Arm Limited

Natalya Bondarenko

Stefano Ghiggini

Geoffray Lacourba

Lucas Garcia

Damien Guillaume Pierre Payet

Cédric Denis Robert Airaud

Abhishek Raja

Frederic Claude Marie Piry

Florent Begon

Nathanael Premillieu

Miles Robert Dooley

Luca Nassi

Huzefa Moiz Sanjeliwala

Ugo Castorina

Devin S Lafford

Damien Matthieu Valentin Cathrine

Luca Maroncelli

Fabian Gruber

Kamil Garifullin

Pierre Marcel Laurent

Apparatuses, methods, systems, and chip-containing products are disclosed, which relate to an arrangement comprising a level N cache level and a level M cache level, where M is greater than N. The level N cache level comprises a plurality of linefill slots and performs a slot allocation procedure in response to a lookup miss in dependence on a linefill slot occupancy criterion. The slot allocation procedure comprises allocation of an available slot of the plurality of slots to a pending linefill request generated in response to the lookup miss. The level N cache level effects a modification of the slot allocation procedure in dependence on the linefill slot occupancy criterion and is responsive to the linefill slot occupancy criterion being fulfilled to cause a linefill delegation action to be instructed to the level M cache level.

Linefill delegation in a cache hierarchy

An apparatus is provided for controlling the operating mode of control circuitry, such that the control circuitry may change between two operating modes. In an allocation mode, data that is loaded in response to an instruction is allocated into storage circuitry from an intermediate buffer, and the data is read from the storage circuitry. In a non-allocation mode, the data is not allocated to the storage circuitry, and is read directly from intermediate buffer. The control of the operating mode may be performed by mode control circuitry, and the mode may be changed in dependence on the type of instruction that calls the data, and whether the data may be used again in the near future, or whether it is expected to be used only once.

Controlling data allocation to storage circuitry

A data processing apparatus comprises receiver circuitry for receiving instructions from each of a plurality of requester devices. Processing circuitry executes the instructions associated with each of a subset of the requester devices at a time and arbitration circuitry determines the subset of the requester devices and causes the instructions associated with each of the subset of the requester devices to be executed next. In response to the receiver circuitry receiving an instruction of a predetermined type from one of the requester devices outside the subset of requester devices, the arbitration circuitry causes the instruction of the predetermined type to be executed next.

Shared unit instruction execution

The invention provides a data processing apparatus and a data processing method for generating prefetches of data for use during execution of instructions by processing circuitry. The prefetches that are generated are based on a nested prefetch pattern. The nested prefetch pattern comprises a first pattern and a second pattern. The first pattern is defined by a first address offset between sequentially accessed addresses and a first observed number of the sequentially accessed addresses separated by the first address offset. The second pattern is defined by a second address offset between sequential iterations of the first pattern and a second observed number of the sequential iterations of the first pattern separated by the second address offset.

Data processing apparatus and method for generating prefetches based on a nested prefetch pattern

An apparatus comprises a cache to store information, items of information in the cache being associated with addresses; cache lookup circuitry to perform lookups in the cache; and a prefetcher to prefetch items of information into the cache in advance of an access request being received for said items of information. The prefetcher selects addresses to train the prefetcher. In response to determining that a cache lookup specifying a given address has resulted in a hit and determining that a cache lookup previously performed in response to a prefetch request issued by the prefetcher for the given address resulted in a hit, the prefetcher selects the given address as an address to be used to train the prefetcher.

Prefetcher training

A technique is provided for training a prediction apparatus. The apparatus has an input interface for receiving a sequence of training events indicative of program instructions, and identifier value generation circuitry for performing an identifier value generation function to generate, for a given training event received at the input interface, an identifier value for that given training event. The identifier value generation function is arranged such that the generated identifier value is dependent on at least one register referenced by a program instruction indicated by that given training event. Prediction storage is provided with a plurality of training entries, where each training entry is allocated an identifier value as generated by the identifier value generation function, and is used to maintain training data derived from training events having that allocated identifier value. Matching circuitry is then responsive to the given training event to detect whether the prediction storage has a matching training entry (i.e. an entry whose allocated identifier value matches the identifier value for the given training event). If so, it causes the training data in the matching training entry to be updated in dependence on the given training event.

Technique for training a prediction apparatus

Examples of the present disclosure relate to an apparatus comprising processing circuitry to perform data processing operations, storage circuitry to store data for access by the processing circuitry, address translation circuitry to maintain address translation data for translating virtual memory addresses into corresponding physical memory addresses, and prefetch circuitry. The prefetch circuitry is arranged to prefetch first data into the storage circuitry in anticipation of the first data being required for performing the data processing operations. The prefetching comprises, based on a prediction scheme, predicting a first virtual memory address associated with the first data, accessing the address translation circuitry to determine a first physical memory address corresponding to the first virtual memory address, and retrieving the first data based on the first physical memory address corresponding to the first virtual memory address. The prefetch circuitry is further arranged, based on the prediction scheme, to predict a second virtual memory address associated with second data in anticipation of the second data being prefetched, and to provide the predicted second virtual memory address to the address translation circuitry to enable the address translation circuitry to obtain the address translation data for the second virtual memory address.

Apparatus and method for performing address translation using buffered address translation data

There is provided an apparatus that includes storage circuitry. The storage circuitry is made up from a plurality of sets, each of the sets having at least one storage location. Receiving circuitry receives an access request that includes an input address. Lookup circuitry obtains a plurality of candidate sets that correspond with an index part of the input address. The lookup circuitry determines a selected storage location from the candidate sets using an access policy. The access policy causes the lookup circuitry to iterate through the candidate sets to attempt to locate an appropriate storage location. The appropriate storage location is accessed in response to the appropriate storage location being found.

Increasing effective cache associativity

Examples of the present disclosure relate to an apparatus comprising processing circuitry to perform data processing operations and a hierarchical cache structure. The cache structure comprises a plurality of cache levels to store data for access by the processing circuitry, and includes a highest cache level arranged to receive data requests directly from the processing circuitry. The apparatus comprises a plurality of prefetch units, each prefetch unit being associated with a cache level and being arranged to prefetch data into the associated cache level in anticipation of the processing circuitry requiring the data, wherein: each cache level has a plurality of entries and is arranged to maintain prefetch tag information for each entry which, when a given entry contains prefetched data, indicates which prefetch unit of that cache level and/or of a lower cache level prefetched that data; and each cache level is arranged, responsive to a data request from a higher cache level, to provide to the higher cache level the requested data and the prefetch tag information corresponding to the requested data. The apparatus further comprises accuracy information storage to: maintain accuracy inferring information for each prefetch unit; and when given data is evicted from a cache level, update the accuracy inferring information based on the prefetch tag information.

Methods and apparatus for monitoring prefetcher accuracy information using a prefetch flag independently accessible from prefetch tag information

A technique is described for managing a cache structure in a system employing transactional memory. The apparatus comprises processing circuitry to perform data processing operations in response to instructions, the processing circuitry comprising transactional memory support circuitry to support execution of a transaction, and a cache structure comprising a plurality of cache entries for storing data for access by the processing circuitry. Each cache entry has associated therewith an allocation tag, and allocation tag control circuitry is provided to control use of a plurality of allocation tags and to maintain an indication of a current state of each of those allocation tags. The transactional memory support circuitry is arranged, when initial data in a chosen cache entry is to be written to during the transaction, to cause a backup copy of the initial data to be stored in a further cache entry and to cause the allocation tag control circuitry to associate with that further cache entry a selected allocation tag selected for the transaction. The current state of that selected allocation tag is updated to a first state which prevents the processing circuitry from accessing that further cache entry. In the event that the transaction is aborted prior to reaching a transaction end point, the transactional memory support circuitry causes the chosen cache entry to be invalidated, and the allocation tag control circuitry changes the state of the selected allocation tag to a second state that allows the processing circuitry to access the further cache entry. As a result, this enables a hit to subsequently be detected within the cache structure for the initial data without a requirement to refetch the initial data into the cache structure. This can give rise to significant performance enhancements.

Technique for managing a cache memory in a system employing transactional memory including storing a backup copy of initial data for quick recovery from a transaction abort

An apparatus comprises processing circuitry, transactional memory support circuitry and a cache. The processing circuitry processes threads of data processing, and the transactional memory support circuitry supports execution of a transaction within a thread, including tracking a read set of addresses, comprising addresses accessed by read instructions within the transaction. A transaction comprises instructions for which the processing circuitry is configured to prevent commitment of the results of speculatively executed instruction until the transaction has completed. The cache has a plurality of entries, each associated with an address and specifying a replaceable-information value for that address that comprises information for which, outside of the transaction, processing would be functionally correct even if the information was incorrect. While the transaction is pending, the transactional memory support circuitry identifies, based on an encoding of the replaceable-information values, read-set information identifying addresses in the read set tracked for the transaction.

Identifying read-set information based on an encoding of replaceable-information values

Data processing circuitry comprises a cache memory to cache a subset of data elements from a main memory; a processing element to execute program code to access data elements having respective memory addresses, the processing element being configured to access the data elements in the cache memory and, in the case of a cache miss, to fetch the data elements from the main memory; prefetch circuitry, responsive to an access to a current data element, to initiate prefetching into the cache memory of a data element at a memory address defined by a current offset value relative to the address of the current data element; and offset value selection circuitry comprising: an address table to store memory addresses for which a data element accessed by the processing element resulted in a cache miss or an access to a previously prefetched data element; and detector circuitry to detect, for each of a group of candidate offset values, one or more respective metrics representing a proportion of a set of data element accesses which resulted in a cache miss or an access to a previously prefetched data element, for which the memory address for that data element access differs by the candidate offset value from a memory address in the address table; in which the detector circuitry is configured to process the group of candidate offset values as successive complementary sub-groups of one or more of the group of candidate offset values and to set a next instance of the current offset value in response to processing each sub-group, in dependence upon the proportions indicated by the one or more detected metrics for that sub-group; and the one or more metrics previously detected for the current offset value.

Prefetching in data processing circuitry

Data processing circuitry comprises a cache memory to cache a subset of data elements from a main memory; a processing element to execute program code to access data elements having respective memory addresses, the processing element being configured to access the data elements in the cache memory and, in the case of a cache miss, to fetch the data elements from the main memory; prefetch circuitry, responsive to an access to a current data element, to initiate prefetching into the cache memory of a data element at a memory address defined by a current offset value relative to the address of the current data element; offset value selection circuitry comprising: an address table to store memory addresses for which a data element accessed by the processing element resulted in a cache miss or an access to a previously prefetched data element; detector circuitry to detect, for each of a group of candidate offset values, one or more respective metrics representing a proportion of a set of data element accesses which resulted in a cache miss or an access to a previously prefetched data element, for which the memory address for that data element access differs by the candidate offset value from a memory address in the address table; in which the detector circuitry is configured to set a next instance of the current offset value in response to the one or more detected metrics; verification circuitry to detect, at one or more predetermined stages with respect to the processing of the group of candidate offset values by the offset value selection circuitry, one or more verification metrics representing a proportion of a set of data element accesses which resulted in a cache miss or an access to a previously prefetched data element, for which the memory address for that data element access differs by the current offset value from a memory address in the address table, to detect whether the one or more verification metrics comply with a predetermined condition; and control circuitry to inhibit prefetching at least until a next selection of a current offset value by the offset value selection circuitry, in response to a detection by the verification circuitry that the one or more verification metrics do not comply with the predetermined condition.

Growing community of inventors

Antibes, France

Natalya Bondarenko