摘要:
A method and system for optimizing image quality while operating an interactive graphics application within a data processing system. First, the image rendering speed for each of the rendering modes available within the interactive graphics application are assessed. Upon initial operation of the interactive graphics system, a default rendering mode is activated. During operation of the interactive graphics application, the processing load imposed on the data processing system is monitored and utilized as a user activity metric. The active rendering mode is updated in accordance with the user activity metric, such that the speed of the selected rendering mode varies inversely with the current processing load.
摘要:
A performance monitor configured to count memory transactions and to issue an interrupt to the computer system if the monitor detects a specified number of transactions associated with a particular segment of the physical address space of the system. The monitor includes an interface suitable for coupling to an interconnect network of a computer system and configured to extract physical address information from a transaction traversing the interconnect network, a translation module adapted for associating the extracted physical address with one of a plurality of memory blocks and, in response thereto, incrementing a memory block counter corresponding to the memory block, and an interrupt unit configured to assert an interrupt if the block counter exceeds a predetermined value. The interface unit is configurable to selectively monitor either incoming or outgoing transactions and the translation unit preferably includes a plurality of region filters each comprising one or more of the memory blocks. In the preferred embodiment, the plurality of block counters are implemented with a random access memory device. In one embodiment useful for simulating operation of the system and for checking the design of the performance monitor, the monitor further includes a transaction generator coupled to the interconnect network and configured to issue specified remote memory transactions at specified intervals if a monitor enable bit of the performance monitor is disabled. In one embodiment, the transaction generator is configurable to issue either outgoing transactions or incoming transactions.
摘要:
A system and method for providing in-server caching of shared data involves a server program that defines a server cache in RAM of a server machine and stores a selected file in the server cache. If a cached file is modified through the file system interface of the operating system of the server machine, the operating system automatically issues an upcall to the server program, the upcall identifying the modified file. In response to receipt of the upcall, the server program removes the modified file from the server cache. In one embodiment, the server program responds to a client request requiring access to a requested file by obtaining the requested file from the server cache if the server cache contains that file. Otherwise, the server program calls the operating system to obtain the requested file and then adds that file to the server cache as a cached file. The server program then generates a result based on the requested file and transmits the result to the remote data processing system.
摘要:
A system for and method of monitoring memory transactions in a data processing system are disclosed. The method includes defining a set of memory transaction attributes with a monitoring system and detecting, on a data processing system connected to the monitoring system, memory transactions that match the defined set of memory transaction attributes. The number of detected memory transactions occurring during a specified duration are then displayed in a graphical format. In one embodiment, the data processing system comprises a non-uniform memory architecture (NUMA) system comprising a set of nodes. In this embodiment, the detected transactions comprise transactions passing through a switch connecting the nodes of the NUMA system. The set of memory transaction attributes may include memory transaction type information, node information, and transaction direction information. The data processing system may operate under a first operating system such as a Unix® based system while the monitoring system operates under a second operating system such as a Windows® operating system. The set of memory transactions may include memory address information. In this embodiment, defining the memory address information may include defining a memory window size, subdividing the memory window into a set of memory grains, and displaying the number of detected memory transactions corresponding to each memory grain in the memory window.
摘要:
A performance monitor for a computer system that includes an interface, a filter module, and an address mapping module. The interface is suitable for coupling to an interconnect network of the computer system. The interconnect network links a local node of the system with at least one remote node of the system. The interface is configured to extract physical address information from a transaction traversing the interconnect network. The filter module associates the physical address with one of several memory blocks, where each memory block comprises a contiguous portion of the system's physical address space. The address mapping module associates the identified memory block with at least one range of virtual addresses associated with at least one of a plurality of concurrently executing programs and increments each of a set of access counters. The association between the selected memory block and the access counters is facilitated by a pointer field corresponding to the memory block.
摘要:
A data processing network and method for conserving energy in which an initial negotiation between a network server and a switch to which the server is connected is performed to establish an initial operating frequency of the server-switch link. An effective data rate of the server is determined based on network traffic at the server. Responsive to determining that the effective data rate is materially different than the current operating frequency, a subsequent negotiation is performed to establish a modified operating frequency where the modified operating frequency is closer to the effective data rate than the initial operating frequency. The determination of the effective date rate and the contingent initiation of a subsequent negotiation may be repeated periodically during the operating of the network. In one embodiment, the initial and subsequent negotiation are compliant with the IEEE 802.3 standard.
摘要:
A data processing system includes a plurality of processing nodes that each contain at least one processor and data storage. The plurality of processing nodes are coupled together by a system interconnect. The data processing system further includes a configuration utility residing in data storage within at least one of the plurality of processing nodes. The configuration utility selectively configures the plurality of processing nodes into either a single non-uniform memory access (NUMA) system or into multiple independent data processing systems through communication via the system interconnect.
摘要:
A system comprising a cluster of diskless servers employing a distributed shared memory abstraction that presents an area of shared memory for two or more processes executing on different servers in the cluster. The invention provides the appearance of a shared memory space between two or more processes thereby potentially reducing disk latency or eliminating redundant computation associated with conventional server clusters. The DSM abstraction may be dynamically alterable such that selectable groups of processes executing on the cluster share a common address space temporarily. The shared memory spaces may be determined empirically or servers may subscribe to a group reactively in response to client requests. Multiple groups may exist simultaneously and a single server may belong to more than one group. The types of objects to which the abstraction is applied may be restricted. Shared memory may be restricted, for example, to read-only objects to alleviate consistency considerations.
摘要:
In information storage systems in which data retrieval requires movement of at least one physical element, a measurable amount of time is required to reposition that physical element in response to each data write or read request. After selecting one or more data requests for dispatch based solely on an approaching or past due time deadline, additional requests are identified for data to be read or written to locations which are in close proximity to previously scheduled requests, previously selected additional requests, or the present position of the moveable physical element, obviating the need to expend the full amount of time required to accelerate the physical element and then decelerate the physical element to position it over the desired area within the information storage system. In this manner, data may be transferred to or retrieved from an information storage system more efficiently with less expenditure of time.
摘要:
A method for a computation agent to acquire a queue lock in a multiprocessor system that prevents deadlock between the computation agent and external interrupts. The method provides for the computation agent to join a queue to acquire a lock. Next, upon receiving ownership of the lock, the computation agent raises its priority level to a higher second priority level. In response to a receipt of an external interrupt having a higher priority level occurring before the computation agent has raised its priority level to the second higher priority level, the computation agent relinquishes ownership of the lock. Subsequent to raising its priority level to the second higher priority level, the computation agent determines if it still has ownership of the lock. If the computation agent determines that it has not acquired possession of the lock after raising its priority level, the computation agent rejoins the queue to reacquire the lock. In one embodiment of the present invention, the computation agent's priority level is restored to its original, i.e., first priority level, when it rejoins the queue to reacquire the lock.