International Business Machines Corporation

Diana Lynn Orf

Michael F Fee

Robert J Sonnelitter, Iii

Christine Comins Jones

Ekaterina M Ambroladze

Deanna Postles Dunn Berger

Deanna P Berger

Arthur J O'Neill

Arthur J O'Neill, Jr

Mark Anthony Check

Gary E Strait

Hong Deng

Christian Jacobi

Patrick James Meaney

Martin Recktenwald

Pak-kin Mak

Jeffrey Haskell Derby

Christopher J Berry

Hans-Werner Tast

Garrett M Drapala

Christian Habermann

Ralf Winkelmann

Lawrence D Curley

Michael J Healy

Hanno Ulrich

Jason A Viehland

Provided are techniques for a hardware accelerator with locally stored macros. A plurality of macros are stored in a lookup memory of a hardware accelerator. In response to receiving an operation code, the operation code is mapped to one or more macros of the plurality of macros, wherein each of the one or more macros includes micro-instructions. Each of the micro-instructions of the one or more macros is routed to a function block of a plurality of function blocks. Each of the micro-instructions is processed with the plurality of function blocks. Data from the processing of each of the micro-instructions is stored in an accelerator memory of the hardware accelerator. The data is moved from the accelerator memory to a host memory.

Hardware accelerator with locally stored macros

Various embodiments mitigate busy time in a hierarchical store-through memory cache structure including a cache directory associated with a memory cache. The cache directory is divided into a plurality of portions each associated with a portion of memory cache. A determination is made that a first subpipe of a shared cache pipeline comprises a non-store request. The shared pipeline is communicatively coupled to the plurality of portions of the cache directory. A store command is prevented from being placed in a second subpipe of the shared cache pipeline based on determining that a first subpipe of the shared cache pipeline comprises a non-store request. Simultaneous cache lookup operations are supported between the plurality of portions of the cache directory and cache write operations. Two or more store commands simultaneously processed in a shared cache pipeline communicatively coupled to the plurality of portions of the cache directory.

Mitigating busy time in a high performance cache

In an approach for backing up designated data located in a cache, data stored within an index of a cache is identified, wherein the data has an associated designation indicating that the data is applicable to be backed up to a higher level memory. It is determined that the data stored to the cache has been updated. A status associated with the data is adjusted, such that the adjusted status indicates that the data stored to the cache has not been changed. A copy of the data is created. The copy of the data is stored to the higher level memory.

Designated cache data backup during system operation

A pipelined processing device includes: a pipeline controller configured to receive at least one instruction associated with an operation from each of a plurality of subcontrollers, and input the at least one instruction into a pipeline; and a pipeline counter configured to receive an active time value from each of the plurality of subcontrollers, the active time value indicating at least a portion of a time taken to process the at least one instruction, the pipeline controller configured to route the active time value to a shared pipeline storage for performance analysis.

Monitoring processing time in a shared pipeline

A method, information processing device, and computer program product mitigate busy time in a hierarchical store-through memory cache structure. A cache directory associated with a memory cache is divided into a plurality of portions each associated with a portion memory cache. Simultaneous cache lookup operations and cache write operations between the plurality of portions of the cache directory are supported. Two or more store commands are simultaneously processed in a shared cache pipeline communicatively coupled to the plurality of portions of the cache directory.

Dynamic allocation of cache buffer slots includes receiving a request to perform an operation that requires a storage buffer slot, the storage buffer slot residing in a level of storage. The dynamic allocation of cache buffer slots also includes determining availability of the storage buffer slot for the cache index as specified by the request. Upon determining the storage buffer slot is not available, the dynamic allocation of cache buffer slots includes evicting data stored in the storage buffer slot, and reserving the storage buffer slot for data associated with the request.

On demand allocation of cache buffer slots

Various embodiments of the present invention manage a hierarchical store-through memory cache structure. A store request queue is associated with a processing core in multiple processing cores. At least one blocking condition is determined to have occurred at the store request queue. Multiple non-store requests and a set of store requests associated with a remaining set of processing cores in the multiple processing cores are dynamically blocked from accessing a memory cache in response to the blocking condition having occurred.

Managing in-line store throughput reduction

An apparatus for controlling operation of a cache includes a first command queue, a second command queue and an input controller configured to receive requests having a first command type and a second command type and to assign a first request having the first command type to the first command queue and a second command having the first command type to the second command queue in the event that the first command queue has not received an indication that a first dedicated buffer is available.

Dynamic cache queue allocation based on destination availability

A pipelined processing device includes: a device controller configured to receive a request to perform an operation; a plurality of subcontrollers configured to receive at least one instruction associated with the operation, each of the plurality of subcontrollers including a counter configured to generate an active time value indicating at least a portion of a time taken to process the at least one instruction; a pipeline processor configured to receive and process the at least one instruction, the pipeline processor configured to receive the active time value; and a shared pipeline storage area configured to store the active time value for each of the plurality of subcontrollers.

A pipelined processing device includes: a processor configured to receive a request to perform an operation; a plurality of processing controllers configured to receive at least one instruction associated with the operation, each of the plurality of processing controllers including a memory to store at least one instruction therein; a pipeline processor configured to receive and process the at least one instruction, the pipeline processor including shared error detection logic configured to detect a parity error in the at least one instruction as the at least one instruction is processed in a pipeline and generate an error signal; and a pipeline bus connected to each of the plurality of processing controllers and configured to communicate the error signal from the error detection logic.

Error detection and recovery in a shared pipeline

A method, information processing device, and computer program product mitigate busy time in a hierarchical store-through memory cache structure. In one embodiment, a cache directory associated with a memory cache is divided into a plurality of portions each associated with a portion memory cache. Simultaneous cache lookup operations and cache write operations between the plurality of portions of the cache directory are supported. Two or more store commands are simultaneously processed in a shared cache pipeline communicatively coupled to the plurality of portions of the cache directory.

Handling corrupted background data in an out of order processing environment. Modified data is stored on a byte of a word having at least one byte of background data. A byte valid vector and a byte store bit are added to the word. Parity checking is done on the word. If the word does not contain corrupted background date, the word is propagated to the next level of cache. If the word contains corrupted background data, a copy of the word is fetched from a next level of cache that is ECC protected, the byte having the modified data is extracted from the word and swapped for the corresponding byte in the word copy. The word copy is then written into the next level of cache that is ECC protected.

Handling corrupted background data in an out of order execution environment

A method that includes providing LRU selection logic which controllably pass requests for access to computer system resources to a shared resource via a first level and a second level, determining whether a request in a request group is active, presenting the request to LRU selection logic at the first level, when it is determined that the request is active, determining whether the request is a LRU request of the request group at the first level, forwarding the request to the second level when it is determined that the request is the LRU request of the request group, comparing the request to an LRU request from each of the request groups at the second level to determine whether the request is a LRU request of the plurality of request groups, and selecting the LRU request of the plurality of request groups to access the shared resource.

Multiple level linked LRU priority

An apparatus for controlling access to a resource includes a shared pipeline configured to communicate with the resource, a plurality of command queues configured to form instructions for the shared pipeline and an arbiter coupled between the shared pipeline and the plurality of command queues configured to grant access to the shared pipeline to a one of the plurality of command queues based on a first priority scheme in a first operating mode. The apparatus also includes interface logic coupled to the arbiter and configured to determine that contention for access to the resource exists among the plurality of command queues and to cause the arbiter to grant access to the shared pipeline based on a second priority scheme in second operating mode.

Growing community of inventors

Somerville, MA, United States of America

Diana Lynn Orf