The present invention extends to methods, systems, and computer program products for indicating parallel operations with user-visible events. Event markers can be used to indicate an abstracted outer layer of execution as well as expose internal specifics of parallel processing systems, including systems that provide data parallelism. Event markers can be used to show a variety of execution characteristics including higher-level markers to indicate the beginning and end of an execution program (e.g., a query). Inside the execution program (query) individual fork/join operations can be indicated with sub-levels of markers to expose their operations. Additional decisions made by an execution engine, such as, for example, when elements initially yield, when queries overlap or nest, when the query is cancelled, when the query bails to sequential operation, when premature merging or re-partitioning are needed can also be exposed.

Indicating parallel operations with user-visible events

The present invention relates to mechanisms for handling and detecting collisions between threads (5, 6, 7) that execute computer program instructions out of program order. According to an embodiment of the present invention each of a plurality of threads (5, 6, 7) are associated with a respective data structure (9, 10, 11) comprising a number of bits (12) that correspond to memory elements (m0, m1, m2, mn) of a shared memory (4). When a thread accesses a memory element in the shared memory, it sets a bit in its associated data structure, which bit corresponds to the accessed memory element. This indicates that the memory element has been accessed by the thread. Collision detection may be carried out after the thread has finished executing by means of comparing the data structure of the thread with the data structures of other threads on which the thread may depend.

Collision handling apparatus and method

A facility is presented for scheduling computer jobs of a computing environment having a plurality of computing nodes. The facility includes scheduling N sequential computer jobs of a job queue until achieving a top job party, wherein N≧2, and the number of jobs in the top job party is dynamically determined by monitoring backfill window coverage of the computing nodes by the N scheduled jobs and automatically closing membership in the top job party once (i) all computing nodes are covered by at least one backfill window, or (ii) a backfill end time of at least one computing node is greater than or equal to a maximum wallclock limit of a set of job classes having the N scheduled jobs. Responsive to determining the top job party, the facility then backfill schedules jobs of the job queue around the top job party for execution.

Scheduling of computer jobs employing dynamically determined top job party

Mechanisms for extracting data dependencies during runtime are provided The mechanisms execute a portion of code having a loop and generate, for the loop, a first parallel execution group comprising a subset of iterations of the loop less than a total number of iterations of the loop The mechanisms further execute the first parallel execution group and determining, for each iteration in the subset of iterations, whether the iteration has a data dependence Moreover, the mechanisms commit store data to system memory only for stores performed by iterations in the subset of iterations for which no data dependence is determined Store data of stores performed by iterations in the subset of iterations for which a data dependence is determined is not committed to the system memory

Runtime Extraction of Data Parallelism

Mechanisms for extracting data dependencies during runtime are provided. With these mechanisms, a portion of code having a loop is executed. A first parallel execution group is generated for the loop, the group comprising a subset of iterations of the loop less than a total number of iterations of the loop. The first parallel execution group is executed by executing each iteration in parallel. Store data for iterations are stored in corresponding store caches of the processor, Dependency checking logic of the processor determines, for each iteration, whether the iteration has a data dependence. Only the store data for stores where there was no data dependence determined are committed to memory.

Parallel execution unit that extracts data parallelism at runtime

The invention generally relates to a processor developed for a service network that provides various services to a plurality of users connected to the network. The processor (30; 50) comprises a job queue (52) with a number of storage positions for storing job signals corresponding to jobs that form part of substantially independent services requested by the users of the network, and a plurality of parallel processing units (54) which independently process job signals from different storage positions of the job queue (52) to execute corresponding jobs in parallel. As a number of jobs are executed speculatively, a unit (56) for checking for possible dependencies between the executed jobs is incorporated into the processor. If a dependency is detected for a speculative job, that job is flushed. To ensure prompt and proper service for the users of the service network, flushed jobs are quickly restarted directly from the job queue.

A job-parallel processor

The present invention presents a processing system (1) comprising a number of memory-sharing processors (10a-e) arranged for parallel processing of jobs, and data consistency means (14) for assuring data consistency. The processing system (1) comprises a scheduler (17), for scheduling jobs for execution on the processors (10a-e) according to a first algorithm. The processing system (1) according to the present invention further uses means for retiring the jobs (18) in an order given by a second algorithm, preferably according to a global order of creation. The second algorithm is different from the first algorithm. The first algorithm may be adjusted to the particular system used, and may base the scheduling to a particular processor on e.g. the source, target, communication channel or creating processor for the job in question. The processing system (1) uses a common job queue (16), and the scheduling is preferably performed adaptively.

Processing system scheduling

In general, the invention is directed towards a multiprocessing system in which jobs are speculatively executed in parallel by multiple processors (30-1, 30-2, ..., 30-N). By speculating on the existence of more coarse-grained parallelism, so-called job-level parallelism, and backing of to sequential execution only in cases where dependencies that prevent parallel execution of jobs are detected, a high degree of parallelism can be extracted. According to the invention a private memory buffer is speculatively allocated for holding results, such as a communication message, an operation system call or a new job signal, of a speculatively executed job, and these results are speculatively written directly into the allocated memory buffer. When commit priority is assigned to the speculatively executed job, a pointer referring to the allocated memory buffer is transferred to an input/output (10) device which may access the memory buffer by means of the transferred pointer. In this way, by speculatively writing messages and signals into private memory buffers, even further parallelism can be extracted.

Kling Lars-Oerjan

Papers

A job-parallel processor

Processing system scheduling

Improved result handling in a multiprocessing system

Improved handling of the results of speculatively executed jobs in a multiprocessing system

A job parallel processing system for a service network