Abstract: A method of runtime analysis of a computer program can include, after each access of a first type to a memory address, incrementing a counter associated with the memory address and, after each access of a second type to a memory address, comparing the counter associated with the memory address with a threshold. The method further can include identifying at least one memory address having a counter that exceeds the threshold and determining a portion of code responsible for accessing the memory address associated with the counter that exceeds the threshold.

Abstract: A method of identifying race conditions in a computer program can include identifying a call to a selected function specifying a memory range. The selected function can cause data to be written to the memory range asynchronously. The method further can include spraying the memory range with a predetermined bit pattern and determining whether a detected write operation at least partially overwrites the predetermined bit pattern in the memory range. A race condition can be selectively indicated according to the determining step.

Abstract: A computer-implemented method of finding portions of a computer program exhibiting irregular performance can include detecting an entry into a function of a computer program under test indicating a current execution of the function and, responsive to detecting an exit from the function, determining a count of a computing resource utilized by the function during the current execution of the function. The count of the computing resource can be compared with a count range determined according to at least one prior execution of the function. The function can be selectively represented within a call graph as a plurality of nodes according to the comparing.

Abstract: Thread starvation is indicated for a program under test while the program is running. Beginning with a least frequently used thread in a list of thread tracking structures, if recorded events are currently listed for the thread, thread starvation is detected based on a percentage of the recorded events that took place while other threads were waiting a synchronization object held by the thread. If no recorded events are listed for the thread, the percentage of recorded of events that take place while other threads were waiting for synchronization object held by the thread does not exceed the predetermined percentage, and there are timestamps currently listed for the thread and the thread is currently waiting for a synchronization object, thread starvation is detected based on a percentage of time the thread has been waiting for the synchronization object.

Abstract: A method of intra-block memory usage analysis for a program can include identifying a memory block that has been allocated to the program and determining at least one intra-memory block usage characteristic for the allocated memory block.

Abstract: A method of detecting inefficient use of virtual memory by an application that is programmed using memory pointers. The method can include selecting a first heap from a list of tracked heaps and identifying allocated memory blocks within the first heap. A first total can be determined as a sum of sizes of the identified memory blocks. A second total can be determined as a sum of unallocated, NULL-filled bytes within the first heap that are not located outside outermost memory block addresses in each range of committed memory in the first heap. A ratio of the first total and the second total can be determined, and a memory profile corresponding to the ratio can be generated. A memory profile suggestion also can be generated.

Abstract: Methods and apparatus, including computer program products, for identifying unnecessary synchronization objects in software applications. A computer-implemented method for determining a usefulness of synchronization objects includes, in a multi-threaded application program, tracking threads as they are created and terminated, intercepting application program interfaces (APIs) used to create and destroy thread synchronization objects, in the API wrappers, tracking existing synchronization objects in a list, and associating a list entry for each of the synchronization objects with a unique bit in a bit field.

Abstract: A method and computer program product for detecting an attempt by a first thread to engage a synchronization object. A determination is made as to whether an earlier thread has already engaged the synchronization object. In response to determining that the earlier thread has already engaged the synchronization object, a determination is made as to whether the number of threads waiting to engage the synchronization object is greater than a highwater mark.

Abstract: A method and computer program product for monitoring the chronological order in which one or more portions of a first instance of a computer program are executed, thus generating a first data file. The chronological order in which one or more portions of a second instance of the computer program are executed is monitored, thus generating a second data file. The first and second data files are compared.

Abstract: A method of identifying race conditions in a computer program can include identifying a call to a selected function specifying a memory range. The selected function can cause data to be written to the memory range asynchronously. The method further can include spraying the memory range with a predetermined bit pattern and determining whether a detected write operation at least partially overwrites the predetermined bit pattern in the memory range. A race condition can be selectively indicated according to the determining step.

Abstract: A method of runtime analysis for a computer program can include generating runtime data relating to memory usage for an instrumented computer program and creating a memory map comprising a plurality of memory ranges of different types according to the runtime data. At least a portion of the memory map can be presented to indicate selected ones of the plurality of memory ranges.

Abstract: A method and computer program product for detecting an attempt to engage a synchronization object. A tracking list for a line of code that attempted to engage the synchronization object is updated.

Abstract: Runtime analysis of a computer program can be performed. An improper memory access to a memory location can be detected during runtime of the computer program. A record of the improper memory access can be stored during runtime. A subsequent error condition in the computer program can be detected during runtime. The error condition can be compared with the record corresponding to the improper memory access. Based upon the comparison, an indication that the error condition is related to the improper memory access can be provided.

Abstract: A method of runtime analysis of a computer program can include detecting allocations of virtual memory regions during runtime of the computer program. An entry can be created in a list of existing virtual memory regions for each allocation of a virtual memory region during runtime. The list can be updated, during runtime, according to usage of the virtual memory regions. Virtual memory regions allocated to, but not used by, the computer program can be indicated according to the list.

Abstract: A method for detecting race conditions involving heap memory access including a plurality of threads being tracked. At runtime a plurality of APIs utilized to create and destroy thread synchronization objects are intercepted, and each synchronization object created via the APIs is tracked. A bit field is created that contains a unique bit for each synchronization object. Heap memory allocations and deallocations are intercepted and tracked. The heap memory access is intercepted, and at that time, the ID of the accessing thread is compared with the last thread ID associated with that memory block when it was last accessed. If the thread IDs do not match, then the current thread synchronization object bit field is compared with the last synchronization object bit field associated with thread memory block. Provided the bit fields are different, a race condition warning is reported that is displayable to the user having the call chains.

Abstract: An automated method for historical analysis of a memory state, including tracking all memory write accesses by an application under test. Then, recording a directed graph having values in a plurality of memory locations, such values derived from values previously stored. The method further includes displaying a portion of the graph that led to the update of a user-specified memory location by the application under test. The method proceeds by traversing backward from any node located on the directed graph via an arc utilized to follow the path of events that led to the content of the memory location represented by that node at the time represented by that node. One end of a first displayed arc ends at the node representing the user-specified first memory location and the other end of the first displayed arc ends at another node representing a second memory location.

Abstract: A method for use of a diagnostic software tool that can allow software developers to track the number of times each memory block is enlarged, and highlight the most frequently enlarged memory blocks. In this regard, in better understanding the performance characteristics of memory reallocation a developer can use this method to identify and implement better coding techniques to improve code efficiency and reduce the processing time utilized for memory reallocations. In addition, graphs can be generated to indicate the time/CPU utilization dedicated to the memory reallocation process.