Abstract: Systems and methods are disclosures for scheduling code in a multiprocessor system. Code is portioned into code blocks by a compiler. The compiler schedules execution of code blocks in nodes. The nodes are connected in a directed acyclical graph with a top node, terminal node and a plurality of intermediate nodes. Execution of the top node is initiated by the compiler. After executing at least one instance of the top node, an instruction in the code block indicates to the scheduler to initiate at least one intermediary node. The scheduler schedules a thread for execution of the intermediary node. The data for the nodes resides in a plurality of data buffers; the index to the data buffer is stored in a command buffer.

Abstract: Disclosed herein is a method and a system for generating a mixed precision quantization model for performing image processing. The method comprises receiving a validation dataset of images to train a neural network model. The method comprises for each image of the validation dataset, generating a union sensitivity list, selecting a group of layers, generating a mixed precision quantization model by quantizing the selected group of layers into a high precision format; computing accuracy of the mixed precision quantization model for comparison with a target accuracy; in response to determining the accuracy is less than the target accuracy, generating another mixed precision model by selecting a next group of layers and computing the accuracy. In response to determining the accuracy is greater than or equal to the target accuracy, storing the mixed precision quantization model as a final mixed precision quantization model for image processing.

Abstract: The present disclosure relates to a system and a method of optimizing register allocation by a processor. The method comprising receiving an intermediate representation (IR) code of a source code and initializing single instruction multiple data (SIMD) width for the IR code. The method comprising analyzing each basic block of the IR code to classify determine one or more instructions of the IR code as vector instructions, wherein each basic block is one of LOAD, STORE and arithmetic logical and multiply (ALM) instructions. The method comprising dynamically setting the SIMD width for each of the vector instructions.

Abstract: Methods, systems, and apparatuses for adaptive power supply voltage transient protection are disclosed. One system includes a system on a chip (SOC). wherein the SOC includes a power supply, a voltage transient sensor, and a power control processing entity. The power supply operates to provide power to one or more processors operating on the SOC. The voltage transient sensor is connected to the power supply and operates to sense voltage transients on the power supply at greater than a predetermined speed or rate. The power control processing entity operates to receive a digital representation of the sensed voltage transients and adjust a power load of the SOC based on the sensed voltage transients.

Abstract: Methods, systems, and apparatuses for graph streaming processing system are disclosed. One system includes a plurality of graph streaming processors operative to process a plurality of threads, wherein the plurality of threads is organized as nodes. The system further includes a scheduler that includes a plurality of stages. Each stage includes a command parser operative to interpret commands within a corresponding input command buffer, an alternate command buffer, and a thread generator coupled to the command parser. The thread generator is operative to generate the plurality of threads, and dispatch the plurality of threads, where the processing of the plurality of thread for each stage includes storing write commands in the corresponding output command buffer or in the alternate command buffer.

Abstract: Disclosed embodiments relate to a method and device for optimizing compilation of source code. The proposed method receives a first intermediate representation code of a source code and analyses each basic block instruction of the plurality of basic block instructions contained in the first intermediate representation code for blockification. In order to blockify the identical instructions, the one or more groups of basic block instructions are assessed for eligibility of blockification. Upon determining as eligible, the group of basic block instructions are blockified using one of one dimensional SIMD vectorization and two-dimensional SIMD vectorization. The method further generates a second intermediate representation of the source code which is translated to executable target code with more efficient processing capacity.

Abstract: Methods, systems and apparatuses for performing walk operations of single instruction, multiple data (SIMD) instructions are disclosed. One method includes initiating, by a scheduler, a SIMD thread, where the scheduler is operative to schedule the SIMD thread. The method further includes fetching a plurality of instructions for the SIMD thread. The method further includes determining, by a thread arbiter, at least one instruction that is a walk instruction, where the walk instruction iterates a block of instructions for a subset of channels of the SIMD thread, where the walk instruction includes a walk size, and where the walk size is a number of channels in the subset of channels of the SIMD thread that are processed in a walk iteration in association with the walk instruction. The method further includes executing the walk instruction based on the walk size.

Abstract: Methods, systems and apparatuses for graph processing are disclosed. One graph streaming processor includes a thread manager, wherein the thread manager is operative to dispatch operation of the plurality of threads of a plurality of thread processors before dependencies of the dependent threads have been resolved, maintain a scorecard of operation of the plurality of threads of the plurality of thread processors, and provide an indication to at least one of the plurality of thread processors when a dependency between the at least one of the plurality of threads that a request has or has not been satisfied. Further, a producer thread provides a response to the dependency when the dependency has been satisfied, and each of the plurality of thread processors is operative to provide processing updates to the thread manager, and provide queries to the thread manager upon reaching a dependency.

Abstract: The present disclosure relates to a system and method of performing quantization of a neural network having multiple layers. The method comprises receiving a floating-point dataset as input dataset and determining a first shift constant for first layer of the neural network based on the input dataset. The method also comprises performing quantization for the first layer using the determined shift constant of the first layer. The method further comprises determining a next shift constant for next layer of the neural network based on output of a layer previous to the next layer, and performing quantization for the next layer using the determined next shift constant. The method further comprises iterating the steps of determining shift constant and performing quantization for all layers of the neural network to generate fixed point dataset as output.

Abstract: Systems, apparatuses, and methods are disclosed for scheduling threads comprising of code blocks in a graph streaming processor (GSP) system. One system includes a scheduler for scheduling plurality of prefetch threads, main threads, invalidate threads. The plurality of prefetch threads includes prefetching data from main memory required for execution of the main threads of the next stage. The plurality of main threads includes a set of instructions operating on the graph streaming processors of GSP system. The plurality of the invalidate threads includes invalidating data location/s consumed by the plurality of the main threads of the previous stage. A portion of the scheduler is implemented in hardware.

Abstract: Systems, apparatuses, and methods are disclosed for scheduling threads comprising of code blocks in a graph streaming processor (GSP) system. One system includes a scheduler for scheduling plurality of prefetch threads, main threads, invalidate threads. The plurality of prefetch threads includes prefetching data from main memory required for execution of the main threads of the next stage. The plurality of main threads includes a set of instructions operating on the graph streaming processors of GSP system. The plurality of the invalidate threads includes invalidating data location/s consumed by the plurality of the main threads of the previous stage. A portion of the scheduler is implemented in hardware.

Abstract: Methods, systems, and apparatuses for graph streaming processing system are disclosed. One system includes a plurality of graph streaming processors operative to process a plurality of threads, wherein the plurality of threads is organized as nodes. The system further includes a scheduler that includes a plurality of stages. Each stage includes a command parser operative to interpret commands within a corresponding input command buffer, an alternate command buffer, and a thread generator coupled to the command parser. The thread generator is operative to generate the plurality of threads, and dispatch the plurality of threads, where the processing of the plurality of thread for each stage includes storing write commands in the corresponding output command buffer or in the alternate command buffer.

Abstract: Methods, systems and apparatuses for reducing operations of Sum-Of-Multiply-Accumulate (SOMAC) instructions are disclosed. One method includes scheduling, by a scheduler, a thread for execution, executing, by a processor of a plurality of processors, the thread, fetching, by the processor, a plurality of instructions for the thread from a memory, selecting, by a thread arbiter of the processor, an instruction of the plurality of instructions for execution in an arithmetic logic unit (ALU) pipeline of the processor, and reading the instruction, and determining, by a macro-instruction iterator of the processor, whether the instruction is a Sum-Of-Multiply-Accumulate (SOMAC) instruction with an instruction size, wherein the instruction size indicates a number of iterations that the SOMAC instruction is to be executed.

Abstract: Methods, systems, and apparatuses for graph stream processing are disclosed. One apparatus includes a cascade of graph streaming processors, wherein each of the graph streaming processor includes a processor array, and a graph streaming processor scheduler. The cascade of graph streaming processors further includes a plurality of shared command buffers, wherein each shared command buffer includes a buffer address, a write pointer, and a read pointer, wherein for each of the plurality of shared command buffers a first graph streaming processor writes commands to the shared command buffer as indicated by the write pointer of the shared command buffer and a second graph streaming processor reads commands from the shared command buffer as indicated by the read pointer, wherein at least one graph streaming processor scheduler operates to manage the write pointer and the read pointer to avoid overwriting unused commands of the shared command buffer.

Abstract: Methods, systems and apparatuses for graph streaming processing system are disclosed. One system includes a plurality of graph streaming processors operative to process a plurality of threads, wherein the plurality of threads is organized as nodes. The system further includes a scheduler that includes a plurality of stages. Each stage includes a command parser operative to interpret commands within a corresponding input command buffer, an alternate command buffer, and a thread generator coupled to the command parser. The thread generator is operative to generate the plurality of threads, and dispatch the plurality of threads, where the processing of the plurality of thread for each stage includes storing write commands in the corresponding output command buffer or in the alternate command buffer.

Abstract: Systems, apparatuses and methods are disclosed for scheduling threads comprising of code blocks in a graph streaming processor (GSP) system. One system includes a scheduler for scheduling plurality of threads, the plurality of threads includes a set of instructions operating on the graph streaming processors of GSP system. The scheduler comprises a plurality of stages where each stage is coupled to an input command buffer and an output command buffer. A portion of the scheduler is implemented in hardware and comprises of a command parser operative to interpret commands within a corresponding input command buffer, a thread generator coupled to the command parser operate to generate the plurality of threads, and a thread scheduler coupled to the thread generator for dispatching the plurality of threads for operating on the plurality of graph streaming processors.

Abstract: Methods, systems and apparatuses for graph streaming processing are disclosed. One method includes loading, by a group load register, a subset of a an input tensor from a data cache, wherein the group load register provides the subset of the input tensor to all of a plurality of processors, loading, by a plurality of weight data registers, a plurality of weights of a weight tensor, wherein each of the weight data registers provide an weight to a single of the plurality of processors, and performing, by the plurality of processors, a SOMAC (Sum-Of-Multiply-Accumulate) instruction, including simultaneously determining, by each of the plurality of processors, an instruction size of the SOMAC instruction, wherein the instruction size indicates a number of iterations that the SOMAC instruction is to be executed and is equal to a number of outputs within a subset of a plurality of output tensors.

Abstract: Disclosed embodiments relate to a method and device for optimizing compilation of source code. The proposed method receives a first intermediate representation code of a source code and analyses each basic block instruction of the plurality of basic block instructions contained in the first intermediate representation code for blockification. In order to blockify the identical instructions, the one or more groups of basic block instructions are assessed for eligibility of blockification. Upon determining as eligible, the group of basic block instructions are blockified using one of one dimensional SIMD vectorization and two-dimensional SIMD vectorization. The method further generates a second intermediate representation of the source code which is translated to executable target code with more efficient processing capacity.

Abstract: Methods, systems and apparatuses for graph stream processing are disclosed. One apparatus includes a cascade of graph streaming processors, wherein each of the graph streaming processor includes a processor array, and a graph streaming processor scheduler. The cascade of graph streaming processors further includes a plurality of shared command buffers, wherein each shared command buffer includes a buffer address, a write pointer, and a read pointer, wherein for each of the plurality of shared command buffers a first graph streaming processor writes commands to the shared command buffer as indicated by the write pointer of the shared command buffer and a second graph streaming processor reads commands from the shared command buffer as indicated by the read pointer, wherein at least one graph streaming processor scheduler operates to manage the write pointer and the read pointer to avoid overwriting unused commands of the shared command buffer.

Abstract: Systems, apparatuses and methods are disclosed for efficient management of registers in a graph stream processing (GSP) system. The GSP system includes a thread scheduler module operative to initiate a Single Instruction Multiple Data (SIMD) thread, the SIMD thread including a dispatch mask with an initial value. A thread arbiter module operative to select an instruction from the instructions and provide the instruction to each of one or more compute resources, and an instruction iterator module, associated with the each of one or more compute resources operative to determine a data type of the instruction. The instruction iterator module iteratively executes the instruction based on the data type and the dispatch mask.