Sifive, Inc.

Wesley Waylon Terpstra

Henry Cook

Eric Andrew Gouldey

Michael Klinglesmith

John Ingalls

Yunsup Lee

Ryan Macdonald

Derek E Pappas

Richard Xia

Mark Nugent

Pin Hung Chen

Megan Anneke Wachs

Leigang Kou

Ernest L Edgar

Richard Van

Erik Arthur Danie

A method and apparatus for a cache coherency state request vector is described. A method includes selecting, by a first agent, one or more bits in a cache coherency state request vector, where a selected bit in the cache coherency state request vector indicates an acceptable cache coherency state for a cache block indicated in a request message, transmitting, by the first agent to a second agent, the request message for the cache block, the request message including the cache coherency state request vector, and receiving, by the first agent from the second agent, a response message with a cache coherency response state, wherein the cache coherency response state indicates a cache coherency state responsive to the cache coherency state request vector.

Cache coherency state request vector encoding and use thereof

Systems and methods are disclosed for integrated circuit design using integrated circuit shells. For example, a system may generate an integrated circuit core design expressed in a hardware description language. The integrated circuit core design may express circuitry that describes one or more functions to be included in an application specific integrated circuit (ASIC). The one or more functions may have connection points providing first inputs and outputs to the one or more functions. The system may query an integrated circuit shell expressed in a hardware description language. The integrated circuit shell may express circuitry that describes a limited set of pads to be implemented in the ASIC. The limited set of pads may provide second inputs and outputs to the integrated circuit. The query may determine availability of pads of the limited set of pads to connect to the connection points of the one or more functions.

Integrated circuit generation using an integrated circuit shell

Systems and methods for data storage in a non-inclusive cache. An integrated circuit includes a cache that includes a databank with multiple entries configured to store respective cache lines and an array of cache tags. Each cache tag includes a data pointer that points to an entry in the databank. A method includes allocating the entry in the databank to the cache including the array of cache tags from amongst multiple caches in the integrated circuit by writing the data pointer to the cache tag in the array of cache tags.

Data storage in non-inclusive cache

A data responder may determine a selection between granting a request for a priority byte to be prioritized for transmission ahead of other bytes via a bus and ignoring the request. Granting the request may include transferring a block of bytes of data across multiple clock cycles with the priority byte transferred in a first clock cycle before other clock cycles of the multiple clock cycles. Ignoring the request may include transferring the block across multiple clock cycles with the priority byte transferred in a clock cycle after the first clock cycle. The data responder may receive the request from a data requestor. The data responder may assert a signal on a wire, connected to the data requestor, to indicate a grant of the request and a transfer of the priority byte in the first clock cycle.

Selective transfer of data including a priority byte

A method and apparatus for a speculative request indicator is described. A method includes providing, for a cache hierarchy, a messaging protocol used for transfer operations among agents in the cache hierarchy, the messaging protocol indicating acceptable cache coherency states for a cache block indicated in a request message and providing, in the messaging protocol for selection by an agent, a speculative request indicator when sending the request message, wherein the speculative request indicator differentiates between a demand request and a speculative request with respect to the cache block.

Speculative request indicator in request message

A method for managing orders of operations between one or more clients and one or more servers is disclosed. The method includes partitioning addressable regions of logical servers on or within an interconnect link into multiple regions including a first orderable region, and providing logical client an ability to push ordering responsibility within the first orderable region to a server. Over the first orderable region, two request messages for access to memory-mapped sources including two respective operations are transmitted, and the two request messages originate from a same logical client. The ordering responsibility can include a first rule for order of operations between the two request messages.

Orderability of operations

Systems and techniques are disclosed for relative age tracking for entries in a buffer. For example, some techniques may include pre-computing age matrix entries of an age matrix corresponding to invalid entries of a data buffer based on a validity indication (e.g., a valid bit mask), wherein the validity indication identifies valid entries in the data buffer and the age matrix tracks relative ages of the entries in the data buffer; responsive to data being received for storage in the data buffer, selecting an entry corresponding to an index value in the data buffer from among a set of invalid entries of the data buffer; storing the data in the entry corresponding to the index value; and updating the validity indication to indicate that the entry corresponding to the index value is valid.

Relative age tracking for entries in a buffer

Systems and methods are disclosed for concurrent support for multiple cache inclusivity schemes using low priority evict operations. For example, some methods may include, receiving a first eviction message having a lower priority than probe messages from a first inner cache; receiving a second eviction message having a higher priority than probe messages from a second inner cache; transmitting a third eviction message, determined based on the first eviction message, having the lower priority than probe messages to a circuitry that is closer to memory in a cache hierarchy; and, transmitting a fourth eviction message, determined based on the second eviction message, having the lower priority than probe messages to the circuitry that is closer to memory in the cache hierarchy.

Concurrent support for multiple cache inclusivity schemes using low priority evict operations

An agent may be configured to invoke a first eviction operation that is interruptible by probe operations when receiving a first type of eviction message and invoke a second eviction operation in which probe operations are interruptible by the second eviction operation when receiving a second type of eviction message. In some implementations, the agent may maintain a data storage that is inclusive of at least one of unique or dirty cache blocks in a cache maintained by an agent that transmits the second type of eviction message. In some implementations, the agent may prevent a cache block from transitioning from a modified state to an exclusive state when the agent invokes the second eviction operation to evict the cache block. In some implementations, the agent may convert from the second eviction operation to the first eviction operation when receiving the second type of eviction message.

Eviction operations based on eviction message types of different priorities

Cache circuitry may be configured to receive a first message to downgrade a permission associated with data stored in a current level cache. For example, the current level cache could be a level two (L2) cache. The cache circuitry could receive the first message from a processor core having a level one (L1) cache. The cache circuitry may forward the first message to a higher level cache. For example, the higher level cache could be a level three (L3) cache. The cache circuitry may downgrade the permission associated with data stored in the current level cache based on receiving a second message from the higher level cache. The cache circuitry may forward the first message before receiving the second message and downgrading the permission. The second message may cause downgrade of the permission in multiple caches (e.g., the L1, L2, and L3 caches).

Downgrading a permission associated with data stored in a cache

Prefetch circuitry may be configured to transmit a message to cancel a prefetch of one or more cache blocks of a group. The message may correspond to a prefetch message by indicating an address for the group and a bit field for the one or more cache blocks of the group to cancel. In some implementations, the message may target a higher level cache to cancel prefetching the one or more cache blocks, and the message may be transmitted to the higher level cache via a lower level cache. In some implementations, the message may target a higher level cache to cancel prefetching the one or more cache blocks, the message may be transmitted to a lower level cache via a first command bus, and the lower level cache may forward the message to the higher level cache via a second command bus.

Canceling prefetch of cache blocks based on an address and a bit field

First agent circuitry may receive from a second agent a first request and a first set of one or more bits. The first request may be part of a data operation. The first agent circuitry may transmit to the second agent a message including a first response to the first request, the first set of one or more bits, a second request, and a second set of one or more bits. The second set of one or more bits may be generated by the first agent circuitry to transmit state information about the second request. In some implementations, a set of one or more wires may be generated for transmission of the second set of one or more bits. The first agent circuitry may receive from the second agent a second response to the second request and the second set of one or more bits.

Transmitting a response with a request and state information about the request

Described are methods and a system for atomic memory operations with contended cache lines. A processing system includes at least two cores, each core having a local cache, and a lower level cache in communication with each local cache. One local cache configured to request a cache line to execute an atomic memory operation (AMO) instruction, receive the cache line via the lower level cache, receive a probe downgrade due to other local cache requesting the cache line prior to execution of the AMO, and send the AMO instruction to the lower level cache for remote execution in response to the probe downgrade.

Method for executing atomic memory operations when contested

Systems and methods are disclosed for automated generation of integrated circuit designs and associated data. These allow the design of processors and SoCs by a single, non-expert who understands high-level requirements; allow the en masse exploration of the design-space through the generation processors across the design-space via simulation, or emulation; allow the easy integration of IP cores from multiple third parties into an SoC; allow for delivery of a multi-tenant service for producing processors and SoCs that are customized while also being pre-verified and delivered with a complete set of developer tools, documentation and related outputs. Some embodiments, provide direct delivery, or delivery into a cloud hosting environment, of finished integrated circuits embodying the processors and SoCs.

Integrated circuits as a service

Described is a data cache implementing hybrid writebacks and writethroughs. A processing system includes a memory, a memory controller, and a processor. The processor includes a data cache including cache lines, a write buffer, and a store queue. The store queue writes data to a hit cache line and an allocated entry in the write buffer when the hit cache line is initially in at least a shared coherence state, resulting in the hit cache line being in a shared coherence state with data and the allocated entry being in a modified coherence state with data. The write buffer requests and the memory controller upgrades the hit cache line to a modified coherence state with data based on tracked coherence states. The write buffer retires the data upon upgrade. The data cache writebacks the data to memory for a defined event.

Growing community of inventors

San Mateo, CA, United States of America

Wesley Waylon Terpstra