Mips Technologies, Inc.

Mips Tech Limited

Imagination Technologies Limited

Ryan C Kinter

Darren Miller Jones

Sanjay Vishin

Michael Gottlieb Jensen

Gideon Intrater

Radhika Thekkath

Matthias Knoth

Soumya Banerjee

Kevin D Kissell

Thomas Albert Petersen

David A Courtright

Scott Michael McCoy

Anders M Jagd

G Michael Uhler

Thomas Benjamin Berg

John L Kelley

Era Kasturia Nangia

Chinh Nguyen Tran

Jaidev Prasad Patwardhan

Timothy Charles Mace

Rivka Shenhav

Thomas A Petersen

Fault tolerant and fault detecting multi-threaded processors are described. Instructions from a program are executed by both a master thread and a slave thread and execution of the master thread is prioritized. If the master thread stalls or reaches a memory write after having executed a sequence of instructions, the slave thread executes a corresponding sequence of instructions, where at least the first and last instructions in the sequence are the same as the sequence executed by the master thread. When the slave thread reaches the point at which execution of the master thread stopped, the contents of register banks for both the threads are compared, and if they are the same, execution by the master thread is allowed to continue, and any buffered speculative writes are committed to the memory system.

Fault detecting and fault tolerant multi-threaded processors

A multiprocessor system maintains cache coherence among processors in a coherent domain. Within the coherent domain, a first processor can receive a command to perform a cache maintenance operation. The first processor can determine whether the cache maintenance operation is a coherent operation. For coherent operations, the first processor sends a coherent request message for distribution to other processors in the coherent domain and can cancel execution of the cache maintenance operation pending receipt of intervention messages corresponding to the coherent request. The intervention messages can reflect a global ordering of coherence traffic in the multiprocessor system and can include instructions for maintaining a data cache and an instruction cache of the first processor. Cache maintenance operations that are determined to be non-coherent can be executed at the first processor without sending the coherent request.

Coherent instruction cache utilizing cache-op execution resources

A computer readable storage medium includes executable instructions to characterize a coherency controller. The executable instructions define ports to receive processor trace information from a set of processors. The processor trace information from each processor includes a processor identity and a condensed coherence indicator. Circuitry produces a trace stream with trace metrics and condensed coherence indicators.

Apparatus and method for condensing trace information in a multi-processor system

A multithreading processor for concurrently executing multiple threads is provided. The processor includes an execution pipeline and a thread scheduler that dispatches instructions of the threads to the execution pipeline. The execution pipeline execution pipeline is configured for generating a thread context (TC) flush indicator associated with a thread context when one or more instructions of the thread context would stall in the execution pipeline. One or more instructions in the pipeline of the thread context associated with the thread context flush signal can be flushed or nullified.

Multithreading microprocessor with optimized thread scheduler for increasing pipeline utilization efficiency

A number of coherence domains are maintained among the multitude of processing cores disposed in a microprocessor. A cache coherency manager defines the coherency relationships such that coherence traffic flows only among the processing cores that are defined as having a coherency relationship. The data defining the coherency relationships between the processing cores is optionally stored in a programmable register. For each source of a coherent request, the processing core targets of the request are identified in the programmable register. In response to a coherent request, an intervention message is forwarded only to the cores that are defined to be in the same coherence domain as the requesting core. If a cache hit occurs in response to a coherent read request and the coherence state of the cache line resulting in the hit satisfies a condition, the requested data is made available to the requesting core from that cache line.

Support for multiple coherence domains

A three-tiered TLB architecture in a multithreading processor that concurrently executes multiple instruction threads is provided. A macro-TLB caches address translation information for memory pages for all the threads. A micro-TLB caches the translation information for a subset of the memory pages cached in the macro-TLB. A respective nano-TLB for each of the threads caches translation information only for the respective thread. The nano-TLBs also include replacement information to indicate which entries in the nano-TLB/micro-TLB hold recently used translation information for the respective thread. Based on the replacement information, recently used information is copied to the nano-TLB if evicted from the micro-TLB.

Three-tiered translation lookaside buffer hierarchy in a multithreading microprocessor

A modular subtraction instruction for execution on a microprocessor having at least one register. The instruction includes opcode bits for designating the instruction and operand bits for designating at least one register storing an offset index, a decrement value, and an address index. When the modular subtraction instruction is executed on the microprocessor, the address index is modified by the decrement value if the address index is not zero and is modified by the offset index if the address index is zero. For example, the address index is repeatedly decremented using the decrement value until it reaches zero, and then the address index is reset back to the offset index. The operand bits may include multiple fields identifying multiple registers selected from the general purpose registers of the microprocessor. The modular subtraction instruction enables access to a buffer in memory in circular fashion by virtue of its operation.

Microprocessor instruction using address index values to enable access of a virtual buffer in circular fashion

An apparatus for reducing instruction re-fetching in a multithreading processor configured to concurrently execute a plurality of threads is disclosed. The apparatus includes a buffer for each thread that stores fetched instructions of the thread, having an indicator for indicating which of the fetched instructions in the buffer have already been dispatched for execution. An input for each thread indicates that one or more of the already-dispatched instructions in the buffer has been flushed from execution. Control logic for each thread updates the indicator to indicate the flushed instructions are no longer already-dispatched, in response to the input. This enables the processor to re-dispatch the flushed instructions from the buffer to avoid re-fetching the flushed instructions. In one embodiment, there are fewer buffers than threads, and they are dynamically allocatable by the threads. In one embodiment, a single integrated buffer is shared by all the threads.

Instruction/skid buffers in a multithreading microprocessor that store dispatched instructions to avoid re-fetching flushed instructions

Livelocks are prevented in multiple core processors by canceling data access requests upon determining that they conflict with other data access requests. A requesting processor core sends a data access request potentially causing livelock to a cache coherency manager. A cache coherency manager receives data access requests from multiple processor. The cache coherency manager sends intervention messages to all of the processor cores in response to all data access requests that may cause livelock. Upon receiving an intervention message from the cache coherency manager, the processor core determines if the intervention message corresponds with any of its own pending data access requests. If the intervention message is associated with a data access request conflicting with one of its own pending data access requests, the processor core responds to the invention message by directing the cache coherency manager to cancel its own conflicting pending data access request.

Avoiding livelock using intervention messages in multiple core processors

A leaky-bucket style thread scheduler for scheduling concurrent execution of multiple threads in a microprocessor is provided. The execution pipeline notifies the scheduler when it has completed instructions. The scheduler maintains a virtual water level for each thread and decreases it each time the execution pipeline executes an instruction of the thread. The scheduler includes an instruction execution rate for each thread. The scheduler increases the virtual water level based on the requested rate per a predetermined number of clock cycles. The scheduler includes virtual water pressure parameters that define a set of virtual water pressure ranges over the height of the virtual water bucket. When a thread's virtual water level moves from one virtual water pressure range to the next higher range, the scheduler increases the instruction issue priority for the thread; conversely, when the level moves down, the scheduler decreases the instruction issue priority for the thread.

Leaky-bucket thread scheduler in a multithreading microprocessor

Livelocks are prevented in multiple core processors by verifying that a data access request is still valid before sending messages to processor cores that may cause other data access requests to fail. A cache coherency manager receives data access requests from multiple processor cores. Upon receiving a data access request that may cause a livelock, the cache coherency manager first sends an intervention message back to the requesting processor core to confirm that this data access request will succeed. If the requesting processor core determines that the data access request is still valid, it directs the cache coherency manager to proceed with the data access request. The cache coherency manager may then send intervention messages to other processor cores to complete the data access request. If the requesting processor core determines that the data access request is invalid, it directs the cache coherency manager to abandon the data access request.

Avoiding livelock using a cache manager in multiple core processors

A method, cache controller, and computer processor provide a parallel mapping system whereby a plurality of mappers processes several inputs simultaneously. The plurality of mappers are disposed in a pipelined processor upstream from a multiplexor. Mapping, tag comparison, and selection by the multiplexor all occur in a single pipeline stage. Data does not wait idly to be selected by the multiplexor. Instead, each instruction of a first instruction set is read in parallel into a corresponding one of the plurality of mappers. This parallel mapping system implementation reduces processor cycle time and results in improved processor efficiency.

Mapping system and method for instruction set processing

A method and apparatus for recoding one or more instruction sets. An expand instruction and an expandable instruction are read from an instruction cache. A tag compare and way selection unit checks to verify each instruction is a desired instruction. An instruction staging unit dispatches the expand instruction to a first recoder and the expandable instruction to a second recoder of a recoding unit. At least one information bit based on the expand instruction is generated at the first recoder. The second recoder uses the at least one information bit generated at the first recoder to recode the expandable instruction, and the recoded expandable instruction is placed in an instruction buffer.

Multi-ISA instruction fetch unit for a processor, and applications thereof

An instruction dispatching apparatus in a multi threading microprocessor that concurrently executes N threads each in one of G groups each having one of P priorities. G round-robin vectors each have N bits corresponding to the threads, each being a 1-bit left-rotated and subsequently sign-extended version of an N-bit vector with a single bit true of the last thread selected for dispatching in the group. Each of N G-input muxes receive a corresponding one of the N bits of each of the round-robin vectors and selects for output one of the inputs specified by the corresponding thread's group. Selection logic selects for dispatching one of the N instructions corresponding to the thread whose dispatch value is greater than or equal to any of the N threads left thereof. Each dispatch value comprises a least-significant bit of the corresponding mux output, a most-significant dispatchable instruction bit, and middle thread group priority bits.

Multithreading instruction scheduler employing thread group priorities

An apparatus for scheduling dispatch of instructions among a plurality of threads being concurrently executed in a multithreading processor is provided. The apparatus includes an instruction decoder that generate register usage information for an instruction from each of the threads, a priority generator that generates a priority for each instruction based on the register usage information and state information of instructions currently executing in an execution pipeline, and selection logic that dispatches at least one instruction from at least one thread based on the priority of the instructions. The priority indicates the likelihood the instruction will execute in the execution pipeline without stalling. For example, an instruction may have a high priority if it has little or no register dependencies or its data is known to be available; or may have a low priority if it has strong register dependencies or is an uncacheable or synchronized storage space load instruction.

Growing community of inventors

Sammamish, WA, United States of America

Ryan C Kinter