Nxp Usa, Inc.

Freescale Semiconductor,inc.

Nxp B.v.

Analog Devices,inc.

Sourav Roy

Sneha Mishra

Arvind Kaushik

Joseph Gergen

Prokash Ghosh

Michael Brian Schinzler

Ashish Mathur

Howard Dewey Owens

Rajan Srivastava

Vikas Ahuja

Neha Raj

Shourjo Banerjee

Tushar Prakash Ringe

A system on chip (SoC) architecture includes an integrated branch and cache hit-miss trace circuit operably coupled to a CPU core, a first trace circuit, and a cache hit-miss trace circuit. Following an occurrence of a cache-fetch instruction: the cache hit-miss trace circuit identifies whether the fetch instruction is a cache-missed instruction, and, in response thereto, sends a cache miss report message that includes a fetch instruction address to the first trace circuit. The first trace circuit is configured to identify whether the fetch instruction is a taken-branch instruction and creates a modified branch trace response message (BTM) that includes the fetch instruction address and sends the modified BTM to a create trace messages circuit. The modified BTM indicates an instruction address of the cache miss.

System and method for tracing instruction cache misses

In a hardware-accelerated computing systems, calls are made from a processor to an accelerator core. The hardware-accelerated computing system includes a processor core having a stack, an accelerator, and an accelerator scheduler. The computing system is configured to process an accelerator command by the processor core issuing an accelerator command to the accelerator scheduler during execution of a task the accelerator scheduler receiving the accelerator command and requesting data from the stack, the processor sending the requested data from the stack to the accelerator scheduler, the accelerator scheduler sending the requested data to the accelerator and sending a write response to the processor, the accelerator processing the accelerator command, and the processor continuing execution of the task. The processor pauses execution of the task upon issuing the accelerator command and resumes execution of the task upon receiving the write response from the accelerator scheduler.

Hardware-accelerated computing system

A thread management circuit of a processing system stores a thread identifier table and a thread completion table. The thread management circuit receives, from a processor core, a request for execution of a portion of an application by an accelerator circuit. The thread management circuit allocates a thread identifier available in the thread identifier table to the processor core for the execution of the portion by the accelerator circuit. The thread management circuit communicates a response and an acceleration request, both including the allocated thread identifier, to the processor core and the accelerator circuit, respectively. The thread management circuit communicates a thread joining response to the processor core based on a received thread joining request and an indication by the thread completion table that the execution of the portion by the accelerator circuit is complete. The executed portion is integrated with the application based on the thread joining response.

Method and system for accelerator thread management

A memory controller for a (DRAM) memory processes an (access) command for a target row in the memory, increments a count value for each victim row associated with the target row, and issues a (dummy activate) command for a victim row whose count value reaches a specified threshold. By tracking victim rows instead of target rows, the memory controller can thwart both single-sided and double-sided row-hammer attacks. The memory controller maintains the victim-row addresses and corresponding command counts in a TCAM memory to detect rows that may be prone to row-hammer attacks. If so, then the memory controller issues dummy activate commands to the corresponding memory rows to thwart such row-hammer attacks.

Technique for detecting and thwarting row-hammer attacks

A technique for operating a system including a plurality of processors and a shared resource includes executing a first instruction by a first processor of the plurality of processors. The first instruction generates a reservation of the shared resource for the first processor. The technique includes, after generating the reservation of the shared resource for the first processor, executing a spin lock by the first processor until successful execution of a second instruction acquires a lock of the shared resource. The technique includes disabling interrupts of the first processor in response to an indicator of the successful execution of the second instruction. The first instruction may be a load and reserve instruction and the second instruction may be a conditional store instruction.

Reduction of interrupt service latency in multi-processor systems

An error-resilient memory device includes sets of memory blocks and redundant memory blocks for storing a set of data bits. A memory block includes a set of memory cells, each memory cell is adjacent to at least two other memory cells, and a memory block is formed by a matrix of the set of memory cells. In a row-folded implementation, a word line is connected to each memory cell, and a set of bit lines is connected to the corresponding set of memory cells. In a column-folded implementation, a bit line is connected to each memory cell, and a set of word lines is connected to the corresponding set of memory cells. A redundant memory block is used to store the set of data bits when the memory block includes a fault.

Error-resilient memory device with row and/or column folding with redundant resources and repair method thereof

A system for executing an accelerator call function includes a processor, a register context memory, an accelerator scheduler, multiple accelerator cores, and a stack memory. The processor executes a program task. The processor includes a register that stores task context information of the program task. The accelerator call function includes an accelerator operation. The processor forwards the accelerator operation to the accelerator scheduler. Concurrently, the processor stores the task context information in the register context memory. The accelerator scheduler identifies one of the accelerator cores and forwards the accelerator operation to the identified core. The identified core executes the accelerator operation, generates a return value, and stores the return value in the register context memory, which in turn provides the return value and the task context information to the processor.

System and method for executing accelerator call

A processor includes a processing unit, a memory, a data cache, an One Block Look-ahead (OBL) prefetch engine, a Stride-Allocate on Miss (AoM) prefetch engine and a prefetch back-off module. The prefetch back-off module assigns and sets a status bits to a prefetched cache line and resets the status bit when the cache line is used by the processing unit. The back-off module also decrements a count value when at least two cache lines are used consecutively by the processing unit, increments the count value when at least two unused cache lines are evicted consecutively from the data cache, and disables cache line prefetching when the count value is greater than zero. The stride-AoM prefetch engine includes a reference pattern table (RPT) that stores details of only those instructions that have undergone a cache miss.

Data cache prefetch controller

A method for reducing memory latency in a processor includes identifying an independent instruction (or cache miss instruction) and corresponding dependent instructions from a re-circulating issue window (RIW) when a cache miss is encountered. The cache miss instruction and corresponding dependent instructions are moved to a re-circulating issue buffer (RIB) and moved back to the RIW from the RIB for processing when the cache miss is resolved.

Method for reducing memory latency in processor

A method for estimating energy usage of a program code executed by a very long instruction word (VLIW) processor. The program code includes multiple instructions, which are organized as groups referred to as execution sets. The instructions of an execution set are executed simultaneously. A no operation (NOP) energy and incremental instruction energy for each execution set of the program code are determined to compute a base energy of each execution set. An inter-execution set energy of each execution set is then computed. The estimated energy usage of each execution set is determined by adding the corresponding inter-execution set energies to the base energies. A stall energy consumed on account of executing the program code is then computed. Finally, the estimated energy usage of the program code is determined by adding the stall energy to the determined energies of all the execution sets.

Method for estimating processor energy usage

A method and system for simultaneously reading data from multiple indexed arrays, where each indexed array includes one or more memory locations and is coupled to a multiplexing circuit. Each multiplexing circuit includes one or more multiplexers and is driven by a set of input selector signals. The method includes enabling each multiplexing circuit with a distinct combination of the set of input selector signals. The distinct combinations of the set of input selector signals cause each input selector signal to drive a comparable number of multiplexers. Each multiplexing circuit selects a memory location from the coupled indexed array. Further, the method includes reading the data at the selected memory locations through the output of each multiplexing circuit.

Growing community of inventors

Kolkata, India

Sourav Roy