The present invention relates to a layered semi-parallel LDPC decoder system having a single permutation network, and belongs to the field of decoder hardware design. The system comprises a layered decoding architecture of the single permutation network, a layered semi-parallel decoding architecture of the single permutation network, a pipeline design for layered semi-parallel decoding and a hardware framework of a layered semi-parallel LDPC decoder. The present invention removes a permutation network module between a check node and a variable node by modifying the cyclic shift value of each information block transferred from the variable node to the check node, i.e., the cyclic shift operation of the decoder can be completed through the single permutation network so as to reduce hardware resources of the decoder. A semi-parallel decoding structure is adopted, and meanwhile, a pipeline is added between half layers. Compared with a decoder with a layered full-parallel structure, a decoder with a semi-parallel structure has the degree of parallelism of a variable node equal to only half of the code length but can achieve ¾ of the throughput as well as reduce hardware resources by half.

A bit interleaving method involves applying a bit permutation process to a QC LDPC codeword made up of N cyclic blocks each including Q bits, and dividing the codeword, after the bit permutation process, into a plurality of constellation words each imade up of M bits, the codeword being divided into N/M sections, each constellation word being associated with one of the N/M sections, and the bit permutation process being performed such that each of the constellation words includes one bit from each of M different cyclic blocks associated with a given section.

A decoding method and device are provided. The method includes: decoding grouped original data in parallel by a first decoding unit to obtain grouped decoded data; decoding merged grouped decoded data by a second decoding unit to obtain decoded data; and if the sum of the lengths of the decoded data is an integer multiple of an upper limit of the decoding times of the second decoding unit, updating the first decoding unit and the second decoding unit, and if the sum of the lengths of the decoded data is not an integer multiple of the upper limit of the decoding times of the second decoding unit, updating the second decoding unit to obtain the decoded data again, until the sum of the lengths of the decoded data is equal to a decoding length, and merging the decoded data to serve as a decoding result of the original data.

Systems and methods which provide parallel processing of multiple message bundles for a codeword undergoing a decoding process are described. Embodiments provide low-latency segmented quasi-cyclic low-density parity-check (QC-LDPC) decoder configurations in which decoding process tasks are allocated to different segments of the low-latency segmented QC-LDPC decoder for processing multiple bundles of messages in parallel. A segmented shifter of a low-latency segmented QC-LDPC decoder implementation may be configured to process multiple bundles of a plurality of edge paths in parallel. Multiple bundles of messages of a same check node cluster (CNC) are processed in parallel. Additionally, multiple bundles of messages of a plurality of CNCs are processed in parallel.

Systems and methods which provide parallel processing of multiple message bundles for a codeword undergoing a decoding process are described. Embodiments provide low-latency segmented quasi-cyclic low-density parity-check (QC-LDDC) decoder configurations in which decoding process tasks are allocated to different segments of the low-latency segmented QC-LDPC decoder for processing multiple bundles of messages in parallel. A segmented shifter of a low-latency segmented QC-LDPC decoder implementation may be configured to process multiple bundles of a plurality of edge paths in parallel. Multiple bundles of messages of a same check node cluster (CNC) are processed in parallel. Additionally, multiple bundles of messages of a plurality of CNCs are processed in parallel.

A decoding method includes a selection step of reading reception data from a storage unit in units of P words, of reproducing data based on a column weight of a P-column unit of a check matrix, of writing reproduced data into an intermediate value storage unit, and of reading data from as many applicable register files as a row weight on a row block-by-row block basis for row blocks generated by row-wise division of the check matrix; a first shifting step of shifting the data read; a parallel row operation step of performing a row operation in parallel on a word-by-word basis using data shifted; a second shifting step of shifting as many operational results as the row weight, obtained by the row operation, to undo the shifting; and a first update step of updating values in the intermediate value storage unit with operational results.

A low density parity check (LDPC) decoder initializing variable nodes with a value of a codeword and outputting the updated variable nodes as decoded messages with reference to an irregular parity check matrix. The LDPC decoder includes a plurality of unit logic circuits operating in a single mode in which all the unit logic circuits update one variable node group including at least one variable node, or a multi-mode in which each of the unit logic circuits updates a plurality of variable node groups in parallel by updating different variable nodes, and a mode controller controlling the plurality of unit logic circuits to update a high-degree variable node group having a degree greater than a threshold degree among the variable node groups in the single mode, and update a low-degree variable node group having a degree less than or equal to the threshold degree among the variable node groups in the multi-mode.

A memory system includes a memory device and a controller. The memory device includes a plurality of non-volatile memory groups individually storing a plurality of data segments, each data segment corresponding to a codeword. The controller is configured to perform hard decision decoding to correct an error when the error is included in a first data segment among the plurality of data segments, determine whether other data segments associated with the first data segment, among the plurality of data segments, are readable when the hard decision decoding fails, and perform chipkill decoding based on the first data segment and the other data segments when the other data segments are readable.

A bit interleaving method applying a bit permutation process to a QC LDPC codeword made up of N cyclic blocks of Q bits each, dividing the processed codeword into constellation words of M bits each, and applying an intra-cyclic-block permutation process to the cyclic blocks, where the codeword is divided into F×N/M folding sections of M/F cyclic blocks each and the constellation words are each associated with one of the folding sections, and the bit permutation process is applied such that the constellation words are each made up of F bits from each of M/F different cyclic blocks in the associated section, after the permutation process.

A bit interleaving method involves applying a bit permutation process to a QC LDPC codeword made up of N cyclic blocks each including Q bits, and dividing the codeword, after the bit permutation process, into a plurality of constellation words each made up of M bits, the codeword being divided into N/M sections, each constellation word being associated with one of the N/M sections, and the bit permutation process being performed such that each of the constellation words includes one bit from each of M different cyclic blocks associated with a given section.