System and method for rotation vector input

By introducing rotating vector register files and multiplication and accumulation circuits into the processor, the problem of data transfer delay in vector processing is solved, and more efficient data processing and computing capabilities are achieved.

CN119032342APending Publication Date: 2024-11-26QUALCOMM INC
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202380034226.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2022-05-02
Filing Date
2023-04-25
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

In the prior art, vector processors have processing bottlenecks in the data transfer process, especially the delay caused by multiple data transfers from memory to scalar registers, which affects the computing efficiency.

Method used

The rotation vector register file and multiplication and accumulation circuit are used to load the data in the rotation vector register into a dedicated part in a single pass, reducing the number of data transfers, and parallel processing is performed in the multiplication and accumulation circuit to reduce the circuit complexity.

Benefits of technology

It improves the efficiency of vector processing, reduces data transfer delay, reduces circuit complexity, and improves the processing capabilities of computing devices.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119032342A_ABST
    Figure CN119032342A_ABST
Patent Text Reader

Abstract

An apparatus includes a processor including a rotation vector register file, a second vector register file, and a multiply-accumulate circuit (MAC). The rotation vector register file includes rotation vector registers. The rotation vector register file is configured to rotate data in the rotation vector register. The second vector register file includes a source vector register. The MAC is configured to receive first input data from the rotation vector register file and second input data from the source vector register.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] I. Cross-Reference to Related Applications

[0002] This application claims the benefit of priority from commonly owned U.S. non-provisional patent application No. 17 / 661,707, filed on May 2, 2022, the contents of which are expressly incorporated herein by reference in their entirety.

[0003] II. Technical field

[0004] The present disclosure generally relates to data in vector registers for vector processing.

[0005] III. Description of Related Art

[0006] Technological advances have led to smaller and more powerful computing devices. For example, there are currently a variety of portable personal computing devices, including small, lightweight and user-friendly wireless phones (such as mobile and smart phones, tablet computers and laptop computers). These devices can communicate voice and data packets over wireless networks. In addition, many such devices incorporate additional functionality, such as digital still cameras, digital video cameras, digital recorders, and audio file players. In addition, such devices can process executable instructions, including software applications, such as web browser applications, which can be used to access the Internet. As such, these devices can include significant computing capabilities.

[0007] The computing device may include one or more digital signal processors (DSPs), network processing units (NPUs), network signal processors (NSPs), image processors, or other processing devices that perform vector processing, which includes performing multiple instances of common operations (e.g., multiplication operations) to process multiple elements of vector data in parallel. For example, a vector may include multiple sub-vector values, such as 16 sub-vector values ​​each including 4 half-word values. In an exemplary multiplication operation, for each sub-vector value of the vector, the first half-word is multiplied by the first one-byte value, the second half-word is multiplied by the second one-byte value, the third half-word is multiplied by the third one-byte value, and the fourth half-word is multiplied by the fourth one-byte value. The four multiplication products are added together, and the resulting sum is added to the accumulator vector register.

[0008] In some examples, each sub-vector value (e.g., 4 half-word values) may be read from a scalar register and provided as an input to a multiplication circuit. However, the data transfer of the half-word values ​​from a memory (e.g., a dynamic random access memory (DRAM), a static random access memory (SRAM), or another type of memory) to the scalar register may cause a processing bottleneck due to the scalar register being loaded via conventional processor operations involving multiple transfers of data (e.g., loading the sub-vector values ​​from the memory to a level 2 (L2) cache, from the L2 cache to a level 1 (L1) cache, and from the L1 cache to a scalar register in a register file).

[0009] IV. Summary of the Invention

[0010] According to a specific implementation of the present disclosure, a device includes a processor, the processor includes a rotation vector register file, a second vector register file and a multiply-accumulate circuit (MAC). The rotation vector register file includes a rotation vector register. The rotation vector register file is configured to rotate data in the rotation vector register. The second vector register file includes a source vector register. The MAC is configured to receive first input data from the rotation vector register file and receive second input data from the source vector register.

[0011] According to another specific implementation of the present disclosure, a processor-implemented method includes using a rotated vector register file to rotate data in a rotated vector register of the rotated vector register file. The processor-implemented method also includes receiving first input data from the rotated vector register file at a multiply-accumulate circuit (MAC). The processor-implemented method also includes receiving second input data from a source vector register of a second vector register file at the MAC.

[0012] According to another specific implementation of the present disclosure, a non-transitory computer-readable medium stores instructions that, when executed by a processor, cause the processor to use a rotation vector register file to rotate data in a rotation vector register of the rotation vector register file. When executed by the processor, the instructions also cause the processor to receive first input data from the rotation vector register file at a multiply-accumulate circuit (MAC). When executed by the processor, the instructions also cause the processor to receive second input data from a source vector register of a second vector register file at the MAC.

[0013] According to another specific implementation of the present disclosure, an apparatus includes a component for rotating data in a rotated vector register of a rotated vector register file. The apparatus also includes a component for receiving first input data from the rotated vector register file at a multiply-accumulate circuit (MAC). The apparatus also includes a component for receiving second input data from a source vector register of a second vector register file at the MAC.

[0014] Other aspects, advantages, and features of the present disclosure will become apparent upon review of the entire application, including the Brief Description of the Drawings, Detailed Description, and Claims.

[0015] V. Description of the Figures

[0016] Figure 1A is a block diagram of certain illustrative aspects of a system operable to rotate a vector input according to some examples of the present disclosure.

[0017] Figure 1B According to some examples of the present disclosure Figure 1A A diagram of illustrative aspects of rotation of data of a rotation vector register of a system.

[0018] Figure 2A is operable to rotate a vector input for executing instructions according to some examples of the present disclosure Figure 1A A diagram of illustrative aspects of components of a system.

[0019] Figure 2B According to some examples of the present disclosure Figure 1A FIG. 1 is a diagram of illustrative aspects of changes in performance (eg, operation count) of a system relative to changes in operating intensity (eg, operations per byte).

[0020] Figure 3 According to some examples of the present disclosure Figure 2A Before the first processing stage of the execution of the instruction Figure 1A A diagram of illustrative aspects of a first state of a vector register of a system.

[0021] Figure 4 According to some examples of the present disclosure Figure 2A During the first sub-stage of the first processing stage of execution of an instruction Figure 1A A diagram of illustrative aspects of the operation of components of a system.

[0022] Figure 5 According to some examples of the present disclosure Figure 2A After the first sub-stage of the first processing stage of the execution of the instruction Figure 1A A diagram of illustrative aspects of a second state of a vector register of a system.

[0023] Figure 6 According to some examples of the present disclosure Figure 2A During the second sub-stage of the first processing stage of execution of the instruction Figure 1A A diagram of illustrative aspects of the operation of components of a system.

[0024] Figure 7 According to some examples of the present disclosure Figure 2A After the first processing stage of the execution of the instruction Figure 1A A diagram of illustrative aspects of a third state of a vector register of a system of FIG.

[0025] Figure 8 According to some examples of the present disclosure Figure 2A During the execution of the instruction Figure 1A A diagram of illustrative aspects of rotation of data of a rotation vector register of a system.

[0026] Fig. 9 According to some examples of the present disclosure Figure 2A Before the second processing stage of the execution of the instruction Figure 1A A diagram of illustrative aspects of a fourth state of a vector register of a system of FIG.

[0027] Fig.10 According to some examples of the present disclosure Figure 2A During the first sub-stage of the second processing stage of execution of the instruction Figure 1A A diagram of illustrative aspects of the operation of components of a system.

[0028] Fig.11 According to some examples of the present disclosure Figure 2A After the first sub-stage of the second processing stage of the execution of the instruction Figure 1A A diagram of illustrative aspects of a fifth state of a vector register of a system of FIG.

[0029] Fig.12 According to some examples of the present disclosure Figure 1A A diagram of illustrative aspects of the operation of a selection circuit of a system.

[0030] Fig.13 According to some examples of the present disclosure Figure 1A A diagram of illustrative aspects of the operation of a rotation circuit of a system.

[0031] Fig.14 An example of an integrated circuit including a rotated vector register file operable to rotate a vector input according to some examples of the present disclosure is illustrated.

[0032] Fig.15is a diagram of a mobile device including a rotation vector register file operable to rotate a vector input according to some examples of the present disclosure.

[0033] Fig.16 is a diagram of a headset including a rotation vector register file operable with a rotation vector input according to some examples of the present disclosure.

[0034] Fig.17 is a diagram of a wearable electronic device including a rotation vector register file operable with a rotation vector input according to some examples of the present disclosure.

[0035] Fig.18 is a diagram of a voice controlled speaker system including a rotation vector register file operable with a rotation vector input according to some examples of the present disclosure.

[0036] Fig.19 is a diagram of a camera including a rotation vector register file operable with a rotation vector input according to some examples of the present disclosure.

[0037] Fig. 20 is a diagram of a headset, such as a virtual reality, mixed reality, or augmented reality headset, including a rotation vector register file operable with a rotation vector input, according to some examples of the present disclosure.

[0038] Fig.21 is a diagram of a first example of a vehicle including a rotation vector register file operable with a rotation vector input according to some examples of the present disclosure.

[0039] Fig. 22 is a diagram of a second example of a vehicle including a rotation vector register file operable with a rotation vector input according to some examples of the present disclosure.

[0040] Fig.23 According to some examples of the present disclosure, Figure 1A A diagram of a specific implementation of a method of rotating a vector input performed by a device.

[0041] Fig.24 is a block diagram of a specific illustrative example of a device including a rotated vector register file operable to rotate a vector input according to some examples of the present disclosure.

[0042] VI. Specific Implementation Methods

[0043] A vector register file including a rotated vector register and a rotate circuit is disclosed. The rotate circuit is operable to rotate vector data in the rotated vector register before providing sub-vector values ​​of the rotated vector data as input to another circuit (such as a multiply-accumulate circuit (MAC)). In conventional systems, multiple passes of data for loading sub-vector values ​​from a memory into a scalar register cause a processing bottleneck. Such processing bottlenecks are reduced (e.g., eliminated) by using a vector register file instead of a scalar register because the sub-vector values ​​can be loaded into the rotated vector register with fewer data passes (e.g., a single pass).

[0044] According to some aspects, a vector load operation may be performed to load vector data into a rotated vector register. In some examples, vector data is loaded from a memory into a cache memory (e.g., an L2 cache) and loaded from the cache memory into a rotated vector register. Any intermediate cache memory (e.g., an L1 cache) between the cache memory (e.g., an L2 cache) and the rotated vector register is bypassed to reduce the overhead associated with loading data from a cache memory (e.g., an L2 cache) into an intermediate cache memory (e.g., an L1 cache) and loading from the intermediate cache memory into a rotated vector register. The vector data may include a plurality of sub-vector values ​​(e.g., 16 sub-vector values). Each sub-vector value may include a plurality of elements (e.g., 4 half-words).

[0045] Artificial neural network processing for machine learning tasks is usually computationally intensive, requiring a large number of multiplication and addition operations. In some examples, the processor may execute a vector instruction (e.g., a multiply-accumulate instruction) to multiply each element of each sub-vector value stored in the rotation vector register with the element of the vector data stored in the source vector register. For example, during the first processing stage of executing the vector instruction, the processor broadcasts the elements (e.g., 4 half-words) of the sub-vector value (v0-v3) of the rotation vector register to the corresponding multiplier. For illustration, the first half-word (v0), the second half-word (v1), the third half-word (v2), and the fourth half-word (v3) of the sub-vector value are broadcasted to the first multiplier, the second multiplier, the third multiplier, and the fourth multiplier in parallel.

[0046] The source vector register stores sub-vector values ​​(s0-s3), which include a first one-byte value (s0), a second one-byte value (s1), a third one-byte value (s2), and a fourth one-byte value (s3). As used herein, "v" is used as a prefix to indicate a sub-vector value read from a rotation vector register, and "s" is used as a prefix to indicate a sub-vector value read from a source vector register.

[0047] During a first sub-stage of the first processing stage, each element of the sub-vector value of the rotation vector register is multiplied by a corresponding value of the sub-vector value of the source vector register. For example, the first multiplier generates a first multiplication product (v0s0) by multiplying a first halfword (v0) by a first one-byte value (s0), generates a second multiplication product (v1s1) by multiplying a second halfword (v1) by a second one-byte value (s1), generates a third multiplication product (v2s2) by multiplying a third halfword (v2) by a third one-byte value (s2), and generates a fourth multiplication product (v3s3) by multiplying a fourth halfword (v3) by a fourth one-byte value (s3). The four multiplication products are added together, and the resulting sum is added to the accumulation vector register.

[0048] During a second sub-stage of the first processing stage, each element of the sub-vector values ​​(v0-v3) of the rotated vector register is multiplied by the corresponding value of the next sub-vector value (s4-s7) of the source vector register. The four multiplication products are added together, and the resulting sum is added to the accumulator vector register. In some aspects, additional sub-stages of the first processing stage are performed until the sub-vector values ​​(v0-v3) of the rotated vector register have been multiplied by all sub-vector values ​​of the source vector register.

[0049] After the first processing stage, the next sub-vector value (v4-v7) of the rotated vector register will be used for multiplication. Conventionally, enabling broadcasting of elements of sub-vector values ​​from different parts of the vector register can increase circuit complexity. For example, the selection circuit must be able to read each element of each of the sub-vector values ​​of the rotated vector register (e.g., each of the 4 half-word values) to provide the selected sub-vector value as input to the broadcast circuit for broadcasting to the appropriate multiplier.

[0050] In order to reduce such complexity, a rotation circuit is used to rotate the data in the rotation vector register. For example, a specific portion of the rotation vector register is dedicated to broadcasting. During the first processing stage, the sub-vector value (v0-v3) is read from the dedicated portion of the rotation vector register and broadcasted to the multiplier. During the rotation stage after the first processing stage, the sub-vector value is rotated in the rotation vector register so that the next sub-vector value (v4-v7) is moved to the dedicated portion of the rotation vector register. During the second processing stage after the rotation stage, the next sub-vector value (v4-v7) is broadcasted to the multiplier, and the multiplier generates a multiplication product based on the next sub-vector value and updates the accumulation vector register.

[0051] In some aspects, additional rotation stages and processing stages may be performed until all sub-vector values ​​of the rotated vector register have been rotated to the dedicated portion at least once, and then the next batch of vector data may be loaded into the rotated vector register. Enabling the selection circuit to read data from elements of the dedicated portion of the rotated vector register may reduce circuit complexity compared to enabling the selection circuit to read data from all portions of the rotated vector register.

[0052] As an illustrative example, a sub-vector value of a rotated vector register including four elements (e.g., 4 half-word values) broadcasted to four multipliers is provided. In other examples, the sub-vector value of a rotated vector register may include less than four or more than four elements broadcasted to corresponding multipliers.

[0053] Certain aspects of the present disclosure are described below with reference to the accompanying drawings. In this specification, common features are designated by common reference numerals. As used herein, various terms are used only for the purpose of describing a particular implementation and are not intended to limit the implementation. For example, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. In addition, some features described herein are singular in some implementations and plural in other implementations. For illustration, Figure 1A Depicts a system including one or more processors ( Figure 1A 190), indicating that in some implementations, the device 102 includes a single processor 190, and in other implementations, the device 102 includes multiple processors 190. For ease of reference herein, such features are generally introduced as "one or more" features and are subsequently referred to in the singular unless aspects are described that relate to multiple ones of the features.

[0054] In some of the drawings, multiple instances of a particular type of feature are used. Although the features are physically and / or logically different, the same reference number is used for each feature, and the different instances are distinguished by adding a letter to the reference number. For example, reference numeral 1 is used for each feature. Figure 1A , one or more rotators are illustrated and associated with reference numerals 172A and 172M. When referring to a particular one of these rotators (such as rotator 172A), a distinguishing letter "A" is used. However, when referring to any one of these rotators, reference numeral 172 is used without a distinguishing letter.

[0055] As used herein, the term "including" can be used interchangeably with "comprising". In addition, the term "wherein" can be used interchangeably with "wherein". As used herein, "exemplary" indicates examples, specific implementations and / or aspects, and should not be interpreted as restrictive or indicating a preference or preferred specific implementation. As used herein, ordinal terms (e.g., "first", "second", "third", etc.) used to modify elements (such as, structures, components, operations, etc.) do not themselves indicate any priority or order of the element relative to another element, but only distinguish the element from another element with the same name (but using ordinal terms). As used herein, the term "set" refers to one or more specific elements in a specific element, and the term "plurality" refers to multiple (e.g., two or more) specific elements.

[0056] As used herein, "coupling" may include "communicatively coupled", "electrically coupled" or "physically coupled", and may also (or alternatively) include any combination thereof. Two devices (or components) may be directly or indirectly coupled (e.g., communicatively coupled, electrically coupled or physically coupled) via one or more other devices, components, wires, buses, networks (e.g., wired networks, wireless networks, or combinations thereof), etc. As illustrative, non-limiting examples, two electrically coupled devices (or components) may be included in the same device or in different devices, and may be connected via electronic devices, one or more connectors, or inductive coupling. In some specific implementations, two devices (or components) that are communicatively coupled (such as electrically connected) may directly or indirectly transmit and receive signals (e.g., digital signals or analog signals) via one or more wires, buses, networks, etc. As used herein, "direct coupling" may include two devices that are coupled (e.g., communicatively coupled, electrically coupled or physically coupled) without intermediate components.

[0057] In the present disclosure, terms such as "determine", "calculate", "estimate", "shift", "adjust", etc. may be used to describe how to perform one or more operations. It should be noted that such terms should not be interpreted as limiting, and other technologies may be used to perform similar operations. In addition, as mentioned herein, "generate", "calculate", "estimate", "use", "select", "access" and "determine" may be used interchangeably. For example, "generating", "calculating", "estimating" or "determining" a parameter (or signal) may refer to actively generating, estimating, calculating or determining the parameter (or signal), or may refer to using, selecting or accessing a parameter (or signal) that has already been generated (such as by another component or device).

[0058] refer to Figure 1A, showing certain illustrative aspects of a system 100 configured for rotation vector input. The system 100 includes a device 102 including a memory 132 coupled to one or more processors 190.

[0059] One or more processors 190 include a multiply-accumulate circuit (MAC) 160 coupled to a vector register (VR) file 150 and a rotate VR file 140. VR file 150 includes a source VR 154, an accumulation VR 156, one or more additional VRs, or a combination thereof. Rotate VR file 140 includes one or more rotate VRs 144 coupled to a rotate circuit 142. One or more rotate VRs 144 include rotate VR 144A, one or more additional rotate VRs, rotate VR 144N, or a combination thereof, and are referred to herein as rotate VRs 144A-144N. In some examples, rotate VRs 144A-144N include a single rotate VR 144A. In other examples, rotate VRs 144A-144N include rotate VR 144A and one or more additional rotate VRs.

[0060] Rotation circuit 142 includes one or more rotators, such as rotator 172A, one or more additional rotators, rotator 172M, or a combination thereof, and is referred to herein as rotators 172A-172M. In some examples, rotators 172A-172M include a single rotator 172A. In other examples, rotators 172A-172M include rotator 172A and one or more additional rotators. The count of rotators 172A-172M may be less than, equal to, or greater than the count of rotation VRs 144A-144N.

[0061] Each of the rotators 172A-172M is configured to rotate the data by a certain rotation amount. For example, the rotator 172A is configured to receive the data 183 from one or more of the rotation VRs 144A-144N, perform a data rotation corresponding to the rotation amount 174A to generate the rotated data 177, and store the rotated data 177 in one or more of the rotation VRs 144A-144N, as described with reference to FIG. Fig.13 For illustration, performing data rotation includes rotating at least a portion of data 183 by rotation amount 174A.

[0062] In some implementations, the rotation VR 144A and the rotator 172A together act as a shift register. For example, the rotator 172A may include a multiplexer, dedicated wiring, or other bit shifting hardware to enable the bit value read from the rotation VR 144A to be written back to the rotation VR 144A at a shift position corresponding to the rotation amount 174A. For illustration, the rotator 172A performs a circular shift of the bit value read from the rotation VR 144A based on the rotation amount 174A.

[0063] In some alternative implementations, the rotation VRs 144A-144N include a shift register, and the rotators 172A-172M provide inputs (e.g., data advance signals) to the rotation VRs 144A-144N based on the amount of rotation to perform data rotation. As an illustrative example, the rotation VR 144A includes a shift register having a cascade of flip-flops, where the output of one flip-flop is connected to the input of the next flip-flop and the last flip-flop is connected to the first flip-flop. When the data advance signal is received, the bit value stored at each flip-flop is shifted to the next flip-flop and the bit value from the last flip-flop is shifted to the first flip-flop. In some aspects, the rotator 172A provides the data advance signal to the rotation VR 144A, where the count of the data advance signal provided to the rotation VR 144A is based on the amount of rotation 174A.

[0064] In some implementations, a variable target rotation can be achieved by cascading multiple fixed rotations, such as Fig.13Further description. In an illustrative example, rotator 172A reads a bit value from rotation VR 144A, generates a first rotated bit value by rotating the bit value based on rotation amount 174A, and stores the first rotated bit value in a pipeline register. A second rotator in rotators 172A-172M reads a first rotated bit value from a pipeline register, generates a second rotated bit value based on the first rotated bit value, and writes the second rotated bit value to rotation VR 144A. In one example, the second rotator is configured to selectively rotate the bit value based on the second rotation amount. For illustration, if the target rotation is the same as the rotation amount 174A, the second rotator outputs the first rotated bit value (without rotation) as the second rotated bit value. If the target rotation is the same as the sum of the rotation amount 174A and the second rotation amount, the second rotator rotates the first rotated bit value based on the second rotation amount. Therefore, a variable target rotation (e.g., the rotation amount 174A or the sum of the rotation amount 174A and the second rotation amount) can be implemented using two cascaded rotations. The fixed amount of rotation to be cascaded may be selected based at least in part on an instruction set architecture (ISA) specification associated with MAC 160, which may describe the various amounts of rotation to be supported. Using a pair of rotators each configured to perform a predetermined rotation may reduce hardware cost (e.g., power consumption, area, or both), improve performance (e.g., faster execution), or a combination thereof, compared to a single rotator configured to perform a variable rotation.

[0065] Rotational VRs 144A-144N are also coupled to selection circuit 146. In some implementations, selection circuit 146 is configured to read data from a dedicated portion of each of rotational VRs 144A-144N, and is not configured to read data from the remaining portion of each of rotational VRs 144A-144N. As an illustrative example, rotational VR 144A includes a dedicated portion 134A that is readable by selection circuit 146. In some implementations, selection circuit 146 is configured to read data from dedicated portion 134A, and is not configured to read data from the remaining portion of rotational VR 144A. For example, in these implementations, selection circuit 146 is coupled to dedicated portion 134A of rotational VR 144A, and is not coupled to the remaining portion of rotational VR 144A.

[0066] In some implementations, the selection circuit 146 is configurable. For example, the selection circuit 146 can read data 145 from one or more sub-portions of the dedicated portion of the rotating VR 144A-144N based on the configuration 148, as shown in FIG. Fig.12 Further description.

[0067] MAC 160 is configured to receive data 145 from rotation VR file 140 and data 155 from source VR 154. MAC 160 is configured to generate output data 165 based on data 145 and data 155. For example, MAC 160 is configured to generate a multiplication product based on elements of data 145 and elements of data 155, generate output data 165 based on the multiplication product, and update accumulation data 167 of accumulation VR 156 based on output data 165. For illustration, MAC 160 may be used in conjunction with an artificial neural network process that includes multiplying a weight value by a corresponding activation value. In an artificial neural network example, data 145 may correspond to an activation value and data 155 may correspond to a weight value. In some examples, MAC 160 is configured to store output data 165 as accumulation data 167 in accumulation VR 156. In some examples, MAC 160 is configured to update accumulation data 167 by adding the value indicated by output data 165 to the value indicated by accumulation data 167 and storing the sum as accumulation data 167 in accumulation VR 156 .

[0068] In some implementations, device 102 corresponds to or is included in one of various types of devices. In the illustrative example, one or more processors 190 are integrated in a headphone device, such as a headset. Fig.16 In other examples, the one or more processors 190 are integrated into at least one of the following: Fig.15 The mobile phone or tablet computer device, as referenced Fig.17 The wearable electronic device, as shown in reference Fig.18 The voice controlled speaker system, as shown in reference Fig.19 The camera device or as referenced Fig. 20 In another illustrative example, one or more processors 190 are integrated into a vehicle, such as a vehicle such as a vehicle in which the one or more processors 190 are integrated. Fig.21 and Fig. 22 As further described.

[0069] During operation, vector 151 is loaded from memory 132 into source VR 154, and vector 171 is loaded from memory 132 into rotation VRs 144A-144N, as shown in FIG. Figure 2A In certain aspects, vector 171 includes a plurality of sub-vector values, such as sub-vector value (SV) 173A, SV 173B, one or more additional SVs, SV 173T, or a combination thereof, and is referred to herein as SVs 173A-173T. SVs 173A-173T are Figure 1B144A, which is illustrated as being rotatable, as further described below. In certain aspects, vector 151 includes multiple sub-vector values, such as SV 175A, SV 175B, one or more additional SVs, SV 175S, or a combination thereof, and are referred to herein as SVs 175A-173S. The count of SVs 173A-173T may be less than, equal to, or greater than the count of SVs 175A-175S. In some examples, vector 151 is loaded into a single source vector register (e.g., source VR 154) and vector 171 is loaded into a single rotation VR (e.g., rotation VR 144A), as described in reference to FIG. Figure 3 In some examples, vector 171 is loaded into multiple rotation vector registers. For example, a portion of vector 171 is loaded into rotation VR 144A and another portion of vector 171 is loaded into rotation VR 144B, as shown in FIG. Figure 12 to Figure 13 Further described. In some examples, vector 151 may be loaded into a plurality of source vector registers including source VR 154 and one or more additional source VRs.

[0070] Elements of vector 171 from rotation VRs 144A-144N are multiplied with elements of vector 151 from source VR 154 to generate an output to be stored in accumulation VR 156. In the illustrative example, vector 171 is stored in rotation VR 144A, and MAC 160 multiplies elements of vector 171 stored in rotation VR 144A with elements of vector 151 stored in source VR 154. For example, during each of a plurality of processing stages, a different SV stored in rotation VR 144A is multiplied with an SV stored in source VR 154.

[0071] In the illustrative example, during the first processing stage, selection circuit 146 reads SV 173A from dedicated portion 134A of rotation VR 144A based on configuration 148, as shown in reference Fig.12 The selection circuit 146 provides the SV 173A as the data 145 to the MAC 160 .

[0072] In some implementations, each processing stage includes multiple sub-stages, and in each sub-stage, a different SV is provided from the source VR 154 to the MAC 160 as the data 155. For example, during the first sub-stage of the first processing stage, the MAC 160 receives the SV 175A from the source VR 154 as the data 155. The MAC 160 generates the output data 165 based on the data 145 (received from the rotated VR file 140) and the data 155 (received from the source VR 154), as shown in FIG. Figure 4175A. MAC 160 stores output data 165 as accumulated data 167 in accumulation VR 156. MAC 160 stores output data 165 as accumulated data 167 in accumulation VR 156. MAC 160 stores output data 165 as accumulated data 167 in accumulation VR 156. MAC 160 stores output data 165 as accumulated data 167 in accumulation VR 156. MAC 160 stores output data 165 as accumulated data 167 in accumulation VR 156. MAC 160 stores output data 165 as accumulated data 167 in accumulation VR 156. MAC 160 stores output data 165 as accumulated data 167 in accumulation VR 156. MAC 160 stores output data 165 as accumulated data 167 in accumulation VR 156. MAC 160 stores output data 165 as accumulated data 167 in accumulation VR 156. MAC 160 stores output data 165 as accumulated data 167 in accumulation VR 156.

[0073] If vector 151 stored in source VR 154 includes multiple sub-vector values, each processing stage may include multiple sub-stages. For example, during the second sub-stage of the first processing stage, MAC 160 receives SV 175B (e.g., the next sub-vector value) from source VR 154 as data 155. For illustration, elements of SV 173A received from rotation VR 144A are multiplied in parallel with elements of SV 175B received from source VR 154. MAC 160 generates output data 165 based on data 145 (e.g., SV 173A) and data 155 (e.g., SV 175B). In some aspects, additional sub-stages of the first processing stage are performed (in which MAC 160 processes each of SV 173A and SV 175B of source VR 154) until SV 173A and SV 175S have been processed by MAC 160.

[0074] During the rotation phase following the first processing phase, rotators 172A-172M rotate the data in one or more of rotation VRs 144A-144N by a rotation amount, as shown in reference Fig.13 For example, rotator 172A receives data 183 from rotation VR 144A, rotates data 183 by rotation amount 174A to generate rotation data 177, and stores rotation data 177 in rotation VR 144A. Figure 1B , in illustrative example 180, SV 173A is stored in dedicated portion 134A before rotation. For illustration, data 183 corresponds to SV 173A, followed by SV 173B, one or more additional SVs, and SV 173T. After rotation, SV 173B is stored in dedicated portion 134A of rotated VR 144A. For illustration, rotated data 177 corresponds to SV 173B, followed by one or more additional SVs, SV 173T, and SV 173A.

[0075] Return to Figure 1A In some aspects, during the initial rotation after vector 171 is loaded from memory 132 into rotation VR 144A, data 183 is the same as vector 171 retrieved from memory 132. In some aspects, in subsequent rotations, data 183 corresponds to a rotated version of vector 171. For example, in these aspects, data 183 corresponds to rotated data 177 generated in a previous rotation.

[0076] In some implementations, multiple rotators are used to generate the rotated data 177. For example, rotator 172A rotates data 183 by rotation amount 174A to generate first rotated data, a second rotator among rotators 172A-172M rotates the first rotated data by a second rotation amount to generate second rotated data, and so on, until a rotator among rotators 172A-172M rotates the rotated data received from the previous rotator to generate the rotated data 177. The rotation circuit 142 stores the rotated data 177 in the rotation VR 144A.

[0077] During the second processing stage, selection circuit 146 retrieves data 145 (e.g., SV 173B) from dedicated portion 134A and provides data 145 to MAC 160. During the first sub-stage of the second processing stage, MAC 160 receives SV 175A as data 155 from source VR 154. MAC 160 generates output data 165 based on SV 173B (e.g., data 145) and SV 175A (e.g., data 155) and updates accumulated data 167 based on output data 165.

[0078] During a second sub-stage of the second processing stage, MAC 160 receives SV 175B from source VR 154 as data 155. MAC 160 generates output data 165 based on SV 173B (e.g., data 145) and SV 175B (e.g., data 155) and outputs accumulation data 167 based on output data 165. In some aspects, additional sub-stages of the second processing stage are performed in which MAC 160 processes each of SV 173B and SV 175S of source VR 154 until SV 173B and SV 175S have been processed by MAC 160. In some aspects, additional rotation stages and processing stages are performed until all SVs 173A-173T of vector 171 have been processed by MAC 160 to update accumulation VR 156.

[0079] During each processing stage, MAC 160 multiplies elements of a single SV retrieved from rotate vector register 144A with elements of a single SV retrieved from source VR 154 as provided as an illustrative example. In some implementations, one or more processors 190 may include multiple copies of the components of MAC 160 described herein, which may be used to multiply elements of an SV retrieved from one of rotate VRs 144A-144N with elements of multiple SVs retrieved from source VR 154. For example, during a first processing stage, a first copy of the component may multiply elements of SV 173A with elements of SV 175A, and a second copy of the component may multiply elements of SV 173A with elements of SV 175B in parallel. In some implementations, MAC 160 may multiply elements of SV 173A with elements of two or more of SVs 175A-175S in parallel. During the rotation phase following the first processing phase, SV 173A is rotated to the end of rotation VR 144A and SV 173B is rotated to dedicated portion 134A. During the second processing phase, the first copy of the component may multiply the elements of SV 173B with the elements of SV 175A, and the second copy of the component may multiply the elements of SV 173B with the elements of SV 175B in parallel. In some implementations, MAC 160 may multiply the elements of SV 173B with the elements of two or more of SVs 175A-175S in parallel.

[0080] In some implementations, one or more processors 190 may include multiple copies of a component of MAC 160 described herein that may be used to multiply elements of multiple SVs retrieved from rotation VRs 144A-144N with elements of SVs retrieved from source VR 154. For example, during a first processing stage, a first copy of the component may multiply elements of SV 173A with elements of SV 175A, and a second copy of the component may multiply elements of SV 173B with elements of SV 175A in parallel. In a first aspect, SV 173A and SV 173B are retrieved from the same rotation VR (e.g., rotation VR 144A). In a second aspect, SV 173A is retrieved from rotation VR 144A and SV 173B is retrieved from rotation VR 144B. During a rotation stage that follows the first processing stage, SVs (e.g., SV 173A, SV 173B, one or more additional SVs, or a combination thereof) that were provided to the component of MAC 160 during the first processing stage are rotated. In a first aspect, SV 173A, SV 173B, one or more additional SVs, or a combination thereof are rotated to the end of rotation VR 144A. In a second aspect, SV 173A is rotated to the end of rotation VR 144A, SV 173B is rotated to the end of rotation VR 144B, each of the one or more additional SVs is rotated to the end of a corresponding rotation VR, or a combination thereof.

[0081] System 100 thus enables vector processing with reduced complexity of selection circuit 146. For example, selection circuit 146 is enabled to read data from dedicated portion 134A of rotated VR 144A, and does not have to include circuitry to support reading from the remainder of rotated VR 144A.

[0082] refer to Figure 2A , showing Figure 1A FIG200 is a diagram of illustrative aspects of one or more processors 190 and memory 132 of system 100. The one or more processors 190 are operable to rotate a vector input for executing instructions.

[0083] The one or more processors 190 include an instruction buffer 202 coupled to an instruction selector 204. The instruction selector 204 is coupled to the MAC 160 and a load / store unit 208. The load / store unit 208 is coupled to the memory 132. In certain aspects, the load / store unit 208 is coupled to the memory 132 via one or more caches. For example, the load / store unit 208 is coupled to the L2 cache 234 via the L1 cache 232, and the L2 cache 234 is coupled to the memory 132. In some aspects, the load / store unit 208 is also coupled to the L2 cache 234 (bypassing the L1 cache 232). Each of the MAC 160 and the load / store unit 208 is coupled to the source VR 154, the rotation VR file 140, and the accumulation VR 156.

[0084] In certain aspects, a plurality of instructions are added to an instruction set architecture (ISA). The ISA may support instructions corresponding to various configurations of the selection circuit 146, the rotation circuit 142, or both, various SV sizes of the rotation VRs 144A-144N, various SV sizes of the source VRs 154, SVs with signed values, SVs with unsigned values, one or more source VRs for storing vector 151, one or more of the rotation VRs 144A-144N for storing vector 171, or combinations thereof. In some implementations, the one or more processors 190 correspond to vector processors implementing the ISA. For example, the one or more processors 190 are configured to efficiently operate on vectors. In certain aspects, the one or more processors 190 are configured to efficiently copy vectors (e.g., large one-dimensional data arrays) from the memory 132 to vector registers and vice versa, and perform parallel processing of multiple data values ​​from the vector registers (such as using multiple parallel computation paths) in a single instruction multiple data (SIMD) configuration.

[0085] The instruction selector 204 uses various instruction selection techniques to select an instruction 280 (e.g., a multiply-accumulate instruction) from the instruction buffer 202 for processing. The instruction 280 includes an opcode 282 and a parameter 284. The opcode 282 indicates the type of instruction 280 (e.g., a multiply-accumulate). In a first implementation, the parameter 284 indicates a first memory address of the vector 171 and a second memory address of the vector 151. In the first implementation, during execution of the instruction 280, the vector 171 is loaded into at least one of the rotation VRs 144A-144N, and the vector 151 is loaded into the source VR 154. In a second implementation, before execution of the instruction 280, the vector 171 is loaded into at least one of the rotation VRs 144A-144N, and the vector 151 is loaded into the source VR 154. In the second implementation, the parameter 284 indicates at least one of the rotation VRs 144A-144N and the source VR 154.

[0086] Instruction selector 204 provides instructions 280 (or data representing aspects of instructions 280) to MAC 160, load / store unit 208, or both. In some aspects, instruction selector 204 initiates execution of instruction 280 by providing opcode 282 to MAC 160.

[0087] In a first implementation, parameter 284 indicates a first memory address of vector 171 and a second memory address of vector 151, and instruction selector 204 provides the first memory address and the second memory address to load / store unit 208. Load / store unit 208 performs a load operation 273 of vector 171 to rotated VR file 140 in response to receiving the first memory address. In a particular implementation, load / store unit 208 loads vector 171 from L2 cache 234 to rotated VR file 140 in response to receiving the first memory address of vector 171. If vector 171 is not available in L2 cache 234, load / store unit 208 copies vector 171 from the received first memory address of memory 132 to L2 cache 234 and from L2 cache 234 (e.g., bypassing L1 cache 232) to rotated VRs 144A-144N. Similarly, the load / store unit 208 performs a load operation 275 of the vector 151 to the source VR 154 in response to receiving the second memory address. In a particular implementation, the load / store unit 208 loads the vector 151 from the L2 cache 234 to the source VR 154 in response to receiving the second memory address. If the vector 151 is not available in the L2 cache 234, the load / store unit 208 copies the vector 151 from the received second memory address of the memory 132 to the L2 cache 234 and from the L2 cache 234 (e.g., bypassing the L1 cache 232) to the source VR 154.

[0088] The one or more processors 190 execute instructions 280 to retrieve data 145 from the rotation VR file 140, retrieve data 155 from the source VR 154, process data 145 and data 155 at the MAC 160 to generate output data 165, store output data 165 in the accumulation VR 156, and rotate the data in the rotation VR 144A after retrieving data 145 from the rotation VR file 140. For example, executing instructions 280 includes performing a plurality of processing stages, one or more rotation stages, or a combination thereof to update the accumulation data 167, as described with reference to FIG. Figure 1A described.

[0089] In certain aspects, load / store unit 208 stores accumulated data 167 to memory 132 in response to determining that SVs 173A-173T of vector 171 have been processed by MAC 160. In some examples, vector 171 corresponds to a portion of the vector data, and load / store unit 208 loads the next portion of the vector data as vector 171 to rotate VR 144A after MAC 160 processes vector 171. In these examples, additional processing stages, rotate stages, and loads are performed until all of the vector data has been processed.

[0090] refer to Figure 2B , graph 290 illustrates changes in performance (eg, operation count) of MAC 160 relative to changes in operation intensity (eg, operations per byte).

[0091] During the first processing stage, the MAC 160 generates output data 165 based on the vector 171 and the vector 151 loaded to the rotation VR file 140 and the source VR 154, respectively. In some implementations, the MAC 160 generates the output data 165 once the data 155 (e.g., SV 175A) and the data 145 (e.g., SV 173A) are loaded. In these implementations, the output data 165 is generated concurrently with loading the remaining SVs of the vector 171 and the vector 151. In other implementations, the MAC 160 generates the output data 165 once all SVs of the vector 151 and all SVs of the vector 171 are loaded to the rotation VR file 140 and the source VR 154, respectively. The performance of the MAC 160 corresponds to the bandwidth used to load the vector 171 and the vector 151 from the memory 132 (or L2 cache).

[0092] Once vector 171 and vector 151 are loaded, the "optimal" operating intensity of MAC 160 is achieved. For example, the performance of MAC 160 is not limited by bandwidth because the input to MAC 160 is available in rotated VR file 140 and source VR 154. Therefore, rotating the data in rotated VR file 140 to make the next sub-vector value available to MAC 160 enables MAC 160 to execute at a high (e.g., optimal) operating intensity compared to retrieving the next sub-vector value (e.g., a scalar value) from memory 132 as input to MAC 160.

[0093] Figures 3 to 11 An example of executing instruction 280 is illustrated. Instruction 280 corresponds to a "vmpyaddhbz_x4" instruction, where "vmpyadd" indicates that a vector multiply-accumulate (e.g., add) algorithm is implemented. Instruction 280 may have the following form: Vxx+=vmpyaddhbz_x4(Vuu, Zy4):sat:rot, where Vxx is a target VR (e.g., accumulation VR 156), Vuu is a source VR (e.g., source VR 154), Zy4 is a rotation VR (e.g., rotation VR 144A), "sat" indicates that the value is saturated, and "rot" indicates that a rotation stage will be performed between processing stages.

[0094] Although the examples herein describe instruction 280 that performs a vector multiply-accumulate operation that reads data from a single source VR (e.g., source VR 154), in other examples, instruction 280 of the form Vxx+=vmpyaddzt_x8(Vuu, Vvv, Zy4):sat:rot indicates that data may be read from multiple source VRs (such as a first source VR represented by Vuu and a second source VR represented by Vvv).

[0095] refer to Figure 3 , a diagram 300 showing illustrative aspects of a first state of the rotation VR 144A, the source VR 154, and the accumulation VR 156. In certain aspects, the rotation VR 144, the source VR 154, and the accumulation VR 156 are in Figure 2A Before the first processing stage of execution of instruction 280 is in the first state.

[0096] In a particular implementation, rotated VR 144A includes 64 elements (e.g., v0 to v63), each of elements v0 to v63 includes a halfword value, and SV includes 8 elements. For example, SV 173A includes elements v0-v7, SV 173B includes elements v8-v15, and so on. In a first processing state, SV 173A is stored in dedicated portion 134A of rotated VR 144A. In a particular aspect, SV 173A may include multiple SVs. For example, SV 173A includes SV 312A (e.g., v0-v3) and SV 312B (e.g., v4-v7).

[0097] In a particular implementation, source VR 154 includes 256 elements (e.g., s0-s255), and each of elements s0-s255 includes a byte value. In a particular aspect, the SVs of source VR 154 include 4 elements. For example, SV 320A includes elements s0-s3, and SV 320B includes elements s128-s131. In a particular aspect, SV 175A includes SV 320A and SV 320B.

[0098] In a particular implementation, accumulation VR 156 includes 64 elements (eg, a0-a63). For example, accumulation VR 156 includes SV 340A (eg, a0), SV 340B (eg, a32), one or more additional SVs, or a combination thereof.

[0099] In the first state illustrated in diagram 300, before the first processing stage, selection circuit 146 is configured to select SV 173A stored in dedicated portion 134A as data 145 for the first processing stage, which will refer to Figure 4 Further description.

[0100] refer to Figure 4 , shows during the first sub-stage of the first processing stage of execution of instruction 280 Figure 1A 400 of illustrative aspects of the operation of components of system 100.

[0101] In a specific example, instruction 280 has the following form: Vxx+=vmpyaddhbz_x4(Vuu, Zy4):sat:rot, and the processing stage of execution of instruction 280 corresponds to the following pseudo-code for loop:

[0102] fHIDE(int i;)for(i=0;i<32;i++){

[0103] fHIDE(size8s_t acc[2];)acc[0]=Vxx.V32s[i];

[0104] acc[0]+=fMPY8SS(Vuu.V8s[4*i+0],ZyV.V16s[0]);

[0105] acc[0]+=fMPY8SS(Vuu.V8s[4*i+1],ZyV.V16s[1]);

[0106] acc[0]+=fMPY8SS(Vuu.V8s[4*i+2],ZyV.V16s[2]);

[0107] acc[0]+=fMPY8SS(Vuu.V8s[4*i+3],ZyV.V16s[3]);

[0108] acc[1] = Vxx.V32s[32+i];

[0109] acc[1]+=fMPY8SS(Vuu.V8s[128+4*i+0],ZyV.V16s[4]);

[0110] acc[1]+=fMPY8SS(Vuu.V8s[128+4*i+1],ZyV.V16s[5]);

[0111] acc[1]+=fMPY8SS(Vuu.V8s[128+4*i+2],ZyV.V16s[6]);

[0112] acc[1]+=fMPY8SS(Vuu.V8s[128+4*i+3],ZyV.V16s[7]);

[0113] Vxx.V32s[i]=fVSATN(32,acc[0]);

[0114] Vxx.V32s[32+i]=fVSATN(32,acc[1]);

[0115] }

[0116] Where Vxx corresponds to accumulate VR 156, acc[0] corresponds to SV 340A, acc[1] corresponds to SV 340B, Vuu corresponds to source VR 154, ZyV corresponds to rotate VR 144A, V8s indicates a byte value, V16s indicates a halfword value, and fMPY8SS indicates a signed floating point multiplication. Each iteration of the for loop corresponds to a substage of the processing stage.

[0117] During the first sub-stage of the first processing stage, selection circuit 146 selects SV 173A as data 145, as shown in FIG. Figure 1A The one or more processors 190 include a broadcast circuit 402 coupled to the MAC 160. The MAC 160 includes a plurality of multipliers, such as multiplier 410A, multiplier 410B, multiplier 410D, multiplier 410E, one or more additional multipliers, multiplier 410H, or combinations thereof, which are referred to herein as multipliers 410A-410H.

[0118] MAC 160 includes a plurality of inputs, a pair of inputs of the plurality of inputs being associated with corresponding multipliers. In some aspects, MAC 160 includes inputs 407A-H configured to be coupled to broadcast circuit 402, and inputs 409A-H configured to be coupled to source VR 154. For example, input 407A of MAC 160 corresponds to (e.g., includes or is coupled to) a first input of multiplier 410A, input 407B of MAC 160 corresponds to (e.g., includes or is coupled to) a first input of multiplier 410B, and so on. As another example, input 409A of MAC 160 corresponds to (e.g., includes or is coupled to) a second input of multiplier 410A, input 409B of MAC 160 corresponds to (e.g., includes or is coupled to) a second input of multiplier 410B, and so on. In some aspects, inputs 407A-H and inputs 409A-H implement parallel multiplication of elements of data 145 with elements of data 155. For example, multiplier 410A multiplies data elements received from input 407A with data elements received from input 409A, multiplier 410B multiplies data elements received from input 407B with data elements of data 155 received from input 409B, and so on.

[0119] In a particular implementation, the broadcast circuit 402 includes a plurality of outputs, and each of the plurality of outputs is coupled to a corresponding input of the MAC 160 associated with a single multiplier. For example, a first output of the broadcast circuit 402 is coupled to an input 407A associated with the multiplier 410A and is not coupled to an input associated with any of the multipliers 410B-410H. As another example, a second output of the broadcast circuit 402 is coupled to an input 407B associated with the multiplier 410B and is not coupled to inputs associated with other multipliers 410A-410H. In a particular aspect, one or more of the plurality of outputs of the broadcast circuit 402, one or more of the plurality of inputs of the MAC 160 (e.g., inputs 407A-H and inputs 409A-H), or a combination thereof include one or more bus interfaces, one or more latches, one or more flip-flops, one or more buffers, other data buffering circuits, or a combination thereof.

[0120] In certain aspects, the selection circuit 146 provides the data 145 (e.g., SV 173A) to the broadcast circuit 402. The data 145 (e.g., SV 173A) includes a plurality of elements (e.g., v0-v7). The broadcast circuit 402 provides the data 145 to the multipliers 410A-410H. For example, for each of the plurality of elements (e.g., v0-v7), the broadcast circuit 402 provides the element to a respective different one of the inputs 407A-H of the MAC 160. For illustration, the first element (e.g., v0) is provided to the input 407A associated with the multiplier 410A via the first output of the broadcast circuit 402, and is not provided by the broadcast circuit 402 to any input associated with the other multipliers of the MAC 160. Similarly, the second element (eg, v1) is provided to input 407B associated with multiplier 410B via the second output of broadcast circuit 402 and is not provided by broadcast circuit 402 to any inputs associated with other multipliers of MAC 160 .

[0121] Data 155 (e.g., SV 175A) includes a plurality of elements (e.g., s0-s3 and s128-s131). MAC 160 receives data 155 (e.g., SV 175A) from source VR 154, as shown in FIG. Figure 1A For example, for each element in the plurality of elements (e.g., s0-s3 and s128-s131), the element is provided to a corresponding different one of the inputs 409A-H of the MAC 160. For illustration, the first element (e.g., s0) is provided to the input 409A associated with the multiplier 410A, the second element (e.g., s1) is provided to the input 409B associated with the multiplier 410B, and so on.

[0122] Multipliers 410A-410H generate output data 165 based on data 145 (e.g., SV 173A) and data 155 (e.g., SV 175A). For example, multipliers 410A-410H generate multiplier data 455A-455H as output data 165. For illustration, multiplier 410A receives a first element (e.g., v0) of data 145 (e.g., SV 173A) via input 407A and receives a first element (e.g., s0) of data 155 (e.g., SV 175A) via input 409A. Multiplier 410A generates multiplier data 455A (e.g., m0, output SV) by multiplying the first element (e.g., v0) of data 145 (e.g., SV 173A) with the first element (e.g., s0) of data 155 (e.g., SV 175A). Similarly, multiplier 410B receives the second element (e.g., v1) of data 145 (e.g., SV 173A) via input 407B and receives the second element (e.g., s1) of data 155 (e.g., SV 175A) via input 409B. Multiplier 410B generates multiplier data 455B (e.g., m1, output SV) by multiplying the second element (e.g., v1) of data 145 (e.g., SV 173A) by the second element (e.g., s1) of data 155 (e.g., SV 175A).

[0123] MAC 160 provides output data 165 to accumulation VR 156. In certain aspects, output data 165 is added to accumulation data 167 and the result is stored in accumulation VR 156. For example, MAC 160 includes a plurality of adders, such as adder 450A, adder 450B, adder 450D, adder 450E, adder 450H, one or more additional adders, or combinations thereof, which are referred to herein as adders 450A-450H. Adders 450A-450H receive at least a portion of accumulation data 167 from accumulation VR 156, generate a sum by adding output data 165 to accumulation data 167, and overwrite portions of accumulation data 167 with the sum in accumulation VR 156.

[0124] For example, adder 450A receives SV 340A (e.g., a value represented by a0) from accumulation VR 156 and receives multiplier data 455A (e.g., a value represented by m0) from multiplier 410A. Adder 450A generates adder data 457A (e.g., a value represented by p0) by adding SV 340A (e.g., a0) and multiplier data 455A (e.g., m0) and stores adder data 457A (e.g., p0) as an updated value of SV 340A in accumulation VR 156. Similarly, adder 450B receives SV 340A (e.g., p0) from accumulation VR 156 and receives multiplier data 455B (e.g., a value represented by m1) from multiplier 410B. The adder 450B generates adder data 457B (e.g., a value represented by p1) by adding the SV 340A (e.g., p0) and the multiplier data 455B (e.g., m1), and stores the adder data 457B (e.g., p1) as the updated value of SV 340A in the accumulation VR 156. At the end of the first sub-stage, the SV 340A has been updated from the first value (e.g., a0) to the second value (e.g., the value represented by b0), and the SV 340B has been updated from the first value (e.g., the value represented by a32) to the second value (e.g., the value represented by b32).

[0125] refer to Figure 5 , a diagram 500 showing illustrative aspects of a second state of the rotation VR 144A, the source VR 154, and the accumulation VR 156. In certain aspects, the rotation VR 144A, the source VR 154, and the accumulation VR 156 are in Figure 2A The instruction 280 is in the second state after the first sub-stage of the first processing stage of execution and before the second sub-stage.

[0126] In the second state, SV 173A remains in dedicated portion 134A. SV 175B of source VR 154 includes SV 520A and SV 520B. Accumulation VR 156 includes SV 540A (e.g., a1) and SV 540B (e.g., a33). SV 540A (e.g., a1) follows SV 340A (e.g., b0) updated in the first sub-stage of the first processing stage. SV 540B (e.g., a33) follows SV 340B (e.g., b32) updated in the first sub-stage of the first processing stage.

[0127] refer to Figure 6 , shows during the second sub-stage of the first processing stage of execution of instruction 280 Figure 1A 600 of illustrative aspects of the operation of components of system 100.

[0128] During the second sub-phase of the first processing phase, MAC 160 receives SV 175B from source VR 154 as data 155, as shown in FIG. Figure 1A SV 175B includes a plurality of elements (e.g., s4-s7 and s132-s135). For each of the plurality of elements (e.g., s4-s7 and s132-s135), the element is provided to a corresponding different one of inputs 409A-H of MAC 160. For purposes of illustration, a first element (e.g., s4) is provided to input 409A associated with multiplier 410A, a second element (e.g., s5) is provided to input 409B associated with multiplier 410B, and so on.

[0129] The multipliers 410A-410H generate output data 165 based on the data 145 (e.g., SV 173A) and the data 155 (e.g., SV 175B). For example, the multipliers 410A-410H generate multiplier data 455A-455H as the output data 165. For illustration, the multiplier 410A generates the multiplier data 455A (e.g., m0, output SV) by multiplying the first element (e.g., v0) of the data 145 (e.g., SV 173A) with the first element (e.g., s4) of the data 155 (e.g., SV 175B). Similarly, the multiplier 410B generates the multiplier data 455B (e.g., m1, output SV) by multiplying the second element (e.g., v1) of the data 145 (e.g., SV 173A) with the second element (e.g., s5) of the data 155 (e.g., SV 175B).

[0130] MAC 160 provides output data 165 to accumulation VR 156. For example, adder 450A receives SV 540A (e.g., a value represented by a1) from accumulation VR 156 and receives multiplier data 455A (e.g., a value represented by m0) from multiplier 410A. Adder 450A generates adder data 457A (e.g., a value represented by p0) by adding SV 540A (e.g., a1) and multiplier data 455A (e.g., m0) and stores adder data 457A (e.g., p0) as an updated value of SV 540A in accumulation VR 156. Similarly, adder 450B receives SV 540A (e.g., p0) from accumulation VR 156 and receives multiplier data 455B (e.g., a value represented by m1) from multiplier 410B. The adder 450B generates adder data 457B (e.g., a value represented by p1) by adding the SV 540A (e.g., p0) and the multiplier data 455B (e.g., m1), and stores the adder data 457B (e.g., p1) as the updated value of the SV 540A in the accumulation VR 156. At the end of the second sub-stage, the SV 540A has been updated from the first value (e.g., a1) to the second value (e.g., the value represented by b1), and the SV 540B has been updated from the first value (e.g., the value represented by a33) to the second value (e.g., the value represented by b33).

[0131] refer to Figure 7 , a diagram 700 showing illustrative aspects of a third state of the rotation VR 144A, the source VR 154, and the accumulation VR 156. In certain aspects, the rotation VR 144A, the source VR 154, and the accumulation VR 156 are in Figure 2A The instruction 280 is in a third state after the first processing stage and before the second processing stage of execution.

[0132] In the third state, SV 173A remains in dedicated portion 134A. SV 175S of source VR 154 includes SV 720A and SV 720B. All elements of accumulation VR 156 have been updated based on SV 173A and SVs 175A-175S. For example, accumulation VR 156 includes SV 740A updated from a first value (e.g., a31) to a second value (e.g., b31) and SV 740B updated from a first value (e.g., a63) to a second value (e.g., b63).

[0133] refer to Figure 8 , diagram 800 showing illustrative aspects of the rotation of data of rotation VR 144A of FIG. 1 during execution of instruction 280. For example, Figure 1AThe rotation circuit 142 rotates the data of the rotation VR 144A, as referenced Figure 1A to Figure 1B In certain aspects, the rotation is performed during a rotation phase following the first processing phase. During the rotation, SV 173A is moved to the end of rotation VR 144A and the remaining SVs are shifted so that SV 173B is stored in dedicated portion 134A.

[0134] In a specific example, instruction 280 has the following form: Vxx+=vmpyaddhbz_x4(Vuu, Zy4):sat:rot, and the rotate phase of execution of instruction 280 corresponds to the following pseudo-code for loop:

[0135] fHIDE(size1u_t tmp;)fHIDE(int k;)fHIDE(int a;)fHIDE(int b;)for(k=0;k<128-2*8;k++){

[0136] a=k%128;

[0137] b = (k + 128 - 2 * 8) % 128;

[0138] tmp = ZyV.V8u[a];

[0139] ZyV.V8u[a]=ZyV.V8u[b];

[0140] ZyV.V8u[b]=tmp;

[0141] }

[0142] where ZyV corresponds to rotation VR 144A.

[0143] refer to Fig. 9 , a diagram 900 showing illustrative aspects of a fourth state of the rotation VR 144A, the source VR 154, and the accumulation VR 156. In certain aspects, the rotation VR 144A, the source VR 154, and the accumulation VR 156 are in Figure 8 After the rotation phase and in Figure 2A Before the second processing stage of execution of instruction 280 is in the fourth state.

[0144] In the second processing state, SV 173B is stored in dedicated portion 134A of rotation VR 144A. In certain aspects, SV 173B may include multiple SVs. For example, SV 173B includes SV 912A (e.g., v8-v11) and SV 912B (e.g., v12-v15).

[0145] Source VR 154 includes SV 175A, and SV 175A includes SV 320A and SV 320B. Accumulation VR 156 includes SV 340A (eg, b0) and SV 340B (eg, b32). In certain aspects, selection circuit 146 is configured to select SV 173B stored in dedicated portion 134A as data 145.

[0146] refer to Fig.10 , showing that in Figure 2A During the first sub-stage of the second processing stage of execution of instruction 280 Figure 1A 1000 of illustrative aspects of the operation of components of system 100.

[0147] During the first sub-phase of the second processing phase, the following is performed as in reference Figure 4 4. The broadcast circuit 402 may be configured to receive the data 145 (e.g., SV 173B) from the broadcast circuit 402. The broadcast circuit 402 may be configured to receive the data 145 (e.g., SV 173A) from the broadcast circuit 402. The broadcast circuit 402 may be configured to receive the data 145 (e.g., SV 173B) from the broadcast circuit 402. For example, the selection circuit 146 provides the data 145 (e.g., SV 173B) to the broadcast circuit 402. The data 145 (e.g., SV 173A) includes a plurality of elements (e.g., v8-v15). The broadcast circuit 402 provides the data 145 (e.g., SV 173B) to the multipliers 410A-410H. For example, the first element (e.g., v8) of the data 145 (e.g., SV 173B) is provided to the input 407A associated with the multiplier 410A via the first output of the broadcast circuit 402, the second element (e.g., v9) of the data 145 (e.g., SV 173B) is provided to the input 407B associated with the multiplier 410B via the second output of the broadcast circuit 402, and so on.

[0148] Data 155 (e.g., SV 175A) includes a plurality of elements (e.g., s0-s3 and s128-s131). MAC 160 receives data 155 (e.g., SV 175A) from source VR 154, as shown in FIG. Figure 1A For example, a first element (eg, s0) is provided to input 409A associated with multiplier 410A, a second element (eg, s1) is provided to input 409B associated with multiplier 410B, and so on.

[0149] Multipliers 410A-410H generate output data 165 based on data 145 (e.g., SV 173B) and data 155 (e.g., SV 175A). For example, multiplier 410A generates multiplier data 455A by multiplying a first element (e.g., v8) of data 145 (e.g., SV 173B) with a first element (e.g., s0) of data 155 (e.g., SV 175A). Similarly, multiplier 410B generates multiplier data 455B by multiplying a second element (e.g., v9) of data 145 (e.g., SV 173B) with a second element (e.g., s1) of data 155 (e.g., SV 175A).

[0150] MAC 160 provides output data 165 to accumulation VR 156. Adder 450A receives SV 340A (e.g., a value represented by b0) from accumulation VR 156 and receives multiplier data 455A (e.g., a value represented by m0) from multiplier 410A. Adder 450A generates adder data 457A (e.g., a value represented by p0) by adding SV 340A (e.g., b0) and multiplier data 455A (e.g., m0) and stores adder data 457A (e.g., p0) as an updated value of SV 340A in accumulation VR 156. Similarly, adder 450B receives SV 340A (e.g., p0) from accumulation VR 156 and receives multiplier data 455B (e.g., a value represented by m1) from multiplier 410B. At the end of the first subphase, SV 340A has been updated from a first value (e.g., b0) to a second value (e.g., the value represented by c0), and SV 340B has been updated from a first value (e.g., the value represented by b32) to a second value (e.g., the value represented by c32).

[0151] refer to Fig.11 , diagram 1100 showing illustrative aspects of a fifth state of spin VR 144A, source VR 154, and accumulate VR 156. In a particular aspect, spin VR 144A, source VR 154, and accumulate VR 156 are in the fifth state after a first sub-stage and before a second sub-stage of a second processing stage of execution of instruction 280.

[0152] In the fifth state, SV 173B remains in dedicated portion 134A. SV 175B of source VR 154 includes SV 520A and SV 520B. Accumulated VR 156 includes SV 540A (e.g., b1) and SV 540B (e.g., b33). SV 540A (e.g., b1) follows SV 340A (e.g., c0) updated in the first sub-stage of the second processing stage. SV 540B (e.g., b33) follows SV 340B (e.g., c32) updated in the first sub-stage of the second processing stage. In the second sub-stage of the second processing stage, SV 540A and SV 540B are updated based on SV 173B and SV 175B. In a particular aspect, additional sub-stages of the second processing stage are performed until accumulated data 167 has been updated based on SV 173B and all SVs 175A-175S. In certain aspects, additional rotation stages and processing stages are performed until all SVs 173A- 173T have been processed by MAC 160 to update accumulated data 167 .

[0153] refer to Fig.12 , diagram 1200 showing illustrative aspects of the operation of selection circuit 146. Selection circuit 146 is coupled to rotation VRs 144A-144N. For example, selection circuit 146 is coupled to rotation VR 144A and rotation VR 144B. Selection circuit 146 coupled to two rotation VRs is provided as an illustrative example. In other examples, selection circuit 146 may be coupled to less than two or more than two rotation VRs.

[0154] In certain aspects, selection circuit 146 is configured to access data stored in the dedicated portion of rotation VRs 144A-144N and not from the rest of rotation VRs 144A-144N. For example, selection circuit 146 is configured to access data from dedicated portion 134A of rotation VR 144A. As another example, selection circuit 146 is configured to access data from dedicated portion 134B of rotation VR 144B.

[0155] As an illustrative example, in Figure 2A Before or during the first processing stage of execution of instruction 280, write data 1271 (e.g., by Figure 2A144B) stores a first portion of vector 171 as data portion 1211 in rotated VR 144A, and stores a second portion of vector 171 as data portion 1213 in rotated VR 144B. In a particular aspect, the first portion of vector 171 includes alternating SVs of vector 171, and the second portion of vector 171 includes the remaining SVs of vector 171. For example, data portion 1211 includes SVs 173A (e.g., v0-v7) of vector 171, and data portion 1213 includes SVs 173B (e.g., v8-15) of vector 171. Similarly, data portion 1211 includes SVs 173C (e.g., v16-v23) and data portion 1213 includes SVs 173D (e.g., v24-v31), and so on. Thus, data portion 1211 includes half of vector 171 (e.g., 32 elements: v0-v7, v16-v23, v32-v39, and v48-v55), and data portion 1213 includes the remaining half of vector 171 (e.g., another 32 elements: v8-v15, v24-v31, v40-v47, and v56-v63). Thus, rotation VR 144A is illustrated as storing 32 elements of vector 171, and rotation VR 144B is illustrated as storing another 32 elements of vector 171. In some implementations, such as Figure 3 , each of Rotate VR 144A and Rotate VR 144B is sized to store 64 elements, and thus some storage capacity of each of Rotate VR 144A and Rotate VR 144B is unused. In some implementations, rather than leaving portions of Rotate VR 144A and Rotate VR 144B unused, data for the next vector to be processed may also be loaded into Rotate VR 144A and Rotate VR 144B (e.g., using the next write data 1271 operation). In other implementations, each of Rotate VR 144A and Rotate VR 144B has storage capacity of 32 elements, and thus no storage capacity is unused.

[0156] It should be appreciated that for ease of illustration, each of rotation VR 144A and rotation VR 144B is illustrated as a vector storing data (e.g., 1 column by 32 rows of elements). In other examples, each of rotation vector 144A or rotation VR 144B may store data in various logical and / or physical arrangements, such as an array representation (e.g., 8 columns by 4 rows of elements or 4 columns by 8 rows of elements).

[0157] In a particular aspect, SV 1273A corresponding to the first 8 elements of data portion 1213 is stored in dedicated portion 134A of rotated VR 144A, and SV 1273B corresponding to the first 8 elements of data portion 1211 is stored in dedicated portion 134B of rotated VR 144B.

[0158] In the illustrated example, the first 8 elements of data portion 1211 correspond to SV 173A (e.g., v0-v7) and the first 8 elements of data portion 1213 correspond to SV 173B (e.g., v8-v15). In this example, SV 1273A corresponds to SV 173A and SV 1273B corresponds to SV 173B. In the example in which the rotated data portion is written to rotated VR 144A, as shown in FIG. Fig.13 Described further, SV 1273A includes the first 8 elements (eg, v16-v23) of the rotated data portion.

[0159] 144A, the dedicated portion 134B of the rotating VR 144B, or both, can be output by the selection circuit 146 based on one or more control signals. In a particular aspect, the selection circuit 146 includes a multiplexer 1230 coupled to the rotating VR 144B. A multiplexer 1232 is coupled to the rotating VR 144A and the multiplexer 1230. A delay element 1234 is coupled to the multiplexer 1232. A combiner 1236 is coupled to the rotating VR 144A and the multiplexer 1230. A multiplexer 1238 is coupled to the rotating VR 144A, the delay element 1234, the combiner 1236, and the multiplexer 1230.

[0160] In some implementations, concurrently with providing the data portion 1213 to the rotation VR 144B, the write data 1271 provides the SV 1273B to be stored in the dedicated portion 134B of the rotation VR 144B to the multiplexer 1230. In a particular aspect, the multiplexer 1230 is configured to select the SV 1273B retrieved from the dedicated portion 134B of the rotation VR 144B or the SV 1273B received via the write data 1271. For example, the multiplexer 1230 selects the SV 1273B from the dedicated portion 134B or from the write data 1271 based on one or more control signals, such as the register output indicator 1224, the rotation indicator 1226, one or more additional control signals, or a combination thereof.

[0161] In certain aspects, a first value (eg, 1) of the rotation indicator 1226 indicates that the write data 1271 (eg, Figure 1A144B is being written to at least one of rotation VR 144A or rotation VR 144B. A second value (e.g., 0) of rotation indicator 1226 indicates that rotated data is not being written to either rotation VR 144A or rotation VR 144B. In a particular aspect, a first value (e.g., 1) of register output indicator 1224 indicates that an SV from write data 1271 is to be selected. A second value (e.g., 0) of register output indicator 1224 indicates that an SV from dedicated portion 134B is to be selected.

[0162] In a particular implementation, multiplexer 1230 selects SV 1273B from write data 1271 in response to rotation indicator 1226 having a first value, register output indicator 1224 having a first value, or both. Alternatively, multiplexer 1230 selects SV 1273B from dedicated portion 134B in response to rotation indicator 1226 having a second value, register output indicator 1224 having a second value, or both. In some aspects, selecting SV 1273B from write data 1271 enables multiplexer 1230 to generate an output without waiting for write data 1271 to complete writing data portion 1213 to rotation VR 144B.

[0163] Multiplexer 1230 provides SV 1231A (e.g., initial 4 elements) of SV 1273B (e.g., selected SV 1273B) to multiplexer 1238, and provides SV 1231B (e.g., last 4 elements) of SV 1273B to multiplexer 1232. In the illustrated example, SV 1273B corresponds to SV 173B, SV 1231A corresponds to SV 912A (e.g., v8-v11), and SV 1231B corresponds to SV 912B (e.g., v12-v15).

[0164] In certain aspects, multiplexer 1230 provides SV 1216 (e.g., initial 2 elements) of SV 1273B (e.g., selected SV 1273B) to combiner 1236. In the illustrated example, SV 1273B corresponds to SV 173B, and SV 1216 corresponds to initial 2 elements (e.g., v8-v9) of SV 173B.

[0165] Multiplexer 1232 receives SV 1212B (e.g., the last 4 elements) of SV 1273A from rotation VR 144A, and receives SV 1231B (e.g., the last 4 elements) of SV 1273B (e.g., the selected SV 1273B) from multiplexer 1230. Multiplexer 1232 outputs SV 1212B or SV 1231B based on a control signal (e.g., register indicator 1228). For example, multiplexer 1232 selects SV 1212B (e.g., the last 4 elements) of SV 1273A as multiplexer output 1233 in response to register indicator 1228 having a first value (e.g., 0). Alternatively, multiplexer 1232 selects SV 1231B (eg, last 4 elements) of SV 1273B as multiplexer output 1233 in response to register indicator 1228 having the second value (eg, 1).

[0166] Combiner 1236 is configured to receive SV 1214 (e.g., initial 2 elements) of SV 1273A and SV 1216 (e.g., initial 2 elements) of SV 1273B. Combiner 1236 generates SV 1218 by combining SV 1214 and SV 1216. In the illustrated example, SV 1273A corresponds to SV 173A, SV 1273B corresponds to SV 173B, and SV 1218 corresponds to a combination of initial 2 elements (e.g., v0-v1) of SV 173A and initial 2 elements (e.g., v8-v9) of SV 173B.

[0167] Multiplexer 1238 receives SV 1212A (e.g., the initial 4 elements of SV 1273A), SV 1218 (e.g., the initial 2 elements of SV 1273A and the initial 2 elements of SV 1273B), and SV 1231A (e.g., the initial 4 elements of SV 1273B) from rotation VR 144A. Multiplexer 1238 selects one of SV 1212A, SV 1218, and SV 1231A to output as an initial portion of multiplexer output 1239 based on a control signal (e.g., patterning control 1242).

[0168] Delay element 1234 receives multiplexer output 1233 (e.g., the last 4 elements of 1273A or the last 4 elements of 1273B) and provides multiplexer output 1233 from delay element 1234 to multiplexer 1238 after output of the initial portion of multiplexer output 1239. Multiplexer 1238 outputs multiplexer output 1233 (e.g., the last 4 elements of 1273A or the last 4 elements of 1273B) as the second portion of multiplexer output 1239. Data 145 includes an initial portion of multiplexer output 1239 (e.g., the initial 4 elements of SV 1273A, the initial 2 elements of SV 1273A and the initial 2 elements of SV 1273B, or the initial 4 elements of SV 1273B) and a second portion of multiplexer output 1239 (e.g., the last 4 elements of 1273A or the last 4 elements of 1273B).

[0169] In the illustrated example, SV 1273A corresponds to SV 173A, and multiplexer 1238 outputs the initial 4 elements (e.g., v0-v3) of SV 173A as the initial portion of multiplexer output 1239, and outputs the last 4 elements (e.g., v4-v7) of SV 173A as the second portion of multiplexer output 1239. During the first processing stage, data 145 includes the initial 4 elements (e.g., v0-v3) of SV 173A and the last 4 elements (e.g., v4-v7) of SV 173A. After the first processing stage, rotation circuit 142 rotates the data in VR 144A during the first rotation stage, as described in reference to FIG. Fig.13 Further description.

[0170] In the example with two rotated VRs, during the second processing stage, selection circuit 146 outputs SV from dedicated portion 134B of rotated VR 144B. For example, multiplexer 1238 outputs the initial 4 elements (e.g., v8-v11) of SV 173B as the initial portion of multiplexer output 1239, and outputs the last 4 elements (e.g., v12-v15) of SV 173B as the second portion of multiplexer output 1239. During the second processing stage, data 145 includes the initial 4 elements (e.g., v8-v11) of SV 173B and the last 4 elements (e.g., v12-v15) of SV 173B. After the second processing stage, rotation circuit 142 rotates the data in rotated VR 144B during the second rotation stage, as shown in reference to FIG. Fig.13 Further description.

[0171] During the third processing stage, selection circuit 146 outputs SV from dedicated portion 134A of rotated VR 144A including rotated data. For example, multiplexer 1238 outputs the initial 4 elements (e.g., v16-v19) of SV 1273A as the initial portion of multiplexer output 1239, and outputs the last 4 elements (e.g., v20-v23) of SV 1273A as the second portion of multiplexer output 1239. During the third processing stage, data 145 includes the initial 4 elements (e.g., v16-v19) of SV 1273A and the last 4 elements (e.g., v20-v23) of SV 1273A. After the third processing stage, rotation circuit 142 rotates the data in rotated VR 144A during the third rotation stage, as shown in reference to FIG. Fig.13 In some aspects, additional processing stages and rotation stages are performed until all elements of data portion 1211 and data portion 1213 have been output by selection circuit 146 and processed by MAC 160 .

[0172] As an illustrative example, a particular pattern of outputting the first 8 elements of rotation VR 144A followed by the first 8 elements of rotation VR 144B during each processing stage is provided. In other examples, elements stored in dedicated portion 134A, elements stored in dedicated portion 134B, or a combination thereof may be output by selection circuit 146 in various combinations based on configuration 148. For example, configuration 148 specifies the value of a control signal (e.g., register output indicator 1224, rotation indicator 1226, register indicator 1228, patterning control 1242, or a combination thereof).

[0173] In some aspects, selection circuit 146 may output data 145 including one or more elements from dedicated portion 134A and one or more elements from dedicated portion 134B in the same processing stage based on configuration 148. For example, when patterning control 1242 of configuration 148 indicates that data 145 is to include SV 1218, data 145 includes an initial element stored in dedicated portion 134A and an initial element stored in dedicated portion 134B. As another example, based on register indicator 1228 and patterning control 1242 of configuration 148, selection circuit 146 may output an initial element of one of dedicated portion 134A (e.g., SV 1212A) or dedicated portion 134B (e.g., SV 1231A) in a first time step of the processing stage, and output a last element of the other of dedicated portion 134A (e.g., SV 1212B) or dedicated portion 134B (e.g., SV 1231B) in a second time step of the same processing stage.

[0174] In certain aspects, configuration 148 is based on default data, configuration settings, user input, etc. In certain aspects, configuration 148 is based on data of configuration 148 that maps instructions 280 (eg, opcodes 282, parameters 284, or both) to specific values ​​of control signals.

[0175] refer to Fig.13 , diagram 1300 showing illustrative aspects of the operation of rotation circuit 142. Rotation circuit 142 is coupled to rotation VRs 144A-144N. For example, rotation circuit 142 is coupled to rotation VR 144A and rotation VR 144B. Rotation circuit 142 coupled to two rotation VRs is provided as an illustrative example. In other examples, rotation circuit 142 may be coupled to less than two or more than two rotation VRs.

[0176] The rotation circuit 142 is configured to rotate the data in the rotation VR 144A or the rotation VR 144B based on one or more control signals (e.g., register indicator 1324 and register indicator 1326) and based on one or more rotation amounts (e.g., rotation amount 174A and rotation amount 174B). The rotation circuit 142 includes a logic gate 1302 (e.g., a two-input AND gate) and a logic gate 1304 (e.g., a two-input AND gate). An input of the logic gate 1302 is coupled to the output of the rotation VR 144A, and another input of the logic gate 1302 is configured to receive the register indicator 1324. An input of the logic gate 1304 (e.g., a two-input AND gate) is coupled to the output of the rotation VR 144B, and another input of the logic gate 1304 is configured to receive the register indicator 1326.

[0177] Logic gate 1302 and logic gate 1304 are coupled to logic gate 1306 (e.g., a two-input OR gate). For example, an input of logic gate 1306 is coupled to the output of logic gate 1302, and another input of logic gate 1306 is coupled to the output of logic gate 1304. Logic gate 1306 is coupled to Figure 1A Rotators 172A-172M.

[0178] In some implementations, each of the rotators 172A-172M is configured to rotate the data by a predetermined amount. In these implementations, the rotation circuit 142 may include one or more of the rotators 172A-172M. For example, the output of the logic gate 1306 is coupled to the input of the rotator 172A, and the output of the rotator 172A is coupled to the input of the rotator 172B. In some examples, the output of the rotator 172A is coupled to the input of the rotator 172B via the pipeline register. In a particular aspect, the rotator 172A is configured to rotate the data by a rotation amount 174A and store the rotated data in the pipeline register, and the rotator 172B is configured to selectively rotate the data retrieved from the pipeline register based on the control signal by a rotation amount 174B. For example, when the control signal has a first value (e.g., 1), the rotator 172B rotates the data by a rotation amount 174B. When the control signal has a second value (e.g., 0), the rotator 172B suppresses the rotation data. The output of rotator 172B is coupled to the input of rotation VR 144A and the input of rotation VR 144B. In some alternative implementations, rotation circuit 142 includes a single programmable rotator (e.g., rotator 172A). In these implementations, the output of logic gate 1306 is coupled to the input of rotator 172A, and the output of rotator 172A is coupled to the input of rotation VR 144A and the input of rotation VR 144B. In some examples, rotator 172A and rotation VR 144A together act as a variable shift register.

[0179] Logic gate 1302 provides data portion 1211 received from rotation VR 144A to logic gate 1306 in response to register indicator 1324 having a first value (e.g., 1). Logic gate 1304 provides data portion 1213 received from rotation VR 144B to logic gate 1306 in response to register indicator 1326 having a first value (e.g., 1). In some aspects (e.g., after a processing stage), one of register indicator 1324 or register indicator 1326 has a first value (e.g., 1) and the other of register indicator 1324 or register indicator 1326 has a second value (e.g., 0). In these aspects, data portion 1211 or data portion 1213 is passed to logic gate 1306 for rotation. In some aspects (eg, during a processing stage), register indicator 1324 and register indicator 1326 both have a second value (eg, 0) such that neither data portion 1211 nor data portion 1213 is passed to logic gate 1306 and no rotation is performed.

[0180] Logic gate 1306 passes data portion 1211 or data portion 1213 received from logic gate 1302 or logic gate 1304, respectively, to rotator 172A as output 1307. For example, after the first processing stage, register indicator 1324 has a first value (e.g., 1), logic gate 1302 passes data portion 1211 to logic gate 1306, and logic gate 1306 provides data portion 1211 as output 1307 to rotator 172A. Rotator 172A rotates output 1307 by rotation amount 174A (e.g., 2 halfwords) to generate rotated data 1377 and provides rotated data 1377 to rotator 172B. In some examples, rotator 172A provides rotated data 1377 to a pipeline register, and rotator 172B retrieves rotated data 1377 from the pipeline register. The rotator 172B generates the rotated data 1377 by rotating the rotated data 177 based on the rotation amount 174B (e.g., 6 half words). The rotation circuit 142 stores the rotated data 177 in the rotated VR 144A. For example, the rotation circuit 142 stores the rotated data 177 in the rotated VR 144A in response to the register indicator 1324 having a first value (e.g., 1). Alternatively, the rotation circuit 142 stores the rotated data 177 in the rotated VR 144B in response to the register indicator 1326 having a first value (e.g., 1).

[0181] The rotation mechanism including pipeline registers is provided as an illustrative example. In other specific implementations, the rotator 172A and the rotation VR 144A (or the rotation VR 144B) together act as a first shift register, and the rotator 172B and the rotation VR 144A (or the rotation VR 144B) together act as a second shift register. In one example, the rotator 172A rotates the bit values ​​stored in the rotation VR 144A independently of (e.g., without) any additional registers, so that the rotation data 1377 is stored in the rotation VR 144A after the rotation. In another example, the rotator 172B rotates the bit values ​​corresponding to the rotation data 1377 stored in the rotation VR 144A independently of (e.g., without) any additional registers, so that the rotation data 177 is stored in the rotation VR 144A after the rotation.

[0182] In some examples, rotation circuit 142 rotates data in a single one of rotation VRs 144A-144N during the rotation phase. In some examples, rotation circuit 142 rotates data in multiple ones of rotation VRs 144A-144N during the rotation phase. For example, one of register indicator 1324 or register indicator 1326 has a first value (e.g., 1) during a first pass, and the other of register indicator 1324 or register indicator 1326 has a first value (e.g., 1) during a second pass.

[0183] In some aspects, the rotation circuit 142 is configurable. For example, one or more control signals, one or more rotation amounts, or a combination thereof are based on a configuration 1348 of the rotation circuit 142. The configuration 1348 may be based on default data, user input, configuration settings, etc. In a particular aspect, the configuration 1348 is based on data that maps the instruction 280 (e.g., opcode 282, parameter 284, or both) to a configuration 1348 indicating a specific value of one or more control signals, one or more rotation amounts, or both. In some aspects, at least one of the rotation amounts is selectable. In one example, the rotation amount 174A is fixed (e.g., 2 halfwords), and the rotation amount 174B is selectable. For illustration, the rotation amount 174B has a value based on the configuration 1348 (e.g., a selection of 0 halfwords or 6 halfwords).

[0184] In some aspects, the rotation circuit 142 enables the data in a single rotation VR to be rotated, such as reference Figures 3 to 11 For example, register indicator 1324 has a first value (e.g., 1) after each processing stage of the instruction. In some aspects, rotation circuit 142 enables data in an alternate rotation VR to be rotated after a processing stage, as described in reference to Fig.12 For example, register indicator 1324 has a first value (e.g., 1) after the first processing stage and each odd-numbered processing stage after the first processing stage, and register indicator 1326 has a first value (e.g., 1) after the second processing stage and each even-numbered processing stage after the second processing stage.

[0185] In a particular example, when configuration 148 of selection circuit 146 indicates that dedicated portion 134A of rotate VR 144A is to be selected for a particular processing stage, configuration 1348 indicates that data in rotate VR 144A (e.g., and not data in any other rotate VRs) is to be rotated during a rotate stage following the particular processing stage. For example, register indicator 1324 has a first value (e.g., 1) and rotate amount 174B is selected to indicate 6 halfwords to have a total rotation of 8 halfwords including the 2 halfwords indicated by rotate amount 174A.

[0186] In some examples, rotation circuitry 142 enables data in multiple rotation VRs to be rotated during a single rotation phase. For example, when configuration 148 of selection circuitry 146 indicates that two initial elements of dedicated portion 134A of rotation VR 144A and two initial elements of dedicated portion 134B of rotation VR 144B are to be selected during a particular processing phase (e.g., as described in reference Fig.12 1218 of SV 1218), configuration 1348 indicates that the data in rotation VR 144A and the data in rotation VR 144B are to be rotated during a rotation phase following a particular processing phase. For example, rotation amount 174B is selected to indicate 0 halfwords to have a total rotation of 2 halfwords including the 2 halfwords indicated by rotation amount 174A. During a first pass of the rotation phase, register indicator 1324 has a first value (e.g., 1) and the data in rotation VR 144A is rotated by 2 halfwords. During a second pass of the rotation phase, register indicator 1326 has a first value (e.g., 1) and the data in rotation VR 144B is rotated by 2 halfwords.

[0187] Fig.14 The device 102 is depicted as an implementation 1400 of an integrated circuit 1402 that includes one or more processors 190. The one or more processors 190 include a rotating VR file 140. In some aspects, the one or more processors 190 also include a MAC 160, a VR file 150, or both. The integrated circuit 1402 also includes a signal input 1404, such as one or more bus interfaces, one or more latches, one or more flip-flops, one or more buffers, other data buffering circuits, or a combination thereof, to enable receiving data 1428 for processing. In one example, the data 1428 includes Figure 1A The integrated circuit 1402 also includes a signal output 1406 (such as a bus interface) to enable the transmission of output signals, such as the accumulated data 167. The integrated circuit 1402 enables the implementation of the rotation vector input as a component in a system, such as: Fig.15 A mobile phone or tablet computer as depicted in Fig.16 The headset depicted in Fig.17 The wearable electronic device depicted in Fig.18 The voice controlled speaker system described in Fig.19 The camera depicted in Fig. 20 A virtual reality, mixed reality, or augmented reality headset as depicted in Fig.21 or Fig. 22 The means of transport depicted in .

[0188] As an illustrative, non-limiting example, Fig.15 A specific implementation 1500 is depicted in which the device 102 includes a mobile device 1502 (such as a phone or tablet). The mobile device 1502 includes a display screen 1504. Components of the one or more processors 190, including the rotation VR file 140, are integrated into the mobile device 1502 and are shown using dashed lines to indicate internal components that are generally not visible to a user of the mobile device 1502. In some aspects, the one or more processors 190 also include the MAC 160, the VR file 150, or both. In a specific example, the rotation VR file 140 operates to rotate data in the rotation VR during the execution of a vector operation (e.g., a matrix multiplication), and then processes the results of the vector operation to perform one or more operations at the mobile device 1502, such as launching a graphical user interface or otherwise displaying other information associated with the vector operation at the display screen 1504 (e.g., via an integrated "smart assistant" application).

[0189] In some implementations, as illustrative non-limiting examples, device 102 includes one or more other sensors or components that generate data that can be operated on by vector operations using instructions 280, such as wireless network signal data, global positioning data or other location data, video data or image data from one or more cameras, inertial measurement data or other movement data from an inertial measurement unit (e.g., one or more gyroscopes, compasses, accelerometers, etc.), or health data from one or more corresponding sensors, such as heart rate data, oxygen content data, breathing data, etc. The vector operations generate output data that can be output or can be processed to generate processed data, and as illustrative non-limiting examples, either or both of the output data and the processed data can be displayed via display screen 1504, output via a speaker, sent via a wireless network to another device (such as a wearable electronic device (e.g., a smart watch or a headset)), or output via a tactile output signal.

[0190] Fig.16 A specific implementation 1600 is depicted in which the device 102 includes a headset device 1602. Components of the one or more processors 190, including the rotation VR file 140, are integrated in the headset device 1602. In some aspects, the one or more processors 190 also include the MAC 160, the VR file 150, or both. In a specific example, the rotation VR file 140 operates to rotate data in the VR rotation during the execution of a vector operation (e.g., a matrix multiplication), which may cause the headset device 1602 to perform one or more operations at the headset device 1602, send the data to a second device (not shown) for further processing, or a combination thereof.

[0191] Fig.17A specific implementation 1700 is depicted in which the device 102 includes a wearable electronic device 1702 (illustrated as a "smart watch"). The rotation VR file 140 is integrated into the wearable electronic device 1702. In some aspects, the MAC 160, the VR file 150, or both are also integrated into the wearable electronic device 1702. In a specific example, the rotation VR file 140 operates to rotate data in the rotation VR during the execution of a vector operation (e.g., a matrix multiplication), which can cause the wearable electronic device 1702 to perform one or more operations at the wearable electronic device 1702, such as launching a graphical user interface or otherwise displaying other information associated with the vector operation at a display 1704 of the wearable electronic device 1702. For illustration, the wearable electronic device 1702 may include a display that is configured to display a notification based on the vector operation performed by the wearable electronic device 1702. In a particular example, the wearable electronic device 1702 includes a haptic device that provides a tactile notification (e.g., a vibration) in response to the execution of an operation using vector operations, such as a neural network-based voice interface. For example, the tactile notification can enable a user viewing the wearable electronic device 1702 to see a displayed notification indicating the execution of a vector operation (e.g., an action performed in response to a verbal query identifying the user). Thus, the wearable electronic device 1702 can alert a user with hearing impairment or a user wearing a headset about an operation using vector operations.

[0192] Fig.18 1800 is a specific implementation in which the device 102 includes a wireless speaker and a voice activated device 1802. The wireless speaker and the voice activated device 1802 may have wireless network connectivity and be configured to perform auxiliary operations. One or more processors 190 including rotating the VR file 140 are included in the wireless speaker and the voice activated device 1802. In some aspects, the one or more processors 190 also include the MAC 160, the VR file 150, or both. The wireless speaker and the voice activated device 1802 also includes a speaker 1804. During operation, the wireless speaker and the voice activated device 1802 may perform auxiliary operations, such as via execution of a voice activated system (e.g., an integrated auxiliary application). The auxiliary operation may include adjusting the temperature, playing music, turning on the lights, etc. For example, the auxiliary operation is performed in response to receiving a command after a keyword or key phrase (e.g., "Hello, assistant"). In a particular example, the rotation VR file 140 operates to rotate data in the rotation VR during execution of a vector operation (e.g., matrix multiplication), which may cause one or more operations to be performed at the wireless speaker and voice activated device 1802 or may be part of a calculation associated therewith.

[0193] Fig.19An implementation 1900 is depicted in which the device 102 includes a portable electronic device corresponding to a camera device 1902. A rotation VR file 140 is included in the camera device 1902. In some aspects, the MAC 160, the VR file 150, or both are also included in the camera device 1902. As an illustrative example, during operation, the camera device 1902 may perform operations in response to verbal user commands, such as adjusting image or video capture settings, image or video playback settings, or image or video capture instructions. In a specific example, the rotation VR file 140 operates to rotate data in the rotation VR during execution of a vector operation (e.g., a matrix multiplication associated with image filtering), which may cause one or more operations to be performed at the camera device 1902 or may be part of a calculation associated therewith.

[0194] Fig. 20 A specific implementation 2000 is depicted in which the device 102 includes a portable electronic device corresponding to a virtual reality, mixed reality, or augmented reality headset 2002. The rotation VR file 140 is integrated into the headset 2002. In some aspects, the MAC 160, the VR file 150, or both are also integrated into the headset 2002. In a specific example, the rotation VR file 140 operates to rotate data in the rotation VR during the execution of a vector operation (e.g., a matrix multiplication associated with image processing), which may cause one or more operations to be performed at the headset 2002 or may be part of a calculation associated therewith. The visual interface device is positioned in front of the user's eyes to enable the display of an augmented reality, mixed reality, or virtual reality image or scene to the user while wearing the headset 2002. In a specific example, the visual interface device is configured to display a notification indicating the execution of an operation based on a vector operation (e.g., an image processing operation). For illustration, the visual interface device is configured to display one or more images generated by the image processing operation.

[0195] Fig.21 An implementation 2100 is depicted in which the device 102 corresponds to or is integrated within a vehicle 2102, shown as a manned or unmanned aerial device (e.g., a package delivery drone). The rotation VR file 140 is integrated into the vehicle 2102. In some aspects, the MAC 160, the VR file 150, or both are also integrated into the vehicle 2102. In a specific example, the rotation VR file 140 operates to rotate data in the rotation VR during the execution of a vector operation (e.g., a matrix multiplication associated with an image processing operation), which may cause one or more operations to be performed at the vehicle 2102 or may be part of a calculation associated therewith. For example, an image generated by the image processing operation may be displayed by the vehicle 2102 to provide assembly instructions to a package recipient.

[0196] Fig. 22 Another implementation 2200 is depicted in which the device 102 corresponds to or is integrated within a vehicle 2202 (shown as a car). The vehicle 2202 includes one or more processors 190 that include a rotation VR file 140. In some aspects, the one or more processors 190 also include a MAC 160, a VR file 150, or both. In a specific example, the rotation VR file 140 operates to rotate data in the rotation VR during execution of a vector operation (e.g., a matrix multiplication), which may cause the vehicle 2202 to perform one or more operations at the vehicle 2202.

[0197] In certain aspects, the voice activation system initiates one or more operations of the vehicle 2202 based on one or more keywords (e.g., "unlock," "start engine," "play music," "show weather forecast," or another voice command) detected in the output signal of the microphone (such as by providing feedback or information via the display 2220 or one or more speakers). In some examples, the one or more keywords are detected in the output signal by performing operations based on vector operations using the rotation VR file 140, such as neural network calculations in an artificial intelligence (AI)-based automatic speech recognition system.

[0198] refer to Fig.23 , showing a specific implementation of the method 2300 for rotating a vector input. In certain aspects, one or more operations of the method 2300 (e.g., a processor-implemented method) are performed by the rotators 172A-172M, the rotation circuit 142, the rotation VRs 144A-144N, the selection circuit 146, the rotation VR file 140, the MAC 160, the VR file 150, the source VR 154, the accumulation VR 156, one or more processors 190, the device 102, Figure 1A The system 100 or at least one of a combination thereof is executed.

[0199] The method 2300 includes, at 2302, using a rotate vector register file to rotate data in a rotate vector register of the rotate vector register file. For example, Figure 1A The data in the rotation VR file 144A of the rotation VR file 140 is rotated, as shown in FIG. Figure 1A , Figure 1B and Fig.13 described.

[0200] The method 2300 also includes receiving first input data from a rotated vector register file at a multiply-accumulate circuit (MAC) at 2304. For example, Figure 1A The MAC 160 receives data 145 from the rotation VR file 140, as shown in FIG. Figure 1A and Fig.12 described.

[0201] The method 2300 also includes, at 2306, receiving, at the MAC, second input data from a source vector register of the second vector register file. For example, Figure 1A MAC 160 receives data 155 from source VR 154 of VR file 150, as shown in reference Figure 1A described.

[0202] In some examples, method 2300 includes using a MAC to generate an output based on first input data and second input data. For example, Figure 1A MAC 160 generates output data 165 based on data 145 and data 155, as shown in FIG. Figure 1A Method 2300 also includes storing the output in an accumulation vector register of a second vector register file. For example, Figure 1A MAC 160 stores output data 165 in accumulation VR 156 of VR file 150, as shown in FIG. Figure 1A described.

[0203] Method 2300 enables vector processing with reduced complexity of select circuit 146. For example, select circuit 146 is enabled to read data rotated into dedicated portion 134A of rotated VR 144A without having to include circuitry to support reading from the remainder of rotated VR 144A.

[0204] Fig.23 The method 2300 may be implemented by a field programmable gate array (FPGA) device, an application specific integrated circuit (ASIC), a processing unit (such as a central processing unit (CPU)), a DSP, a graphics processing unit (GPU), a controller, another hardware device, a firmware device, or any combination thereof. As an example, Fig.23 The method 2300 may be performed by a processor executing instructions, such as reference Fig.24 described.

[0205] refer to Fig.24 , depicts a block diagram of a particular exemplary implementation of a device and generally designated as 2400. In various implementations, the device 2400 may have more Fig.24 More or fewer components may be illustrated. In an exemplary implementation, device 2400 may correspond to device 102. In an exemplary implementation, device 2400 may perform a reference to Figure 1A-23 Describes one or more operations.

[0206] In certain implementations, the device 2400 includes a processor 2406 (e.g., a CPU). The device 2400 may include one or more additional processors 2410 (e.g., one or more DSPs, one or more GPUs, or a combination thereof). In certain aspects, Figure 1A The one or more processors 190 correspond to the processor 2406, the processor 2410, or a combination thereof. The processor 2410 may include a speech and music coder-decoder (codec) 2408, which includes a speech decoder ("vocoder") encoder 2436, a vocoder decoder 2438, a rotation VR file 140, or a combination thereof.

[0207] The device 2400 may include a memory 2486 and a codec 2434. The memory 2486 may include instructions 2456 that are executable by one or more additional processors 2410 (or processor 2406) to implement the functionality described with reference to rotating the VR file 140. The device 2400 may include a modem 2448 coupled to an antenna 2452 via a transceiver 2450.

[0208] The device 2400 may include a display 2428 coupled to a display controller 2426. One or more speakers 2492 and one or more microphones 2490 may be coupled to a codec 2434. The codec 2434 may include a digital-to-analog converter (DAC) 2402, an analog-to-digital converter (ADC) 2404, or both. In a specific implementation, the codec 2434 may receive an analog signal from the one or more microphones 2490, convert the analog signal to a digital signal using the analog-to-digital converter 2404, and provide the digital signal to the voice and music codec 2408. The voice and music codec 2408 may process the digital signal. In a specific implementation, the voice and music codec 2408 may provide the digital signal to the codec 2434. The codec 2434 may convert the digital signal to an analog signal using the digital-to-analog converter 2402 and may provide the analog signal to the one or more speakers 2492.

[0209] In a particular implementation, device 2400 may be included in a system-in-package or system-on-chip device 2422. In a particular implementation, memory 2486, processor 2406, processor 2410, display controller 2426, codec 2434, and modem 2448 are included in the system-in-package or system-on-chip device 2422. In a particular implementation, input device 2430 and power supply 2444 are coupled to the system-in-package or system-on-chip device 2422. Additionally, in a particular implementation, as shown in FIG. Fig.24As illustrated, the display 2428, input device 2430, one or more speakers 2492, one or more microphones 2490, antenna 2452, and power supply 2444 are external to the system-in-package or system-on-chip device 2422. In a particular implementation, each of the display 2428, input device 2430, one or more speakers 2492, one or more microphones 2490, antenna 2452, and power supply 2444 may be coupled to a component of the system-in-package or chip-on-chip device 2422, such as an interface or a controller.

[0210] Device 2400 may include a smart speaker, a speaker bar, a mobile communication device, a smart phone, a cellular phone, a laptop, a computer, a tablet computer, a personal digital assistant, a display device, a television, a game console, a music player, a radio, a digital video player, a digital video disc (DVD) player, a tuner, a camera, a navigation device, a vehicle, a headset, an augmented reality headset, a mixed reality headset, an extended reality headset, a virtual reality headset, an air vehicle, a home automation system, a voice-activated device, a wireless speaker and voice-activated device, a portable electronic device, an automobile, a computing device, a communication device, an Internet of Things (IoT) device, a virtual reality (VR) device, a base station, a mobile device, or any combination thereof.

[0211] In conjunction with the described implementation, the apparatus includes means for rotating data in a rotated vector register of a rotated vector register file. For example, the means for rotating may correspond to one or more of the following: rotators 172A-172M, rotation circuit 142, rotation VR file 140, one or more processors 190, device 102, Figure 1A The system 100, the processor 2406, the one or more processors 2410, the device 2400, one or more other circuits or components configured to rotate data in a rotated vector register of a rotated vector register file, or any combination thereof.

[0212] The apparatus also includes means for receiving first input data from the rotated vector register file at a multiply-accumulate circuit (MAC). For example, the means for receiving the first input data may correspond to one or more of the following: inputs 407A-H of the MAC 160, one or more processors 190, device 102, Figure 1A The system 100, the processor 2406, the one or more processors 2410, the device 2400, one or more other circuits or components configured to receive input data from the rotated vector register file at the MAC, or any combination thereof.

[0213] The apparatus also includes means for receiving second input data from a source vector register of a second vector register file at the MAC. For example, the means for receiving the second input data may correspond to one or more of the following: inputs 409A-H of the MAC 160, one or more processors 190, device 102, Figure 1A The system 100, the processor 2406, the one or more processors 2410, the device 2400, one or more other circuits or components configured to receive input data from the source vector register at the MAC, or any combination thereof.

[0214] In some implementations, a non-transitory computer-readable medium (e.g., a computer-readable storage device such as memory 2486) includes instructions (e.g., instructions 2456) that, when executed by one or more processors (e.g., one or more processors 2410 or processor 2406), cause the one or more processors to use a rotate vector register file (e.g., rotate VR file 140) to rotate data in a rotate vector register (e.g., rotate VR 144A) of the rotate vector register file. The instructions, when executed by the one or more processors, also cause the one or more processors to receive first input data (e.g., data 145) from the rotate vector register file at a multiply-accumulate circuit (MAC) (e.g., MAC 160). The instructions, when executed by the one or more processors, also cause the one or more processors to receive second input data (e.g., data 155) from a source vector register (e.g., source VR 154) of a second vector register file (e.g., VR file 150) at the MAC.

[0215] Specific aspects of the present disclosure are described below in various sets of related embodiments:

[0216] According to embodiment 1, a device includes a processor, the processor including: a rotation vector register file, the rotation vector register file including: a rotation vector register, the rotation vector register file is configured to rotate data in the rotation vector register; a second vector register file, the second vector register file including a source vector register; and a multiplication-accumulation circuit (MAC), the multiplication-accumulation circuit (MAC) is configured to receive first input data from the rotation vector register file and second input data from the source vector register.

[0217] Embodiment 2 includes an apparatus according to embodiment 1, wherein the first input data comprises a sub-vector value of the rotated vector register, wherein the sub-vector value has multiple elements, and wherein the processor further comprises a broadcast circuit configured to provide, for each element of the multiple elements, the element to a corresponding different one of the multiple inputs of the MAC.

[0218] Embodiment 3 includes the apparatus of embodiment 1 or embodiment 2, wherein the second vector register file comprises an accumulation vector register, and wherein the MAC is configured to: generate an output based on the first input data and the second input data; and store the output in the accumulation vector register.

[0219] Embodiment 4 includes an apparatus according to any one of Embodiments 1 to 3, wherein the first input data includes a first input sub-vector value, wherein the second input data includes a second input sub-vector value, wherein the second vector register file includes an accumulator vector register, and wherein the MAC is configured to: generate a first output sub-vector value based on the first input sub-vector value and the second input sub-vector value; and store the first output sub-vector value in the accumulator vector register.

[0220] Embodiment 5 includes a device according to embodiment 4, wherein the MAC is configured to: receive a third input sub-vector value from the source vector register; generate a second output sub-vector value based on the first input sub-vector value and the third input sub-vector value; and store the second output sub-vector value in the accumulator vector register.

[0221] Embodiment 6 includes the apparatus of any one of embodiments 1 to 5, wherein the rotate vector register file is configured to rotate the data by a rotate amount.

[0222] Embodiment 7 includes the apparatus of embodiment 6, wherein the amount of rotation is selectable.

[0223] Embodiment 8 includes the apparatus of embodiment 6 or embodiment 7, wherein the rotation amount is based on an opcode of a multiply-accumulate instruction or a parameter of the multiply-accumulate instruction.

[0224] Embodiment 9 includes a device according to any one of Embodiments 6 to 8, wherein the rotation vector register file includes a rotation circuit, the rotation circuit including: a first rotator, the first rotator is coupled to the output of the rotation vector register, the first rotator is configured to perform a first data rotation corresponding to a first rotation amount to generate first rotated data; and a second rotator, the second rotator is coupled to the output of the first rotator, the second rotator is configured to perform a second data rotation corresponding to a second rotation amount to generate second rotated data, wherein the output of the second rotator is coupled to the input of the rotation vector register to update the data in the rotation vector register by the second rotated data.

[0225] Embodiment 10 includes the apparatus of Embodiment 9, wherein the second rotation amount is selectable.

[0226] Embodiment 11 includes the apparatus of embodiment 9 or embodiment 10, wherein the MAC is configured to receive the second rotated data from the rotated vector register as the first input data.

[0227] Embodiment 12 includes a device according to any one of embodiments 1 to 11, wherein the rotation vector register file further includes: a second rotation vector register; and a configurable selection circuit, the configurable selection circuit being configured to: in a first configuration, output a first sub-vector value from the rotation vector register as the first input data; and in a second configuration, output a second sub-vector value from the rotation vector register and output a third sub-vector value from the second rotation vector register as the first input data.

[0228] Embodiment 13 includes an apparatus according to any one of Embodiments 1 to 12, wherein the processor is configured to execute a multiply-accumulate instruction to: retrieve the first input data from the rotated vector register file; retrieve the second input data from the source vector register; process the first input data and the second input data at the MAC; and rotate the data in the rotated vector register.

[0229] Embodiment 14 includes the apparatus of embodiment 13, wherein after retrieving the first input data from the rotated vector register file, the data is rotated in the rotated vector register.

[0230] Embodiment 15 includes a device according to any one of Embodiments 1 to 14, wherein the processor is integrated into at least one of a mobile device, a headphone device, a wearable electronic device, a wireless speaker and voice-activated device, a camera device, an extended reality headset, or a vehicle.

[0231] According to embodiment 16, a processor-implemented method includes: using a rotate vector register file to rotate data in a rotate vector register of the rotate vector register file; receiving first input data from the rotate vector register file at a multiply-accumulate circuit (MAC); and receiving second input data from a source vector register of a second vector register file at the MAC.

[0232] Embodiment 17 comprises a processor-implemented method according to embodiment 16, wherein the first input data comprises a sub-vector value of the rotated vector register, wherein the sub-vector value has multiple elements, and further comprising providing the element to a corresponding different input of the multiple inputs of the MAC using a broadcast circuit for each element of the multiple elements.

[0233] Embodiment 18 includes a processor-implemented method according to embodiment 16 or embodiment 17, further comprising: generating an output based on the first input data and the second input data using the MAC; and storing the output in an accumulation vector register of the second vector register file.

[0234] Embodiment 19 includes a processor-implemented method according to any one of Embodiments 16 to 18, further comprising: generating a first output sub-vector value at the MAC based on a first input sub-vector value and a second input sub-vector value, wherein the first input data includes the first input sub-vector value, and wherein the second input data includes the second input sub-vector value; and storing the first output sub-vector value in an accumulation vector register of the second vector register file.

[0235] Embodiment 20 includes a processor-implemented method according to embodiment 19, further comprising: receiving a third input sub-vector value from the source vector register at the MAC; generating a second output sub-vector value based on the first input sub-vector value and the third input sub-vector value at the MAC; and storing the second output sub-vector value in the accumulator vector register.

[0236] Embodiment 21 includes a processor-implemented method according to any one of Embodiments 16 to 20, wherein the rotate vector register file is used to rotate the data by a rotate amount.

[0237] Embodiment 22 comprises the processor-implemented method of embodiment 21, wherein the amount of rotation is selectable.

[0238] Embodiment 23 includes a processor-implemented method according to embodiment 21 or embodiment 22, wherein the rotation amount is based on an opcode of a multiply-accumulate instruction or a parameter of the multiply-accumulate instruction.

[0239] Embodiment 24 includes a method implemented by a processor according to any one of Embodiments 21 to 23, further comprising: using a first rotator of a rotation circuit of the rotation vector register file to perform a first data rotation corresponding to a first rotation amount to generate first rotated data; using a second rotator of the rotation circuit to perform a second data rotation corresponding to a second rotation amount to generate second rotated data; and updating the data in the rotation vector register by the second rotated data.

[0240] Embodiment 25 comprises the processor-implemented method of embodiment 24, wherein the second rotation amount is selectable.

[0241] Embodiment 26 includes a processor-implemented method according to embodiment 24 or embodiment 25, wherein the second rotated data is received at the MAC from the rotated vector register as the first input data.

[0242] Embodiment 27 includes a method implemented by a processor according to any one of Embodiments 16 to 26, further comprising: in a first configuration, outputting a first sub-vector value from the rotation vector register as the first input data; and in a second configuration, outputting a second sub-vector value from the rotation vector register and outputting a third sub-vector value from the second rotation vector register as the first input data, wherein the rotation vector register file includes the second rotation vector register.

[0243] Embodiment 28 includes a processor-implemented method according to any one of Embodiments 16 to 27, further comprising executing a multiply-accumulate instruction, wherein the multiply-accumulate instruction comprises: retrieving the first input data from the rotated vector register file; retrieving the second input data from the source vector register; processing the first input data and the second input data at the MAC; and rotating the data in the rotated vector register.

[0244] Embodiment 29 includes the processor-implemented method of embodiment 28, wherein after retrieving the first input data from the rotated vector register file, the data is rotated in the rotated vector register.

[0245] Embodiment 30 includes a processor-implemented method according to any one of Embodiments 16 to 29, wherein the rotated vector register file, the MAC, and the second vector register file are integrated into at least one of a mobile device, a headphone device, a wearable electronic device, a wireless speaker and voice-activated device, a camera device, an extended reality headset, or a vehicle.

[0246] According to embodiment 31, a device includes: a memory configured to store instructions; and a processor configured to execute the instructions to perform a processor-implemented method according to any one of embodiments 16 to 30.

[0247] According to embodiment 32, a non-transitory computer-readable medium stores instructions, which, when executed by a processor, cause the processor to perform a method implemented by a processor according to any one of embodiments 16 to 30.

[0248] According to embodiment 33, an apparatus includes means for performing the processor-implemented method according to any one of embodiments 16 to 30.

[0249] According to embodiment 34, a non-transitory computer-readable medium storing instructions, which, when executed by a processor, causes the processor to: use a rotate vector register file to rotate data in a rotate vector register of the rotate vector register file; receive first input data from the rotate vector register file at a multiply-accumulate circuit (MAC); and receive second input data from a source vector register of a second vector register file at the MAC.

[0250] Embodiment 35 includes a non-transitory computer-readable medium according to embodiment 34, wherein the first input data includes a sub-vector value of the rotated vector register, wherein the sub-vector value has multiple elements, and wherein the instructions, when executed by the processor, cause the processor to provide, for each element of the multiple elements, the element to a corresponding different one of the multiple inputs of the MAC using a broadcast circuit.

[0251] According to embodiment 36, an apparatus includes: means for rotating data in a rotated vector register of a rotated vector register file; means for receiving first input data from the rotated vector register file at a multiply-accumulate circuit (MAC); and means for receiving second input data from a source vector register of a second vector register file at the MAC.

[0252] Embodiment 37 comprises the apparatus according to embodiment 36, wherein the member for rotating, the member for receiving the first input data and the member for receiving the second input data are integrated into at least one of the following: a smart speaker, a speaker bar, a computer, a tablet computer, a display device, a television, a game console, a music player, a radio, a digital video player, a camera, a navigation device, a vehicle, a headset, an augmented reality headset, a mixed reality headset, a virtual reality headset, an extended reality headset, an air vehicle, a home automation system, a voice-activated device, a wireless speaker and a voice-activated device, a portable electronic device, a communication device, an Internet of Things (IoT) device, a virtual reality (VR) device, a base station or a mobile device.

[0253] It will also be appreciated by the skilled person that the various illustrative logic blocks, configurations, modules, circuits, and algorithmic steps described in conjunction with the specific implementations disclosed herein can be implemented as electronic hardware, computer software executed by a processor, or a combination of the two. Various illustrative components, blocks, configurations, modules, circuits, and steps have been generally described above in terms of their functionality. Whether such functionality is implemented as hardware or processor executable instructions depends on the specific application and the design constraints imposed on the overall system. The skilled person may implement the described functionality in different ways for each specific application, and such specific implementation decisions will not be interpreted as departing from the scope of the present disclosure.

[0254] The steps of the method or algorithm described in conjunction with the specific implementation disclosed herein may be directly embodied in hardware, in a software module executed by a processor, or in a combination of the two. The software module may reside in a random access memory (RAM), a flash memory, a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a register, a hard disk, a removable disk, a compact disc read-only memory (CD-ROM), or any other form of non-transient storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. In an alternative, the storage medium may be integral with the processor. The processor and the storage medium may reside in an application specific integrated circuit (ASIC). The ASIC may reside in a computing device or a user terminal. In an alternative, the processor and the storage medium may reside in a computing device or a user terminal as discrete components.

[0255] The previous description of the disclosed aspects is provided to enable those skilled in the art to make or use the disclosed aspects. Various modifications to these aspects will be apparent to those skilled in the art, and the principles defined herein may be applied to other aspects without departing from the scope of the present disclosure. Thus, the present disclosure is not intended to be limited to the aspects shown herein, but should be granted the broadest scope that may be consistent with the principles and novel features as defined by the following claims.

Claims

1. A device comprising a processor, the processor include: a rotation vector register file, the rotation vector register file comprising a rotation vector register, the rotation vector register file being configured to rotate data in the rotation vector register; a second vector register file, the second vector register file comprising source vector registers; and A multiply-accumulate circuit (MAC) is configured to receive first input data from the rotated vector register file and second input data from the source vector register.

2. The apparatus of claim 1 , wherein the first input data comprises a sub-vector value of the rotated vector register, wherein the sub-vector value has a plurality of elements, and wherein the processor further comprises a broadcast circuit configured to provide, for each element of the plurality of elements, the element to a corresponding different one of the plurality of inputs of the MAC.

3. The apparatus of claim 1 , wherein the second vector register file comprises an accumulation vector register, and wherein the MAC is configured to: generating an output based on the first input data and the second input data; and The output is stored in the accumulation vector register.

4. The apparatus of claim 1 , wherein the first input data comprises a first input sub-vector value, wherein the second input data comprises a second input sub-vector value, wherein the second vector register file comprises an accumulation vector register, and wherein the MAC is configured to: generating a first output sub-vector value based on the first input sub-vector value and the second input sub-vector value; and The first output sub-vector value is stored in the accumulation vector register.

5. The apparatus of claim 4, wherein the MAC is configured to: receiving a third input sub-vector value from the source vector register; generating a second output sub-vector value based on the first input sub-vector value and the third input sub-vector value; and The second output sub-vector value is stored in the accumulation vector register. 6 . The apparatus of claim 1 , wherein the rotate vector register file is configured to rotate the data by a rotate amount.

7. The apparatus of claim 6, wherein the amount of rotation is selectable.

8. The apparatus of claim 6, wherein the rotation amount is based on an opcode of a multiply-accumulate instruction or a parameter of the multiply-accumulate instruction.

9. The apparatus of claim 6, wherein the rotated vector register file further comprises a rotate circuit, the rotate circuit include: a first rotator coupled to an output of the rotation vector register, the first rotator configured to perform a first data rotation corresponding to a first rotation amount to generate first rotated data; and a second rotator coupled to an output of the first rotator, the second rotator configured to perform a second data rotation corresponding to a second rotation amount to generate second rotated data, The output of the second rotator is coupled to the input of the rotation vector register to update the data in the rotation vector register with the second rotation data.

10. The apparatus of claim 9, wherein the second amount of rotation is selectable. 11 . The apparatus of claim 9 , wherein the MAC is configured to receive the second rotated data from the rotated vector register as the first input data.

12. The apparatus of claim 1, wherein the rotated vector register file further comprises: include: A second rotation vector register; and A configurable selection circuit, the configurable selection circuit being configured to: In a first configuration, outputting a first sub-vector value from the rotation vector register as the first input data; as well as In a second configuration, a second sub-vector value is output from the rotated vector register and a third sub-vector value is output from the second rotated vector register as the first input data.

13. The apparatus of claim 1 , wherein the processor is configured to execute a multiply-accumulate instruction to: retrieving the first input data from the rotated vector register file; retrieving the second input data from the source vector register; processing the first input data and the second input data at the MAC; as well as The data in the rotate vector register is rotated.

14. The apparatus of claim 13, wherein after retrieving the first input data from the rotated vector register file, the data is rotated in the rotated vector register.

15. The device of claim 1, wherein the processor is integrated into at least one of a mobile device, a headphone device, a wearable electronic device, a wireless speaker and voice-activated device, a camera device, an extended reality headset, or a vehicle.

16. A processor-implemented method, include: using a rotate vector register file to rotate data in a rotate vector register of the rotate vector register file; receiving first input data from the rotated vector register file at a multiply-accumulate circuit (MAC); as well as Second input data is received at the MAC from a source vector register of a second vector register file.

17. The processor-implemented method of claim 16, wherein the first input data comprises a sub-vector value of the rotated vector register, wherein the sub-vector value has a plurality of elements, and further comprising providing, for each element of the plurality of elements, using a broadcast circuit, the element to a corresponding different one of a plurality of inputs of the MAC.

18. The processor-implemented method of claim 16, further comprising: include: generating an output based on the first input data and the second input data using the MAC; as well as The output is stored in an accumulation vector register of the second vector register file.

19. The processor-implemented method of claim 16, further comprising: include: generating, at the MAC, a first output sub-vector value based on a first input sub-vector value and a second input sub-vector value, wherein the first input data includes the first input sub-vector value, and wherein the second input data includes the second input sub-vector value; and The first output sub-vector value is stored in an accumulation vector register of the second vector register file.

20. The processor-implemented method of claim 19, further comprising: include: receiving, at the MAC, a third input sub-vector value from the source vector register; generating, at the MAC, a second output sub-vector value based on the first input sub-vector value and the third input sub-vector value; as well as The second output sub-vector value is stored in the accumulation vector register.

21. The processor implemented method of claim 16, wherein the rotate vector register file is used to rotate the data by a rotate amount.

22. The processor-implemented method of claim 21, wherein the amount of rotation is selectable.

23. The processor-implemented method of claim 21, wherein the amount of rotation is based on an opcode of a multiply-accumulate instruction or a parameter of the multiply-accumulate instruction.

24. The processor-implemented method of claim 21 , further comprising: include: performing a first data rotation corresponding to a first rotation amount using a first rotator of a rotation circuit of the rotated vector register file to generate first rotated data; performing a second data rotation corresponding to a second rotation amount using a second rotator of the rotation circuit to generate second rotated data; as well as The data in the rotation vector register is updated by the second rotation data.

25. The processor-implemented method of claim 24, wherein the second rotation amount is selectable.

26. The processor-implemented method of claim 16, further comprising executing a multiply-accumulate instruction, the multiply-accumulate instruction include: retrieving the first input data from the rotated vector register file; retrieving the second input data from the source vector register; processing the first input data and the second input data at the MAC; as well as The data in the rotate vector register is rotated.

27. A non-transitory computer readable medium storing instructions that, when executed by a processor, cause the processor to: using a rotate vector register file to rotate data in a rotate vector register of the rotate vector register file; receiving first input data from the rotated vector register file at a multiply-accumulate circuit (MAC); and Second input data is received at the MAC from a source vector register of a second vector register file.

28. The non-transitory computer-readable medium of claim 27, wherein the first input data comprises a sub-vector value of the rotated vector register, wherein the sub-vector value has a plurality of elements, and wherein the instructions, when executed by the processor, cause the processor to provide, for each element of the plurality of elements, to a corresponding different one of a plurality of inputs of the MAC using a broadcast circuit.

29. A device, include: means for rotating data in a rotate vector register of a rotate vector register file; means for receiving first input data from the rotated vector register file at a multiply-accumulate circuit (MAC); and means for receiving, at the MAC, second input data from a source vector register of a second vector register file.

30. An apparatus according to claim 29, wherein the component for rotating, the component for receiving the first input data, and the component for receiving the second input data are integrated into at least one of the following: a smart speaker, a speaker bar, a computer, a tablet computer, a display device, a television, a game console, a music player, a radio, a digital video player, a camera, a navigation device, a vehicle, a headset, an augmented reality headset, a mixed reality headset, a virtual reality headset, an air vehicle, a home automation system, a voice-activated device, a wireless speaker and a voice-activated device, a portable electronic device, a communication device, an Internet of Things (IoT) device, a virtual reality (VR) device, a base station, or a mobile device.

Citation Information

Patent Citations

  • System and method of rotating vector input

    US20230350640A1