Active Messages Using Selective Interrupts without Polling
نویسندگان
چکیده
Active Messages form a user-level communication mechanism that allows a message to extract itself from the network upon arrival at its destination. Delivery of such a message can be realized via an interrupt-driven or a polling-based approach. This paper proposes a new mechanism, Active Messages Using Selective Interrupts Without Polling (AM_SIWOP). It improves upon Active Messages by reducing the number of interrupt dispatch and deeply overlapping communication with computation. The AM_SIWOP uses dual hardware switched receive buffers and an adaptive timer in the network interface. Instead of each message causing an interrupt, a group of messages in a buffer as a whole will cause a single interrupt when the buffer becomes full or when a certain period of time has elapsed for messages waiting in a buffer. The timer is adjustable according to the communication traffic. The simulation results showed that the AM_SIWOP significantly improved the computational efficiency of Active Messages. In an AM_SIWOP with a pair of small buffers, messages would be delayed slightly when messages were arriving quickly. Due to the adaptive timer, communication latency of a message was bounded in light communication traffic. Since message buffering deepens the transmission pipeline of active messages, the AM_SIWOP increases the bandwidth of user-level communication.
منابع مشابه
Polling Eeciently on Stock Hardware
Two strategies for supporting asynchronous interrupts are: the use of the processor's hardware interrupt system and the use of polling. The advantages of polling include: portabil-ity, simplicity, and low cost for handling interrupts. Unfortunately , polling has an overhead for the explicit interrupt checks inserted in the code. This paper describes balanced polling, a method for placing the in...
متن کاملHIP: Hybrid Interrupt-Pol l ing for the Network Interface
The standard way to notify the processor of a network event, such as the arrival or transmission of a packet, is through interrupts. Interrupts are more effective than polling, in terms of the per packet send/receive latency. Interrupts, however, incur a high overhead both during and after the interrupt handling, because modern superscalar processors use long pipelines, out-of-order and specula...
متن کاملReducing Waiting Costs in User-Level Communication
This paper describes a mechanism for reducing the cost of waiting for messages in architectures that allow userlevel communication libraries. We reduce waiting costs in two ways: by reducing the cost of servicing interrupts, and by carefully controlling when the system uses interrupts and when it uses polling. We have implemented our mechanism on the SHRIMP multicomputer and integrated it with ...
متن کاملReducing Waiting Costs in User - Level
This paper describes a mechanism for reducing the cost of waiting for messages in architectures that allow user-level communication libraries. We reduce waiting costs in two ways: by reducing the cost of servicing interrupts, and by carefully controlling when the system uses interrupts and when it uses polling. We have implemented our mechanism on the SHRIMP multicomputer and integrated it with...
متن کاملA Communication Library Using Active Messages to Improve Performance of PVM
We present a communication library to improve performance of PVM. The new library introduces communication primitives based on Active Messages. We propose a hybrid scheme that includes a signal driven message notiication scheme plus controlled polling. The new communication library is tested along with the normal PVM library to assess the improvement in performance.
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 1994