Electronics and Telecommunications Research Institute

Jin-Ho Han

Young-Su Kwon

Joo-Hyun Lee

Seong-Min Kim

Young-Deuk Jeon

Yi-Gyeong Kim

Ju-Yeob Kim

Kyung-Jin Byun

Han-Jin Cho

Byung-Jo Kim

Hye-Ji Kim

Disclosed herein are a method for supporting cache coherency based on virtual addresses for an artificial intelligence processor having large on-chip memory and an apparatus for the same. The method for supporting cache coherency according to an embodiment of the present disclosure includes, by an artificial intelligence processor including multiple processor cores and multiple caches, setting external memory address areas which do not overlap each other for respective multiple caches; and providing virtual addresses with which the multiple processor cores access the multiple caches.

Method for supporting cache coherency based on virtual addresses for artificial intelligence processor having large on-chip memory and apparatus for the same

An apparatus for receiving a strobe signal may include an amplifier for amplifying a strobe signal input thereto, an offset generator for controlling the setting of a threshold for detecting a preamble signal by generating an offset for the amplifier, and a preamble detector for detecting a first preamble signal occurring at a point at which the amplified strobe signal is equal to or greater than the threshold and turning off the offset generator when the first preamble signal is detected.

Apparatus and method for receiving strobe signal

Disclosed herein is an Artificial Intelligence (AI) processor. The AI processor includes multiple NVM AI cores for respectively performing basic unit operations required for a deep-learning operation based on data stored in NVM; SRAM for storing at least some of the results of the basic unit operations; and an AI core for performing an accumulation operation on the results of the basic unit operation.

Artificial intelligence processor and method of processing deep-learning operation using the same

Disclosed herein is a heterogeneous system based on unified virtual memory. The heterogeneous system based on unified virtual memory may include a host for compiling a kernel program, which is source code of a user application, in a binary form and delivering the compiled kernel program to a heterogenous system architecture device, the heterogenous system architecture device for processing operation of the kernel program delivered from the host in parallel using two or more different types of processing elements, and unified virtual memory shared between the host and the heterogenous system architecture device.

Host apparatus, heterogeneous system architecture device, and heterogeneous system based on unified virtual memory

The exemplary embodiments of the invention relates to fault tolerance of a cache memory which recovers an error occurred in the cache memory or reports an error. A cache memory may include a first layer cache configured to store data requested from a processor, together with a tag related to the data and parity check bits for detecting data error and tag error; a second layer cache configured to store data requested from the first layer cache, together with parity check bits and an error correction code(ECC) bit for detecting data error and tag error; and a fault tolerance unit configured to generate an error signal indicating whether the data error or tag error occurred in at least one of the first layer cache and the second layer cache is recoverable.

Cache memory with fault tolerance

Disclosed herein is an apparatus and method for controlling levelcaches, capable of delivering data to a processor without errors and storing error-free data in the caches even when soft errors occur in the processor and caches. The apparatus includes: a levelcache #connected to the load/store unit of a first processor; a levelcache #connected to the load/store unit of a second processor; and a fault detection and recovery unit for reading from and writing to tag memory, data memory, and valid bit memory of the levelcache #and the levelcache #performing the write-back and flush of the levelcache #and the levelcache #based on information stored therein, and instructing the load/store units of the first and second processors to stall a pipeline and to restart an instruction #n.

Apparatus and method for controlling level 0 cache

The present invention relates to an apparatus for computing an error rate comprising: a first circuit interface being connected to a first sub-circuit receiving data and computing output data through a predetermined computation process; a second circuit interface and being connected to a first test circuit receiving the same data, which is inputted to the first sub-circuit, and computing output data through the predetermined computation process; an error injecting part injecting an error to the first test circuit; an error detecting part comparing output data of the first sub-circuit to output data of the first test circuit; and an error rate computing part computing input node error probability of the first sub-circuit by statistic processing of the compared result. The apparatus and method for computing error rate of the present invention is able to shorten the time required to obtain error probability, compared to the direct simulation of the full circuit.

Apparatus for error simulation and method thereof

An apparatus and method for detecting the fault of a processor are disclosed. The apparatus includes a fetch fault control unit, a decoding fault control unit, and an execution fault control unit. The fetch fault control unit detects the fault of each of fetch units of a plurality of processor cores connected to memory. The decoding fault control unit detects the fault of each of decoding units of the plurality of processor cores connected to the memory. The execution fault control unit detects the fault of each of execution units of the plurality of processor cores connected to the memory, executes the same instruction in the plurality of processor cores, determines a processor core where a fault has occurred, and provides notification of the determined processor to the fetch fault control unit and the decoding fault control unit.

Apparatus and method for detecting fault of processor

There is provided a slave network interface circuit for improving the parallelism of an On-Chip network, including: a MUX for selecting one of a Write Address inputted from the On-Chip network and a Read Address to read data from a slave module, which is inputted from a slave network interface (SNI) controller, in response to the control of the SNI controller and inputs the selected address to the slave module; and the SNI controller for controlling writing and reading data at the slave module and generating a Read Address to store data read from the slave module and to transfer to the On-Chip network.

Growing community of inventors

Seoul, South Korea

Jin-Ho Han