Samsung Electronics Co., Ltd.

Hyundai Electronics Industries Co. Ltd.

Gi-ho Park

Shin-dug Kim

Sung-Bae Park

Jung-Wook Park

Tack-don Han

Hoi-Jin Lee

Gun-Ok Jung

Shin-Dug Kim

In-Jik Lee

Jun-Kyu Park

Hoon-Mo Yang

Cheong-ghil Kim

An image search method may include: determining a quadrant of a predicted motion vector; calculating a tilt value of a first reference frame and a tilt value of a second reference frame using the predicted motion vector; deciding a search area for uneven hexagon search in response to the quadrant of the predicted motion vector and the calculated tilt values; performing the uneven hexagon search with respect to the decided search area; and/or comparing a result of the performed uneven hexagon search with a threshold value to determine termination of the uneven hexagon search. The second reference frame is earlier-in-time relative to the first reference frame.

Image search methods for reducing computational complexity of motion estimation

A systolic data processing apparatus includes a processing element (PE) array and control unit. The PE array comprises a plurality of PEs, each PE executing a thread with respect to different data according to an input instruction and pipelining the instruction at each cycle for executing a program. The control unit inputs a new instruction to a first PE of the PE array at each cycle.

Sequentially propagating instructions of thread through serially coupled PEs for concurrent processing respective thread on different data and synchronizing upon branch

A memory may be configured to rearrange and store data to enable a conflict free mode for a memory access pattern required by a coder-decoder(codec) and configured to output a plurality of data from a plurality of banks of the memory in parallel. In addition, a data interconnection unit is configured to shift the plurality of data output from the memory and provide the shifted data to a plurality of operation units as input data. The operation result from each of the plurality of operation units is stored in a region of the memory.

Memory array structure and single instruction multiple data processor including the same and methods thereof

A processor with cache way prediction and method thereof. The processor includes a cache way prediction unit for predicting at least one cache way for selection from a plurality of cache ways. The processor may further include an instruction cache for accessing the selected at least one cache way, where the selected at least one cache way is less than all of the plurality of cache ways. The method includes predicting less than all of a plurality of cache ways for selection and accessing the selected less than all of the plurality of cache ways. In both the process and method thereof, by accessing less than all of the plurality of cache ways, a power consumption and delay may be reduced.

Processor with cache way prediction and method thereof

A semiconductor memory device includes a memory cell array, a decoder, and an access control unit. The decoder generates a word line voltage according to an address for a plurality of memory cells in the memory cell array. The access control unit controls access to the plurality of memory cells according to the word line voltage and additional access information separate from the address.

Accessing semiconductor memory device according to an address and additional access information

An integrated circuit device is provided with a power supply voltage generator therein. This voltage generator is configured to respond to an operating mode control signal by generating first and second power supply voltages at equivalent voltage levels when the operating mode control signal designates a normal mode of operation within the integrated circuit device. The power supply voltage generator is also configured to reduce the first and second power supply voltages to unequal lower voltage levels when the operating mode control signal designates a power saving mode of operation within the integrated circuit device. The power supply voltage generator may also generate a third power supply voltage at a constant level during both the normal and power saving modes of operation.

Integrated circuit devices that support dynamic voltage scaling of power supply voltages

Dynamic Frequency Scaling (DFS) cache memories that can be accessed during an idle time in a single low frequency DFS clock cycle are disclosed. The access can begin during the idle time in the single low frequency DFS clock cycle and may continue during a subsequent low frequency DFS clock cycle. The idle time can be a time interval in the single low frequency DFS clock cycle between completion of a single high frequency DFS clock cycle and completion of the single low frequency DFS clock cycle. Related circuits and systems are also disclosed.

Methods, circuits, and systems for utilizing idle time in dynamic frequency scaling cache memories

We describe a semiconductor memory device including a memory cell array and a storage device to store access data. The memory cell array is accessed responsive to the access data. The memory cell array access is determined by the access data stored in the storage device. The memory cell array is accessed according to access data only if necessary, drastically reducing power dissipation.

Semiconductor memory device including circuit to store access data

Processing circuits that are associated with the operation of threads in an SMT processor can be configured to operate at different performance levels based on a number of threads currently operated by the SMT processor. For example, in some embodiments according to the invention, processing circuits, such as a floating point unit or a data cache, that are associated with the operation of a thread in the SMT processor can operate in one of a high power mode or a low power mode based on the number of threads currently operated by the SMT processor. Furthermore, as the number of threads operated by the SMT operator increases, the performance levels of the processing circuits can be decreased, thereby providing the architectural benefits of the SMT processor while allowing a reduction in the amount of power consumed by the processing circuits associated with the threads. Related computer program products and methods are also disclosed.

Simultaneous multi-threading processor circuits and computer program products configured to operate at different performance levels based on a number of operating threads and methods of operating

A cache memory system reduces the rate of cache misses. The cache memory system includes a first auxiliary storage device which stores first information blocks and a second auxiliary storage device which stores second information blocks fetched from a lower level memory device. Each second block includes a plurality of the first information blocks. A process for fetching information selectively fetches a first or second information block from the lower level memory device and selectively stores the fetched block in the first auxiliary storage device and/or the second auxiliary storage device. Selection of the size of block to fetch and where to store the fetched block is according to whether the data to be referenced by the central controller is in the first auxiliary storage device or the second auxiliary storage device and whether first information blocks that do not include the referenced data are both in the second information block including the referenced data and in the first auxiliary storage device. A control unit that controls the selective fetching and storing maintains state data that includes entries corresponding to second information blocks. Each entry identifies a corresponding second information block and indicates the number of first information blocks that are in the first auxiliary storage device and from the corresponding second information block.

Cache memory system and method for managing the same

A method of and a circuit for instruction/data prefetching using a non-referenced prefetch cache, adapted to store instruction/data blocks prefetched in accordance with a variety of existing prefetchinig machanisms, but not referenced by the central processing unit in an on-chip memory as the non-referenced prefetch cache without discarding them when they are replaced by new ones in a prefetch buffer so that a direct memory reference to the non-referenced prefetch instruction/data blocks can be achieved when they are to be referenced at later times, without any requirement of fetching or prefetching them from the lower memory again. Accordingly, it is possible to not only decrease the number of cache misses and the memory latency due to the fetching of instructions/data from the lower memory for the reference to the instructions/data, but also to reduce memory traffic.

Growing community of inventors

Seoul, South Korea

Gi-ho Park