International Business Machines Corporation

Xiaowei Shen

Balaram Sinharoy

Robert William Wisniewski

Robert Brett Tremaine

Orran Yaakov Krieger

Karin Strauss

David Francis Bacon

Hazim Shafi

Peter Francis Sweeney

Jaehyuk Huh

Elmootazbellah Nabil Elnozahy

John Timothy Robinson

Aaron Christoph Sawdey

Yefim Shuf

Dilma Menezes Da Silva

Zhigang Hu

Man Cheuk Ng

Hua Wang

Alper Buyuktosunoglu

Kun Wang

Ramakrishnan Rajamony

Harold Wade Cain, Iii

Vijayalakshmi Srinivasan

Peter H Hochschild

Jude A Rivers

Rui Hou

Huayong Wang

Jessica Hui-Chun Tseng

Brinda Ganesh

An atomic transaction includes one or more memory access operations that are completed atomically. A Best-Effort Transaction (BET) system makes its best effort to complete each atomic transaction without guaranteeing completion of all atomic transactions. When an atomic transaction is aborted, BET may provide software with appropriate runtime information such as cause of the abortion. With proper coherence layer enhancements, BET can be implemented efficiently for multiprocessor systems, using caches as buffers for data accessed by atomic transactions. Furthermore, with appropriate fairness support, forward progress can be guaranteed for atomic transactions that incur no buffer overflow.

Architecture support of best-effort atomic transactions for multiprocessor systems

A system includes a deterministic system, and a controller electrically coupled to the deterministic system via a link, wherein the controller comprises a transaction scheduling mechanism that allows data responses from the deterministic system, corresponding to requests issued from the controller, to be returned out of order.

Reordering data responses using ordered indicia in a linked list

The present invention broadly contemplates braids and fibers, high-level programming constructs which facilitate the creation of programs that are partially ordered, to address the continuing trend of ever-increasing processor speeds and attendant increases in memory latencies. These partial orders can be used to respond adaptively to memory latencies. It is shown how these constructs can be effectively supported with simple and inexpensive instruction set and micro-architectural extensions.

Systems and methods for indirect register access using status-checking and status-setting instructions

Color-based caching allows each cache line to be distinguished by a specific color, and enables the manipulation of cache behavior based upon the colors of the cache lines. When multiple threads are able to share a cache, effective cache management is critical to overall performance. Color-based caching provides an effective method to better utilize caches and avoid unnecessary cache thrashing and pollution. Hardware maintains color-based counters relative to the cache lines to monitor and obtain feedback on cache line events. These counters are utilized for cache coherence transactions in multiple processor systems.

Cache coherence monitoring and feedback

A computer-implemented method for memory access monitoring, implemented by a managed runtime environment computer system including a controller that monitors application behavior and determines actions to be taken to change a behavior of an application, and a runtime, dynamic compiler that analyzes the application and generates code sequences to access a memory access monitoring (MAM) mechanism, includes determining monitor information of a plurality of fields of a memory block to drive an optimization of the application.

Memory access monitoring

Memory Access Coloring provides architecture support that allows software to classify memory accesses into different congruence classes by specifying a color for each memory access operation. The color information is received and recorded by the underlying system with appropriate granularity. This allows hardware to monitor color-based cache monitoring information and provide such feedback to the software to enable various runtime optimizations. It also enables enforcement of different memory consistency models for memory regions with different colors at the same time.

Architecture support of memory access coloring

Systems and methods for cache replacement monitoring (CRM) are provided. The system includes a monitored cache comprising a monitored cache line set, the monitored cache line set comprising at least one cache line capable of holding data of a monitored address; and a CRM mechanism operatively associated with the monitored cache. The CRM mechanism collects CRM information for the monitored address. The method includes the steps of collecting CRM information for a monitored address in a monitored cache; and recording the CRM information for the monitored address, when at least one of (1) the monitored address is cached in the monitored cache, (2) the monitored address is replaced in the monitored cache, (3) any cache line in a cache line set corresponding to the monitored address is cached in the monitored cache, and (4) any cache line in a cache line set corresponding to the monitored address is replaced in the monitored cache.

Cache line replacement monitoring and profiling

A method for performing a transaction including a transaction head and a transaction tail, includes executing the transaction head, including executing at least one memory reserve instruction to reserve a transactional memory location that are accessed in the transaction and executing the transaction tail, wherein the transaction cannot be aborted due to a data race on that transactional memory location while executing the transaction tail, wherein data of memory write operations to the transactional memory location is committed without being buffered.

Reservation required transactions

There is disclosed a method and apparatus for handling transaction buffer overflow in a multi-processor system as well as a transaction memory system in a multi-processor system. The method comprises the steps of: when overflow occurs in a transaction buffer of one processor, disabling peer processors from entering transactions, and waiting for any processor having a current transaction to complete its current transaction; re-executing the transaction resulting in the transaction buffer overflow without using the transaction buffer; and when the transaction execution is completed, enabling the peer processors for entering transactions.

Handling transaction buffer overflow in multiprocessor by re-executing after waiting for peer processors to complete pending transactions and bypassing the buffer

A method for reconfiguring a cache memory is provided. The method in one aspect may include analyzing one or more characteristics of an execution entity accessing a cache memory and reconfiguring the cache based on the one or more characteristics analyzed. Examples of analyzed characteristic may include but are not limited to data structure used by the execution entity, expected reference pattern of the execution entity, type of an execution entity, heat and power consumption of an execution entity, etc. Examples of cache attributes that may be reconfigured may include but are not limited to associativity of the cache memory, amount of the cache memory available to store data, coherence granularity of the cache memory, line size of the cache memory, etc.

System for reconfiguring cache memory having an access bit associated with a sector of a lower-level cache memory and a granularity bit associated with a sector of a higher-level cache memory

In a computer system with a memory hierarchy, when a high-level cache supplies a data copy to a low-level cache, the shared copy can be either volatile or non-volatile. When the data copy is later replaced from the low-level cache, if the data copy is non-volatile, it needs to be written back to the high-level cache; otherwise it can be simply flushed from the low-level cache. The high-level cache can employ a volatile-prediction mechanism that adaptively determines whether a volatile copy or a non-volatile copy should be supplied when the high-level cache needs to send data to the low-level cache. An exemplary volatile-prediction mechanism suggests use of a non-volatile copy if the cache line has been accessed consecutively by the low-level cache. Further, the low-level cache can employ a volatile-promotion mechanism that adaptively changes a data copy from volatile to non-volatile according to some promotion policy, or changes a data copy from non-volatile to volatile according to some demotion policy.

Adaptive mechanisms and methods for supplying volatile data copies in multiprocessor systems

A system, method, and computer readable article of manufacture for sharing buffer management. The system includes: a predictor module to predict at runtime a transaction data size of a transaction according to history information of the transaction; and a resource management module to allocate sharing buffer resources for the transaction according to the predicted transaction data size in response to beginning of the transaction, to record an actual sharing buffer size occupied by the transaction in response to the successful commitment of the transaction, and to update the history information of the transaction.

Method and system for a sharing buffer

A computer readable medium is provided embodying instructions executable by a processor to perform a method for performing a transaction including a transaction head and a transaction tail, the method includes executing the transaction head, including executing at least one memory reserve instruction to reserve a transactional memory location that are accessed in the transaction and executing the transaction tail, wherein the transaction cannot be aborted due to a data race on that transactional memory location while executing the transaction tail, wherein data of memory write operations to the transactional memory location is committed without being buffered.

Methods, systems, and media for reducing memory latency seen by processors by providing a measure of control over on-chip memory (OCM) management to software applications, implicitly and/or explicitly, via an operating system are contemplated. Many embodiments allow part of the OCM to be managed by software applications via an application program interface (API), and part managed by hardware. Thus, the software applications can provide guidance regarding address ranges to maintain close to the processor to reduce unnecessary latencies typically encountered when dependent upon cache controller policies. Several embodiments utilize a memory internal to the processor or on a processor node so the memory block used for this technique is referred to as OCM.

Growing community of inventors

Hopewell Junction, NY, United States of America

Xiaowei Shen