Improvements for LDPC encoding
AVX512 modifications for LDPC encoding: interleaving, ZC384 BG1, output
formatting for encoder, segmentation (memcpy instead of loop). some
improvement in TX for aarch64 in same places where AVX512 support was
added. Also, rate matching and interleaving are done on bytes containing
8 segments and reformatting of the output is done at the end of segment
processing instead of after ldpc encoding.
This improves the overall performance of the NR DL transmitter in gNB
See the merge request description for numbers.
- modify retrieve_ldpc_enc_op() to store encoded results in a packed format
- add timing measurements for T2 encoder
- add simde implementation for bit reversal
- Introduce new bit reversal implementation using simde_mm_gf2p8affine_epi64_epi8
instruction, based on https://wunkolo.github.io/post/2020/11/gf2p8affineqb-bit-reversal/.
This implementation requires GFNI and AVX512VL support on the target machine.
1. The encoder writes the output in sequence after the encoding tasks completed instead of writing within the tasks.
2. Simplify the output writing to make it more readable.
Co-authored-by: Laurent Thomas <laurent.thomas@open-cells.com>
There was a risk of a race condition between encoding tasks on the output array.
It could be that two task compete in writing a word at the border of their respective segment groups.
This commit offers to solve this issue by retaining the tail word in a struct that is returned after task completion so that segment group border can be fixed after completion of all the tasks.
Comes with additional cleanups of nrLDPC_coding_segment_encoder:
1. Move the section of code dedicated to writing to the encoding library output to a separate function.
2. Remove bit interleave debug in AVX512VBMI case of output writing for improving code readability.
1. nrLDPC_coding_segment_encoding: Clarifying encoder parameters
The way the encoder implementation params was used could be made clearer:
* `macro_num` was always used in order to calculate segment indexes.
It is clearer to pass directly the first segment index.
* The initialisation of a common implementation parameters structure can be made clearer.
2. nrLDPC_coding_segment_encoder: Remove multi segment dimension in encoder output
The output of `LDPCencoder` was a 2D unsigned char array.
But all the segments were actually packed in the first segment, 8 bits of different segments per byte of the first segment.
So the first dimension was useless. This commit removes this dimension.
3. nrLDPC_coding_segment_encoder: Rename `nrLDPC_prepare_TB_encoding` into `nrLDPC_launch_TB_encoding`
4. nrLDPC_coding_segment_encoder: Remove the useless second rate matching for E2, one rate matching with max(E, E2) does the job.
5. nr_interleaving_ldpc: cleanup
Rewrite code to use different SIMD sizes in a linear fashion.
This means cascading preprocessor directives are used to start from a high SIMD size an decrease it progressively.
6. ldpc_encoder_optim8segmulti: linear usage of SIMD
Use cascading preprocessor directives to include different SIMD sizes and types depending on which are available on the system.
The different loops use a common index that is progressively incremented.
Also add AVX512 version of bits distribution from 8 segments into one byte.
Co-authored-by: Laurent Thomas <laurent.thomas@open-cells.com>
Since this branch aims to bring all the encoded segments packed and contiguous,
it doesn't make anymore sense to have one output array per segment.
Therefore this commit moves the output array to the TB instead of the segment and adapts its size.
TODO: Adapt the size of f where it is relevant.
1. nrLDPC_coding_segment_encoder: pointer not aligned on 64 bits
An undefined behavior was happening because of a shift on a __m64 which
was actually a uint32_t which was not aligned on 64 bits.
This commit suggests a fix by properly loading into a __m64 with _mm_set_pi32
before performing the shift.
2. nr_rate_matching: Filler bits detection
Previously, if the bit selection was reaching Ncb and restarting selection from the beginning of encoded data, the filler bits were detected by testing the bytes again NR_NULL which was equal to 2.
But on this branch the bits of up to 8 segments are stored in one byte while doing rate matching so that one byte could be tested to be a filler while it is not.
So this commit changes the way to test filler bits.
Instead of relying on NR_NULL, bit selection now relies on the offset and size of filler bits.
3. nr_ulsch_coding: missing filling f with zeros
The array `f` of the UE UL HARQ process was not filled with zeros between transmissions.
This is a problem if the encoder does not properly overwrite its output.
In order to avoid such situation, this commit adds a memset to fill `f` with zeros before passing it to the encoder.
4. nrLDPC_coding_segment_encoder: missing init of f to zeros
In the encoder, `f` is not filled with zeros before being used.
`f` is instead somehow filled with arbitrary values.
When writing to `output` by 64 bit vectors, some arbitrary bits may be added after the segment.
Then the arbitrary added ones are not overwritten when writing the next segment because of writing with a logical or.
This commit properly fills `f` and `f2` with zeros.
5. nr_ulschsim & nr_dlschsim: Adapt modulation to new encoder output
The QPSK symbols that are feed to the channel model of nr_dlschsim were determined based on the bytes of the encoder output.
With the new output of the encoder, it had to be reworked to determine the QPSK symbols based on the bits of the encoder output.
1. 3GPP Rate-matching/interleaving on 8-segment blocks, deinterleaving at output of multi-segment processing.
2. Bit-packing done in nrLDPC_coding_segment_encoder.c now instead of nr_scrambling.
Interestingly, it improves performance significantly too.
Tested for AVX512 only for now.
Avoid the use of simde_mm_sign_epi8 which is inefficient on aarch64.
Decoding times are significantly improved on aarch64.
Co-authored-by: Romain Beurdouche <romain.beurdouche@eurecom.fr>
Add --allowerasing, which will downgrades some packets in order to
satisfy LLVM dependencies.
When building the docker image for clang, CI fails with
Error:
Problem: llvm-18.1.8-3.el9.i686 from ubi-9-appstream-rpms does not belong to a distupgrade repository
- package llvm-toolset-18.1.8-3.el9.x86_64 from ubi-9-appstream-rpms requires llvm = 18.1.8, but none of the providers can be installed
- package llvm-18.1.8-3.el9.x86_64 from ubi-9-appstream-rpms requires llvm-libs(x86-64) = 18.1.8-3.el9, but none of the providers can be installed
- package llvm-18.1.8-3.el9.x86_64 from ubi-9-appstream-rpms requires libLLVM.so.18.1()(64bit), but none of the providers can be installed
- package llvm-18.1.8-3.el9.x86_64 from ubi-9-appstream-rpms requires libLLVM.so.18.1(LLVM_18.1)(64bit), but none of the providers can be installed
- cannot install both llvm-libs-18.1.8-3.el9.x86_64 from ubi-9-appstream-rpms and llvm-libs-19.1.7-2.el9.x86_64 from @System
- package mesa-dri-drivers-24.2.8-2.el9_6.x86_64 from @System requires libLLVM.so.19.1()(64bit), but none of the providers can be installed
- package mesa-dri-drivers-24.2.8-2.el9_6.x86_64 from @System requires libLLVM.so.19.1(LLVM_19.1)(64bit), but none of the providers can be installed
- package mesa-libGL-24.2.8-2.el9_6.x86_64 from @System requires libgallium-24.2.8.so()(64bit), but none of the providers can be installed
- package mesa-libGL-24.2.8-2.el9_6.x86_64 from @System requires libgallium-24.2.8.so(libgallium-24.2.8.so)(64bit), but none of the providers can be installed
- package mesa-libGL-devel-24.2.8-2.el9_6.x86_64 from @System requires mesa-libGL(x86-64) = 24.2.8-2.el9_6, but none of the providers can be installed
- conflicting requests
- problem with installed package mesa-libGL-devel-24.2.8-2.el9_6.x86_64
(try to add '--allowerasing' to command line to replace conflicting packages or '--skip-broken' to skip uninstallable packages or '--nobest' to use not only best candidate packages)
So when trying to install llvm-toolset from docker/Dockerfile.clang.rhel9,
we have a conflict because this installs 18.1 from ubi-9-appstream-rpms,
but various mesa packets have llvm 19.1.
1. I don't understand why we have mesa installed (no graphics system?)
2. --allowerasing downgrades the relevent mesa package(s)
Alternatively, we would need to select 19.1 when installing
llvm-toolset, but I don't know how to do that.
Integration: `2025.w19`
Closes#852, #952, and #957
See merge request oai/openairinterface5g!3427
* !3409 Correctly handle minimum RB condition in MAC
* !3421 Fix DMRS for PUCCH format 2
* !3424 limit the number of HARQ processes in case of DCI00 and 10
* !3116 ue txData more contextual, but still global as it is entangled with usrp driver
* !3408 Handling 2 search spaces per slot in SIB1
* !3419 PRACH configuration index warning
* !3429 CI: Add `-l` option for the UL iperf test
* !3381 Add 5GMM Authentication Failure enc/dec lib and unit test
CI: Add `-l` option for the UL iperf test
In this MR, we set the packet length to 1428 bytes for UL iperf3 testing using
the -l option to prevent packet fragmentation. Fragmentation is causing packet
drops between the gNB and the UPF, leading to failed UL throughput tests in CI.
Add 5GMM Authentication Failure enc/dec lib and unit test
This MR is introducing the 5GMM Authentication Failure message library as per
8.2.4 of 3GPP TS 24.501. This message is sent from UE to network under certain
conditions. The message is also introduced in stack and is send if the received
ngKSI already in use.
- add new library for 5GMM Authentication Failure enc/dec and unit test
- introduce handler and generate function for NAS Authentication Failure
store ngKSI
- send Authentication Failure if ngKSI is already associated with one of
the 5G security contexts stored in the UE
Closes#852
ue txData more contextual, but still global as it is entangled with usrp driver
- txData buffer is global and entangled even in usrp oai driver because of tx
thread feature
- it also makes weird combination with continuous tx usrp bug workaround
- this MR makes most of the calls contextual but keep the globakl txData
- later a MR should remove the global and tackle it with proper context, but we
need to modify gNb, UE, 4G and probably all oai drivers
Fix DMRS for PUCCH format 2
This MR fixes the DMRS for PUCCH format 2.
While at the OAI-UE we have:
uint16_t startingPRB = pucch_pdu->prb_start + pucch_pdu->bwp_start;
uint idxGold = startingPRB >> 2;
uint32_t *seq = gold_cache(temp_x2, idxGold + pucch_pdu->prb_size);
At gNB side, it was missing pucch_pdu->bwp_start, i.e., it was:
uint32_t *sGold = gold_cache(x2, pucch_pdu->prb_start / 4 + ngroup / 2);
uint8_t *sGold8 = (uint8_t *)(sGold + pucch_pdu->prb_start / 4);
This MR was tested with COTS UE with a bwp_start other than 0, and now PUCCH format 2 is being decoded correctly.
Move this message down, at the end of the function, if no code path
exists under which msg3 allocation might fail despite this log having
been printed.
Assume the minimum grant (min_rb) size to be the bandwidth part size. As
an example: rbStart 0, min_rb=bwpSize=106. This should pass, hence
modify these two conditions.
Co-authored-by: Robert Schmidt <robert.schmidt@openairinterface.org>
Integration: `2025.w18`
Closes#953
See merge request oai/openairinterface5g!3413
* !3385 Add enc/dec lib for E1 Bearer Context Setup Failure
* !3410 Avoid segmentation fault when testing trigger reestablishment and there is no DRB[0] established
* !3411 Remove assert for SRI-PUSCH-PowerControl
* !3291 Improvements in handling UE context in gNB DU
* !3416 Signal gNB-DU config update CellStatus from DU to CU
* patch to fix memory leak on DRB_ToAddModList
* update phytest-timing thresholds
Signal gNB-DU config update CellStatus from DU to CU
Add the cell status to gNB-DU configuration update message, including
test. Use the gNB-DU configuration update to signal the cell status in
the stack.
This IE was not included in a previous refactoring round, but is at
least necessary to be decoded when interoperating with the OSC O-DU.
Robert Schmidt