Abstract: According to an example, persistent memory garbage collection may include determining whether termination of a program is based on a specified termination of the program during execution of the program or an unspecified termination of the program during the execution of the program. In response to a determination that the termination of the program is based on the specified termination of the program during the execution of the program, persistent metadata stored in a persistent memory may be used to restart the program. In response to a determination that the termination of the program is based on the unspecified termination of the program during the execution of the program, the persistent metadata stored in the persistent memory may be used to collect garbage from the persistent memory and to restart the program.

Abstract: Examples herein involve a variable cache. An example variable cache controller obtains cache lines corresponding to accesses of a non-volatile memory of a system, monitors access history of the non-volatile memory, determines a number of distinct objects accessed in the access history during a time period from the object information, and sets a size of a variable cache of the system based on the number of distinct objects accessed in the access history during the time period.

Abstract: Methods, apparatus, systems and articles of manufacture are disclosed to control a cache. An example method includes monitoring cache lines in a cache, the cache lines storing recently written data to the cache, the recently written data corresponding to main memory, comparing a total quantity of the cache lines to a threshold that is less than a cache line storage capacity of the cache, and causing a write back of at least one of the cache lines to the main memory when a store event causes the total quantity of the cache lines to satisfy the threshold.

Abstract: Examples herein involve a variable cache. An example variable cache controller obtains cache lines corresponding to accesses of a non-volatile memory of a system, monitors access history of the non-volatile memory, determines a number of distinct objects accessed in the access history during a time period from the object information, and sets a size of a variable cache of the system based on the number of distinct objects accessed in the access history during the time period.

Abstract: Example implementations may relate to a version controller allocating a copy page in persistent memory upon receiving, from an application executing on a processor, a copy command to version an image page for an atomic transaction. The version controller may receive application data addressed to a cache line of the image page, and may write the application data to a cache line of the copy page corresponding to the addressed cache line of the image page. If the version controller receives a replace-type transaction commit command, the version controller may generate a final page by either forward merging the image page into the copy page or backward merging the copy page into the image page, depending a merge direction policy.

Abstract: According to an example, persistent memory garbage collection may include determining whether termination of a program is based on a specified termination of the program during execution of the program or an unspecified termination of the program during the execution of the program. In response to a determination that the termination of the program is based on the specified termination of the program during the execution of the program, persistent metadata stored in a persistent memory may be used to restart the program. In response to a determination that the termination of the program is based on the unspecified termination of the program during the execution of the program, the persistent metadata stored in the persistent memory may be used to collect garbage from the persistent memory and to restart the program.

Abstract: Example implementations may relate to a version controller allocating a copy page in persistent memory upon receiving, from an application executing on a processor, a copy command to version an image page for an atomic transaction. The version controller may receive application data addressed to a cache line of the image page, and may write the application data to a cache line of the copy page corresponding to the addressed cache line of the image page. If the version controller receives a replace-type transaction commit command, the version controller may generate a final page by either forward merging the image page into the copy page or backward merging the copy page into the image page, depending a merge direction policy.

Abstract: Crash recovery with asynchronous consistent snapshots in persistent memory stores of a processing environment. A processing environment includes a user program and infrastructure-maintained data structures. The infrastructure-maintained data structures include a log of updates made to program data structures and a snapshot of the state of the program data structures. The systems and methods include writing log entries in the log to a transient memory. The log entries correspond to store instructions and memory management instructions operating on a nonvolatile memory (NVM), and input/output (I/O) operations executed by program instructions of the user program. Each of the log entries represents an effect of a corresponding operation in the program instructions. The systems and methods also include creating a snapshot in the NVM after a consistent program point based on the log of updates. The snapshot provides a rollback position during restart following a crash.

Abstract: A technique includes performing an update to a location of a non-volatile memory. The update is created by execution of at least one machine executable instruction of a plurality of machine executable instructions. The technique includes using a processor-based machine to selectively track the update to allow recovery of the execution to a given consistency point based at least in part on whether the machine executable instruction(s) creating the update are located within a synchronized section of the plurality of machine executable instructions.

Abstract: A system includes an initiator and processing nodes. The initiator distributes portions of a transaction among the processing nodes. Each processing node has at least one downstream neighbor to which the processing node sends commit messages. The commit messages include a commit status of the processing node. The downstream neighbor is also a processing node.

Abstract: Methods, apparatus, systems and articles of manufacture are disclosed to control a cache. An example method includes monitoring cache lines in a cache, the cache lines storing recently written data to the cache, the recently written data corresponding to main memory, comparing a total quantity of the cache lines to a threshold that is less than a cache line storage capacity of the cache, and causing a write back of at least one of the cache lines to the main memory when a store event causes the total quantity of the cache lines to satisfy the threshold.

Abstract: Disclosed herein are a system, non-transitory computer readable medium, and method for recovering from an abnormal failure of a program. Changes made by a plurality of threads of the program are undone in a reverse order in which the changes were made.

Abstract: A method for performing memory operations is provided. One or more processors can determine that at least a portion of data stored in a cache memory of the one or more processors is to be stored in the main memory. One or more ranges of addresses of the main memory is determined that correspond to a plurality of cache lines in the cache memory. A set of cache lines corresponding to addresses in the one or more ranges of addresses is identified, so that data stored in the identified set can be stored in the main memory. For each cache line of the identified set having data that has been modified since that cache line was first loaded to the cache memory or since a previous store operation, data stored in that cache line is caused to be stored in the main memory.

Abstract: A technique includes identifying a dependency between a first persistent memory region and at least one other persistent memory region. The technique includes using a process having access to the first persistent memory region to selectively perform garbage collection for the first persistent memory region based at least in part on whether the process has access to the other persistent memory region(s) from which the first persistent memory region depends.

Abstract: Crash recovery with asynchronous consistent snapshots in persistent memory stores of a processing environment. A processing environment includes a user program and infrastructure-maintained data structures. The infrastructure-maintained data structures include a log of updates made to program data structures and a snapshot of the state of the program data structures. The systems and methods include writing log entries in the log to a transient memory. The log entries correspond to store instructions and memory management instructions operating on a nonvolatile memory (NVM), and input/output (I/O) operations executed by program instructions of the user program. Each of the log entries represents an effect of a corresponding operation in the program instructions. The systems and methods also include creating a snapshot in the NVM after a consistent program point based on the log of updates. The snapshot provides a rollback position during restart following a crash.

Abstract: A log entry is created in persistent memory that represents a modification to a variable that resides in persistent memory. A log entry is created in persistent memory that represents a synchronization operation. A program-order based dynamic ordering relationship is created between two successive log entries within an execution entity. A synchronization-order based dynamic ordering relationship is created between two log entries corresponding to synchronization operations in concurrently executing distinct execution entities of said execution instance.

Abstract: In one example, a method for implementing recovery segments includes sending an application message from a parent process executed by a first computing device to a child process executed by a second computing device and identifying a dependency created by the application message. This identified dependency is included in a dependence set of the child process and saved. A checkpoint is generated by the parent process and a checkpoint message that includes dependency information is sent from the parent process to the child process. The child process modifies the dependence set according to the dependency information and generates a second checkpoint that is saved in nonvolatile memory of the second computing device. Upon occurrence of a failure of the parent process, the child process reverts to a most recent checkpoint generated by the child process that does not include the effects of processing an orphan message.

Abstract: A technique includes identifying a dependency between a first persistent memory region and at least one other persistent memory region. The technique includes using a process having access to the first persistent memory region to selectively perform garbage collection for the first persistent memory region based at least in part on whether the process has access to the other persistent memory region(s) from which the first persistent memory region depends.

Abstract: In a method of identifying a globally consistent state in a multithreaded program, a plurality of locally consistent states is identified, in which a locally consistent state of a thread comprises a set of memory locations and their corresponding data values accessed between points in the multithreaded program where no locks are held. Globally consistent states are identified based at least in part on the locally consistent states.

Abstract: A log entry is created in persistent memory that represents a modification to a variable that resides in persistent memory. A log entry is created in persistent memory that represents a synchronization operation. A program-order based dynamic ordering relationship is created between two successive log entries within an execution entity. A synchronization-order based dynamic ordering relationship is created between two log entries corresponding to synchronization operations in concurrently executing distinct execution entities of said execution instance.