International Business Machines Corporation

Gary Bischoff

Leslie Robert Wilson

Marc Segre

Dag R Blokkum

Paul Joseph Milot

David L Peterson

Jeffrey S Spencer

Antonio De Leon Penaloza, Iii

Data transfers between a workstation bus and a graphics adapter bus are handled by a plurality of first-in-first-out (FIFO) buffers, each of which is independently operable to transfer data in a selected direction between the two buses. The FIFOs are accessible either directly by the workstation processor or by means of a DMA operation. Each FIFO is assigned a unique range of addresses in the address space of the workstation processor to permit a workstation process to transfer a block of data to or from a selected FIFO using a single instruction. Workstation writes (reads) to a FIFO are suspended in response to a first status signal indicating that the high (low) threshold for that FIFO has been reached and are restarted in response to a second status signal indicating that the low (high) threshold has been reached. A buffer counter indicating the amount of data in each FIFO is initialized at zero for outbound transfers from the workstation to the adapter or at the maximum buffer count for inbound transfers from the adapter to the workstation. The buffer count is incremented in response to accesses from the workstation side and is decremented in response to accesses from the adapter side, regardless of the direction of transfer.

Bidirectional FIFO buffer for interfacing between two buses of a

A plurality of processors or intelligent controllers separately utilize discrete pages of a large memory. Within each of these pages a processor can address a plurality of subdivisions or blocks utilizing the processors' address lines. Thus, separate processors having access to this memory and having a limited addressing capability can utilize a plurality of different pages of this memory, within an identical address range, and nevertheless remain confined to separate memory environments established for each of the separate processors. This is accomplished by use of a hardware register to point the separate processors to their assigned pages of the memory and a stored translate table to point to particular blocks of memory within the pages in accordance with a portion of an address generated by the processor accessing the memory.

Memory allocation for multiple processors

Growing community of inventors

Saugerties, NY, United States of America

Gary Bischoff