摘要:
An arbitration controller that temporarily raises the priority of a graphic device's HWM request above that of the BitBLT engine while that device's LWM request is being served. In this manner, the BitBLT engine cannot interrupt the transfer of data to the graphics device. Each device capable of issuing memory access requests is categorized into one of four classes. The LWM requests are the highest priority requests, followed by CPU memory access requests, then BitBLT engine requests, and finally by HWM requests. When a LWM request is granted, the requesting device's HWM request is elevated to a priority between the CPU and the BitBLT engine. Once the LWM request is complete, the HWM request is served until either it completes, the CPU issues a memory access request, or another LWM request occurs.
摘要:
A memory bandwidth allocation scheme in a computer system having a unified memory architecture. When a first device requesting access to a resource detects that two higher priority devices are also requesting access to the same resource, the first device shuts itself down. Accordingly, the first device deactivates its access request signals and clears its buffers. Then, the first device reactivates itself when it receives a graphics VSYNC signal. This allocation scheme ensures that the arbiter will not get stuck in a loop serving only high priority requests and not serving lower priority requests.
摘要:
A keyboard controller is used to drive the PDEN pins on DIMMs. A PC has a keyboard controller with a plurality of programmable input/output (I/O) pins. The state of the programmable I/O pins can be set by software. The pins are coupled to individual PDEN pins on the DIMMs. When the programmable I/O pins are activated, the PDEN pins are driven active and each DIMM outputs a signal indicating its characteristics. The signals are latched and stored for use by a memory controller.