Abstract: A junction structure element, a method of manufacturing the same, and an in-memory computing device including the same are disclosed. The junction structure element may include: a first polarization layer containing a material having a ferroelectric characteristic of being polarized only horizontally; a second polarization layer disposed on the first polarization layer and containing a material having a ferroelectric characteristic of being polarized only vertically; a source electrode and a drain electrode each in contact with the first polarization layer and spaced apart from each other; and a gate electrode disposed on the second polarization layer.

Abstract: A method for fabricating semiconductor device includes forming an alternating stack that includes a lower multi-layered stack and an upper multi-layered stack by alternately stacking a dielectric layer and a sacrificial layer over a substrate, forming a vertical trench that divides the upper multi-layered stack into dummy stacks, and forming an asymmetric stepped trench that is extended downward from the vertical trench to divide the lower multi-layered stack into a pad stack and a dummy pad stack, wherein forming the asymmetric stepped trench includes forming a first stepped sidewall that is defined at an edge of the pad stack, and forming a second stepped sidewall that is defined at an edge of the dummy pad stack and occupies less area than the first stepped sidewall.

Abstract: Proposed are a method and apparatus for accelerating data processing in a neural network. The apparatus for accelerating data processing in a neural network may include: a control unit configured to quantize data by at least one method according to a characteristic of data calculated at a node forming at least one layer constituting the neural network, and to separately perform calculation at the node according to the quantized data; and memory configured to store the quantized data.

Abstract: Provided is a method and apparatus for processing instructions using a processing-in-memory (PIM). A PIM management apparatus includes: a PIM directory comprising a reader-writer lock regarding a memory address that an instruction accesses; and a locality tracer configured to figure out locality regarding the memory address that the instruction accesses and determine whether or not an object that executes the instruction is a PIM.

Abstract: Proposed are a method and apparatus for accelerating data processing in a neural network. The apparatus for accelerating data processing in a neural network may include: a control unit configured to quantize data by at least one method according to a characteristic of data calculated at a node forming at least one layer constituting the neural network, and to separately perform calculation at the node according to the quantized data; and memory configured to store the quantized data.

Abstract: A circuit for selecting a row of memory cells of a memory device to be refreshed may include: a cold table suitable for storing as a cold row a row selected as a hammered row when the row selected as the hammered row is neither one of cold rows stored in the cold table nor one of hot rows stored in a hot table; and the hot table suitable for storing, as a hot row, the row selected as the hammered row when the row selected as the hammered row is one of the cold rows stored in the cold table.

Abstract: A circuit for selecting a row of memory cells of a memory device to be refreshed may include: a cold table suitable for storing as a cold row a row selected as a hammered row when the row selected as the hammered row is neither one of cold rows stored in the cold table nor one of hot rows stored in a hot table; and the hot table suitable for storing, as a hot row, the row selected as the hammered row when the row selected as the hammered row is one of the cold rows stored in the cold table.

Abstract: Provided is a method and apparatus for processing instructions using a processing-in-memory (PIM). A PIM management apparatus includes: a PIM directory comprising a reader-writer lock regarding a memory address that an instruction accesses; and a locality tracer configured to figure out locality regarding the memory address that the instruction accesses and determine whether or not an object that executes the instruction is a PIM.

Abstract: A power reduction method for a multi-path receiver including multi-receivers, includes detecting a state of the multi-receivers, and controlling clock gating or power gating of the multi-receivers based on the state.

Abstract: Provided is a reconfigurable processor that may process a first type of operation in first mode using a first group of functional units, and process a second type of operation in second mode using a second group of functional units. The reconfigurable processor may selectively supply power to either the first group or the second group, in response to a mode-switch signal or a mode-switch instruction.

Abstract: An apparatus and method for controlling a low-voltage memory in a mobile communication system are provided. The apparatus includes a memory for storing data including at least one error caused by a low-voltage, and an error correction unit for identifying whether the at least one error exists in the memory according to a first bit set in a local buffer of an error correction code storage, for comparing location information on the error data read from the memory and location information on error data of at least one protection set in the local buffer of the error correction code storage when it is determined that the at least one error exists in the memory, for generating an error correction code as a result of the comparison, and for correcting the error data of the memory according to the error correction code.

Abstract: A method of operating a memory system includes the operations of outputting dirty cache lines from a data cache to a volatile memory device as instructions are executed, and outputting from the volatile memory device to a non-volatile memory device as many dirty cache lines as the size of a page of the non-volatile memory.

Abstract: An apparatus and method for controlling a low-voltage memory in a mobile communication system are provided. The apparatus includes a memory for storing data including at least one error caused by a low-voltage, and an error correction unit for identifying whether the at least one error exists in the memory according to a first bit set in a local buffer of an error correction code storage, for comparing location information on the error data read from the memory and location information on error data of at least one protection set in the local buffer of the error correction code storage when it is determined that the at least one error exists in the memory, for generating an error correction code as a result of the comparison, and for correcting the error data of the memory according to the error correction code.

Abstract: The present disclosure provides a controller which comprises a command generator configured to generate a command to non volatile memory, and buffer configured to receive a first data and a second data and configured to combine the first data and the second data, an ECC unit configured to perform the ECC decoding. And the first page data may include at least one error bit corresponding to an error location table and the second page data may include at least one original bit which can be replaced with the error bit. The buffer may replace the at least one error bit with the said at least one original bit. The error location table may save information of location for the repeated error bit.

Abstract: A power reduction method for a multi-path receiver including multi-receivers, includes detecting a state of the multi-receivers, and controlling clock gating or power gating of the multi-receivers based on the state.

Abstract: A memory controller analyzes read data received from a memory device and first error correction code (ECC) data of the read data. A control unit generates a plurality of sub-data from write data to be written in the memory device where the number of error bits in the read data is greater than a number of error bits that can be corrected using the first ECC data. An ECC block generates the first ECC data and second ECC data by using substantially the same algorithm to correct errors in each of the sub-data. The control unit transmits each of the sub-data, the first ECC data and the second ECC data to the memory device.

Abstract: A method of controlling a non-volatile memory device includes comparing the number of banks that are in operating states with a threshold value. If the number of the banks is smaller than the threshold value, data stored in a standby bank is read. If there is no bank having data to be read, a standby bank is programmed. If the number of the banks is equal to or greater than the threshold value or if the reading or the programming is performed, it is determined whether there is a reading or programming command to be performed. If there is the reading or programming command to be performed, the process is repeated from the comparing step. The programming may include programming of a most significant bit (MSB) page or a least significant bit (LSB) page.

Abstract: Disclosed is a network-on-chip system including an active memory processor for processing increased communication latency by multiple processors and memories. The network-on-chip system includes a plurality of processing elements that request to perform an active memory operation for a predetermined operation from a shared memory to reduce access latency of the shared memory, and an active memory processor connected to the processing elements through a network, storing codes for processing custom transaction in request to the active memory operation, performing an operation addresses or data stored in a shared cache memory or the shared memory based on the codes and transmitting the performed operation result to the processing elements.

Abstract: The present disclosure provides a controller which comprises a command generator configured to generate a command to non volatile memory, and buffer configured to receive a first data and a second data and configured to combine the first data and the second data, an ECC unit configured to perform the ECC decoding. And the first page data may include at least one error bit corresponding to an error location table and the second page data may include at least one original bit which can be replaced with the error bit. The buffer may replace the at least one error bit with the said at least one original bit. The error location table may save information of location for the repeated error bit.

Abstract: Provided is a reconfigurable processor that may process a first type of operation in first mode using a first group of functional units, and process a second type of operation in second mode using a second group of functional units. The reconfigurable processor may selectively supply power to either the first group or the second group, in response to a mode-switch signal or a mode-switch instruction.