International Business Machines Corporation

Vijay Sundaresan

Andrew James Craik

Mark Graham Stoodley

Patrick R Doyle

Kevin Alexander Stoodley

Daniel J Heidinga

Derek Bruce Inglis

Sathiskumar Palaniappan

Marcel Mitran

Andres H Voldman

Marius Pirvu

Nikola Grcevski

Timothy P Ellison

Amit S Mane

Ali Ijaz Sheikh

Angela Lin

Graham Alan Chapman

Ryan Andrew Sciampacone

Trent Gray-Donald

Karl Michael Taylor

Younes Manton

Aruna A Kalagananam

Joseph Devin Micheal Papineau

Gita Koblents

Prashanth S Krishna

Pramod Ramarao

Yi Zhang

Michiaki Tatsubori

Reid T Copeland

Michael H Dawson

Motohiro Kawahito

Tamiya Onodera

Allan Henry Kielstra

Michael S Fulton

Akihiko Tozawa

Mike Stephen Fulton

Daryl James Maier

Joran S C Siu

Zhong Liang Wang

Christopher B Larsson

Ashutosh Mehra

Filip Jeremic

Kishor V Patil

Dina Tal

Anyang Yu

James I Gartley

James Ian Andrew Gartley

Dhruv Chopra

James I A Gartley

Alen Badel

Ning Thomas Wong

Alexander Vaseilevskiy

A computer-implemented method for optimizing memory usage. A processor set classifies data stored in virtual memory into hot data and cold data based on frequency of access for the data stored in the virtual memory. The processor set segregates the hot data and the cold data into different memory regions in the virtual memory to generate hot memory regions comprising the hot data and cold memory regions comprising the cold data. The processor set maps the hot memory regions to large physical pages in memory pages of physical memory and the cold memory regions to small physical pages in the memory pages. The processor set disclaims data in the cold memory regions from the memory pages of physical memory to reduce memory space used in the physical memory.

Optimizing memory usage in physical memory

Vectorization and scalarization of methods are provided. A plurality of node webs is constructed based on traversing an intermediate representation of a program. Transitive closure of the plurality of node webs is performed to form a set of final node webs. It is determined that each respective node in the set of final node webs can be converted into one of vector operation code or a sequence of scalar operation codes based on at least one node including a specified vector length and only one vector length value being specified within the set of final node webs. Each respective node in the set of final node webs is converted into one of corresponding vector operation code or a corresponding sequence of scalar operation codes to accelerate execution of supported and unsupported methods of the program.

High-level vectorization and scalarization of Java vector API library methods

Improving execution of application program instructions by receiving code having a security classification, determining that the code is untrusted according to the security classification and inserting instructions for a cache flush associated with executing the code.

Controlling processor instruction execution

An embodiment performs escape analysis of a function as a compiler optimization and stack-allocates an object referenced by the function. At runtime, the embodiment includes detecting a hot code replacement of a portion of the function while the referenced object is stored in stack memory. Responsive to detecting the hot code replacement, the embodiment includes allocating heap memory for the object and moving the object from the stack memory to the allocated heap memory. The embodiment also updates references to the object that were pointing to the object in the stack memory to instead point to the object in the heap memory.

Escape analysis support for method redefinition

An embodiment includes requesting, by a compiler responsive to execution of a first code segment, a first profile dataset associated with the first code segment. The embodiment also includes executing, responsive to receiving an indication that the first profile dataset is not available, a querying process that searches other code segments based on specified criteria relating to an attribute of the first code segment. The embodiment also includes receiving a search result from the querying process, where the search result includes a second code segment. The embodiment also includes generating an extrapolated profile dataset based at least in part on the second code segment. The embodiment stores the extrapolated profile dataset in memory associated with the first code segment, and the compiler performs an optimization process on the first code segment using the extrapolated profile dataset.

Deriving profile data for compiler optimization

A method list is built for a currently executing application within a process virtual machine at a snapshot point, the method list comprising a set of methods capable of being executed by the currently executing application after the snapshot point, the snapshot point comprising an execution state of the currently executing application when a snapshot process is triggered. Profiling data of the currently executing application, collected prior to reaching the snapshot point, is committed, to a designated storage location. Using the profiling data and a just-in-time compiler of the process virtual machine, a method in the method list is compiled. Snapshot data comprising data of the execution state of the currently executing application at the snapshot point, including a result of the compiling, is stored.

Compilation strategy for a sharable application snapshot

Embodiments of the present invention comprise a class sharing orchestrator (CSO) operating as a subsystem of a container orchestrator. The CSO manages sharing of class data among containerized applications to improve startup performance, CPU consumption and memory footprint. The CSO stores application class data in a shared class cache (SCC). The CSO provides a compatible SCC to every deployed application running within the CSO's purview. The CSO collects SCC data from running applications, and processes the data offline to continuously improve the quality of SCCs provided to each deployment, while causing minimum impact to running applications. The CSO combines updates from running applications of a given type to generate an improved quality SCC for use by newly launched applications of the same type. The CSO comprises an SCC update mechanism that optimizes the size, and network traffic, associated with SCC data exchange, and guarantees SCC quality improves over time.

Sharing cached class data in a containerized environment

An approach is provided for optimizing reference counting. Responsive to receiving code representing a program by a just-in-time compiler, one or more processors in computing machinery supporting transactional memory identify regions of the code having respective sets of reference counting operations executed dynamically. Identifying the regions of the code uses an analysis of semantics of the code. The identified regions are enclosed in respective transactions. The code that was to perform atomic operations, including the reference counting operations in the identified regions, is transformed into new code that performs non-atomic operations that are variants of the atomic operations. Fallback code sequences are inserted into the transformed code. In a non-transactional manner and in response to detections of failures in respective transactions, the fallback code sequences execute original code sequences that were in the code prior to the transformation of the code. The original code sequences include respective multiple atomic operations.

Using hardware transactional memory to optimize reference counting

A data-serialization system initially uses a recursive serialization algorithm to serialize a hierarchy of nested data objects by translating those objects into a serial stream of data. The system determines that a stack-overflow error is likely to occur whenever the number of objects serialized by the system exceeds a threshold value, or whenever the stack has reached an unacceptable level of utilization. When the system determines that a stack-overflow error is likely or if the system detects that a stack-overflow error will definitely occur if another object is serialized, the system either transfers control to a nonrecursive algorithm that does not require a stack data structure or reduces stack utilization by transferring contents of the stack to a variable-size queue-like data structure.

Recursive stack-based serialization of hierarchical data

An approach is provided in which an information handling system selects an assumption point in a software program corresponding to a compile-time assumption made by a compiler, and selects an assumption violation point in the software program corresponding to a location at which the compile-time assumption can be violated at runtime. The information handling system propagates backwards in the software program from the assumption point and reaches the assumption violation point. The information handling system determines that the assumption point corresponds to a first method and the assumption violation point corresponds to a second method that is different from the first method, and inserts a conditional transition in the software program at the assumption violation point. The information handling system executes a compiled version of the software program that includes the conditional transition.

Dataflow analysis to reduce the overhead of on stack replacement

An enhanced object allocation optimization selectively traverses an intermediate representation detecting on-stack replacement transitions, which when found are analyzed to determine whether a control flow-edge from a first block to a second block that is marked as an OSR resumption block exists. Responding to when the second block is marked, a pseudo call including arguments of all live local variables holding pointers to objects is inserted into the intermediate representation while optimization opportunities exist and executing a modified escape analysis on a modified intermediate representation examining each pseudo call as an escape point for all object references received by the pseudo call as arguments; ignoring uses of local variables dominated by these pseudo calls; and stack allocating objects to handle the non-local control flow due to on-stack replacement control flow using the pseudo call. Responsive to when no further optimization opportunities exist, removing respective inserted pseudo calls needing removal.

Escape analysis supporting on-stack replacement

Determining whether to perform atomic or conventional reference counting is provided. A single bit of a reference count corresponding to a reference-counted resource is read to determine whether to atomic reference counting is to be performed. It is determined whether the single bit of the reference count is set. In response to determining that the single bit is set, an atomic operation for atomic maintenance of the reference count corresponding to the reference-counted resource is performed by adjusting the reference count by a value of two. In response to determining that the single bit is not set, a conventional adjustment operation is performed for maintenance of the reference count corresponding to the reference-counted resource.

Dynamic resource visibility tracking to avoid atomic reference counting

An approach is provided in which an information handling system selects a first point in a software program corresponding to a compile-time assumption made by a compiler. The information handling system then selects a set of second points in the software program corresponding to a set of locations at which the compile-time assumption can be violated at runtime. Next, the information handling system starts at the first point and propagates backwards in the software program to identify one or more of the second points that are reached from the backwards propagation. The information handling system then inserts conditional transitions in the software program at each of the identified assumption violation points and executes a compiled version of the software program, resulting in an evaluation of the compile-time assumption at the conditional transitions.

A method and system for partial connection of iterations during loop unrolling during compilation of a program by a compiler. Unrolled loop iterations of a loop in the program are selectively connected during loop unrolling during the compilation, including redirecting, to the head of the loop, undesirable edges of a control flow from one iteration to a next iteration of the loop. Merges on a path of hot code are removed to increase a scope for optimization of the program. The head of the loop and a start of a replicated loop body of the loop are equivalent points of the control flow.

Partial connection of iterations during loop unrolling

Recursive serialization of a hierarchy of data objects

Growing community of inventors

North York, Canada

Vijay Sundaresan