44875e1c580e05fbbfb6dfa9de13a36da0fbef42
This commit introduces a new, configurable benchmark, `test_channel_scalability`, designed to measure the performance and scalability of the CUDA channel simulation pipeline under various conditions. This benchmark provides several key features: - It supports multiple GPU execution models, including a high-throughput "batch" mode, an asynchronous "stream" mode, and a "serial" mode for baseline measurements. - It includes an optional interference simulation mode, activated by the `--sum-outputs` flag, which sums the outputs of multiple parallel channels. - Simulation parameters (MIMO configuration, signal length, channel length, and batch size) are configurable via command-line arguments. - It runs the full simulation pipeline on both the CPU and the GPU, reporting the average execution time for each and calculating the overall speedup.
Revert "CI: upgrade RAN-SA-Handover-CN5G F1 handover test to test Reestablishment on a different DU"
Merge remote-tracking branch 'origin/prevent-p7-msg-after-stop-indication' into integration_2026_w05 (!3880)
OpenAirInterface License
It is distributed under OAI Public License V1.1.
The license information is distributed under LICENSE file in the same directory.
Please see NOTICE file for third party software that is included in the sources.
Where to Start
- General overview of documentation
- The implemented features
- System Requirements for Using OAI Stack
- How to build
- How to run the modems
Not all information is available in a central place, and information for specific sub-systems might be available in the corresponding sub-directories. To find all READMEs, this command might be handy:
find . -iname "readme*"
RAN repository structure
The OpenAirInterface (OAI) software is composed of the following parts:
openairinterface5g
├── charts
├── ci-scripts : Meta-scripts used by the OSA CI process. Contains also configuration files used day-to-day by CI.
├── CMakeLists.txt : Top-level CMakeLists.txt for building
├── cmake_targets : Build utilities to compile (simulation, emulation and real-time platforms), and generated build files.
├── common : Some common OAI utilities, some other tools can be found at openair2/UTILS.
├── doc : Documentation
├── docker : Dockerfiles to build for Ubuntu and RHEL
├── executables : Top-level executable source files (gNB, eNB, ...)
├── maketags : Script to generate emacs tags.
├── nfapi : (n)FAPI code for MAC-PHY interface
├── openair1 : Layer 1 (3GPP LTE Rel-10/12 PHY, NR Rel-15 PHY)
├── openair2 : Layer 2 (3GPP LTE Rel-10 MAC/RLC/PDCP/RRC/X2AP, LTE Rel-14 M2AP, NR Rel-15+ MAC/RLC/PDCP/SDAP/RRC/X2AP/F1AP/E1AP), E2AP
├── openair3 : Layer 3 (3GPP LTE Rel-10 S1AP/GTP, NR Rel-15 NGAP/GTP)
├── openshift : OpenShift helm charts for some deployment options of OAI
├── radio : Drivers for various radios such as USRP, AW2S, RFsim, 7.2 FHI, ...
├── targets : Some configuration files; only historical relevance, and might be deleted in the future
└── tools : Tools for use by the developers/ci machines: code analysis and formatting
How to get support from the OAI Community
You can ask your question on the mailing lists.
Your email should contain below information:
- A clear subject in your email.
- For all the queries there should be [Query] in the subject of the email and for problems there should be [Problem].
- In case of a problem, add a small description.
- Do not share any photos unless you want to share a diagram.
- OAI gNB/DU/CU/CU-CP/CU-UP configuration file in
.confformat only. - Logs of OAI gNB/DU/CU/CU-CP/CU-UP in
.logor.txtformat only. - In case your question is related to performance, include a small description of the machine (Operating System, Kernel version, CPU, RAM and networking card) and diagram of your testing environment.
- Known/open issues are present on GitLab, so keep checking.
Always remember a structured email will help us understand your issues quickly.
Languages
C
91.8%
CMake
2.9%
C++
1.9%
TeX
1%
Python
1%
Other
1.1%
