The present invention provides a parallel memory scheduler for execution on a high speed highly parallel multiprocessor architecture. The operating system software provides intelligence and efficiency in swapping out process images to facilitate swapping in another process. The splitting and coalescing of data segments are used to fit segments in to current free memory even though a single contiguous space of sufficient size does not exist. Mapping these splits through data control register sets retains the user's contiguous view of the address space. The existence of dual images and partial swapping allows efficient, high speed swapping. Candidates for swap out are chosen in an intelligent fashion, selecting only those candidates which will most efficiently aLlow the swapin of another process.
RELATED APPLICATIONS
This application is a continuation-in-part of an application filed in the U.S. Pat. and Trademark Office on Jun. 11, 1990, entitled INTEGRATED SOFTWARE ARCHITECTURE FOR A HIGHLY PARALLEL MULTIPROCESSOR SYSTEM, Ser. No. 07/537,466 and assigned to the assignee of the present invention, the disclosure of which is hereby incorporated by reference in the present application. The application is also related to co-pending application filed concurrently herewith, entitled DUAL LEVEL SCHEDULING OF PROCESSES TO MULTIPLE PARALLEL REGIONS OF A MULTITHREADED PROGRAM ON A TIGHTLY COUPLED MULTIPROCESSOR COMPUTER SYSTEM, Ser. No. 07/571,995 and to pending application, entitled GLOBAL REGISTERS FOR A MULTIPROCESSOR SYSTEM, Ser. No. 07/536,198, both of which are assigned to the assignee of the present invention, the disclosure of which is hereby incorporated by reference in the present application.
A computer method and system for coalescing free space in a heap data structure. In preferred embodiments, the heap comprises a plurality of sequential pages (segments), and each page is logically divided into blocks of memory that are either free or allocated. Each block resides in a single page. According to the present invention, a source pointer and a destination pointer advance through the pages. The source pointer is advanced to point to an allocated block to be copied. The destination pointer is advanced to point to an area in which the block currently pointed to by the source pointer can be copied. The block pointed to by the source pointer is then copied to the area pointed to by the destination pointer, and the source and destination pointers are advanced. When the destination pointer advances to an area that is not large enough to hold the block pointed to by the source pointer, a free block is created out of the area.
A processing unit for carrying out specified data processing operations while performing read/write operations on data in an internal memory is coupled to a memory control unit for performing read/write operations on data in an external memory. Data exchange is carried out between the internal and external memories through the memory control unit. Data requiring a longer processing time or data frequently accessed is mapped into the internal memory in accordance with the data exchange, thereby improving overall memory system performance.
A method for dynamic management of a free store of a computer system, the free store being subdivided into at least two logical store areas with different access characteristics. Each logical store area of the at least two logical stores is separately managed. A segment is assigned size to a logical store area of the free store, the segment size corresponding to a maximum size described as a maximum in a buddy algorithm. In the event of a store request regarding a first logical store area of the at least two logical store areas, a segment from the free store is assigned to the first logical store area if there is no store segment of sufficient size available. In the event of a store request regarding a second logical store area of the at least two logical store areas, a segment from the free store which has not been assigned to any other logical store area is assigned to the second logical store area and which is adjacent to already assigned segments if no store segment of sufficient size is available. If it is no longer possible to satisfy a store request from a logical store area of the at least two logical store areas, which requires an assignment from the free store, since the free store has already been completely assigned, store managements of the other store areas are evaluated via a store management of this store area to ascertain whether there are other segments in the other store areas with the maximum size. If this condition is met, it takes over such other segment from a respective other store area and the other segment is assigned to this store area.
The amount of real memory space used by descriptor tables of a computer system is reduced. On system initialization, sufficient contiguous virtual memory space for each table is allotted for descriptors in the tables to define all memory space in both central storage and expanded storage. However, only descriptors for on-line storage locations, relative to a particular partition, are placed in these tables. As the system operates, the tables are continuously updated, removing descriptors for storage locations taken off-line and adding descriptors for blocks of memory brought on-line so that the tables reflect only the storage locations then on-line in the particular partition.
A method and system for allocating blocks of disk in persistent storage to requesting threads. A primary data structure is provided for organizing and categorizing blocks of disk space. In addition, a secondary data structure is provided for maintaining a list of all active file system processes and blocks of disk space used by those processes. Blocks of disk space are assigned to pages. At such time as a thread may request allocation of disk space, both data structures are reviewed to determine if the requested disk space is available and to limit access of available disk space to a single page of memory to a single thread at any one time.
