Sun Microsystems, Inc.

Wisconsin Alumni Research Foundation

Advanced Micro Devices Corporation

Mark Donald Hill

Erik E Hagersten

David Allen Wood

Michael Mansfield Swift

Gabriel H Loh

Robert Cypher

Yasuko Eckert

James Michael O'Connor

Bradford Michael Beckmann

Vilas K Sridharan

Srilatha Manne

Christopher J Jackson

Arkaprava Basu

Lena E Olson

Milo M K Martin

Steven Karl Reinhardt

Thomas M Wicki

Michael Roger Marty

John R Feehrer

Thomas P Webber

Benedict R Gaster

Daniel R Cassiday

Shuai Che

Robert Zak

Monica C Wong-Chan

Marc S Orr

Derek R Hower

Blake A Hechtman

Ayse Yilmazer

Robert W Pfile

Pazhani Pillai

Boris Ostrovsky

Jason Lowe-Power

Daniel J Sorin

Jayneel Gandhi

Babak Falsafi

Shubhendu S Mukherjee

Steven K Reinhardt

Luke Yen

Kevin E Moore

Jayaram Bobba

Haris Volos

A high-bandwidth adaptive cache reduces unnecessary cache accesses by providing a laundry counter indicating whether a given adaptive cache region has any dirty frames to allow write-back without a preparatory adaptive cache read. An optional laundry list allows the preparatory adaptive cache read to also be avoided if the tags of the data being written back match all tags of the data in the adaptive cache that is dirty.

Adaptive computer cache architecture

Described herein is an apparatus and method for remote scoped synchronization, which is a new semantic that allows a work-item to order memory accesses with a scope instance outside of its scope hierarchy. More precisely, remote synchronization expands visibility at a particular scope to all scope-instances encompassed by that scope. Remote scoped synchronization operation allows smaller scopes to be used more frequently and defers added cost to only when larger scoped synchronization is required. This enables programmers to optimize the scope that memory operations are performed at for important communication patterns like work stealing. Executing memory operations at the optimum scope reduces both execution time and energy. In particular, remote synchronization allows a work-item to communicate with a scope that it otherwise would not be able to access. Specifically, work-items can pull valid data from and push updates to scopes that do not (hierarchically) contain them.

Remote scoped synchronization for work stealing and sharing

A system and method for efficient cache data access in a large row-based memory of a computing system. A computing system includes a processing unit and an integrated three-dimensional (3D) dynamic random access memory (DRAM). The processing unit uses the 3D DRAM as a cache. Each row of the multiple rows in the memory array banks of the 3D DRAM stores at least multiple cache tags and multiple corresponding cache lines indicated by the multiple cache tags. In response to receiving a memory request from the processing unit, the 3D DRAM performs a memory access according to the received memory request on a given cache line indicated by a cache tag within the received memory request. Rather than utilizing multiple DRAM transactions, a single, complex DRAM transaction may be used to reduce latency and power consumption.

DRAM cache with tags and data jointly stored in physical rows

A method of way prediction for a data cache having a plurality of ways is provided. Responsive to an instruction to access a stack data block, the method accesses identifying information associated with a plurality of most recently accessed ways of a data cache to determine whether the stack data block resides in one of the plurality of most recently accessed ways of the data cache, wherein the identifying information is accessed from a subset of an array of identifying information corresponding to the plurality of most recently accessed ways; and when the stack data block resides in one of the plurality of most recently accessed ways of the data cache, the method accesses the stack data block from the data cache.

Methods and apparatus for data cache way prediction based on classification as stack data

The translation of virtual guest addresses to host physical addresses in a virtualized computer system provides a compound page table that may simultaneously support nested-paging and shadow-paging for different memory regions. Memory regions with stable address mapping, for example, holding program code, may be treated using shadow-paging while memory regions with dynamic address mapping, for example, variable storage, may be treated using nested-paging thereby obtaining the benefits of both techniques.

Efficient memory management system for computers supporting virtual machines

A memory management unit for I/O devices uses page table entries to translate virtual addresses to physical addresses. The page table entries include removal rules allowing the I/O memory management unit to delete page table entries without CPU involvement significantly reducing the CPU overhead involved in virtualized I/O data transactions.

I/O memory management unit providing self invalidated mapping

A method, computer program product, and system is described that determines the correctness of using memory operations in a computing device with heterogeneous computer components. Embodiments include an optimizer based on the characteristics of a Sequential Consistency for Heterogeneous-Race-Free (SC for HRF) model that analyzes a program and determines the correctness of the ordering of events in the program. HRF models include combinations of the properties: scope order, scope inclusion, and scope transitivity. The optimizer can determine when a program is heterogeneous-race-free in accordance with an SC for HRF memory consistency model. For example, the optimizer can analyze a portion of program code, respect the properties of the SC for HRF model, and determine whether a value produced by a store memory event will be a candidate for a value observed by a load memory event. In addition, the optimizer can determine whether reordering of events is possible.

Method for memory consistency among heterogeneous computer components

A method of partitioning a data cache comprising a plurality of sets, the plurality of sets comprising a plurality of ways, is provided. Responsive to a stack data request, the method stores a cache line associated with the stack data in one of a plurality of designated ways of the data cache, wherein the plurality of designated ways is configured to store all requested stack data.

Methods and apparatus for soft-partitioning of a data cache for stack data

A computer system using virtual memory provides hybrid memory access either through a conventional translation between virtual memory and physical memory using a page table possibly with a translation lookaside buffer, or a high-speed translation using a fixed offset value between virtual memory and physical memory. Selection between these modes of access may be encoded into the address space of virtual memory eliminating the need for a separate tagging operation of specific memory addresses.

Virtual memory management system with reduced latency

The present disclosure relates to a method and system for mapping cache lines to a row-based cache. In particular, a method includes, in response to a plurality of memory access requests each including an address associated with a cache line of a main memory, mapping sequentially addressed cache lines of the main memory to a row of the row-based cache. A disclosed system includes row index computation logic operative to map sequentially addressed cache lines of a main memory to a row of a row-based cache in response to a plurality of memory access requests each including an address associated with a cache line of the main memory.

System and method for cache organization in row-based memories

A system and method for efficiently determining whether a requested memory location is in a large row-based memory of a computing system. A computing system includes a processing unit that generates memory requests on a first chip and a cache (LLC) on a second chip connected to the first chip. The processing unit includes an access filter that determines whether to access the cache. The cache is fabricated on top of the processing unit. The processing unit determines whether to access the access filter for a given memory request. The processing unit accesses the access filter to determine whether given data associated with a given memory request is stored within the cache. In response to determining the access filter indicates the given data is not stored within the cache, the processing unit generates a memory request to send to off-package memory.

Hardware filter for tracking block presence in large caches

A coherence controller in hardware of an apparatus in an example detects conflicts on coherence requests through direct, non-broadcast employment of signatures that: summarize read-sets and write-sets of memory transactions; and provide false positives but no false negatives for the conflicts on the coherence requests. The signatures comprise fixed-size representations of a substantially arbitrary set of addresses for the read-sets and the write-sets of the memory transactions.

Non-broadcast signature-based transactional memory

A computer system may include a system memory, an active device configured to access data stored in the system memory, where the active device includes a cache configured to store data accessed by the active device, an address network for conveying address packets between the active device and the system memory, and a data network for conveying data packets between the active device and the system memory. An access right corresponding to a given block allocated in the cache transitions in response to a corresponding data packet being received by the cache. An ownership responsibility for the given block transitions in response to a corresponding address packet being received by the cache. The access right transitions at a different time than the ownership responsibility transitions. The cache is configured to inhibit receipt of the corresponding data packet based on a value of a timestamp associated with the corresponding data packet.

Computer system implementing synchronized broadcast using timestamps

An apparatus and method for resending a request in a computer system using a delay value is provided. In response to receiving a request, a target device in a computer system may detect that it is temporarily unable to process the request. The target device can send a response to the sending device to indicate that it is temporarily unavailable. The response can include a delay value that can provide a hint to the sending device as to when to resend the request. The target device may generate the delay value according to the type of condition that is causing it to be temporarily unavailable. The delay value may be generated according to a static heuristic or a dynamic algorithm based on previous temporarily unavailable conditions. The delay value may also be used by an error recovery mechanism where a sending device exceeds a retry limit for a particular request.

System and method for enhancing communication between devices in a computer system

A mechanism and method for maintaining cache consistency in computer systems that implements synchronized broadcasts using skew control and queuing. An access right corresponding to a given block allocated in a first active device may be configured to transition in response to a corresponding data packet being received through a data network. Additionally, transitions in ownership of the given block may occur at a different time than the time at which the access right to the given block is changed. To implement synchronized broadcasts, the address and data networks are configured such that a maximum amount of time from when a given broadcast packet conveyed on the address network arrives at a first active device to a time when the given broadcast packet arrives at a second active device is less than or equal to a minimum amount of time from when a data packet sent on the data network from the first active device arrives at the second active device. Each of the active devices may further comprise a queue control circuit coupled to an address-in queue and a data-in queue. The queue control circuit may be configured to prevent processing of a particular data packet that arrived in the data-in queue until all address packets that arrived earlier in the address-in queue are processed.

Computer system implementing synchronized broadcast using skew control and queuing

Growing community of inventors

Madison, WI, United States of America

Mark Donald Hill