Option `-q` of the NR softmodem can now take an argument to choose the
NR L1 statistics to display:
* no `-q` show no statistics
* `-q` or `-q 1` shows the usual statistics with averaging from gNB start
* `-q 2` shows the advanced statistics with distribution with a reset every second
When timers are started/stopped/merged on mixed slot,
the measured times are not homogeneous as the mixed slots are much
lighter than full UL/DL slots.
Then, to make the analysis of results easier,
we do not measure the mixed slots.
Changed tinput, tprep, tparity and toutput timers
from pointers at the slot to parts of the struct at the segment.
They can then be merged like the other timers of the encoder.
This solution enables an easier usage.
Timers were relying on rdtsc which gives a number of ticks which should
be proportionated by the clock frequency.
But the clock frequency can vary.
Then this way of calculating time is not reliable.
This commit offers to use the posix clock monotonic raw instead.
* Standard Deviation
* Optional Sorted list enabling:
* min
* median
* q1
* q3
* d1
* d9
The sorted list is optional.
It is enabled by initializing the sorted list which allocates the list.
The size of the list that is allocated is an argument of the initialization.
Integration `2026.w16`
* !4040 fix: issue 911 - replace timer oneshot to periodic.
* !4041 Compilation without T: move to physims
* !4046 fix(record_db doc): correct the path for T_messages.txt and mention use of ninja/make
* !3994 Update CN5G images to release v2.2.0 and update traffic generator image
* !4024 Fix fill_srs_channel_matrix
* !3983 Add SIB 3,4 support, configurable SIB2 and refactor CU/DU SIB management
* !4030 Miscellaneous code improvements
* !4039 Verify consistency of CSI report in L2 instead of RRC
* !3991 L1 RX: use queues instead of arrays and linear search for PUCCH, PUSCH, SRS, PRACH
* !4011 CI: remove cppcheck, cleanup in build and license checks
* !4051 Fix EMA cold-start for noise and SNR/RSSI measurements
Closes#1063
See merge request oai/openairinterface5g!4048
Fix EMA cold-start for noise and SNR/RSSI measurements
Initialize n0_subband_power and power control avg_snr/avg_rssi directly
from the first measurement instead of starting from zero.
Starting from zero causes the EMA to converge slowly, leading to
underestimated noise power at gNB startup. This results in inflated SNR
estimates, which triggers UE uplink power ramp-up, antenna saturation,
and a positive feedback loop of increasing noise.
CI: remove cppcheck, cleanup in build and license checks
Remove cppcheck. For those who want to use it, it's now in tools/cppcheck/,
and likely easier to use locally.
Clean up some code for CI build and license check. Remove global variable.
We don't enforce cppcheck through the CI, although it's there since
years. It runs on Ubuntu 18/20, so it's old. For folks, it's likely not
discoverable on how to run it locally. Let's make a fresh start.
This removes cppcheck from all CI-related code. Instead, it adds it
under tools/cppcheck/, including documentation on how to run it locally,
bare-matel or in docker.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
L1 RX: use queues instead of arrays and linear search for PUCCH, PUSCH, SRS, PRACH
This MR is an attempt to reduce the time L1 RX searches in array for the
next job to process, and instead use a queue. This queue is a FIFO,
because the various jobs (FAPI messages) to process come in order, and
need to be processed in order. The MR (hopefully, to be measured)
reduces the amount of time spent searching for the next UE (because the
next job is always at the beginning of the queue), and should scale
better for many UEs.
It does the following:
- introduce two helper libraries for (1) Frame.Slot calculation (sfn_t),
already introduced in !3521 (merged) commit 3102068e, and (2) a ring
buffer with fixed size
- use sfn_t and ring buffer (gNB->pucch_queue) to remove the linear
array for PUCCH (gNB->pucch)
- use sfn_t and ring buffer (gNB->pusch_queue) for some PUSCH lookups.
Because we need to still store PUSCH contexts, gNB->pusch is still
there
- use sfn_t and ring buffer (gNB->srs_queue) to remove the linear array
for SRS (gNB->srs)
- use sfn_t and ring buffers (gNB->prach_ru_queue and
gNB->prach_l1rx_queue) to remove the linear array for PRACH
(gNB->prach_list)
- some minor cleanups, e.g., additional loops over the PUSCH array,
using const, using pointers instead of indices, etc
Miscellaneous code improvements
- A fix in a yaml config file as reported by @Abdo-Gaber
- proper positioning of static functions in a couple of gNB scheduler
files
- some effort to split NR from LTE code in compilation
- harmonization of macros for unused variables
Add SIB 3,4 support, configurable SIB2 and refactor CU/DU SIB management
This MR makes neighbour and inter-frequency configuration drive how
SIB2/SIB3/SIB4 are built and sent from CU to DUs. It standardizes SIB
payloads as byte_array_t with typed SIB IDs across RRC, F1AP and MAC,
reducing ad‑hoc buffer handling. Neighbour parsing, validation and
lookup are tightened.
Changes
- Minor refactor to gNB neighbour parsing and storage (shared PLMN
extraction, safer allocation, etc).
- Represent SIB containers uniformly as byte_array_t plus nr_sib_type_t,
and adapt F1AP, MAC and RRC users to the new container API.
- Make SIB2 cell-reselection information fully config-driven with
explicit bounds checking and SIB2 ASN.1 building from that config.
- Generate SIB3 intra-frequency neighbours from the per-cell neighbour
list and propagate them from CU to DU over F1, with MAC
decoding/attaching them to SystemInformation.
- Generate SIB4 inter-frequency neighbours from a new frequency_list
plus neighbour SIB3/SIB4 offsets, and propagate them from CU to DU
over F1, with MAC decoding/attaching them.
- Add basic ASN.1 round-trip tests for SIB2/SIB3/SIB4 and SIB4 range
checks, and update RRC docs to describe the neighbour/inter-frequency
configuration model.
Testing:
1. in gNB conf file:
cu_sibs = (2, 3, 4);
2. Update neighbour config file with SIB3/SIB4 conf:
(see documentation)
3. run gNB and UE as usual
---
Logs & configs: see MR on Gitlab
Use two queues for PRACH processing:
- prach_ru_queue: pass jobs from TX thread (scheduler) to RU thread. For
split 8, the RU thread itself handles PRACH processing; for split 7.2,
it is the library that is responsible for handling PRACH messages (see
oran_fh_if4p5_south_in())
- prach_l1rx_queue: after jobs have been handled in RU thread, use this
queue to pass jobs to the RX thread. There, preamble detection is
performed, and RACH.indication FAPI messages are filled if a preamble
has been detected.
Together, these queues replace a linear search in a global array that
has been modified by three threads at the same time. The design ensures
that access now is thread-safe, and with less overhead.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
Expand RRC usage and handover documentation to describe the two-level
neighbour configuration layout, lookup-key semantics,
and config-time SIB4 grouping behaviour. Expand the example config
to include a `frequency_list` block and per-neighbour SIB3/SIB4
offset fields. Add SIB2 config example.
Update handover-tutorial.md to describe the same nested model and
note that F1 and N2 handover share the same serving-cell keyed
mapping.
Add documentation about SIB3/SIB4 and MeasGaps implementation
in OAI with stress on the shared neighbour configuration data model.
Update FEATURE_SET.
Use the spsc queue to pass jobs from TX thread (scheduler) to RX thread
(handling those jobs), reducing the amount spent searching for SRS jobs,
clarifying the design, and making it thread-safe.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
The structure in which was snr will be a "SRS job" in a follow-up
commit, so the snr parameter cannot be stored in there (and it is also
not necessary to do so, as it can be local to the function handling
nr_srs_rx_procedures()).
Correctly read SRS SNR through FAPI message in nr_ulsim. Switch to SRS
BEAMMANAGEMENT as only this message returns the actual SNR (unlike
CODEBOOK). Add asserts to check what is returned is valid data in the
sense of the test.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
Remove useless logging defines. Move srs_estimated_channel_time to where
it is used. Prepare for the next commit by moving UL_INFO to earlier
inside phy_procedures_gNB_uespec_RX().
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
- calculation in float: should be enough precision (it'll go into 16bit
ints later...)
- avoid round: the lower 15bits will be shifted away, the rounding will
have no effect
- group some multiplications to avoid re-computation of the same
results.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
This means that as of this commit, the signal has to be permanently
regenerated. In fact, with the SRS queue to be introduced in a follow-up
commit, we don't have an easy way to cache the sequence generation, as
there is no "index" to which we could refer to (hence, caching would be
tricky).
Instead, the next commits further refactor this to reduce the time for
SRS sequence generation as well as the memory footprint.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
Use the spsc queue to pass jobs from TX thread (scheduler) to RX thread
(handling those jobs), reducing the amount spent searching for PUSCH
jobs, clarifying the design, and making it thread-safe.
Since the PUSCH contexts have to be stored somewhere, there is still a
single iteration over all contexts. Removing that is left for future
work. Thus, this commit (together with previous commits) reduces the
number of iterations over all PUSCH entries, only iterating over the
actually requested PUSCH jobs. On the other hand, because only the RX
thread iterates over PUSCH contexts, it should be thread-safe.
Remove flag "handled", which is always 0.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
Pass pointers directly instead of ULSCH_id. Put const where possible and
applicable.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
Set the FAPI message in nr_ulschsim correctly in line with what
nr_ulsch_procedures() expected. Add an AssertFatal() to verify that the
assumed length of DMRS in the simulator and what is actually set in the
FAPI message matches. Finally, the internal nr_get_G() code (looping
over all PUSCH) does not double the DMRS because Nl == 3 || Nl == 4, so
harmonize in the simulator.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
It is pointless to iterate ulsch_to_decode to find the number of jobs if
we can simply pass it to the function.
Use an additional {} block to limit the amount of indentation.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
Amidst all these unnecessary pointers, the pointer on
pusch_vars->ptrs_symbols is necessary to reset the PTRS number of
symbols. For clarity, remove this pointer and use
pusch_vars->ptrs_symbols directly.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
Pass pointer to data. This necessitates to move delay from
NR_gNB_ULSCH_t (ULSCH data, including transport channels etc) to
NR_gNB_PUSCH (lower-PHY PUSCH data):
1. This allows to pass a single pointer to nr_pusch_channel_estimation()
2. To me, delay estimation seems to be a lower-PHY calculation, so it
fits better anyway.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
Use the spsc queue to pass jobs from TX thread (scheduler) to RX thread
(handling those jobs), reducing the amount spent searching for PUCCH
jobs in an array, clarifying the design, and making it thread-safe.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
Because it is SPSC, atomic variables are enough to synchronize the two
threads. Concretely put() is modified to ensure that read_idx is
"acquired" so that the read index has been written, including memory, by
the other thread. "Release" ensures that the write operation (including
the memory to the buffer) is written before it is visible to the other
thread in get() (which in turn "acquires" it). For more information, see
also [1].
The rest of the library has been simplified to work only with put() and
get(), reducing total code and the surface for possible bugs.
This (and the previous) version has been tested with the
threadSanitizer:
TSAN_OPTIONS=halt_on_error=1 ./common/utils/ds/tests/test_spsc_q_perf
On my machine, using Google Benchmark, I measure a considerable 5x speed
improvement:
$ /tmp/benchmark/tools/compare.py benchmarks pthread.json atomic.json
Comparing pthread.json to atomic.json
Benchmark Time CPU Time Old Time New CPU Old CPU New
--------------------------------------------------------------------------------------------------------------
BM_spsc_q/10 -0.8201 -0.0740 266779912 47989020 52387 48512
BM_spsc_q/16 -0.8301 -0.0520 249656592 42428540 51462 48784
BM_spsc_q/32 -0.8003 -0.0841 230248798 45972155 53841 49311
BM_spsc_q/64 -0.7995 -0.0506 210429791 42199674 50690 48124
BM_spsc_q/128 -0.7930 -0.1101 205212273 42483155 52745 46936
BM_spsc_q/160 -0.7880 -0.1663 216644738 45938247 53400 44518
OVERALL_GEOMEAN -0.8057 -0.0904 0 0 0 0
[1] https://en.cppreference.com/w/c/atomic/memory_order.html
Assisted-By: Claude:claude-sonnet-4-6
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
Fix fill_srs_channel_matrix
Fix incorrect SRS subcarrier mapping in fill_srs_channel_matrix
The previous implementation of fill_srs_channel_matrix assumed an incorrect
SRS subcarrier mapping, leading to wrong indexing of the estimated channel
in frequency domain.
Update CN5G images to release v2.2.1 and update traffic generator image
- Updated all CN5G image tags to v2.2.1
- Updated image oaisoftwarealliance/trf-gen-cn5g to latest tag supporting
multi-architecture platforms See here
- Enabled iperf3 server in daemon mode in oai-ext-dn
- Handover: Since MR oai/cn5g/oai-cn5g-amf!372 has been merged, the AMF
image tag is updated to the latest release v2.2.1
- upgrade MySQL image to 9.6
