International Business Machines Corporation

Weiming Gu

Robert Francis Berry

William Henry Hartner

Michael Wayne Brown

Frank Eliot Levine

Robert John Urquhart

Kelvin Roderick Lawrence

Riaz Yousuf Hussain

Robert Tod Dimpsey

William Preston Alexander, Iii

William Robert Reynolds

Chendong Zou

Raymond Morris Bryant

John Day Howard

Edwin V Sapugay

Pranta Das

Paul Jerome Kilpatrick

Wai Yee Peter Wong

William Presten Alexander, Iii

Michael Thomas Collins

Maurice T Franklin

Maurice P Franklin

Raghuram Thiagarajan

A business integration (BI) system receives business objects from a source application at an event partitioner. Event partitioner partitions received business objects into independent sets based on interdependent events, and allocates each independent group of business events to a separate event queue based a partitioning modulo reduction transformation of the independent sets to determine the correct event queue for storing a particular received business object. This transformation reduces the number of independent sets to a number of event groups equal to the number of allocated event queues. Events in different groups are delivered to an interchange server through the separate event queues so that they may be processed independently and in parallel, thus improving the overall performance of the interchange server, while guaranteeing that dependent events are processed in the correct serial order.

Efficient and scalable event partitioning in business integration applications using multiple delivery queues

A process and system for dynamically compiling a partially interpreted method is provided. A set of bytecodes for a method is interpreted within a virtual machine. During the interpretation of the method, it is determined, according to the satisfaction of predetermined criteria, that the method contains an execution hot spot and should be just-in-time compiled (JITed) in order to increase the processing speed of the method. The interpretation of the method is halted with a halted execution state and at a halted execution location. Another method is constructed using information from the partially interpreted method and its execution state. The newly constructed method is just-in-time compiled and invoked in such a manner that the newly constructed method recreates the execution state of the partially interpreted method. Once the newly constructed method recreates the execution state of the partially interpreted method, the execution flow follows the bytecode sequence of the partially interpreted method.

Process and system for dynamically compiling a partially interpreted method

The present invention is directed to a system and method for modifying a class file for the purpose of instrumenting without requiring separate files to correlate the instrumentation. A class file is instrumented with hooks. Each hook is injected in a method at a critical point in the code for tracking path flow, such as where the method will be entered or exited. Each hook includes an identifier to identify the method in which it is injected. Rather than using the method's name, hooks use unique major and minor codes to identify the method. Static initializers are declared for the class to output other hooks identifying the methods being instrumented. When a class is loaded, the static initializers are executed and hooks identifying the method name and the major and minor codes for each instrumented method are output to, for instance, a trace record. Then, when a method is entered or exited, the hooks identifying the entry or exit are also outputted to a trace record. When postprocessing the trace records, class and instrumentation correlation information is available for merging from trace records in the trace stream.

System and method for instrumenting application class files with correlation information to the instrumentation

A process and system for optimizing an invocation of a method is provided. A determination is made to compile a calling method, and a call to a callee method is detected within the first method. The callee method may be a non-final, virtual method, and a determination may be made that the callee method has not been previously overridden. The callee method is then inlined within the first method. In addition, no conditional statements are inserted into the calling method along with the inlined method. The determination to compile and optimize these methods may be made by a just-in-time compiler, and if the methods are Java methods, then a Java just-in-time compiler performs the optimization process. If a determination is made to load a class that contains a method that overrides the callee method, then the calling method is recompiled or patched.

Process and system for Java virtual method invocation

A method and apparatus for instrumenting garbage collection in a data processing system is provided. During garbage collection, a pass is made through the object heap and a plurality of heap data is retrieved and written to a trace file. The data may include the type of object, the size of the object on the heap, the class of the object, if appropriate, and a plurality of pointers, including live set pointers, and one or more tags representing attributes that may be associated with the particular object. The instrumentation may controlled to occur at a preselected phase of the garbage collection process. Additionally, a user may dynamically initiate a garbage collection instrumentation if the user determines that an event has occurred during the operation of the data processing system about which the user wishes further data.

Method of instrumenting garbage collection generating a trace file making a single pass analysis of object heap

The present invention is directed to a system, method, and computer-readable medium for providing a highly efficient locking mechanism for an object's shared data. The locking mechanism is bound to an object during program execution (i.e. during run-time) when synchronization is first requested for the object's shared data. Thus, there are no changes to the actual code, either source code or binary code, which defines the object. A locking mechanism is bound to an object by defining a memory area within the object's header, which either contains the locking mechanism or a pointer to a locking mechanism. The locking mechanism remains bound to the object for the life of the object. Efficiency is gained by limiting the use of operating system semaphores (i.e. kernel semaphores). Operating system semaphores are not used unless blocking (i.e. contention) occurs. Rather, the locking mechanism bound to the object is used during non-blocking situations. Additional efficiency is gained by not assigning or initializing an object's locking mechanism until the first synchronization request is received for the object's shared data. In one embodiment of the present invention, the object's header memory area contains the actual locking mechanism. In another embodiment of the present invention, the object's header memory area contains a pointer or index to a locking mechanism.

System and method for adding highly optimized synchronization capability to objects at run-time

The present invention is directed to a system and method for accurately and efficiently synchronizing and then mapping, or normalizing, processor clocks in a multiprocessor information handling system. The system and method of the present invention provide sufficient granularity for subcycle variations between processors, while taking into account the problem of clock drifts. A plurality of processors are selected for the purpose of synchronization. The clocks located on the processors are synchronized, and then time values between synchronization points are mapped from each secondary processor to an equivalent, or normalized, time value in a primary processor. To accomplish this mapping, three clock differences are calculated. The first clock difference is the time between the first and second synchronization points for the primary processor, and the second clock difference is the time between the first and second synchronization points for the secondary processor. The third clock difference is the time difference between the time value to be mapped in the secondary processor and the first synchronization point in the secondary processor. The third time difference is multiplied by the ratio of the first time difference to the second time difference, and then the result is added to the time value of the primary clock at the first synchronization point. The final result is the mapped, or normalized, time value.

System and method for mapping processor clock values in a multiprocessor

The present invention is directed to a system and method of measuring performance data utilizing state transitions within a computer system. A number of system states are defined, and the transitions from one state to another are tracked. At each state transition, performance properties related to the computer system may be checked or calculated, and performance data added to a table or tables. The present invention allows performance data to be measured in a way that is highly precise and has minimal effects on the system performance being measured. The act of measuring performance data utilizing state transitions does not create a misleading measure of performance nor does it adversely impact system performance. Furthermore, the present invention requires minimal changes to the operating system and no changes to application code.

System and method for low overhead, high precision performance

The system and method of the present invention allows synchronization of processor clocks in a multiprocessor information handling system. The present invention calculates an average processor clock value for each processor being synchronized. All processors being synchronized read their clocks a predetermined number of times. The clock reading is done simultaneously by all the processors being synchronized. Each processor then calculates an average processor clock value, which is equivalent, for synchronization purposes, to the average processor clock values of the other processors. When more than two processors are being synchronized, a processor may be chosen as the primary processor. The other, secondary processors are synchronized one at a time with the primary processor. An adjustment is then made to the average processor clock values obtained, so that all average processor clock values are based on the same average processor clock value of the primary processor. Processor clocks in multiprocessor information handling systems may be efficiently synchronized within a small number of clock cycles. The synchronization may be accurately performed on any number of processors in a multiprocessor information handling system.

Growing community of inventors

Austin, TX, United States of America

Weiming Gu