International Business Machines Corporation

Gordon Bernard Bell

Anil Krishna

Srinivasan Ramani

Ganesh Balakrishnan

Ken Van Vu

Gordon Taylor Davis

Jeffrey Haskell Derby

Brian Michael Rogers

Steve Woolet

Timothy Hume Heil

Mvv Anil Krishna

Reclaiming checkpoints in a system in an order that differs from the order when the checkpoints are created. Reclaiming the checkpoints includes: creating one or more checkpoints, each of which having an initial state using system resources and holding the checkpoints state; identifying the completion of all the instructions associated with the checkpoint; reassigning all the instructions associated with the identified checkpoint to an immediately preceding checkpoint; and freeing the resources associated with the identified checkpoint. The checkpoint is created when the instruction that is checked is a conditional branch having a direction that cannot be predicted with a predetermined confidence level.

Out-of-order checkpoint reclamation in a checkpoint processing and recovery core microarchitecture

A system and method to optimize runahead operation for a processor without use of a separate explicit runahead cache structure. Rather than simply dropping store instructions in a processor runahead mode, store instructions write their results in an existing processor store queue, although store instructions are not allowed to update processor caches and system memory. Use of the store queue during runahead mode to hold store instruction results allows more recent runahead load instructions to search retired store queue entries in the store queue for matching addresses to utilize data from the retired, but still searchable, store instructions. Retired store instructions could be either runahead store instructions retired, or retired store instructions that executed before entering runahead mode.

Store-to-load forwarding mechanism for processor runahead mode operation

Embodiments of the invention are directed to optimizing the performance of a split disk cache. In one embodiment, a disk cache includes a primary region having a read portion and write portion and one or more smaller, sample regions also including a read portion and a write portion. The primary region and one or more sample region each have an independently adjustable ratio of a read portion to a write portion. Cached reads are distributed among the read portions of the primary and sample region, while cached writes are distributed among the write portions of the primary and sample region. The performance of the primary region and the performance of the sample region are tracked, such as by obtaining a hit rate for each region during a predefined interval. The read/write ratio of the primary region is then selectively adjusted according to the performance of the one or more sample regions.

Dynamic adjustment of read/write ratio of a disk cache

A processing system includes a memory and a first core configured to process applications. The first core includes a first cache. The processing system includes a mechanism configured to capture a sequence of addresses of the application that miss the first cache in the first core and to place the sequence of addresses in a storage array; and a second core configured to process at least one software algorithm. The at least one software algorithm utilizes the sequence of addresses from the storage array to generate a sequence of prefetch addresses. The second core issues prefetch requests for the sequence of the prefetch addresses to the memory to obtain prefetched data and the prefetched data is provided to the first core if requested.

Prefetching with multiple processors and threads via a coherency bus

Effective prefetching with multiple processors and threads

The illustrative embodiments provide a method, apparatus, and computer program product for managing a number of cache lines in a cache. In one illustrative embodiment, it is determined whether activity on a memory bus in communication with the cache exceeds a threshold activity level. A least important cache line is located in the cache responsive to a determination that the threshold activity level is exceeded, wherein the least important cache line is located using a cache replacement scheme. It is determined whether the least important cache line is clean responsive to the determination that the threshold activity level is exceeded. The least important cache line is selected for replacement in the cache responsive to a determination that the least important cache line is clean. A clean cache line is located within a subset of the number of cache lines and selecting the clean cache line for replacement responsive to an absence of a determination that the least important cache line is not clean, wherein the each cache line in the subset is examined in ascending order of importance according to the cache replacement scheme.

Cache management through delayed writeback

Embodiments that dynamically reorganize data of cache lines in non-uniform cache access (NUCA) caches are contemplated. Various embodiments comprise a computing device, having one or more processors coupled with one or more NUCA cache elements. The NUCA cache elements may comprise one or more banks of cache memory, wherein ways of the cache are horizontally distributed across multiple banks. To improve access latency of the data by the processors, the computing devices may dynamically propagate cache lines into banks closer to the processors using the cache lines. To accomplish such dynamic reorganization, embodiments may maintain 'direction' bits for cache lines. The direction bits may indicate to which processor the data should be moved. Further, embodiments may use the direction bits to make cache line movement decisions.

Growing community of inventors

Madison, WI, United States of America

Gordon Bernard Bell