International Business Machines Corporation

Wheelock, Inc.

Craig Raymond Walters

Pak-kin Mak

Garrett M Drapala

Michael Andrew Blake

Vesselina K Papazova

Timothy Carl Bronson

Jane Helen Bartik

Ram Sai Manoj Bamdhamravuri

Robert J Sonnelitter, Iii

Deanna Postles Dunn Berger

Arthur J O'Neill

Peter Grimm Sutton

Matthias Klein

Charles Franklin Webb

Reinhard Theodor Buendgen

William J Rooney

James A O'Connor

Khary J Alexander

K Paul Muller

Tiberiu (Jim) Suto

Kenneth Douglas Klapproth

Chung-Lung K Shum

Alper Buyuktosunoglu

Patrick James Meaney

Michael F Fee

Gary E Strait

Hieu Trong Huynh

Kathryn Marie Jackson

Nicholas C Matsakis

Hanno Ulrich

Dominic Ditomaso

Sean M Swehla

Timothy John Slegel

Christian Jacobi

James J Bonanno

Vijayalakshmi Srinivasan

Gregory William Alexander

Utz Bacher

Adam B Collura

John C Dayka

Michael J Cadigan, Jr

Patrick Joseph Callaghan

Daniel Lipetz

Christine Comins Jones

Thomas B Mathias

Peter G Spera

Rebecca Marie Gott

Guy G Tracy

Kurt Norman Schroeder

Deanna P D Berger

Mark R Hodges

Stefan Usenbinz

Deanna Postles Dunn

Joshua D Massover

David Trilla Rodriguez

Mark S Farrell

Steven R Carlough

Elpida Tzortzatos

Gary A Van Huben

Ekaterina M Ambroladze

Glenn David Gilda

David S Hutton

Arthur J O'Neill, Jr

David D Cadigan

William J Lewis

Lawrence D Curley

Deanna P Berger

Thomas J Dewkett

Nathan J Lee

Jason D Kohl

Keith N Langston

Harmony Lynn Helterhoff

Richard Joseph Branciforte

Tim Bronson

Sebastian Burckhardt

Kevin Lopes

Edward J Kaminski

Amy S Purdy

Adrian O Robinson

Chunggeon Rhee

Peter B Tarlton

Robert J Sonnelltier, Iii

Scott M Blackledge

Daniel P Millerick

Robert Sonnelitter

Processing within a computing environment is facilitated by using cache hit-miss predictions. A cache hit-miss prediction is determined for a memory access instruction using a predictor. One or more speculative actions are controlled, based on determining the cache hit-miss prediction is a miss. The controlling is further based on a type of cache design.

Controlling speculative actions based on a hit/miss predictor

Examples described herein provide a computer-implemented method that includes generating an extended hot line table that tracks cross-core contended cache lines for multiple processors of a processing system based on cache requests, the extended hot line table storing at least metadata for a cross-core contended cache line. The method further includes polling, using firmware, the extended hot line table in each of the multiple processors of the processing system to identify contention information. The method further includes aggregating the contention information from each of the multiple processors to generate aggregated contention information. The method further includes processing subsequent cache requests using the aggregated contention information.

Detecting and mitigating false structure sharing within a cache line

A cache hit-miss prediction is determined for a memory access instruction using a predictor. The predictor includes a tracker for the memory access instruction. The tracker is used to provide a prediction confidence level of the cache hit-miss prediction for the memory access instruction. Using the tracker, the prediction confidence level of the cache hit-miss prediction is ascertained. Based on the prediction confidence level indicating the cache hit-miss prediction is to be used, the cache hit-miss prediction is provided to be used in instruction execution scheduling.

Instruction execution scheduling using a hit/miss predictor

Aspects of the invention include defining one or more processor units having a plurality of caches, each processor unit comprising a processor having at least one cache, and wherein each of the one or more processor units are coupled together by an interconnect fabric, for each of the plurality of caches, arranging a plurality of cache lines into one or more congruence classes, each congruence class comprises a chronology vector, arranging each cache in the plurality of caches into a cluster of caches based on a plurality of scope domains, determining a first cache line to evict based on the chronology vector, and determining a target cache for installing the first cache line based on a scope of the first cache line and a saturation metric associated with the target cache, wherein the scope of the first cache line is determined based on lateral persistence tag bits.

Lateral persistence directory states

Aspects of the invention include computer-implemented methods, systems, and computer program products that access a multi-copy scope directory state of a cache memory that indicates a scope of sharing of a cache line in a cache memory system and determine a scope of sharing of the cache line in the cache memory system based on the multi-copy scope directory state, where the multi-copy scope directory state enumerates a plurality of scopes within the cache memory system. The scope of sharing is used to reduce a number of queries to one or more cache memories having a larger scope than a shared scope identified in the scope of sharing. The multi-copy scope directory state of the cache memory is updated based on detecting a change in shared scope of the cache line within the cache memory system.

Multiple copy scoping bits for cache memory

Embodiments of the present invention are directed to a computer-implemented method for controller address contention assumption. A non-limiting example computer-implemented method includes a shared controller receiving a fetch request for data from a first requesting agent, the receiving via at least one intermediary controller. The shared controller performs an address compare using a memory address of the data. In response to the memory address matching a memory address stored in the shared controller, the shared controller acknowledges the at least one intermediary controller's fetch request, wherein upon acknowledgement, the at least one intermediary controller resets, wherein the acknowledging comprises exchanging tokens by the shared controller and the at least one intermediary controller, wherein the at least one intermediary controller transmits an identity of the first requesting agent and a type of operation associated with the requested data, and wherein the shared controller transmits an acceptance.

Controller address contention assumption

Aspects of the invention include efficient generation of instrumentation data for direct memory access operations. A non-limiting example apparatus includes an instrumentation component, residing in a cache in communication with a plurality of processing units, an accelerator, and a plurality of input output interfaces. The cache includes a direct memory access monitor that receives events from the accelerator its respective I/O interface and stores DMA state and latency for each event. The cache also includes a bucket including a DMA counter and a latency counter in communication with the DMA monitor, wherein the bucket stores in the DMA counter a count of DMAs coming from a source and stores in the latency counter the latency measured for each DMA coming from the source.

Efficient generation of instrumentation data for direct memory access operations

Embodiments of the present invention are directed to a computer-implemented method for controller address contention assumption. A non-limiting example computer-implemented method includes a shared controller receiving a fetch request for data from a first requesting agent, the receiving via at least one intermediary controller. The shared controller performs an address compare using a memory address of the data. In response to the memory address matching a memory address stored in the shared controller, the shared controller acknowledges the at least one intermediary controller's fetch request, wherein upon acknowledgement, the at least one intermediary controller resets. In response to release of the data by a second requesting agent, the shared controller transmits the data to the first requesting agent.

Embodiments of the present invention are directed to a computer-implemented method for cache eviction. The method includes detecting a first data in a shared cache and a first cache in response to a request by a first processor. The first data is determined to have a mid-level cache eviction priority. A request is detected from a second processor for a same first data as requested by the first processor. However, in this instance, the second processor has indicated that the same first data has a low-level cache eviction priority. The first data is duplicated and loaded to a second cache, however, the data has a low-level cache eviction priority at the second cache.

Multilevel cache eviction management

Methods, systems and computer program products for measuring hardware performance are provided. Aspects include receiving an indication of a start to a hardware operation. A number of clock cycles are counted from the start of a hardware operation to the completion of the hardware operation. A first region comprising a first set of bit location is defined. A second and third region is defined each including a set of bit locations. Based on the first set of bit locations being equal to zero, a granularity flag is set to zero in the sample buffer and the second and third set of bit locations are written to the sample buffer. And based on the first set of bit locations being greater than zero, the granularity flag in the sample buffer is set to one and the first and second set of bit locations are written to the sample buffer.

Accuracy sensitive performance counters

Aspects include a computer-implemented method includes receiving an instruction at a processor to perform an operation on a memory block having an address and accessing a state indicator by the processor without altering a value of the state indicator. The state indicator is stored in a memory location independent of the memory block, and accessing includes sending a request to an operator to return the value of the state indicator to the processor. The method also includes determining based on the value of the state indicator whether the memory block is in a pre-defined state.

Memory state indicator check operations

Aspects include a computer-implemented method that includes receiving an instruction at a processor to perform an operation on a memory block having an address and accessing a state indicator by the processor without altering a value of the state indicator. The state indicator is stored in a memory location independent of the memory block, and accessing includes sending a request to an operator to return the value of the state indicator to the processor. The method also includes determining based on the value of the state indicator whether the memory block is in a pre-defined state.

Speculative data return in parallel with an exclusive invalidate request. A requesting processor requests data from a shared cache. The data is owned by another processor. Based on the request, an invalidate request is sent to the other processor requesting the other processor to release ownership of the data. Concurrent to the invalidate request being sent to the other processor, the data is speculatively provided to the requesting processor.

Speculative data return concurrent to an exclusive invalidate request

A shared cache line is concurrently modified by multiple processors of a computing environment. The concurrent modification is performed based, at least, on receiving one or more architected instructions (Fetch due to Non-Coherent Store instructions) that permit multiple processors to concurrently update the shared cache line absent obtaining a lock or having exclusive ownership of the data.

Concurrent modification of shared cache line by multiple processors

A processor implemented method for spreading data traffic across memory controllers with respect to conditions is provided. The processor implemented method includes determining whether the memory controllers are balanced. The processor implemented method includes executing a conditional spreading with respect to the conditions when the memory controllers are determined as unbalanced. The processor implemented method includes executing an equal spreading when the memory controllers are determined as balanced.

Conditional memory spreading for heterogeneous memory sizes

Growing community of inventors

Highland, NY, United States of America

Craig Raymond Walters