| 摘要 |
<p>1,231,287. Data processing. INTERNATIONAL BUSINESS MACHINES CORP. 15 Aug., 1968 [27 Sept., 1967], No. 39071/68. Heading G4A. A data processing system comprises a main store having word positions for storing variable length multi-word operands which start and end at any byte position within a word position, means to transfer at least one operand from the main store to an auxiliary store a word position at a time, and a processing unit for identically processing successive words of at least one operand stored in the auxiliary store, a byte at a time or a word at a time according to predetermined conditions, under the control of control data from the main store. The auxiliary store can hold a word (four 8-bit bytes) of each of two operands from the main store, the main store addresses (word and byte) of the first byte of each operand, and a count field. The count field specifies the number of bytes per operand, except in decimal operations when the two halves of the field specify the numbers of bytes in the two operands respectively. The count field can be decremented by the arithmetic and logic unit as processing proceeds. The byte portions of the addresses stored in the auxiliary store can be transferred to and stored in a sub-unit (byte) accessing and modifier circuit which can increment and decrement them (according to whether bytes are taken going leftwards or rightwards along the words). This circuit also stores a 4-bit mask which is updated as each byte is processed to indicate which byte positions are to be overwritten when results of the processing are returned to overwrite one operand in the main store. Processing involving two operands is done a byte at a time unless the circuitry detects that bytes to be combined are correspondingly positioned in their respective words when complete words are processed a word at a time, but incomplete words (e.g. the last word position, if the operand does not end on a word boundary) are still processed a byte at a time. Processing involving a single operand can be done a word at a time for incomplete words. Processing is serial by byte even in the "word at a time" processing. In "word at a time" processing, the system remains in execute mode during the processing of a complete word, using the same control data (a microprogramme word read from main store into a control register) repetitively, and only then transfers to another mode, whereas with "byte at a time" processing, transfer to another mode occurs after each byte. The other modes provided are for transfer of words between the main and auxiliary stores and updating of the count field. Modes are changed by microprogramme branching under control of the incremented/ decremented byte addresses in the sub-unit accessing and modifier circuit and the decremented count field from the auxiliary store via a status register.</p> |
