Patent Grant
10379772
The present disclosure relates generally to semiconductor memory and methods, and more particularly, to apparatuses and methods for operations using compressed and decompressed data.
Memory devices are typically provided as internal, semiconductor, integrated circuits in computers or other electronic systems. There are many different types of memory including volatile and non-volatile memory. Volatile memory can require power to maintain its data (e.g., host data, error data, etc.) and includes random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), synchronous dynamic random access memory (SDRAM), and thyristor random access memory (TRAM), among others. Non-volatile memory can provide persistent data by retaining stored data when not powered and can include NAND flash memory, NOR flash memory, and resistance variable memory such as phase change random access memory (PCRAM), resistive random access memory (RRAM), and magnetoresistive random access memory (MRAM), such as spin torque transfer random access memory (STT RAM), among others.
Electronic systems often include a number of processing resources (e.g., one or more processors), which may retrieve and execute instructions and store the results of the executed instructions to a suitable location. A processor can comprise a number of functional units such as arithmetic logic unit (ALU) circuitry, floating point unit (FPU) circuitry, and a combinatorial logic block, for example, which can be used to execute instructions by performing an operation on data (e.g., one or more operands). As used herein, an operation can be, for example, a Boolean operation, such as AND, OR, NOT, NOT, NAND, NOR, and XOR, and/or other operations (e.g., invert, shift, arithmetic, statistics, among many other possible operations). For example, functional unit circuitry may be used to perform the arithmetic operations, such as addition, subtraction, multiplication, and division on operands, via a number of operations.
A number of components in an electronic system may be involved in providing instructions to the functional unit circuitry for execution. The instructions may be executed, for instance, by a processing resource such as a controller and/or host processor. Data (e.g., the operands on which the instructions will be executed) may be stored in a memory array that is accessible by the functional unit circuitry. The instructions and/or data may be retrieved from the memory array and sequenced and/or buffered before the functional unit circuitry begins to execute instructions on the data. Furthermore, as different types of operations may be executed in one or multiple clock cycles through the functional unit circuitry, intermediate results of the instructions and/or data may also be sequenced and/or buffered. A sequence to complete an operation in one or more clock cycles may be referred to as an operation cycle. Time consumed to complete an operation cycle costs in terms of processing and computing performance and power consumption, of a computing apparatus and/or system.
In many instances, the processing resources (e.g., processor and associated functional unit circuitry) may be external to the memory array, and data is accessed via a bus between the processing resources and the memory array to execute a set of instructions. Processing performance may be improved in a processor-in-memory device, in which a processor may be implemented internally and/or near to a memory (e.g., directly on a same chip as the memory array). A processing in memory device may save time by reducing and eliminating external communications and may also conserve power.
The present disclosure includes apparatuses and methods for operations using compressed and decompressed data (e.g., for operations using processing in memory (PIM) structures). In at least one embodiment, a method includes receiving compressed data to a PIM device and decompressing the compressed data on the PIM device.
In previous approaches, data may be transferred from a memory array and sensing circuitry (e.g., via a bus comprising input/output (I/O) lines) to a processing resource such as a processor, microprocessor, and/or compute engine, which may comprise ALU circuitry and/or other functional unit circuitry configured to perform the appropriate operations. However, transferring data from the memory array and sensing circuitry to such processing resource(s) can involve significant time and/or power consumption. Even if the processing resource is located on a same chip as the memory array, significant power can be consumed in moving data out of the array to the compute circuitry, which can involve performing a sense line (which may be referred to herein as a digit line or data line) address access (e.g., firing of a column decode signal) in order to transfer data from sense lines onto I/O lines (e.g., local I/O lines), transferring the data peripheral to the memory array, which may be transfer to a cache in a host, and providing the data to the peripheral compute circuitry (e.g., associated with a central processing unit (CPU) of the host).
Furthermore, the circuitry of the processing resource(s) (e.g., a compute engine) may not conform to pitch rules associated with a memory array. For example, the memory cells of a memory array may have a 4F2 or 6F2 cell size, where “F” is a feature size corresponding to the cells. As such, the devices (e.g., logic gates) associated with ALU circuitry of previous PIM systems may not be capable of being formed on pitch with the memory cells, which can affect chip size and/or memory density, for example. A number of embodiments of the present disclosure can include the sensing circuitry (e.g., including sense amplifiers and/or compute components) and/or an operations component, as described herein, being formed on pitch with the memory cells of the array and being configured to (e.g., being capable of performing) compute functions (e.g., operations), such as those described herein, on pitch with the memory cells.
For example, the sensing circuitry 150 described herein can be formed on a same pitch as a pair of complementary sense lines. As an example, a pair of complementary memory cells may have a cell size with a 6F2 pitch (e.g., 3F×2F). If the pitch of a pair of complementary sense lines for the complementary memory cells is 3F, then the sensing circuitry being on pitch indicates the sensing circuitry (e.g., a sense amplifier and corresponding compute component per respective pair of complementary sense lines) is formed to fit within the 3F pitch of the complementary sense lines.
Furthermore, the circuitry of the processing resource(s) (e.g., a compute engine, such as an ALU) of various prior systems may not conform to pitch rules associated with a memory array. For example, the memory cells of a memory array may have a 4F2 or 6F2 cell size. As such, the devices (e.g., logic gates) associated with ALU circuitry of previous systems may not be capable of being formed on pitch with the memory cells (e.g., on a same pitch as the sense lines), which can affect chip size and/or memory density, for example. In the context of some computing systems and subsystems (e.g., a central processing unit (CPU)), data may be processed in a location that is not on pitch and/or on chip with memory (e.g., memory cells in the array), as described herein. The data may be processed by a processing resource associated with a host, for instance, rather than on pitch with the memory.
In contrast, a number of embodiments of the present disclosure can include the sensing circuitry 150 (e.g., including sense amplifiers and/or compute components) being formed on pitch with the memory cells of the array. The sensing circuitry 150 can be configured for (e.g., capable of) performing compute functions (e.g., logical operations).
In the context of some computing systems and subsystems (e.g., a CPU, a graphics processing unit (GPU), and/or a frame buffer, etc.), data may be processed in a location that is not on pitch with memory cells in the array and/or corresponding pairs of complementary sense lines and/or on chip with memory (e.g., formed on the same chip as the memory cells in the array) and/or a controller for the memory, as described herein. The data may be processed by a processing resource associated with a host, for instance, rather than on pitch and/or on chip with the memory and/or on chip with the controller. As such, the data may be sent to and from memory in the same format as the data was in before or after being processed elsewhere.
In graphics, for instance, data may be sent in compressed form to a frame buffer. Although imaging coordinates (e.g., textures, depths, etc.) may be stored in memory in compressed form, the memory may then send the compressed data to a GPU in the compressed form, with the GPU responsible for decompressing the data and sending uncompressed data (e.g., which may also have been depth textured and/or depth buffered, as described herein, by the GPU) back to the memory and/or to the frame buffer. Such movement (e.g., copying, transferring, and/or transporting) of data off pitch for processing, including decompression of compressed data, may increase memory bandwidth requirements, which also may increase time, power, and/or cost relative to an ability to perform such operations by sensing circuitry on pitch with the memory array.
A memory device configured for on pitch data processing operations, such as a PIM device, can perform operations on data that have been received (e.g., stored) in its original format, as directed by an operations component (e.g., programmed for bitwise logical operations). PIM capable device operations can use bit vector based operations. As used herein, the term “bit vector” is intended to mean a physically contiguous number of bits on a bit vector memory device (e.g., a PIM device) stored physically contiguous in a row of an array of memory cells. Thus, as used herein a “bit vector operation” is intended to mean an operation that is performed on a bit vector that is a contiguous portion of virtual address space (e.g., used by a PIM device). For example, a row of virtual address space in the PIM device may have a bit length of 16K bits (e.g., corresponding to 16K complementary pairs of memory cells in a DRAM configuration). Sensing circuitry, as described herein, for such a 16K bit row may include a corresponding 16K processing elements (e.g., compute components, as described herein) formed on pitch with the sense lines selectably coupled to corresponding memory cells in the 16 bit row. A compute component in the PIM device may operate as a one bit processing element on a single bit of the bit vector of the row of memory cells sensed by the sensing circuitry (e.g., sensed by and/or stored in a sense amplifier paired with the compute component, as described herein).
Among various formats, the data may be stored in a compressed form (e.g., using either a lossy or lossless compression algorithm). For example, image data may be sent to the memory in the form of subsampled YUV compressed data, which reduces resolution of color data compared to standard RGB formats. The memory device itself, as described herein, can decompress the data into another format and can then perform operations on the decompressed data (e.g., alpha blending, as directed by the operations component, among other operations described herein).
Alpha blending as used herein is intended to mean a compositing operation in which one image is combined (e.g., overlaid) with at least one other image and where, through partial or full transparency, an image can, for example, serve as a background image and another can serve as a foreground image. Data also may be received and/or stored at a reduced image size and rescaled to a larger size (e.g., by increasing a number of pixels that the image contains), as directed by the operations component. The operations component can be configured to direct compensation for a reduced bit depth in a received and/or stored image, (e.g., by increasing an 8 bit image having 256 possible colors to a 24 bit image having about 16 million possible colors). The operations component also can be configured to decompress, for example, a highly compressed format (e.g., Joint Photographic Experts Group (JPEG), Moving Pictures Experts Group (MPEG), among other such formats), in addition to other possible operations. As such, the operations component 172 described herein may contribute to an “intelligent frame buffer” (e.g., in combination with the sensing circuitry 150 and/or arrays 130 of memory cells, as shown in and described in connection with
The operations being performed on pitch in association with the intelligent frame buffer may reduce memory traffic, and thus time, power, and/or cost. For example, the memory traffic may be reduced relative to sending the data off pitch to be decompressed (e.g., by a GPU associated with the host). The memory traffic also may be reduced by the operations component being configured to direct operations that affect perceived quality of data output (e.g., alpha blending, among other compositing operations, and operations on texture, depth, etc.), which can convert read-modify-write data movement operations involved with exchange of data between the memory and, for example, a GPU into a one-way movement of data into the memory for processing. Data in a reduced-size format can be sent to a memory device, as described herein, to be expanded into a larger format on pitch with the memory. For example, image data can be stored in a subsampled YUV format and converted to non-subsampled RGB format, as directed by the operations component and performed on pitch. The reformatted image data then can be stored as such in memory and/or can have one or more additional operations performed thereon, as described herein. Using a compressed format for input to memory and/or the memory device being configured to decompress and perform other operations on pitch with the memory may reduce bandwidth requirements into and out of the memory device, thereby increasing performance and/or reducing time, usage of electrical power, and/or cost based on the reduced memory traffic.
As described herein, the embodiments can allow a host system to allocate a number of locations (e.g., sub-arrays (or “subarrays”)) and portions of subarrays, in one or more DRAM banks to receive (e.g., store) and/or process data. A host system and/or a controller, as described herein, may perform the address resolution on an entire block of program instructions (e.g., PIM command instructions) and data and direct (e.g., control) allocation, storage, and/or movement (e.g., flow) of data and commands into allocated locations (e.g., subarrays and portions of subarrays) within a destination (e.g., target) bank. Writing data and/or executing commands (e.g., issuing instructions to perform operations, as described herein) may utilize a normal DRAM write path to the DRAM device. As the reader will appreciate, while a DRAM-style PIM device is discussed with regard to examples presented herein, embodiments are not limited to a PIM DRAM implementation.
The memory devices described herein can use a number of controllers for a bank of subarrays, controllers for individual subarrays, and/or controllers for operations components (e.g., each controller being a sequencer, a state machine, a microcontroller, a sub-processor, ALU circuitry, or some other type of controller) to execute a set of instructions to perform an operation on data (e.g., one or more operands). As used herein, an operation can be, for example, a Boolean logical operation (e.g., AND, OR, NOT, NOT, NAND, NOR, and/or XOR) and/or other operations (e.g., invert, shift, arithmetic, statistics, and various operations on image and/or audio files, among many other possible operations). For example, functional unit circuitry may be used to perform the arithmetic operations, such as addition, subtraction, multiplication, and division on operands, via a number of logical operations.
The present disclosure describes enablement of operations (e.g., PIM operations, such as AND, OR, refresh, row copy, shift, add, multiply, decompression, compression, alpha blending, operations affecting texture, depth, etc.) to be performed on data values as received by (e.g., stored in) sensing circuitry and/or memory cells (e.g., when directed by the operations component 172, as described herein). For example, the operations can be performed on data values substantially simultaneously with the data values being received (e.g., at least temporarily stored and/or cached) by sensing circuitry 150 (e.g., a sensing component stripe 124) of a subarray 125 and/or upon subsequent retrieval of the data values as stored in a row 119 of memory cells of the subarray, as shown in and described in connection with
Implementations of PIM DRAM architecture may perform processing (e.g., in a sensing component stripe 124) at the level of the sense amplifier 206 and/or compute component 231 (e.g., as shown in and described in connection with
Moreover, the sense amplifiers 206 and/or compute components 231 in a sensing component stripe 124 may be directed by instructions sent from the operations component 172 (e.g., via coupled I/O lines), to perform particular operations on data stored (cached) in the sensing component stripe. A number of factors may influence how an operation is performed in the sensing component stripe 124. Comparative complexity of some operations may affect the choice. For example, Boolean operations (e.g., AND, OR, NOT, NOT, NAND, NOR, and/or XOR) and/or shift operations may be effectively performed using functionality resident in and/or local to the sensing component stripe (e.g., as described in connection with
Whereas the sensing circuitry 150 (e.g., sense amplifiers and/or compute components of a sensing component stripe) can be directly coupled to the memory cells of a column of memory cells in a subarray, the operations component 172 may, in some embodiments, not be directly coupled to the memory cells of the subarrays. For example, the operations component 172 described herein can be indirectly coupled to the memory cells of a column via a selectably coupled sensing component stripe and a number of selectably coupled I/O lines.
In the following detailed description of the present disclosure, reference is made to the accompanying drawings that form a part hereof, and in which is shown by way of illustration how one or more embodiments of the disclosure may be practiced. These embodiments are described in sufficient detail to enable those of ordinary skill in the art to practice the embodiments of this disclosure, and it is to be understood that other embodiments may be utilized and that process, electrical, and structural changes may be made without departing from the scope of the present disclosure.
As used herein, designators such as “X”, “Y”, “N”, “M”, etc., particularly with respect to reference numerals in the drawings, indicate that a number of the particular feature so designated can be included. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting. As used herein, the singular forms “a”, “an”, and “the” can include both singular and plural referents, unless the context clearly dictates otherwise. In addition, “a number of”, “at least one”, and “one or more” (e.g., a number of memory arrays) can refer to one or more memory arrays, whereas a “plurality of” is intended to refer to more than one of such things. Furthermore, the words “can” and “may” are used throughout this application in a permissive sense (i.e., having the potential to, being able to), not in a mandatory sense (i.e., must). The term “include,” and derivations thereof, means “including, but not limited to”. The terms “coupled” and “coupling” mean to be directly or indirectly connected physically or for access to and movement (transmission) of commands and/or data, as appropriate to the context. The terms “data”, “data units”, and “data values” are used interchangeably herein and can have the same meaning, as appropriate to the context.
As described herein, an I/O line can be selectably shared by a plurality of subarrays, rows, and/or particular columns of memory cells via the sensing component stripe coupled to each of the subarrays. For example, the sense amplifier and/or compute component of each of a selectable subset of a number of columns (e.g., eight column subsets of a total number of columns) can be selectably coupled to each of a plurality of shared I/O lines for data values stored (cached) in the sensing component stripe to be moved (e.g., copied, transferred, and/or transported) to each of the plurality of shared I/O lines. The I/O line can, in some embodiments, be further shared by the operations component. Because the singular forms “a”, “an”, and “the” can include both singular and plural referents herein, for example, “a shared I/O line” can be used to refer to “a plurality of shared I/O lines”, unless the context clearly dictates otherwise. Moreover, for example, “shared I/O lines” is an abbreviation of “plurality of shared I/O lines”.
The figures herein follow a numbering convention in which the first digit or digits correspond to the figure number and the remaining digits identify an element or component in the figure. Similar elements or components between different figures may be identified by the use of similar digits. For example, 108 may reference element “08” in
System 100 in
For clarity, the system 100 has been simplified to focus on features with particular relevance to the present disclosure. The memory array 130 can be a DRAM array, SRAM array, STT RAM array, PCRAM array, TRAM array, RRAM array, NAND flash array, and/or NOR flash array, among other types of arrays. The array 130 can include memory cells arranged in rows coupled by access lines (which may be referred to herein as word lines or select lines) and columns coupled by sense lines (which may be referred to herein as data lines or digit lines). Although a single array 130 is shown in
The memory device 120 can include address circuitry 142 to latch address signals provided over a data bus 156 (e.g., an I/O bus from the host 110) by I/O circuitry 144 (e.g., provided to external ALU circuitry and to DRAM data lines (DQs) via local I/O lines and global I/O lines). As used herein, DRAM DQs can enable input of data to and output of data from a bank (e.g., from and to the controller 140 and/or host 110) via a bus (e.g., data bus 156). During a write operation, voltage and/or current variations, for instance, can be applied to a DQ (e.g., a pin). These variations can be translated into an appropriate signal and stored in a selected memory cell. During a read operation, a data value read from a selected memory cell can appear at the DQ once access is complete and the output is enabled. At other times, DQs can be in state such that the DQs do not source or sink current and do not present a signal to the system. This also may reduce DQ contention when two or more devices (e.g., banks) share the data bus.
Status and exception information can be provided from the controller 140 of the memory device 120 to a channel controller 143, for example, through a high speed interface (HSI) out-of-band (OOB) bus 157, which in turn can be provided from the channel controller 143 to the host 110. The channel controller 143 can include a logic component 160 to allocate a plurality of locations (e.g., controllers for subarrays) in the arrays of each respective bank to store bank commands, application instructions (e.g., for sequences of operations), and arguments (PIM commands) for the various banks associated with operations of each of a plurality of memory devices (e.g., 120-1, . . . , 120-N as shown in
Address signals are received through address circuitry 142 and decoded by a row decoder 146 and a column decoder 152 to access the memory array 130. Data can be sensed (read) from memory array 130 by sensing voltage and/or current changes on sense lines (digit lines) using a number of sense amplifiers 206, as described herein, of the sensing circuitry 150. A sense amplifier can read and latch a page (e.g., a row) of data from the memory array 130. Additional compute circuitry, as described herein in connection with the operations component 172, can be coupled to the sensing circuitry 150 and can be used in combination with the sense amplifiers to sense, store (e.g., cache and/or buffer), perform compute functions (e.g., operations), and/or move data. The I/O circuitry 144 can be used for bi-directional data communication with host 110 over the data bus 156 (e.g., a 64 bit wide data bus). The write circuitry 148 can be used to write data to the memory array 130.
Controller 140 (e.g., bank control logic and sequencer) can decode signals (e.g., commands) provided by control bus 154 from the host 110. These signals can include chip enable signals, write enable signals, and/or address latch signals that can be used to control operations performed on the memory array 130, including data sense, data store, data movement (e.g., copying, transferring, and/or transporting data values), data decompression, data write, and/or data erase operations, among other operations described herein. In various embodiments, the controller 140 can be responsible for executing instructions from the host 110 and accessing the memory array 130. The controller 140 can be a state machine, a sequencer, or some other type of controller. The controller 140 can, for example, control performance of Boolean logical operations and/or shifting data (e.g., right or left) in a row of an array (e.g., memory array 130).
Examples of the sensing circuitry 150 are described further below (e.g., in connection with
In a number of embodiments, the sensing circuitry 150 can be used to perform on pitch operations using data stored in memory array 130 as input and participate in movement of the data for copy, transfer, transport, writing, logic, and/or storage operations to a different location in the memory array 130 without transferring the data via a sense line address access (e.g., without firing a column decode signal). As such, various compute functions may, in some embodiments, be performed on pitch using, and within, sensing circuitry 150 rather than (or in association with) being performed by processing resources external to the sensing circuitry 150 (e.g., by a processor associated with host 110 and/or other processing circuitry, such as ALU circuitry, located on device 120, such as on controller 140 or elsewhere).
In various previous approaches, data associated with an operand, for instance, would be read from memory via sensing circuitry and provided to external ALU circuitry via I/O lines (e.g., via local I/O lines and/or global I/O lines). The external ALU circuitry could include a number of registers and would perform compute functions using the operands, and the result would be transferred back to the array via the I/O lines.
In contrast, in a number of embodiments of the present disclosure, sensing circuitry 150 is configured to perform operations on data stored in memory array 130 and store the result back to the memory array 130 without enabling an I/O line (e.g., a local I/O line) coupled to the sensing circuitry 150. The sensing circuitry 150 may be formed on pitch with the memory cells of the array. The operations component 172 can be coupled to the sensing circuitry 150 (e.g., via an I/O line), but may be distinct from the sensing circuitry 150. The memory array 130, the sensing circuitry 150, and the operations component 172, in various embodiments, can cooperate in performing operations, according to embodiments described herein, and can collectively be termed an intelligent frame buffer. Additional peripheral sense amplifiers and/or logic 170 (e.g., subarray controllers that each execute instructions for performing a respective operation) can, in some embodiments, be coupled to the memory array 130, the sensing circuitry 150, and/or the operations component 172. The peripheral sense amplifiers and/or logic 170 can cooperate with and/or be part of the intelligent frame buffer in performance of some operations described herein.
As such, in a number of embodiments, circuitry external to array 130, sensing circuitry 150, and operations component 172 is not needed to perform compute functions because the sensing circuitry 150 and/or the operations component 172 can perform the appropriate operations to perform such compute functions without the use of an external processing resource. Therefore, the sensing circuitry 150 and operations component 172 may be used to complement or to replace, at least to some extent, such an external processing resource (or at least reduce the bandwidth consumption of transfer of data to and/or from such an external processing resource).
However, in a number of embodiments, the sensing circuitry 150 and operations component 172 may be used to perform operations (e.g., to execute instructions) in addition to operations performed by an external processing resource (e.g., host 110). For instance, host 110 and/or sensing circuitry 150 and operations component 172 each may be limited to performing only certain operations and/or a certain number of operations.
Enabling an I/O line can include enabling (e.g., turning on, activating) a transistor having a gate coupled to a decode signal (e.g., a column decode signal) and a source/drain coupled to the I/O line. However, embodiments are not limited to not enabling an I/O line. For instance, in a number of embodiments, the sensing circuitry 150 can be used to perform operations without enabling column decode lines of the array. However, the local I/O line(s) may be enabled in order to transfer a result to a suitable location other than back to the array 130 (e.g., to an external register).
As shown in
For example, each of the plurality of banks (e.g., banks 121-0, 121-1, . . . , 121-7 in a plurality of memory devices 120-1, 120-2, . . . , 120-N as shown in
In some embodiments, the channel controller 143 can dispatch commands to the plurality of banks (e.g., Bank 0, . . . , Bank 7) and field return results and/or data from such operations. The return results and/or data can be returned to the channel controller 143 via the OOB bus 157 associated with the status channel interface on each of the plurality of banks.
Embodiments described herein provide a method for operating a memory device 120 to implement operations using compressed and decompressed data. The operations may be performed by execution of non-transitory instructions (e.g., from operations component 172) by a processing resource (e.g., sensing circuitry 150 in sensing component stripes 124-0, 124-1, . . . , 124-N−1). The method can include receiving compressed data to a PIM device and decompressing the compressed data on the PIM device.
As described herein, the compressed data can be received to an intelligent frame buffer on the PIM device. In various embodiments, the intelligent frame buffer can include an operations component 172 and sensing circuitry 150 associated with an array 130 of memory cells, as described herein. For example, the compressed data can be received to (e.g., at least temporarily stored and/or cached in) sensing circuitry 150 associated with a DRAM array on the PIM device.
The compressed data can, in some embodiments, be image data received in, for example, a subsampled YUV image format. The subsampled YUV image format can be received to (e.g., at least temporarily stored and/or cached in) the intelligent frame buffer associated with a random access memory (RAM) array of memory cells on the PIM device. The compressed data can be decompressed by, for example, converting the subsampled YUV image format to a non-subsampled RGB image format using the intelligent frame buffer on the PIM device.
In some embodiments, the method can include storing an image in, for example, the non-subsampled RGB image format to the RAM array. The method can include compositing a plurality of images in the non-subsampled RGB image format into a different image using an alpha blending technique (e.g., in which a first image may be overlaid with at least a second image to form a composite image, as described further herein). The image data may, for example, be received as having a reduced bit depth compared to a bit depth stored on and/or sent from a host 110, as described further herein. The image data may, for example, be received in JPEG format to the intelligent frame buffer on the PIM device. Alpha blending and/or other compositing operations, rescaling, depth texturing, and/or depth buffering, among other possible operations, can be performed on the decompressed data using the sensing circuitry 150 of the intelligent frame buffer associated with, for example, the DRAM array on the PIM device.
Compressed data can, in some embodiments, be received into (e.g., stored in) selected memory cells of an array 130 of the memory cells. The method can, in some embodiments, include decompressing the compressed data to produce decompressed data using the sensing circuitry 150 selectably coupled to the selected memory cells. In some embodiments, the sensing circuitry 150 may be on pitch with the selected memory cells. The method can include directing decompression of the compressed data using an operations component 172 selectably coupled via the sensing circuitry 150 to data values stored in the selected memory cells of the array 130. In some embodiments, the operations component 172 can be associated with a controller 140 on a bank 121 of the array 130.
As indicated in
As shown in
The channel controller 143 can include one or more local buffers 159 to store program instructions and can include logic 160 to allocate a plurality of locations (e.g., subarrays or portions of subarrays) in the arrays of each respective bank to store bank commands, and arguments (e.g., PIM commands) for the various banks associated with operation of each of the plurality of memory devices 120-1, . . . , 120-N. The channel controller 143 can dispatch commands (e.g., PIM commands) to the plurality of memory devices 120-1, . . . , 120-N to store those program instructions within a given bank of a memory device.
As in
Each column 122 is configured to be coupled to sensing circuitry 150, as described in connection with
Each of the of the subarrays 125-0, 125-1, . . . , 125-N−1 can include a plurality of rows 119 shown vertically as Y (e.g., each subarray may include 256, 512, 1024 rows, among various possibilities, in an example DRAM bank). Example embodiments are not limited to the example horizontal and vertical orientation of columns and rows described herein or the example numbers thereof. The sense amplifiers 206 used in sensing data values in memory cells of the subarrays 125 and/or the compute components 231 used in performance of compute operations are located in the plurality of sensing component stripes 124 that are each physically associated with a subarray 125 of memory cells in the bank section 123 shown in
In contrast, the operations component 172 configured to direct operations performed on the data values may, in some embodiments, be located on a periphery of the bank section 123. For example, whereas the operations component 172 may, in various embodiments, be located on chip with memory (e.g., formed on the same chip as the memory cells in the array) and/or a controller 140 for the memory, as described herein, the operations component 172 may not be physically associated with a subarray 125 of memory cells, as shown in
As shown in
Although not shown, memory cells are coupled to the pairs of complementary sense lines 205-1 and 205-2 (e.g., columns). For example, a memory cell can comprise a transistor and a capacitor. The memory cells can be, for example, 1T1C DRAM cells each comprising a storage element (e.g., capacitor) and an access device (e.g., transistor), although other embodiments of configurations can be used (e.g., 2T2C with two transistors and two capacitors per memory cell). In a number of embodiments, the memory cells may be destructive read memory cells (e.g., reading the data stored in the cell destroys the data such that the data originally stored in the cell may be refreshed after being read). The cells of the memory array can be arranged in rows coupled by word lines and columns coupled by pairs of complementary data lines DIGIT(n−1)/DIGIT(n−1)_, DIGIT(n)/DIGIT(n)_, DIGIT(n+1)/DIGIT(n+1)_. The individual data lines corresponding to each pair of complementary data lines can also be referred to as data lines 205-1 (D) and 205-2 (D_) respectively. Although only three pairs of complementary data lines (e.g., three columns) are shown in
Memory cells can be coupled to different data lines and/or word lines. For example, a first source/drain region of an access transistor of a memory cell can be coupled to a data line 205-1 (D), a second source/drain region of the access transistor of the memory cell can be coupled to a capacitor of the memory cell, and a gate of the access transistor of the memory cell can be coupled to a word line of the memory array.
As shown in
In the example illustrated in
The gates of the pass gates 207-1 and 207-2 can be controlled by a logical operation selection logic signal, Pass. For example, an output of the logical operation selection logic 213 can be coupled to the gates of the pass gates 207-1 and 207-2, as shown in
The sensing circuitry 250 illustrated in
In various embodiments, the logical operation selection logic 213 can include four logic selection transistors: logic selection transistor 262 coupled between the gates of the swap transistors 242 and a TF signal control line, logic selection transistor 252 coupled between the gates of the pass gates 207-1 and 207-2 and a TT signal control line, logic selection transistor 254 coupled between the gates of the pass gates 207-1 and 207-2 and a FT signal control line, and logic selection transistor 264 coupled between the gates of the swap transistors 242 and a FF signal control line. Gates of logic selection transistors 262 and 252 are coupled to the true sense line through isolation transistor 251-1 (having a gate coupled to an ISO signal control line). Gates of logic selection transistors 264 and 254 are coupled to the complementary sense line through isolation transistor 251-2 (also having a gate coupled to an ISO signal control line).
Data units present on the pair of complementary sense lines 205-1 and 205-2 can be loaded into the compute component 231 via the pass gates 207-1 and 207-2. When the pass gates 207-1 and 207-2 are OPEN, data units on the pair of complementary sense lines 205-1 and 205-2 are passed to the compute component 231 and thereby loaded into the loadable shift register. The data unit on the pair of complementary sense lines 205-1 and 205-2 can be the data unit stored at least temporarily in the sense amplifier 206 when the sense amplifier is enabled (e.g., fired). The logical operation selection logic signal, Pass, is activated to OPEN (e.g., turn on) the pass gates 207-1 and 207-2.
The ISO, TF, TT, FT, and FF control signals can operate to select a logical operation to implement based on the data unit (“B”) in the sense amplifier 206 and the data unit (“A”) in the compute component 231 (e.g., as used herein, the data unit stored in a latch of a sense amplifier is referred to as a “B” data unit, and the data unit stored in a latch of a compute component is referred to as an “A” data unit). In particular, the ISO, TF, TT, FT, and FF control signals are configured to select the logical operation (e.g., function) to implement independent from the data unit present on the pair of complementary sense lines 205-1 and 205-2 (although the result of the implemented logical operation can be dependent on the data unit present on the pair of complementary sense lines 205-1 and 205-2). For example, the ISO, TF, TT, FT, and FF control signals can select the logical operation to implement directly because the data unit present on the pair of complementary sense lines 205-1 and 205-2 is not passed through logic to operate the gates of the pass gates 207-1 and 207-2.
Additionally,
As an example, the logical operation selection logic signal Pass can be activated (e.g., high) to OPEN (e.g., turn on) the pass gates 207-1 and 207-2 when the ISO control signal line is activated and either the TT control signal is activated (e.g., high) with the data unit on the true sense line being “1” or the FT control signal is activated (e.g., high) with the data unit on the complement sense line being “1.”
The data unit on the true sense line being a “1” OPENs logic selection transistors 252 and 262. The data unit on the complementary sense line being a “1” OPENs logic selection transistors 254 and 264. If the ISO control signal or either the respective TT/FT control signal or the data unit on the corresponding sense line (e.g., sense line to which the gate of the particular logic selection transistor is coupled) is not high, then the pass gates 207-1 and 207-2 will not be OPENed by a particular logic selection transistor.
The logical operation selection logic signal Pass* can be activated (e.g., high) to OPEN (e.g., turn on) the swap transistors 242 when the ISO control signal line is activated and either the TF control signal is activated (e.g., high) with data unit on the true sense line being “1,” or the FF control signal is activated (e.g., high) with the data unit on the complement sense line being “1.” If either the respective control signal or the data unit on the corresponding sense line (e.g., sense line to which the gate of the particular logic selection transistor is coupled) is not high, then the swap transistors 242 will not be OPENed by a particular logic selection transistor.
The sensing circuitry 250 illustrated in
Although not shown in
As noted above, the compute components 231 can comprise a loadable shift register. In this example, each compute component 231 is coupled to a corresponding pair of complementary data lines 205-1/205-2, with a node ST2 being coupled to the particular data line (e.g., DIGIT(n)) communicating a “true” data unit and with node SF2 being coupled to the corresponding complementary data line (e.g., DIGIT(n)_) communicating the complementary data unit (e.g., “false” data unit).
In this example, the loadable shift register comprises a first right-shift transistor 281 of a particular compute component 231 having a gate coupled to a first right-shift control line 282 (e.g., PHASE 1R), and a second right-shift transistor 286 of the particular compute component 231 having a gate coupled to a second right-shift control line 283 (e.g., PHASE 2R). Node ST2 of the particular control component is coupled to an input of a first inverter 287, whose output (e.g., node SF1) is coupled to a first source/drain region of transistor 286. The second source/drain region of transistor 286 is coupled to the input (e.g., node SF2) of a second inverter 288. The output (e.g., node ST1) of inverter 288 is coupled to a first source/drain region of transistor 281, and a second source/drain region of transistor 281 the particular compute component 231 is coupled to an input (e.g., node ST2) of a first inverter 287 of an adjacent compute component 231. The loadable shift register shown in
In operation, a data unit on a pair of complementary data lines (e.g., 205-1/205-2) can be loaded into a corresponding compute component 231 (e.g., by operating logical operation selection logic as described above). As an example, a data unit can be loaded into a compute component 231 via overwriting of the data unit currently stored in the compute component 231 with the data unit stored in the corresponding sense amplifier 206. Alternatively, a data unit may be loaded into a compute component by deactivating the control lines 282, 283, 291, and 292.
Once a data unit is loaded into a compute component 231, the “true” data unit is separated from the complement data unit by the first inverter 287. Shifting data to the right (e.g., to an adjacent compute component 231) can include alternating operation of the first right-shift transistor 281 and the second right-shift transistor 286, for example, via the PHASE 1R and PHASE 2R control signals being periodic signals that go high out of phase from one another (e.g., non-overlapping alternating square waves 180 out of phase). The transistor 290 can be turned on to latch the shifted data unit.
An example of shifting data left via the shift register shown in
Embodiments of the present disclosure are not limited to the shifting capability described in association with the compute components 231. For example, a number of embodiments can include shift circuitry in addition to and/or instead of the shift circuitry described in association with a loadable shift register.
The sensing circuitry 250 in
In a number of examples, the sense amplifier 206 and the compute component 231 can be in at least one of two states associated with the first mode and the second mode. As used herein, a state of a sense amplifier 206 and/or the compute component 231 can describe a movement (e.g., transfer) of data between the sense amplifier 206 and/or the compute component 231. The state of the sense amplifier 206 and/or the compute component 231 can also be described as whether the sense amplifier 206 and/or the compute component 231 is in an equilibration state or is storing a data unit (e.g., a binary 0 or 1 data value). For example, a sense amplifier can be configured to be in an initial state, wherein the initial state is one of an equilibration state and a data storage state.
A data storage state can include the sense amplifiers 206 storing a data unit. As used herein, a data unit can be referred to as a bit and/or a digit value. Data can be moved (e.g., transferred) from a compute component 231 to a sense amplifier 206 in response to enabling a pass gate (e.g., activating the PASS and/or PASS* control signals via the TF 262, TT 252, FT 254, and/or FF 264 control signals that are referred to herein as a logical operation selection logic) and the sense amplifier 206 being in a equilibration state. Data can be moved from a sense amplifier 206 to a compute component 231 in response to enabling the pass gate (e.g., activating the PASS and/or PASS* control signals via the TF 262, TT 252, FT 254, and/or FF 264 control signals that are referred to herein as a logical operation selection logic) and the sense amplifier 206 being in a data storage state. The direction of the movement of data between the sense amplifier 206 and the compute component 231 is determined by whether the sense amplifier 206 is in an equilibration state or stores a data unit before the PASS and/or PASS* control signals are activated and by a particular operation selected via the logical operation selection logic (e.g., TF 262, TT 252, FT 254, and FF 264 control signals).
For example, if the sense amplifier 206 is equilibrated and the PASS and/or PASS* control signals are activated to provide a conduction path (e.g., electrical continuity) between the sense amplifier 206 and the compute component 231, then a data unit stored in the compute component 231 can be moved from the compute component 231 to the sense amplifier 206.
If the sense amplifier 206 is configured to store a first bit (e.g., first data unit) and the PASS and/or PASS* control signals are activated to provide a conduction path between the sense amplifier 206 and the compute component 231, then a second bit (e.g., second data unit) that is stored in the compute component 231 before the activation of the PASS and/or PASS* control signals can be replaced by the first bit and the sense amplifier 206 retains the first bit. Furthermore, a number of operations can be performed using the first bit and the second bit using the logical operation selection logic and the result of the operation can be stored in the compute component 231.
Using an equilibration signal to direct the movement of data between the sense amplifier 206 and the compute component 231 can provide the ability to selectively perform an operation in sense amplifiers that are not equilibrated without performing the operation in sense amplifiers that are equilibrated. For example, a PASS and/or a PASS* control signal can be activated in a plurality of sensing components to move data between a first group of a plurality of sense amplifiers that are equilibrated and a first group of a plurality of compute components. The PASS and/or PASS* control signals can also be activated to move data between a second group of the plurality of sense amplifiers and a second group of the plurality of component components that are not equilibrated to selectively perform an operation in a second group of sense components while not performing the operation on a first group of sense components.
In a number of embodiments, a sense amplifier 206 can comprise a number of transistors formed on pitch with the transistors of the corresponding compute component 231 and/or the memory cells of an array (e.g., memory array 130 shown in
The voltages or currents on the respective data lines D and D— can be provided to the respective latch inputs 233-1 and 233-2 of the cross coupled latch 215 (e.g., the input of the primary latch). In this example, the latch input 233-1 is coupled to a first source/drain region of transistors 227-1 and 229-1 as well as to the gates of transistors 227-2 and 229-2. Similarly, the latch input 233-2 can be coupled to a first source/drain region of transistors 227-2 and 229-2 as well as to the gates of transistors 227-1 and 229-1. The compute component 231, which may be referred to herein as an accumulator, can be coupled to latch inputs 233-1 and 233-2 of the cross coupled latch 215 as shown. However, embodiments are not limited to the example shown in
In this example, a second source/drain region of transistors 227-1 and 227-2 can be commonly coupled to a negative control signal (RnIF) 228. A second source/drain region of transistors 229-1 and 229-2 can be commonly coupled to an active positive control signal (ACT) 265. The ACT signal 265 can be a supply voltage (e.g., VDD) and the RnIF signal can be a reference voltage (e.g., ground). RnIF signal 228 and ACT signal 265 can function as activating signals that enable the cross coupled latch 215.
The enabled cross coupled latch 215 can operate to amplify a differential voltage between latch input 233-1 (e.g., first common node) and latch input 233-2 (e.g., second common node) such that latch input 233-1 is driven to one of the ACT signal voltage and the RnIF signal voltage (e.g., to one of VDD and ground), and latch input 233-2 is driven to the other of the ACT signal voltage and the RnIF signal voltage.
The sense amplifier 206 can also include circuitry configured to equilibrate the data lines D and D— (e.g., in association with preparing the sense amplifier for a sensing operation). In this example, the equilibration circuitry comprises a transistor 224 having a first source/drain region coupled to a first source/drain region of transistor 225-1 and data line D 205-1. A second source/drain region of transistor 224 can be coupled to a first source/drain region of transistor 225-2 and data line D— 205-2. A gate of transistor 224 can be coupled to gates of transistors 225-1 and 225-2.
The second source drain regions of transistors 225-1 and 225-2 can be coupled to an equilibration voltage 238, which can be equal to VDD/2, where VDD is a supply voltage associated with the array. The gates of transistors 224, 225-1, and 225-2 can be coupled to control signal 226 (EQ). As such, activating EQ can enable the transistors 224, 225-1, and 225-2, which can effectively short data line D to data line D— such that the data lines D and D— are equilibrated to equilibration voltage VDD/2. As described herein, a number of logical operations and/or shift operations can be performed using the sense amplifier 206 and compute component 231, and the result can be at least temporarily stored in the sense amplifier and/or compute component.
The sensing circuitry 250 in
As described herein, the sense amplifier 206 can, in conjunction with the compute component 231, be operated to perform various logical operations and/or shift operations (e.g., using data from an array as input). In a number of embodiments, the result of a logical operation and/or shift operation can be stored back to the array without transferring the data via a data line address access (e.g., without firing a column decode signal such that data is transferred to circuitry external to the array and sensing circuitry via local I/O lines). As such, embodiments of the present disclosure can enable performing various operations (e.g., logical operations, shift operations, mathematical operations, etc.) that contribute to performance of, for example, compression, decompression, alpha blending, and other graphics operations using less power than various previous approaches. Additionally, because various embodiments can reduce or eliminate moving (e.g., transferring) data across I/O lines in order to perform operations (e.g., between memory and a discrete processor, which may be off pitch), such embodiments may enable an increased parallel processing capability as compared to previous approaches.
Logic Table 213-1 illustrated in
The logic tables illustrated in
Via selective control of the state of the pass gates 207-1 and 207-2 and the swap transistors 242, each of the three columns of the upper portion of Logic Table 213-1 can be combined with each of the three columns of the lower portion of Logic Table 213-1 to provide 3×3=9 different result combinations, corresponding to nine different logical operations, as indicated by the various connecting paths shown at 275. The nine different selectable logical operations that can be implemented by the sensing circuitry 250 are summarized in Logic Table 213-2 illustrated in
The columns of Logic Table 213-2 illustrated in
For example, the results for the values of FF, FT, TF, and TT of “0000” are summarized as “A” because the result (initially stored in the compute component after the sense amplifier fires) is the same as the starting value in the compute component. Other columns of results are similarly annotated in row 247, where “A*B” intends A AND B, “A+B” intends A OR B, and “AXB” intends A XOR B. By convention, a bar over a data unit or a logical operation indicates an inverted value of the quantity shown under the bar. For example, AXB bar intends not A XOR B, which is also A XNOR B.
As shown in
In the example of
In some embodiments, the control logic 331-1, . . . , 331-7 may be responsible for fetching machine instructions (e.g., microcode instructions) from an array of memory cells (e.g., a DRAM array) in each bank 321-1, . . . , 321-7. The control logic 331-1, . . . , 331-7 may decode the microcode instructions into calls (e.g., microcode functions), implemented by the sequencers 332-1, . . . , 332-7. The microcode functions can be the operations that the sequencers 332-1, . . . , 332-7 receive and operate on to cause the PIM device 320 to perform particular operations, which may include the operations described herein.
For example, the control logic 331 can fetch machine instructions, which when executed, direct performance of operations by the intelligent frame buffer comprising the sensing circuitry 150 (e.g., sensing component stripes 124-0, . . . , 124-N−1 in
The Atto state machines 333-1, . . . , 333-7 may provide timing and be responsible for providing conflict free access to the arrays. In some embodiments, the sequencers 332-1, . . . , 332-7, and atomic state machines (Atto state machines) 333-1, . . . , 333-7 may be state machines and the control logic 331-1, . . . , 331-7 may be a very large instruction word (VLIW) type processing resource (e.g., containing a program counter, instruction memory, etc.)
For example, operations may be received to and operated on by the sequencers 332-1, . . . , 332-7 to cause sensing circuitry 250 shown in
In some embodiments, the sequencers 332-1, . . . , 332-7 may generate sequences of operation cycles for a DRAM array. For example, each sequence may be designed to perform operations, such as a Boolean logic operation (AND, OR, XOR, etc.), which together achieve a specific function, such as repetitively calculating the logic equations for a one (1) bit add in order to calculate a multiple bit sum. Each of these operations may be fed into a first in/first out (FIFO) buffer provided by the Atto state machines 333-1, . . . , 333-7 for providing timing coordination with the sensing circuitry 350 and/or logic 370 associated with the array of memory cells (e.g., DRAM arrays).
In the example embodiment shown in
As described herein, an operations component 172 (e.g., in combination with the controller 140) can direct that data values from a selectable row 119 of a selectable subarray 125 be at least temporarily stored (cached) in a sensing component stripe 124 coupled to that subarray and that a number of operations be performed on at least one of the data values in the sensing component stripe 124 by the sense amplifier 206 and/or the compute component 231. The operations component 172 can direct that these data values be subsequently returned for storage in the memory cells of the same or a different row of the subarray and/or that these data values be subsequently moved (e.g., via a shared I/O line) to selectably coupled sensing circuitry 150 (e.g., in a different sensing component stripe 124) for further processing (e.g., operations) to be performed on the data values.
As described herein, the operations component 172 can be selectably coupled to the sensing circuitry 150 of the plurality of subarrays. The operations component can, in various embodiments, be configured to direct performance of an operation by the sensing circuitry with respect to data stored in a selectable subarray of the respective plurality of subarrays. As such, the memory device can be configured to, as described herein, store a data value as received from a source (e.g., store a data set before an operation has been performed on the data) in a memory cell in a row of the selectable subarray (e.g., store at least part of the data set in the row). The memory device can be further configured to output the data value to an output processing component 171 (e.g., an I/O processing component to perform CPU/GPU type functions) associated with a controller subsequent to performance of the operation thereon in the sensing circuitry, as directed by the operations component 172.
In some embodiments, the controller with which output processing component 171-1 is associated may be a controller 140 for a bank 121 of the respective plurality of subarrays, as shown in
The memory device can be configured to receive (e.g., store) a plurality of compressed data values, as received from the source, to the selectable subarray. In various embodiments, the compressed data values can be from a compressed video recording having a number of video images, a compressed static image, and/or a compressed audio file, as described herein, among other possibilities. The memory device can be configured, in some embodiments, to decompress, as directed by the operations component, a plurality of the compressed data values, as received from the source to the selectable subarray, prior to performance of the operation thereon (e.g., prior to performance of an operation other than the decompression).
The memory device can be configured, in some embodiments, to perform an alpha blending operation, as directed by the operations component, on a plurality of data values as received from the source to (e.g., stored in) the selectable subarray. The memory device can be configured, in some embodiments, to perform a rescale operation, as directed by the operations component, on a plurality of resized data values received from the source to the selectable subarray.
The memory device can be configured, in some embodiments, to perform a depth texturing operation, as directed by the operations component, on a plurality of data values received from the source to the selectable subarray. A depth texture, also known as a shadow map, as used herein is intended to mean a texture that contains data from a depth buffer for a particular image. Each pixel value may contain a depth of the geometry in that pixel. A depth texture may be used to render appropriate shadows in the image. A depth texturing operation may be used to adjust the shadows that indicate image depth.
The memory device can be configured, in some embodiments, to perform a depth buffering operation, as directed by the operations component, on a plurality of data values received from the source to the selectable subarray. Depth buffering, also known as z-buffering, as used herein is intended to mean management of image depth coordinates in a three-dimensional graphics context. Depth buffering can be used to affect visibility, by determining which elements of a rendered image are visible and which are hidden. When an image is rendered, the depth of a generated pixel can be stored in a buffer (e.g., the z-buffer and/or a depth buffer). If a second object of the image is to be rendered in the same pixel, the depths of the two objects can be compared and one object can override the other object in the pixel based on which object is closer to an observer. For example, depth buffering may allow an image to appropriately reproduce depth perception where a close object can hide an object that is farther away. In addition, depth buffering may allow rendering a surface from a point of view of a light source through adjusting of shadows by the depth texturing technique.
In some embodiments, the alpha blending operation, the rescale operation, the depth texturing operation, and/or the depth buffering operation, among other possible operations, can be performed on a plurality of data values decompressed, as directed by the operations component, from compressed data values received from the source to the selectable subarray. The operations component can be further configured to direct storage of data values to a selectable subarray after decompression and a second operation (e.g., alpha blending, among possible others) have been performed thereon. In various embodiments, the selectable subarray for storage may be the same selectable subarray to which the data values were originally stored or a different selectable subarray.
The source that provides the data values to the selectable subarrays may, in some embodiments, be a controller 140 for a bank 121 of the respective plurality of subarrays, as shown in
A host may be selected from a group of devices that can provide data suitable for digital processing, as described herein. As such, the host may, for example, be a personal laptop computer, a desktop computer, a tablet computer, a digital camera, a smart phone, a memory card reader, a digital versatile disc player, a JPEG compressed image, a MPEG compressed audio recording, a MPEG compressed video recording, and/or an audio input, among various other types of hosts.
As such, the memory device can be configured to receive a plurality of data values from a source, where each of the plurality of data values can be received to (e.g., stored by) at least one selectable subarray in the respective plurality of subarrays. The memory device also may be configured to output the plurality of data values to an output processing component 171-1 and/or 171-2 associated with a controller subsequent to performance of an operation thereon, as directed by the operations component 172 (e.g., using instructions stored in the operations component).
The memory device can be configured to receive the data values from an input processing component 171-1 and/or 171-2. The I/O processing component 171-1 and/or 171-2 may be associated with a controller 140 for a bank, or a channel controller 143 associated with a plurality of banks, a plurality of memory devices, and/or a host. The input processing component 171-1 and/or 171-2 can be configured to perform various CPU/GPU type functions (operations), including to select a row of the selectable subarray for storage of the data values inputted from the source to the selectable subarray. The plurality of data values can be stored (cached) in at least one selectable sensing component stripe (e.g., 124) coupled to at least one selectable subarray. As such, the plurality of data values can be stored in at least one selectable row of the at least one selectable subarray.
The memory device can be configured to receive the data values from an input processing component 171-1 and/or 171-2 configured to compress the data values to correspond to a storage capacity of (e.g., a number of memory cells in) a selected row of the selectable subarray. For example, in a video and/or audio file having more than 16,384 data values (bits), a compression algorithm may be directed and/or performed to compress the data in the video and/or audio file to fit a row of a subarray with 16,384 memory cells. The compressed data values may be sent into (e.g., stored in) the selected row. In some embodiments, the compression and/or sending of the data values may be directed by the operations component 172 described herein.
In some embodiments, the input processing component 171-1 and/or 171-2 may be separate from a memory device 120 and/or a bank 121 that includes an array of memory cells 130. For example, the input processing component 171-1 and/or 171-2 may not be on pitch and/or on chip with the array of memory cells 130 and may be associated with and/or part of the host 110 or a separate component.
As described herein, the operations component 172 can be configured to store a set of instructions to direct the performance of the operation by the sensing circuitry with respect to the data stored in the selectable subarray of the respective plurality of subarrays. The operations component can be configured to direct the sensing circuitry in a sensing component stripe (e.g., 124) to perform the operation. The sensing circuitry can include a sense amplifier and/or a compute component (e.g., 206 and 231, respectively, in
The circuitry described herein is configured such that the operation performed in the sensing component stripe can be a first operation performed on the data value subsequent to storage of the data value in the selected row of the selectable subarray. The operations component 172 can be configured to move a data value (e.g., via a shared I/O line) corresponding to a result of an operation with respect to data stored in a first selectable subarray of the respective plurality of subarrays to a memory cell in a row of a second selectable subarray of the respective plurality of subarrays.
As such, the operations component can be configured to direct performance of a first operation by the sensing circuitry of the first subarray with respect to data values stored in the first subarray and direct performance of a second operation by the sensing circuitry of a second subarray with respect to data values stored in the second subarray. The first and second operations can be at least part of a sequence of operations. In some embodiments, the second operation can be different than the first operation (e.g., in a sequence of operations). The operations component can be configured to execute a first set of instructions to direct performance of the first operation and execute a second set of instructions to direct performance of the second operation. The sensing circuitry can be configured to couple to the respective plurality of subarrays to implement parallel movement of data values stored in a first subarray, upon which a first operation has been performed, to a plurality of memory cells in a second subarray. In various embodiments, the memory device can include a plurality of I/O lines shared by the respective plurality of subarrays and configured to couple to the sensing circuitry of a first subarray and a second subarray to selectably implement parallel movement of a plurality of data values stored in the first subarray, upon which a first operation has been performed, to a plurality of memory cells in the second subarray.
The apparatus can include a bank arbiter (e.g., as shown at 145 and described in connection with
As described herein, the memory device can include a sensing component stripe (e.g., 124) configured to include a number of a plurality of sense amplifiers and compute components that corresponds to a number of the plurality of columns of the memory cells (e.g., where each column of memory cells is coupled to a sense amplifier and a compute component). The number of a plurality of sensing component stripes in the bank section (e.g., 124-0 through 124-N−1) can correspond to a number of a plurality of subarrays in the bank section (e.g., 125-0 through 125-N−1).
As described herein, an array of memory cells can include a column of memory cells having a pair of complementary sense (digit) lines (e.g., 205-1 and 205-2 in
An operation performed in the sensing circuitry 150 of the intelligent frame buffer may, in some embodiments, be a first operation performed on the data value subsequent to storage of the data value in the selected row of the selectable subarray. The operation performed in the sensing circuitry 150 of the intelligent frame buffer may be at least a second operation (e.g., second, third, fourth, etc.) performed on the data value subsequent to storage of the data value in the selectable subarray. For example, the first operation may have been decompression performed on the data by the sensing circuitry in a sensing component stripe of a selectable subarray and the second operation may be a different operation (e.g., alpha blending, among other possible operations) performed on the decompressed data.
The controller 140 can be configured to provide a set of instructions to the operations component to direct performance of a number of operations with respect to data stored in the selectable subarray of the respective plurality of subarrays. In some embodiments, the host 110, as described herein, can provide the data to the controller for the controller, in combination with the operations component 172, to issue the commands for operations using compressed and decompressed data.
Embodiments described herein provide a method for operating a memory device 120 to implement the operations using compressed and decompressed data. The operations may be performed by execution of non-transitory instructions (e.g., from operations component 172) by a processing resource (e.g., sensing circuitry 150 in sensing component stripes 124-0, 124-1, . . . , 124-N−1). The method can include storing data, as received from a source, as data values in a selected row of a selected subarray (e.g., in one or more rows selected in any one of, any combinations of, or all of subarrays 125-0, . . . , 125-N−1, for example) of the memory device and performing a digital processing operation on the data values, as directed by an operations component 172 coupled to the selected subarray. The data values can be output to an output processing component 171-1 and/or 171-2 associated with a controller subsequent to performance of the digital processing operation thereon, as directed by the operations component.
As described herein, the method can include receiving (e.g., in parallel via a number of I/O lines) the data values to (e.g., storing the data values in) either a sensing component stripe (e.g., 124) and/or a row (e.g., 319) of the selected subarray. The data values can be stored as at least two thousand bits (e.g., prior to performing an operation thereon). Storing a file size of at least two thousand bits, corresponding to some or all of the number of memory cells (e.g., 16,384) in a row of a subarray, as described herein, may be too many bits for a cache save for processing in a GPU associated with a host.
The method can further include compressing the received data values (e.g., in an image frame or a plurality of received image frames) to a number of bits that corresponds to a number of sense amplifiers and/or compute components in a sensing component stripe and/or a number of memory cells in a row of the selected subarray. In some embodiments, a compressed image frame can be stored as received from the source. The compressed image frame can be decompressed, as directed by the operations component, and the digital processing operation can be performed on the decompressed data values, as directed by the operations component.
The method can further include receiving (e.g., at least temporarily storing and/or caching) the data values (e.g., as received either from the source or as sensed from the selected subarray) to a sensing component stripe coupled to the selected subarray. In some embodiments, the digital processing operation can be performed on the data values by sensing circuitry in the sensing component stripe, as directed by the operations component. The method can further include performing the digital processing operation on the data values in the sensing component stripe coupled to the selected subarray either as received from the source or as sensed from the selected subarray.
For example, the method can include receiving the data values to (e.g., storing the data values in) a sensing component stripe selectably coupled to the selected subarray and performing the operation on the data values using the sensing circuitry in the sensing component stripe, as directed by the operations component. Alternatively, the method can include receiving the data values to a selected row of the selected subarray, moving the data values from the selected row to the sensing component stripe selectably coupled to the selected row, and performing the operation on the data values using sensing circuitry in the sensing component stripe. In some embodiments, the sense amplifiers and/or compute components of the sensing component stripe may be formed on pitch with the selected subarray and/or the selected row (e.g., the corresponding pairs of complementary sense lines thereof). The data values can be received to (e.g., stored in) a selected subarray of the memory device subsequent to performance of the digital processing operation thereon (e.g., in the sensing component stripe) and prior to outputting the data values to the output processing component 171-1 and/or 171-2.
The method can further include receiving the data values from the source (e.g., a source device). The source device can, in various embodiments, be a controller 140 for a bank 121 of a plurality of subarrays and/or a channel controller 143 that is associated with a plurality of banks (e.g., 121-1, 121-2, . . . , 121-N) of the respective plurality of subarrays and/or is connected to a host 110 coupled to the memory device. In some embodiments, the source can be on chip with the sensing circuitry 150 and memory 130 described herein. For example, the source may be a controller 140 for a bank 121 of a respective plurality of subarrays 125, as shown in
While example embodiments including various combinations and configurations of controllers, operations components, sensing circuitry, sense amplifiers, compute components, and/or sensing component stripes, etc., have been illustrated and described herein, embodiments of the present disclosure are not limited to those combinations explicitly recited herein. Other combinations and configurations of the controllers, operations components, sensing circuitry, sense amplifiers, compute components, and/or sensing component stripes, etc., disclosed herein are expressly included within the scope of this disclosure.
Although specific embodiments have been illustrated and described herein, those of ordinary skill in the art will appreciate that an arrangement calculated to achieve the same results can be substituted for the specific embodiments shown. This disclosure is intended to cover adaptations or variations of one or more embodiments of the present disclosure. It is to be understood that the above description has been made in an illustrative fashion, and not a restrictive one. Combination of the above embodiments, and other embodiments not specifically described herein will be apparent to those of skill in the art upon reviewing the above description. The scope of the one or more embodiments of the present disclosure includes other applications in which the above structures and processes are used. Therefore, the scope of one or more embodiments of the present disclosure should be determined with reference to the appended claims, along with the full range of equivalents to which such claims are entitled.
In the foregoing Detailed Description, some features are grouped together in a single embodiment for the purpose of streamlining the disclosure. This method of disclosure is not to be interpreted as reflecting an intention that the disclosed embodiments of the present disclosure have to use more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive subject matter lies in less than all features of a single disclosed embodiment. Thus, the following claims are hereby incorporated into the Detailed Description, with each claim standing on its own as a separate embodiment.
| Number | Name | Date | Kind |
|---|---|---|---|
| 4380046 | Fung | Apr 1983 | A |
| 4435792 | Bechtolsheim | Mar 1984 | A |
| 4435793 | Ochii | Mar 1984 | A |
| 4496944 | Collmeyer | Jan 1985 | A |
| 4727474 | Batcher | Feb 1988 | A |
| 4843264 | Galbraith | Jun 1989 | A |
| 4958378 | Bell | Sep 1990 | A |
| 4977542 | Matsuda et al. | Dec 1990 | A |
| 5023838 | Herbert | Jun 1991 | A |
| 5034636 | Reis et al. | Jul 1991 | A |
| 5201039 | Sakamura | Apr 1993 | A |
| 5210850 | Kelly et al. | May 1993 | A |
| 5253308 | Johnson | Oct 1993 | A |
| 5276643 | Hoffmann et al. | Jan 1994 | A |
| 5325519 | Long et al. | Jun 1994 | A |
| 5367488 | An | Nov 1994 | A |
| 5379257 | Matsumura et al. | Jan 1995 | A |
| 5386379 | Ali-Yahia et al. | Jan 1995 | A |
| 5398213 | Yeon et al. | Mar 1995 | A |
| 5440482 | Davis | Aug 1995 | A |
| 5446690 | Tanaka et al. | Aug 1995 | A |
| 5467087 | Chu | Nov 1995 | A |
| 5473576 | Matsui | Dec 1995 | A |
| 5481500 | Reohr et al. | Jan 1996 | A |
| 5485373 | Davis et al. | Jan 1996 | A |
| 5506811 | McLaury | Apr 1996 | A |
| 5615404 | Knoll et al. | Mar 1997 | A |
| 5638128 | Hoogenboom | Jun 1997 | A |
| 5638317 | Tran | Jun 1997 | A |
| 5654936 | Cho | Aug 1997 | A |
| 5678021 | Pawate et al. | Oct 1997 | A |
| 5706451 | Lightbody et al. | Jan 1998 | A |
| 5724291 | Matano | Mar 1998 | A |
| 5724366 | Furutani | Mar 1998 | A |
| 5751987 | Mahant-Shetti et al. | May 1998 | A |
| 5787458 | Miwa | Jul 1998 | A |
| 5854636 | Watanabe et al. | Dec 1998 | A |
| 5867429 | Chen et al. | Feb 1999 | A |
| 5870504 | Nemoto et al. | Feb 1999 | A |
| 5915084 | Wendell | Jun 1999 | A |
| 5935263 | Keeth et al. | Aug 1999 | A |
| 5986942 | Sugibayashi | Nov 1999 | A |
| 5991209 | Chow | Nov 1999 | A |
| 5991785 | Alidina et al. | Nov 1999 | A |
| 6005799 | Rao | Dec 1999 | A |
| 6009020 | Nagata | Dec 1999 | A |
| 6092186 | Betker et al. | Jul 2000 | A |
| 6122211 | Morgan et al. | Sep 2000 | A |
| 6125071 | Kohno et al. | Sep 2000 | A |
| 6134164 | Lattimore et al. | Oct 2000 | A |
| 6147514 | Shiratake | Nov 2000 | A |
| 6151244 | Fujino et al. | Nov 2000 | A |
| 6157578 | Brady | Dec 2000 | A |
| 6163862 | Adams et al. | Dec 2000 | A |
| 6166942 | Vo et al. | Dec 2000 | A |
| 6172918 | Hidaka | Jan 2001 | B1 |
| 6175514 | Henderson | Jan 2001 | B1 |
| 6181698 | Hariguchi | Jan 2001 | B1 |
| 6208544 | Beadle et al. | Mar 2001 | B1 |
| 6226215 | Yoon | May 2001 | B1 |
| 6301153 | Takeuchi et al. | Oct 2001 | B1 |
| 6301164 | Manning et al. | Oct 2001 | B1 |
| 6304477 | Naji | Oct 2001 | B1 |
| 6389507 | Sherman | May 2002 | B1 |
| 6418498 | Martwick | Jul 2002 | B1 |
| 6466499 | Blodgett | Oct 2002 | B1 |
| 6510098 | Taylor | Jan 2003 | B1 |
| 6563754 | Lien et al. | May 2003 | B1 |
| 6578058 | Nygaard | Jun 2003 | B1 |
| 6704022 | Aleksic | Mar 2004 | B1 |
| 6731542 | Le et al. | May 2004 | B1 |
| 6754746 | Leung et al. | Jun 2004 | B1 |
| 6768679 | Le et al. | Jul 2004 | B1 |
| 6807614 | Chung | Oct 2004 | B2 |
| 6816422 | Hamade et al. | Nov 2004 | B2 |
| 6819612 | Achter | Nov 2004 | B1 |
| 6894549 | Eliason | May 2005 | B2 |
| 6943579 | Hazanchuk et al. | Sep 2005 | B1 |
| 6948056 | Roth | Sep 2005 | B1 |
| 6950771 | Fan et al. | Sep 2005 | B1 |
| 6950898 | Merritt et al. | Sep 2005 | B2 |
| 6956770 | Khalid et al. | Oct 2005 | B2 |
| 6961272 | Schreck | Nov 2005 | B2 |
| 6965648 | Smith et al. | Nov 2005 | B1 |
| 6985394 | Kim | Jan 2006 | B2 |
| 6987693 | Cernea et al. | Jan 2006 | B2 |
| 7020017 | Chen et al. | Mar 2006 | B2 |
| 7028170 | Saulsbury | Apr 2006 | B2 |
| 7045834 | Tran et al. | May 2006 | B2 |
| 7054178 | Shiah et al. | May 2006 | B1 |
| 7061817 | Raad et al. | Jun 2006 | B2 |
| 7079407 | Dimitrelis | Jul 2006 | B1 |
| 7173857 | Kato et al. | Feb 2007 | B2 |
| 7187585 | Li et al. | Mar 2007 | B2 |
| 7196928 | Chen | Mar 2007 | B2 |
| 7260565 | Lee et al. | Aug 2007 | B2 |
| 7260672 | Gamey | Aug 2007 | B2 |
| 7372715 | Han | May 2008 | B2 |
| 7400532 | Aritome | Jul 2008 | B2 |
| 7406494 | Magee | Jul 2008 | B2 |
| 7447720 | Beaumont | Nov 2008 | B2 |
| 7454451 | Beaumont | Nov 2008 | B2 |
| 7457181 | Lee et al. | Nov 2008 | B2 |
| 7535769 | Cernea | May 2009 | B2 |
| 7546438 | Chung | Jun 2009 | B2 |
| 7562198 | Noda et al. | Jul 2009 | B2 |
| 7574466 | Beaumont | Aug 2009 | B2 |
| 7602647 | Li et al. | Oct 2009 | B2 |
| 7663928 | Tsai et al. | Feb 2010 | B2 |
| 7685365 | Rajwar et al. | Mar 2010 | B2 |
| 7692466 | Ahmadi | Apr 2010 | B2 |
| 7752417 | Manczak et al. | Jul 2010 | B2 |
| 7791962 | Noda et al. | Sep 2010 | B2 |
| 7796453 | Riho et al. | Sep 2010 | B2 |
| 7805587 | Van Dyke et al. | Sep 2010 | B1 |
| 7808854 | Takase | Oct 2010 | B2 |
| 7827372 | Bink et al. | Nov 2010 | B2 |
| 7869273 | Lee et al. | Jan 2011 | B2 |
| 7898864 | Dong | Mar 2011 | B2 |
| 7924628 | Danon et al. | Apr 2011 | B2 |
| 7937535 | Ozer et al. | May 2011 | B2 |
| 7957206 | Bauser | Jun 2011 | B2 |
| 7979667 | Allen et al. | Jul 2011 | B2 |
| 7996749 | Ding et al. | Aug 2011 | B2 |
| 8042082 | Solomon | Oct 2011 | B2 |
| 8045391 | Mokhlesi | Oct 2011 | B2 |
| 8059438 | Chang et al. | Nov 2011 | B2 |
| 8095825 | Hirotsu et al. | Jan 2012 | B2 |
| 8117462 | Snapp et al. | Feb 2012 | B2 |
| 8164942 | Gebara et al. | Apr 2012 | B2 |
| 8208328 | Hong | Jun 2012 | B2 |
| 8213248 | Moon et al. | Jul 2012 | B2 |
| 8223568 | Sec | Jul 2012 | B2 |
| 8238173 | Akerib et al. | Aug 2012 | B2 |
| 8274841 | Shimano et al. | Sep 2012 | B2 |
| 8279683 | Klein | Oct 2012 | B2 |
| 8310884 | Iwai et al. | Nov 2012 | B2 |
| 8332367 | Bhattacherjee et al. | Dec 2012 | B2 |
| 8339824 | Cooke | Dec 2012 | B2 |
| 8339883 | Yu et al. | Dec 2012 | B2 |
| 8347154 | Bahali et al. | Jan 2013 | B2 |
| 8351292 | Matano | Jan 2013 | B2 |
| 8356144 | Hessel et al. | Jan 2013 | B2 |
| 8417921 | Gonion et al. | Apr 2013 | B2 |
| 8462532 | Argyres | Jun 2013 | B1 |
| 8484276 | Carlson et al. | Jul 2013 | B2 |
| 8495438 | Roine | Jul 2013 | B2 |
| 8503250 | Demone | Aug 2013 | B2 |
| 8526239 | Kim | Sep 2013 | B2 |
| 8533245 | Cheung | Sep 2013 | B1 |
| 8555037 | Gonion | Oct 2013 | B2 |
| 8599613 | Abiko et al. | Dec 2013 | B2 |
| 8605015 | Guttag et al. | Dec 2013 | B2 |
| 8625376 | Jung et al. | Jan 2014 | B2 |
| 8644101 | Jun et al. | Feb 2014 | B2 |
| 8650232 | Stortz et al. | Feb 2014 | B2 |
| 8873272 | Lee | Oct 2014 | B2 |
| 8958257 | Yoon | Feb 2015 | B2 |
| 8964496 | Manning | Feb 2015 | B2 |
| 8971124 | Manning | Mar 2015 | B1 |
| 9015390 | Klein | Apr 2015 | B2 |
| 9047193 | Lin et al. | Jun 2015 | B2 |
| 9165023 | Moskovich et al. | Oct 2015 | B2 |
| 9628111 | Henry et al. | Apr 2017 | B2 |
| 9768803 | Henry et al. | Sep 2017 | B2 |
| 20010007112 | Porterfield | Jul 2001 | A1 |
| 20010008492 | Higashiho | Jul 2001 | A1 |
| 20010010057 | Yamada | Jul 2001 | A1 |
| 20010028584 | Nakayama et al. | Oct 2001 | A1 |
| 20010043089 | Forbes et al. | Nov 2001 | A1 |
| 20020059355 | Peleg et al. | May 2002 | A1 |
| 20030046485 | Zitlaw | Mar 2003 | A1 |
| 20030167426 | Slobodnik | Sep 2003 | A1 |
| 20030222879 | Lin et al. | Dec 2003 | A1 |
| 20040012599 | Laws | Jan 2004 | A1 |
| 20040073592 | Kim et al. | Apr 2004 | A1 |
| 20040073773 | Demjanenko | Apr 2004 | A1 |
| 20040085840 | Vali et al. | May 2004 | A1 |
| 20040095826 | Perner | May 2004 | A1 |
| 20040154002 | Ball et al. | Aug 2004 | A1 |
| 20040205289 | Srinivasan | Oct 2004 | A1 |
| 20040240251 | Nozawa et al. | Dec 2004 | A1 |
| 20050015557 | Wang et al. | Jan 2005 | A1 |
| 20050078514 | Scheuerlein et al. | Apr 2005 | A1 |
| 20050097417 | Agrawal et al. | May 2005 | A1 |
| 20060047937 | Selvaggi et al. | Mar 2006 | A1 |
| 20060069849 | Rudelic | Mar 2006 | A1 |
| 20060092681 | Kawakubo | May 2006 | A1 |
| 20060146623 | Mizuno et al. | Jul 2006 | A1 |
| 20060149804 | Luick et al. | Jul 2006 | A1 |
| 20060181917 | Kang et al. | Aug 2006 | A1 |
| 20060215432 | Wickeraad et al. | Sep 2006 | A1 |
| 20060225072 | Lari et al. | Oct 2006 | A1 |
| 20060291282 | Liu et al. | Dec 2006 | A1 |
| 20070103986 | Chen | May 2007 | A1 |
| 20070171747 | Hunter et al. | Jul 2007 | A1 |
| 20070180006 | Gyoten et al. | Aug 2007 | A1 |
| 20070180184 | Sakashita et al. | Aug 2007 | A1 |
| 20070195602 | Fong et al. | Aug 2007 | A1 |
| 20070263876 | De Waal et al. | Nov 2007 | A1 |
| 20070285131 | Sohn | Dec 2007 | A1 |
| 20070285979 | Turner | Dec 2007 | A1 |
| 20070291532 | Tsuji | Dec 2007 | A1 |
| 20080025073 | Arsovski | Jan 2008 | A1 |
| 20080037333 | Kim et al. | Feb 2008 | A1 |
| 20080052711 | Forin et al. | Feb 2008 | A1 |
| 20080137388 | Krishnan et al. | Jun 2008 | A1 |
| 20080165601 | Matick et al. | Jul 2008 | A1 |
| 20080178053 | Gorman et al. | Jul 2008 | A1 |
| 20080215937 | Dreibelbis et al. | Sep 2008 | A1 |
| 20090066840 | Karlov | Mar 2009 | A1 |
| 20090067218 | Graber | Mar 2009 | A1 |
| 20090154238 | Lee | Jun 2009 | A1 |
| 20090154273 | Borot et al. | Jun 2009 | A1 |
| 20090254697 | Akerib | Oct 2009 | A1 |
| 20100007678 | Tsai | Jan 2010 | A1 |
| 20100067296 | Li | Mar 2010 | A1 |
| 20100091582 | Vali et al. | Apr 2010 | A1 |
| 20100172190 | Lavi et al. | Jul 2010 | A1 |
| 20100210076 | Gruber et al. | Aug 2010 | A1 |
| 20100226183 | Kim | Sep 2010 | A1 |
| 20100231616 | Park | Sep 2010 | A1 |
| 20100308858 | Noda et al. | Dec 2010 | A1 |
| 20100332895 | Billing et al. | Dec 2010 | A1 |
| 20110051523 | Manabe et al. | Mar 2011 | A1 |
| 20110063919 | Chandrasekhar et al. | Mar 2011 | A1 |
| 20110093662 | Walker et al. | Apr 2011 | A1 |
| 20110103151 | Kim et al. | May 2011 | A1 |
| 20110119467 | Cadambi et al. | May 2011 | A1 |
| 20110122695 | Li et al. | May 2011 | A1 |
| 20110140741 | Zerbe et al. | Jun 2011 | A1 |
| 20110219260 | Nobunaga et al. | Sep 2011 | A1 |
| 20110267883 | Lee et al. | Nov 2011 | A1 |
| 20110317496 | Bunce et al. | Dec 2011 | A1 |
| 20120005397 | Lim et al. | Jan 2012 | A1 |
| 20120017039 | Margetts | Jan 2012 | A1 |
| 20120023281 | Kawasaki et al. | Jan 2012 | A1 |
| 20120120705 | Mitsubori et al. | May 2012 | A1 |
| 20120134216 | Singh | May 2012 | A1 |
| 20120134226 | Chow | May 2012 | A1 |
| 20120135225 | Chow | May 2012 | A1 |
| 20120140540 | Agam et al. | Jun 2012 | A1 |
| 20120182798 | Hosono et al. | Jul 2012 | A1 |
| 20120195146 | Jun et al. | Aug 2012 | A1 |
| 20120198310 | Tran et al. | Aug 2012 | A1 |
| 20120210048 | Chang | Aug 2012 | A1 |
| 20120219236 | Ali | Aug 2012 | A1 |
| 20120246380 | Akerib et al. | Sep 2012 | A1 |
| 20120265964 | Murata et al. | Oct 2012 | A1 |
| 20120281486 | Rao et al. | Nov 2012 | A1 |
| 20120303627 | Keeton et al. | Nov 2012 | A1 |
| 20130003467 | Klein | Jan 2013 | A1 |
| 20130061006 | Hein | Mar 2013 | A1 |
| 20130107623 | Kavalipurapu et al. | May 2013 | A1 |
| 20130107652 | Yoon | May 2013 | A1 |
| 20130117541 | Choquette et al. | May 2013 | A1 |
| 20130124783 | Yoon et al. | May 2013 | A1 |
| 20130132702 | Patel et al. | May 2013 | A1 |
| 20130138646 | Sirer et al. | May 2013 | A1 |
| 20130163362 | Kim | Jun 2013 | A1 |
| 20130173888 | Hansen et al. | Jul 2013 | A1 |
| 20130205114 | Badam et al. | Aug 2013 | A1 |
| 20130219112 | Okin et al. | Aug 2013 | A1 |
| 20130227361 | Bowers et al. | Aug 2013 | A1 |
| 20130283122 | Anholt et al. | Oct 2013 | A1 |
| 20130286705 | Grover et al. | Oct 2013 | A1 |
| 20130326154 | Haswell | Dec 2013 | A1 |
| 20130332707 | Gueron et al. | Dec 2013 | A1 |
| 20140137119 | Hyde | May 2014 | A1 |
| 20140185395 | Seo | Jul 2014 | A1 |
| 20140215185 | Danielsen | Jul 2014 | A1 |
| 20140250279 | Manning | Sep 2014 | A1 |
| 20140344934 | Jorgensen | Nov 2014 | A1 |
| 20150029798 | Manning | Jan 2015 | A1 |
| 20150042380 | Manning | Feb 2015 | A1 |
| 20150063052 | Manning | Mar 2015 | A1 |
| 20150078108 | Cowles et al. | Mar 2015 | A1 |
| 20150279466 | Manning | Mar 2015 | A1 |
| 20150120987 | Wheeler | Apr 2015 | A1 |
| 20150134713 | Wheeler | May 2015 | A1 |
| 20150270015 | Murphy et al. | Sep 2015 | A1 |
| 20150324290 | Leidel | Nov 2015 | A1 |
| 20150325272 | Murphy | Nov 2015 | A1 |
| 20150356009 | Wheeler et al. | Dec 2015 | A1 |
| 20150356022 | Leidel et al. | Dec 2015 | A1 |
| 20150357007 | Manning et al. | Dec 2015 | A1 |
| 20150357008 | Manning et al. | Dec 2015 | A1 |
| 20150357019 | Wheeler et al. | Dec 2015 | A1 |
| 20150357020 | Manning | Dec 2015 | A1 |
| 20150357021 | Hush | Dec 2015 | A1 |
| 20150357022 | Hush | Dec 2015 | A1 |
| 20150357023 | Hush | Dec 2015 | A1 |
| 20150357024 | Hush et al. | Dec 2015 | A1 |
| 20150357047 | Tiwari | Dec 2015 | A1 |
| 20160062672 | Wheeler | Mar 2016 | A1 |
| 20160062673 | Tiwari | Mar 2016 | A1 |
| 20160062692 | Finkbeiner et al. | Mar 2016 | A1 |
| 20160062733 | Tiwari | Mar 2016 | A1 |
| 20160063284 | Tiwari | Mar 2016 | A1 |
| 20160064045 | La Fratta | Mar 2016 | A1 |
| 20160064047 | Tiwari | Mar 2016 | A1 |
| 20170127068 | Lapicque | May 2017 | A1 |
| Number | Date | Country |
|---|---|---|
| 102141905 | Aug 2011 | CN |
| 0214718 | Mar 1987 | EP |
| 2026209 | Feb 2009 | EP |
| H0831168 | Feb 1996 | JP |
| 2009259193 | Mar 2015 | JP |
| 10-0211482 | Aug 1998 | KR |
| 10-2010-0134235 | Dec 2010 | KR |
| 10-2013-0049421 | May 2013 | KR |
| 201640386 | Nov 2016 | TW |
| 201643758 | Dec 2016 | TW |
| 2001065359 | Sep 2001 | WO |
| 2010079451 | Jul 2010 | WO |
| 2013062596 | May 2013 | WO |
| 2013081588 | Jun 2013 | WO |
| 2013095592 | Jun 2013 | WO |
| Entry |
|---|
| Dybdahl, et al., “Destructive-Read in Embedded DRAM, Impact on Power Consumption,” Apr. 2006, (10 pgs.), vol. 2, Issue 2, Journal of Embedded Computing-Issues in embedded single-chip multicore architectures. |
| Kogge, et al., “Processing in Memory: Chips to Petaflops,” May 23, 1997, (8 pgs.), retrieved from: http://www.cs.ucf.edu/courses/cda5106/summer02/papers/kogge97PIM.pdf. |
| Draper, et al., “The Architecture of the DIVA Processing-In-Memory Chip,” Jun. 22-26, 2002, (12 pgs.), ICS '02, retrieved from: http://www.isi.edu/˜draper/papers/ics02.pdf. |
| Adibi, et al., “Processing-In-Memory Technology for Knowledge Discovery Algorithms,” Jun. 25, 2006, (10 pgs.), Proceeding of the Second International Workshop on Data Management on New Hardware, retrieved from: http://www.cs.cmu.edu/˜damon2006/pdf/adibi06inmemory.pdf. |
| U.S. Appl. No. 13/449,082, entitled, “Methods and Apparatus for Pattern Matching,” filed Apr. 17, 2012, (37 pgs.). |
| U.S. Appl. No. 13/743,686, entitled, “Weighted Search and Compare in a Memory Device,” filed Jan. 17, 2013, (25 pgs.). |
| U.S. Appl. No. 13/774,636, entitled, “Memory as a Programmable Logic Device,” filed Feb. 22, 2013, (30 pgs.). |
| U.S. Appl. No. 13/774,553, entitled, “Neural Network in a Memory Device,” filed Feb. 22, 2013, (63 pgs.). |
| U.S. Appl. No. 13/796,189, entitled, “Performing Complex Arithmetic Functions in a Memory Device,” filed Mar. 12, 2013, (23 pgs.). |
| Boyd et al., “On the General Applicability of Instruction-Set Randomization”, Jul.-Sep. 2010, (14 pgs.), vol. 7, Issue 3, IEEE Transactions on Dependable and Secure Computing. |
| Stojmenovic, “Multiplicative Circulant Networks Topological Properties and Communication Algorithms”, (25 pgs.), Discrete Applied Mathematics 77 (1997) 281-305. |
| “4.9.3 Minloc and Maxloc”, Jun. 12, 1995, (5pgs.), Message Passing Interface Forum 1.1, retrieved from http://www.mpi-forum.org/docs/mpi-1.1/mpi-11-html/node79.html. |
| Derby, et al., “A High-Performance Embedded DSP Core with Novel SIMD Features”, Apr. 6-10, 2003, (4 pgs), vol. 2, pp. 301-304, 2003 IEEE International Conference on Accoustics, Speech, and Signal Processing. |
| Debnath, Biplob, Bloomflash: Bloom Filter on Flash-Based Storage, 2011 31st Annual Conference on Distributed Computing Systems, Jun. 20-24, 2011, 10 pgs. |
| Pagiamtzis, Kostas, “Content-Addressable Memory Introduction”, Jun. 25, 2007, (6 pgs.), retrieved from: http://www.pagiamtzis.com/cam/camintro. |
| Pagiamtzis, et al., “Content-Addressable Memory (CAM) Circuits and Architectures: A Tutorial and Survey”, Mar. 2006, (16 pgs.), vol. 41, No. 3, IEEE Journal of Solid-State Circuits. |
| International Search Report and Written Opinion for PCT Application No. PCT/US2013/043702, dated Sep. 26, 2013, (11 pgs.). |
| Elliot, et al., “Computational RAM: Implementing Processors in Memory”, Jan.-Mar. 1999, (10 pgs.), vol. 16, Issue 1, IEEE Design and Test of Computers Magazine. |
| U.S. Appl. No. 13/774,636, entitled, “Memory as a Programmable Logic Device,” filed Feb. 22, 2013. |
| U.S. Appl. No. 13/774,553, entitled, “Neural Network in a Memory Device,” filed Feb. 2, 2013, (63 pgs). |
| International Search Report and Written Opinion for related PCT Application No. PCT/US2017/021758, dated May 24, 2017, 14 pages. |
| Office Action for related Taiwan Patent Application No. 106108692, dated Jan. 9, 2018, 25 pages. |
| Number | Date | Country | |
|---|---|---|---|
| 20170269865 A1 | Sep 2017 | US |
