GB1431163A - Computer memory system - Google Patents

Abstract

1431163 Memory systems HONEYWELL INFORMATION SYSTEMS Inc 23 April 1974 [23 April 1973] 17664/74 Heading G4A An unused area of memory (bounded by regions containing stored data to be preserved) is divided into blocks of at least two sizes by forming first the maximum number of contiguous blocks of the largest size and then repeating this process for successively smaller block sizes. As described division is into 32 word, 16 word and 8 word blocks. The memory is divided into sectors between which and in which unused areas may exist. A "free memory-area" table is stored in the memory, listing starting and ending addresses of each gap and terminating in a sentinel word. The memory also includes software "initialization" routines for dividing the gaps. In a first phase (routine of Fig. 8, not shown) the table is accessed to get the start and end addresses of the first free areas which are saved in memory location. If the end address does not follow the start address for the initialization routine the area is divided into blocks (routine of Fig. 9, not shown) and the table then accessed for the next free area. This process is repeated until the terminal sentinel word is accessed. A check is next made to see if the minimum storage requirements are met (routine of Fig. 11, not shown) and if they are not the free area of memory between the table and the declared end of memory is also divided (routine of Fig. 12, not shown). Divide routine (Fig. 10, not shown).-From the start address N for the free area to be divided, a start address J for the first block of largest size S is derived by letting and checking that J+S is less than the end address M of the free area. This process is then repeated by increasing the address J by the size S until J + S#M when the next smaller block size is processed. A pool parameter table of the information derived i.e. the addresses of the various block sizes, the number of blocks available &c. In a second phase (routine of Fig. 13, not shown) after all the free area has been divided, a fragment routine (Fig. 14, not shown) is affected in which the number of largest block sizes (held in the pool parameter table) is first examined to see if it exceeds the required number of blocks for an operation. If it does the next block size is examined. If a block is not large enough, e.g. if the number of 16 word blocks is insufficient a routine to divide one of the 32 word blocks (routine of Fig. 15, not shown) is effected (if there is an excess of the blocks), incrementing of the number of 16 word blocks and decrementing of the number of 32 word blocks in the pool parameter table being carried out.