has shown this error:
------DL--------------------------------UL------------------------------------
rnti cqi ri mcs brate ok nok (%) snr phr mcs brate ok nok (%) bsr
ASAN:DEADLYSIGNAL
=================================================================
m==31838==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000000 (pc 0x555d482b6893 bp 0x7f6ac32d1160 sp 0x7f6ac32d0bc0 T21)
==31838==The signal is caused by a READ memory access.
==31838==Hint: address points to the zero page.
#0 0x555d482b6892 in srsenb::metrics_stdout::set_metrics(srsenb::enb_metrics_t const&, unsigned int) /mnt/data/jenkins/workspace/srslte_dev_ogt_zmq_nightly/srsLTE/srsenb/src/metrics_stdout.cc:101
#1 0x555d482865f1 in srslte::metrics_hub<srsenb::enb_metrics_t>::run_period() /mnt/data/jenkins/workspace/srslte_dev_ogt_zmq_nightly/srsLTE/lib/include/srslte/common/metrics_hub.h:88
#2 0x555d482865f1 in srslte::periodic_thread::run_thread() /mnt/data/jenkins/workspace/srslte_dev_ogt_zmq_nightly/srsLTE/lib/include/srslte/common/threads.h:143
#3 0x555d4826813d in srslte:🧵:thread_function_entry(void*) /mnt/data/jenkins/workspace/srslte_dev_ogt_zmq_nightly/srsLTE/lib/include/srslte/common/threads.h:102
#4 0x7f6b0dc546da in start_thread (/lib/x86_64-linux-gnu/libpthread.so.0+0x76da)
#5 0x7f6b0bf0171e in __clone (/lib/x86_64-linux-gnu/libc.so.6+0x12171e)
- fix encoding of TPC command
- use of exponential average with irregular sampling for the ULSNR average estimate.
Turns out using a time-windowed average for the SNR was a bad idea.
If the UL grants are very sporadic, the SNR time window will never have samples
when a TPC is encoded
- update of TPC sched test
- other fixes in accumulators lib
- Use single interface for both DL and UL
- Wrote helper functions for tasks that are common to different algorithms
- created subfolder for all scheduling algorithms
- new stats allow to analyze the average bitrates of each user
- there is a new stat that computes the total of RBs allocated and total bitrate achieved by the scheduler
This optimization has been achieved by:
- implementation of "false position method" root-finding algorithm
- application of the false position method in cc_sched_ue::get_required_prb_dl/ul(...)
The number of iterations in the worst case scenario, reduced from 100 to 5 with
the proposed algorithm.