TRS resources if configured by gnb, is present in two consequtive slots
and two symbols per slot. LS estimates from the first symbol is saved
and in second symbol frequency offset and time offset is estimated. The
estimated FO is saved in UE global structure which is used in the last
TRS slot. If the FO is above a threshold, the radio's center frequency
is adjusted.
Time offset is estimated but not used because time is already tracked
using PBCH DMRS.
Signed-off-by: Sakthivel Velumani <s.velumani@northeastern.edu>
Clean up openair0_device, remove eth_params usage in L2/L3 (#243)
This PR has three goals
- Remove the usage of eth_params in L2/L3 code, and replace with
tailored data structures for IP connectivity where necessary
- Remove many of the includes of common_lib.h where possible
- Slightly clean up openair0_device and in particular move fields to
eth_state_t where appropriate
Reviewed-by: Bartosz Podrygajlo <bartosz.podrygajlo@openairinterface.org>
Reviewed-by: Teodora Vladić <teodora.vladic@openairinterface.org>
After having removed eth_params_t from all L2/L3 files, we can reduce
the scope of common_lib.h, and remove the include from many files.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
In 4G, it's not used.
In 5G, there is no need, as we can pass it directly to the CU task.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
Local ssb detection (#171)
PSS/SSS detection in the UE code uses a global structure that describe
the cell parameters it is not logical because we will populate this
information from what we read in the MIB and the SIB1 later it also
blocks us to make a cell scanner in a logical manner, just calling pss
detection function with a set of detection parameters and a buffer of
digital radio
this MR fixes this and simplifies the code.
Reviewed-by: Sakthivel Velumani <s.velumani@northeastern.edu>
UE DL Channel compensation and LLR optimization (#142)
- Common channel compensation: Extracted nr_channel_compensation() as a
shared function (AVX2/MRC inline) used by both gNB and UE PDSCH/PUSCH
paths, eliminating duplicated compensation logic.
- Common ML/MMSE 2-layer MIMO: Created shared ML LLR and MMSE routines
for 2-layer spatial multiplexing, shared between gNB and UE
demodulation.
- File segregation: Moved common LLR and channel compensation functions
into dedicated source files for better modularity and reuse.
- nr_dlsim -E flag: Added a command-line argument to enable/disable
ML/MMSE equalization at runtime (MMSE default), enabling side-by-side
performance comparison without recompilation.
- Persistent UE buffer allocation: pdsch_dl_ch_estimates, rxdataF_comp,
dl_ch_mag,dl_ch_magb,dl_ch_magr and rho_dl are now allocated once in
PHY_VARS_NR_UE on first use, resize automatically when dimensions
change (resizeAllowed=true), and are freed at UE teardown in
term_nr_ue_signal. Eliminates per-slot heap allocation/free overhead.
- PTRS restricted to one antenna port: PTRS phase tracking is only valid
for a single port; processing now correctly restricted to avoid
unnecessary computation.
- Reduce rxdataF_comp buffer size: MRC combines all Rx antenna
contributions into a per-layer output, so the first dimension was
reduced from Nl * nb_antennas_rx to Nl, reducing buffer footprint
proportionally.
- 256QAM segfault fix: nr_256qam_llr() on gNB side used a VLA for LLRs
without alignment, causing AVX2 load faults. Fixed with
attribute((aligned(32))).
Reviewed-by: Sakthivel Velumani <s.velumani@northeastern.edu>
nr_channel_estimation() produces combined output of all antenna ports for each layer. So reduced the buffer size from no. of layer times no. of antennas to no. of layers.
Signed-off-by: Rupanjali <rupanjali.srivastava@openairinterface.org>
- make internal links relative where applicable
- delete link to the wiki, as the documentation is in the main repo, not
the wiki
- remove some "example in oai code" as the examples either don't exist,
or are not in that place, and we can reasonably expect people to grep
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
wrong usage of extern declaration in C files, and header cleanup
wrong usage of extern declaration in C files, instead of headers, and
fix hidden bugs by this wrong extern declaration duplication.
this fixes pure bugs because we cast variables to the wrong types, even
if the bug has no consequence
it also separates 4G/5G include files, as the error is also in the
headers mixup. Clean up header includes in F1AP.
Further remove some "CU/DUuniqinstances".
Reviewed-by: Bartosz Podrygajlo <bartosz.podrygajlo@openairinterface.org>
This commit addresses several architectural issues in vrtsim, specifically
regarding how peer antenna counts are managed. Some additional changes
were made to allow unit tests and several fixes were delivered.
Key changes:
1. Refactored peer antenna management:
- Removed the redundant 'peer_info_t' structure which overlapped with
ue_config and client_info.
- Added explicit 'peer_tx_ant' and 'peer_rx_ant' fields to vrtsim_state_t.
- Server now pulls peer info from the UE config, while Client pulls it from
the GNB info published by the server.
2. Refactor taps_client to be thread safe
3. Fixed a bug where only the first antenna IQ was read from underlying SHM
mechanism
4. Test / usability related changes:
- Added support for configurable SHM channel names via
'--vrtsim.shm_channel_name' to prevent IPC conflicts between tests.
- Reduced the sleep() calls inside the code to reduce test runtime and
speedup vrtsim connection initialization and cleanup
- Added a unit tests for:
+ transparent channel mode
+ taps_client mode
+ cirdb mode
Signed-off-by: Bartosz Podrygajlo <bartosz.podrygajlo@openairinterface.org> and assisted by Gemini
1. Beam index:
So far rxdataF and txdataF had beam number as first index to separate freq
domain samples of antenna ports when mulitple beams are served in a slot.
This works fine when the beam is applied to entire symbol / slot but not
suitable for digital beamforming. This commit removes the beam number index
from the data buffers and the buffers hold all the antenna ports in the first
dimension. The antenna port index to be used for each UE in a MU-MIMO or
concurrent analog beam scenario is signalled by L2 via dedicated fields in
the FAPI PDU.
2. Beam ID array changes:
- The gNB and RU struct has a 2D array to hold beam IDs for all symbols in a
frame and all antenna ports. The order is now changes to first: symbol index
and second: antenna port because the RU access all ports' beam IDs for each
symbol / slot at a time. This ordering is optimal for maximizing cache hits.
- The type is now changed from int to uint16_t. So far the L1 had to deduce
the beam number index (done in beam_index_allocation()) from the beam ID sent
by L2 and this is done by initializing the beam_id array to -1 which denoted
no beam id set. Now since L2 explicitely passes the port indices, there is no
need for denoting if a beam id is assigned or not to a slot in L1. And
uint16_t matches with the type used by FAPI to carry beam ID. Since the type
is uin16_t, beam_id array is initialized to 0 and the check for -1 when
passing beam ID 0 to xran is also removed.
3. Number of antenna ports:
The removal of beam index from rxdataF and txdataF buffers resulted in all
antenna ports to be in a single vector. The number of logical ports passed to
L1 is not N1*N2*XP*number of beams per period. The buffer initialization
function is updated to reflect this change. DAS reference config file is
updated in accordance with these changes.
4. oaioran.c:
Remove the check of beam ID = -1 for setting nPrbElm = 0. This is a hack for
LiteON to supposedely improve performance according to Mario. If it's true
then a proper fix would be to pass allocation information from L1 and use it
instead of beam ID.
Signed-off-by: Sakthivel Velumani <s.velumani@northeastern.edu>
1. Added new config parameter to read RU port indices to be used via config
file.
2. When scheduling multiple UE on different beams in a slot, the antenna ports
used for each UE is explicitely passed down via fapi.
3. Fix dig_bf_interface_list by filling beam id for all logical ports used.
Signed-off-by: Sakthivel Velumani <s.velumani@northeastern.edu>
Preparatory changes for LDPC CUDA integration
This changeset is preparatory work to merge LDPC CUDA offload in !4097.
Notably, it includes changes for cleaning up some defines, updates to
the LDPC interface, and CI specific changes to prepare for the other MR.
The goal is to merge changes in a first step such that, in a second
step, only the actual LDPC CUDA implementation and ancillary changes
(CUDA memory allocations, CI files, ...) need to be merged.
Reviewed-By: Bartosz Podrygajlo <bartosz.podrygajlo@openairinterface.org>
A later commit (for LDPC CUDA) will reuse add_physim_test(), but is in a
completely different location than the existing physim test definitions.
Hence, move it to global scope for later reuse.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
Rename this option from CUDA_ENABLE, as it might otherwise conflict with
another option for LDPC CUDA. For consistency, name the compile option
to CHANNEL_SIM_CUDA. This way, the LDPC CUDA option (ENABLE_LDPC_CUDA
and compile definition LDPC_CUDA) will not conflict, and the
corresponding code is clearly "labelled".
A further change is the use of target_sources to simplify the definition
of the channel_pipeline executable.
Signed-off-by: Robert Schmidt <robert.schmidt@openairinterface.org>
NR_MAC/PHY: align SRS time_start_position semantics between L2 and L1
Problem
nfapi_nr_srs_pdu_t::time_start_position was being interpreted
inconsistently between OAI MAC (the producer) and the various L1
consumers:
- MAC populated the field with SRS-ResourceMapping.startPosition from
the RRC config — the value 3GPP defines as an offset counted backwards
from the end of the slot (TS 38.331).
- L1 consumers (openair1/PHY/NR_TRANSPORT/srs_rx.c,
openair1/SCHED_NR/phy_procedures_nr_gNB.c) and the MAC's own VRB-map
consumer (gNB_scheduler_srs.c) all re-derived the absolute symbol
index with NR_SYMBOLS_PER_SLOT - 1 - time_start_position.
This works for OAI's in-tree soft L1 (every consumer applies the same
inversion), but external FAPI L1s (NVIDIA Aerial / cuBB) treat
time_start_position as the absolute symbol index per the SCF FAPI 222
PDU semantics. The MAC and external L1 therefore disagreed by one symbol
inversion. Net effect: SRS was scheduled on the wrong OFDM symbol, the
RU sampled noise instead of the SRS, and the channel-matrix returned in
SRS.indication was all zeros for almost every occasion. Fix
Move the inversion to the producer side. MAC populates the PDU with the
absolute symbol index (NR_SYMBOLS_PER_SLOT - 1 - startPosition), and
every consumer uses srs_pdu->time_start_position directly as l0.
The two semantics are mathematically equivalent for OAI's soft L1 (l0 =
NR_SYMBOLS_PER_SLOT - 1 - startPosition either way), so the symbol index
used internally is unchanged on non-Aerial builds. External L1 consumers
now receive the value they expect. Testing status
- [x] OAI L1 + L2
- [x] Aerial L1 in CAT-B / mMIMO mode
- [x] Aerial L1 in CAT-A mode
- [x] RFsim / nrUE end-to-end loopback:
Reviewed-By: Rúben Soares Silva <rsilva@allbesmart.pt>
changes in cudaMemadvise/cudaPrefetch API for v13.
Build fails on DGX spark or any Blackwell GPU target without this fix.
It just checks the CUDA version in a couple of places where GPU offload
is used for channel simulation (nr_dlsim,nr_ulsim and
channel_pipeline.cu) and adapts the usage of the CUDA API accordingly.
Reviewed-By: Bartosz Podrygajlo <bartosz.podrygajlo@openairinterface.org>
Refactor DLSCH/ULSCH scheduler: extract proportional fair policy behind function pointer interface
Motivation
The current `pf_dl()` is a ~600-line monolithic function that mixes
infrastructure concerns (UE iteration, HARQ management, CCE allocation,
MAC PDU generation) with the scheduling policy (PF priority, RB
allocation, MCS selection). This makes it hard to modify the scheduling
strategy, test alternatives, or offload scheduling to a GPU (cuMAC).
This MR refactors the DL and UL schedulers into a clean separation
between **infrastructure** and **policy**, using function pointers for
beam allocation and scheduling decisions. The data structures
(`nr_dl_candidate_t`, `nr_dl_sched_params_t`) are also designed to map
directly to cuMAC's
`cumacCellGrpUeStatus`/`cumacSchdSol`/`CumacCellGrpPrms`, to make future
integration smooth.
Changes
Goodput tracking fix
`dl_thr_ue` now tracks actual goodput in bps (EWMA of SDU byte deltas
per frame) instead of accumulating raw byte counts per slot, which are
not as straightforward to interpret (due to variations in TDD patterns
mostly). The new calculation matches closely with the throughput
measured with e.g. iperf.
Helper extraction from `pf_dl()` and `post_process_dlsch()`
- `find_first_available_rbs()` — first-fit contiguous RB allocation
- `setup_dl_harq_process()` — HARQ process management
- `generate_dl_mac_pdu()` — MAC CE + RLC data + padding
- `fill_dl_tx_request()` — FAPI TX_req filling
New scheduling interface
- **`nr_dl_candidate_t`** — per-UE flat struct with all immutable inputs
(buffer status, BLER, MCS limits, beam, BWP) and outputs (scheduled,
rbStart, rbSize, MCS). Kept relatively minimal for now, but it should
be easy to add more input metrics in the future.
- **`nr_dl_sched_params_t`** — per-beam cell-level context (VRB map,
available RBs, slot bitmap)
Function pointers (DL)
| Pointer | Default implementation | Role |
|---------|----------------------|------|
| `dl_ri_pmi_select` | `nr_dl_ri_pmi_select_default` | Rank/PMI selection |
| `dl_beam_select` | `nr_dl_beam_select_default` | Beam direction assignment |
| `dl_tda_select` | `nr_dl_tda_select_default` | Time-domain allocation |
| `dl_mcs_select` | `nr_dl_mcs_select_default` | MCS from BLER/SINR |
| `dl_rb_alloc` | `nr_dl_proportional_fair` | PRB allocation (PF policy) |
Function pointers (UL)
| Pointer | Default implementation | Role |
|---------|----------------------|------|
| `ul_ri_tpmi_select` | `nr_ul_ri_tpmi_select_default` | Rank/TPMI from SRS feedback |
| `ul_beam_select` | `nr_ul_beam_select_default` | Beam direction assignment |
| `ul_tda_select` | `nr_ul_tda_select_default` | Time-domain allocation |
| `ul_mcs_select` | `nr_ul_mcs_select_default` | MCS from BLER/SINR |
| `ul_rb_alloc` | `nr_ul_proportional_fair` | PRB allocation (retx first, then PF-sorted new-tx) |
All default implementations are in `gNB_scheduler_dlsch.c` (DL) and
`gNB_scheduler_ulsch.c` (UL).
MCS selection flow
The old `get_mcs_from_bler()` entangled two concerns: updating the BLER
estimate from HARQ feedback and deciding the MCS. These are now split:
- **BLER tracking is infrastructure's job**: `collect_dl_candidates()`
calls `update_dl_bler_stats()` which updates the BLER estimate from
HARQ round statistics.
- **MCS selection is the policy's job**: the proportional fair policy
calls `select_mcs_from_bler()` internally to adapt MCS based on the
BLER value. A different policy could use an entirely different MCS
strategy (e.g. cuMAC has its own `mcsSelectionLUT` + OLLA, one could
decide to opportunistically lower the MCS while increasing the PRB
allocation for reliability in some cases, etc).
For retransmissions, MCS/number of PRBs are passed as hints so the
policy can use them as-is if desired, but we don't enforce it (adaptive
HARQ possible too: the standard requires us to maintain TBS but in
theory it could be achieved via changing the MCS and number of RBs if we
wanted to).
Refactored flow
`nr_dl_schedule()` (formerly `pf_dl()`):
```
collect_dl_candidates() → build candidate array from UE list
schedule_dl_ues() → beam alloc + per-beam policy calls
for each scheduled candidate → CCE/PUCCH/TBS validation + post_process
```
Beam allocation and scheduling policy are two separate function
pointers, allowing each to be developed and tested independently (with
the goal in the future to add a parameter in the config file for each,
and telnet commands to hotswap).
`schedule_dl_ues()` wraps both into a single function: it first calls
beam allocation to assign candidates to beams, then iterates over beams
and calls the scheduling policy for each one. cuMAC performs joint beam
+ PRB allocation on the GPU, so when integrating later it will replace
`schedule_dl_ues()`.
Future work
- Channel matrix H from SRS on candidates for beam-aware scheduling
- Per-RB channel magnitude derived from SRS on candidates
- Config file parameters and telnet commands for hotswapping policies
- cuMAC integration via `schedule_dl_ues()` replacement
Reviewed-By: Robert Schmidt <robert.schmidt@openairinterface.org>
3GPP RRC SRS-ResourceMapping.startPosition is defined as an offset
counted backwards from the end of the slot (TS 38.331). OAI MAC was
previously passing the raw RRC value straight into the FAPI SRS PDU's
time_start_position field, and every L1 consumer was then re-deriving
the absolute symbol via NR_SYMBOLS_PER_SLOT - 1 - time_start_position.
External L1 implementations (e.g. NVIDIA Aerial / cuBB) interpret
time_start_position as the absolute symbol index where the SRS lives,
which left the MAC and external L1 disagreeing by one inversion. The
result was SRS being scheduled on the wrong OFDM symbol and the
returned channel matrix being all-zero for most occasions.
Make MAC populate time_start_position with the absolute symbol index
(NR_SYMBOLS_PER_SLOT - 1 - startPosition) and remove the matching
inversion from the L1 consumers so the value flows through unchanged.
The UE side and the shared SRS generator are harmonized to the same
absolute-symbol convention so generate_srs_nr can read nr_srs_info->
l_offset directly regardless of caller (gNB or UE).
* gNB MAC: gNB_scheduler_srs.c invert at PDU-build time; consume the
already-inverted value when computing l0 for the VRB mask.
* gNB PHY: srs_rx.c (fill + get_signal) and SCHED_NR/phy_procedures_
nr_gNB.c use srs_pdu->time_start_position directly as l0.
* UE MAC: nr_ue_scheduler.c invert at PDU-build time.
* UE PHY: phy_procedures_nr_ue.c consumes srs_config_pdu->time_start_
position directly as l0.
* Shared SRS generator: nr_phy_common_srs.c reads nr_srs_info->l_offset
as the absolute symbol index, matching both gNB and UE call sites.
* Simulators: srssim.c and ulsim.c populate both the gNB-side and
UE-side SRS PDUs with the absolute symbol index so the nr_srssim
and nr_ulsim unit tests pass with the new convention.
Signed-off-by: Gabriele Gemmi <g.gemmi@northeastern.edu>
even if 2 codewords are currently not supported and not tested, the way UE handled the possibility of having 2 codewords was messy, in between support and non-support
the intention of this MR is to cleanup the code more than extend the support
Signed-off-by: Francesco Mani <email@francescomani.it>
!4083 improved encoding time in some nr_dlsim tests running in
RAN-PhySim-GraceHopper-5G pipeline. Update the timing thresholds
accordingly.
Signed-off-by: Jaroslava Fiedlerova <jaroslava.fiedlerova@openairinterface.org>
Fix Wt (orthogonal code) multiplication for PUSCH DMRS length 2 and
noise power estimation
This MR:
- Introduce PUSCH DMRS length as a configurable parameter in nr_ulsim.
- Fix bugs related to the use of PUSCH DMRS length 2.
- Fix the noise power estimation.
Note:
- Still cannot validate PUSCH DMRS length 4, since we need more than 5
layers (DMRS type 1) / 6 layers (DMRS type 2) to test the orthogonal
spreading over time.
- Yet to implement time domain averaging for PUSCH DMRS length 2 to
separate orthogonally over time.
- The main bug here is the noise power estimation (removes the duplicate
division with the number of antennas).
Reviewed-by: Roberto Louro Magueta <rmagueta@allbesmart.pt>
1. Refactor channel estimation and freq domain data extraction functions
to take rxdataF of only one symbol.
2. Split the PBCH and PSBCH decoding into two parts. Fisrt part is
called for every symbol and generates LLRs of each symbol. Second
part is called in the last PBCH/PSBCH symbol to decode and send the
payload to MAC.
3. In channel estimation, the loop around rx antennas is taken out to
make the code more modular.
Signed-off-by: Sakthivel Velumani <s.velumani@northeastern.edu>
ctest: fix tests when compiling with sanitizers, ignore build directories,
fix spsc_q build
- Fix the directory to the current build directory to enable running with
sanitizers with ctest
- add build* to .gitignore to close#1074
- fix spsc_q compilation on non-C++23 compilers to close#1073
When compiling with sanitizers that require LD_LIBRARY_PATH (see
documentation), ctest could not drive tests as the executables did not
find shared objects. Fix the directory to the current build directory to
enable running with sanitizers.
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
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>
