More complex implementations will be achieved using the new reworked interface
- t_nrLDPC_dec_params: remove F, Qm, rv and perCB
- encoder_implemparams_t: remove Qm, Tbslbrm, G, rv and perCB
- Remove all variables (like OPENAIR1), they serve no purpose and make
unnecessarily long lines
- Remove linking to T, it does not seem to be necessary
- Avoid intermediate variables to keep sources, and declare libraries
directly, making for shorter code
- removed some whitespace noise.
- cleaned up CMakeLists.txt
- reintegrated task_manager.c into thread-pool.c
- cleaned up some unnecessary code
Co-authored-by: Cedric Roux <cedric.roux@eurecom.fr>
Co-authored-by: Mikel Irazabal <mikel.irazabal@openairinterface.com>
FAPI P7 Refactoring and unitary test creation
This MR is a follow-up to the work done in !2714 this time separating
the P7 FAPI messages into its own library (nr_fapi_p7).
In the same way, utility functions to copy, free and compare have been
added for each message, as well as an addition function to get the
allocated size for a message, which is used for the messages that can be
too big to fit in a buffer that's allocated in the stack.
Unitary tests have been added for each of the messages, testing
pack/unpack and the utility functions added.
Fix error reported by the T2 card when running nr_dlsim with 4 DL layers.
Error can be reproduced by running:
./nr_dlsim -n100 -s30 -e25 -b273 -R273 -x4 -z4 -y4 -P --ldpc_offload.dpdk_dev 41:00.0 -c
So far we set NR_LDPC_MAX_NUM_CB to 72, which is sufficient for max 2 DL layers.
Set max possible number of LDPC segments based on the max number of layers
and max number of DLSCH segments (currently it is 144).
Set SA mode as default
- Set standalone (SA) mode as the default.
- Adjust commands in Docker Compose files by removing --sa flags and
adding --nsa flags where applicable.
- Add a function to verify that only one mode is being used at a time.
- Update all relevant documentation.
the version in develop is difficult to understand, and surprizing design
it loops only when avx2mod == 0 because when this test is false, the
size is always small i also spent time to understand, but i think i have
not changed it by not looping over the array.
Use non-recursive polar decoder after traversing the tree once. Maximum number
of nodes is set to 600 (unsure of the maximum). This is more performant
than the recursive version.
- Ensure packed struct through bitfield
- decoder tree buffer increase: i don't find evidence in the 3GPP
standard of the max size of polar coded PDU. even with this max size,
the way this tree is built is not trivial. so, it is hard to predict,
i just put 8 system pages (32KB) that looks enough [1]
- a lot of reformatting
[1] https://arxiv.org/pdf/1804.04389
according to this paper, the largest polar code in 5G is in UL, so
likely the biggest problem will be in gNB, not the UE. in DL, the max
output coded block is 512, in UL 1024. probably the tree we discuss is
related to this size.
"frozen bits" means only bits we don't encode/decode because polar code
size is only by power of 2.
We see this in OAI code: NR_POLAR_PUCCH_N_MAX is 10, so 2^10 is 1024,
that is the maximum size for 5G polar code length. in the code, we
assert if N > 512: "N = %d, not done yet\n",
polarParams->N". so, the today OAI code can't do PUCCH large blocks.
using current oai limitation of 2⁹ max code size, we can create up to
2^9+2⁸... nodes, so 1023 nodes in the tree. the problem then is that
each node have variable size, depending on the 'level' (the power of 2
we are currently). the algo provision large alpha and beta to be able
to store all lower levels in the tree (int leaf_sz = 1 << level)
therefore upper tree nodes are large (the top level for n=9 uses about
1.5KB) and the bottom nodes uses 64 bytes. if all nodes are populated,
the size will be very large: 1.5KB + 2 * 800 + .... but, we create
nodes in the tree only for the "information bits", so the payload, that
is limited to 128 bits (else we assert, the code is not "finished").
1.5KB * 128 is >> 8KB i allocated. but, we can't create more than one
mode at top level, so there is maximum 1.5KB + 2 * 800, + 4 * 400 + 8 *
200 + 16 * 100 + remaining of 128-32 nodes of 64 bytes. about 14KB.
i allocated 1024*sizeof(__m256i) bytes = 32KB that looks good.
See getopt(3): returns int, so store it in an int variable.
I went through all occurrences, project-wide. All return values of
getopt() should be stored in ints now.
Before this commit, the following will trigger a complete rebuild:
- create another git name (git checkout -b test)
- run cmake again manually (cmake ..)
- run ninja again
The top-level CMakeLists.txt uses add_definitions() to add git
information to all targets; if some git information changes, this will
mark all command lines as dirty, triggering the build.
There is an easy solution: instead of passing this information on the
command line, pass it in a header. This commit uses configure_file() to
generate a file oai_version.h in the build directory with the same
information. Various sources files, e.g., main executables, use this to
output the same information as before. It is then also not necessary to
use header guards, as cmake takes care to create this file.
If cmake is re-run, it will update the header, and trigger only files
including this header will be rebuilt (which can be shown with the above
stpes). The information passed is the same as before, and updates the
information when the old method would have updated it.
- Introduced command line and configuration options to set up PCI address, CPU cores used by DPDK, and DPDK file prefix for T2 offload.
- Made the dpdk_dev parameter (for setting the PCI address of the T2) mandatory.
- Configure dpdk_dev in the CI test cases.
- Replaced bbdev TEST_ASSERTs with the OAI macro AssertFatal() to harmonize error handling.
- Removed unused functions and macros from nrLDPC_offload.h.
- Reworked nrLDPC_decoder_offload.c to process all segments in a TB with a single call to LDPCdecoder() or LDPCencoder().
- Perform CRC checks on the T2.
- Modified nr_ulsch_decoding (gNB decoder) to:
- Count processed segments.
- Properly perform CRC checks.
- Created a constant NR_LDPC_MAX_NUM_CB to define the maximum number of codeblocks.
- Conditional library loading:
- On gNB: Load ldpc_t2 library if offload is enabled using the --ldpc-offload-enable flag.
- On UE: Load both ldpc_t2 and ldpc libraries since only encoder offload is supported.
- General cleanup of nrLDPC_decoder_offload.c for improved readability and maintainability.
- Modified the structure of LDPC encoder/decoder/offload parameters:
- Introduced a structure for code block (CB) related parameters.
- Removed parameter E from the encoder and offload parameter structures for clarity.
- Replaced E with perCB->E_cb in the UE encoder code.
We access G_N at indexes that are not necessarily 32-byte aligned,
leading to segfaults. Use simde_mm256_loadu_si256() to load from memory
at an unaligned memory address to avoid the segfault.
Discovered using clang; it is unclear why gcc did not encounter the same
segfault.
Romain Beurdouche