Thomas E. Anderson. The performance of spin lock alternatives for shared-memory multiprocessors. IEEE Transactions on Parallel and Distributed Systems, 1(1):6--16, January 1990.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....timeout in a queue based lock. 1Introduction Spin locks are widely used for mutual exclusion on shared memory multiprocessors. Traditional test and set based spin locks are vulnerable to memory and interconnect contention, and do not scale well to large machines. Queue based spin locks [2, 5, 7, 13, 15] avoid contention by arranging for every waiting thread to spin on a separate, local flag in memory. Over the past ten years queuebased locks have been incorporated into a variety of academic and commercial operating systems, This work was supported in part by NSF grants numbers EIA 0080124 and ....

T. E. Anderson. The Performance of Spin Lock Alternatives for Shared-Memory Multiprocessors. IEEE Transactions on Parallel and Distributed Systems,1(1):6--16, January 1990.

....an internal program error, an action taken by the operating system or the failure of a processor. We focus on halting failures that occur as a process or thread interacts with a low level concurrency control mechanism known as a spin lock. A variety of spin lock mechanisms have been proposed (see [23, 5, 32]) we present novel schemes to address halting failures for many of these mechanisms. The result of our work is the production of recoverable spin lock mechanisms, where a locking mechanism is considered recoverable if arbitrary halting failures of processes which acquire or attempt to acquire the ....

....the semaphore at a time. Spin locking implementations of binary semaphores have been extensively studied in the parallel computing and operating system communities. We point the reader to operating systems texts [68, 76] a survey by Dinning [23] and surveys with performance analysis by Anderson [5] and by Graunke and Thakkar [32] Spin locks are typically two orders of magnitude faster than operating system semaphores, and are generally acknowledged to be an important factor in the success of multi processor 6 shared memory systems [32] To present concretely the problems we are solving, ....

[Article contains additional citation context not shown here]

T. E. Anderson. The performance of spin lock alternatives for shared-memory multiprocessors. IEEE Transactions on Parallel and Distributed Systems, 1(1):6-- 16, January 1990.

....that supports ne grain parallel computations. Bus based systems such as the Sequent Balance and Encore Multimax are possible examples. These systems encourage the use of shared memory to coordinate the activities of processes. When one process needs to wait for another, busy waiting is often used [1]. If more than one process is mapped to each processor, the busy waiting loop may include an instruction to yield the processor to another ready process. This prevents computation cycles from being wasted without requiring extensive bookkeeping to keep track of which process is waiting for what ....

T. E. Anderson, The performance of spin lock alternatives for shared-memory multiprocessors, IEEE Trans. Parallel & Distributed Syst. 1 (1990) 6-16.

.... Furthermore, the burst of refill traffic whenever a lock is released can be reduced by using the Ethernet style exponential backoff algorithm in which, after a failure to obtain the lock, a requester waits for successively longer periods of time before trying to issue another lock operation [1, 18]. The delay between tas attempts should not be too long; otherwise, processors might remain idle even when the lock becomes free. This is the idea behind the TATAS EXP lock, and one typical implementation is shown below. typedef volatile unsigned long tatas lock; 01 void ....

....to the flag; 3) the thread spins until the prev flag memory location, a pointer which was returned by the swap, contains the value FREE. The release function of the queue based locks writes a FREE value to the flag location. Numerous variations of software queuing lock implementations are known [1, 5, 9, 16, 18, 22]. A unique feature of software queuing locks found in many implementations is an explicit starvation avoidance and maximal fairness; in other words, first come, first served order of lock acquire requests is guaranteed. In addition, software queue based locks provide reasonable latency in the ....

T. E. Anderson. The Performance of Spin Lock Alternatives for Shared-Memory Multiprocessors. IEEE Transactions on Parallel and Distributed Systems, 1(1):6--16, Jan. 1990.

....the test set based locking primitive to a hardware queue based lock using modest extensions to the cache coherence protocol. Unlike QOLB, IQOLB does not require any instruction set support nor does it require any software changes. Software queue based locking schemes were proposed by Anderson [9, 10] and Graunke and Thakkar [55] Mellor Crummey and Scott proposed MCS, an improvement to Anderson s algorithm. The MCS scheme [120, 121] is a software based queued lock scheme. MCS adds requesters for a held lock into a software queue at the time of the request, using atomic operations such as swap ....

....behavior. In spite of extensive research, OCC techniques have not been popular because of key limitations [124] Lock based synchronization. Lock based synchronization has been extensively studied in literature. These techniques attempt to optimize the lock and data transfer operations [10, 50, 81, 120, 141]. The techniques are not lock free. These techniques suffer from locking overhead and serialization due to lock acquisitions. Martnez and Torrellas introduced Speculative Locks, allowing speculative threads to bypass a held lock and enter a critical section [117] At any time the lock is always ....

Thomas E. Anderson. The Performance of Spin Lock Alternatives for Shared-Memory Multiprocessors. IEEE Transactions on Parallel and Distributed Systems, 1(1):6--16, January 1990.

....testing if the event has occurred. On both a superscalar and multiprocessors, spinning can waste processor resources due to the opportunity cost of not context switching to another thread. Researchers have measured and developed techniques addressing the costs of spinning on these processors [37, 4, 40, 42, 5, 50, 10, 36, 90]; we do not investigate them here. Spinning can exact a larger performance cost on SMT, because all threads share pipeline resources. We term pipeline resources as all resources shared between contexts that are necessary to execute instructions that is, all shared resources except the caches ....

....systems has been the subject of many studies. Spinning degrades performance by competing for inter processor shared resources such as the memory bus. Many techniques have been explored to reduce spinning s cost by reducing shared resource consumption; more recent investigations include [5, 50, 84, 36, 10]. For example, 50] proposed distributed locks. Distributed locks allow each processor to spin on a local flag rather than competing for a common flag. A processor releases a lock by clearing the flag of another processor, passing the lock to it. On a bus based multiprocessor, no bus transactions ....

ANDERSON, T. E. The performance of spin lock alternatives for sharedmemory multiprocessors. IEEE Transactions on Parallel and Distributed Systems 1, 1 (January 1990).

....to get around this uncomfortable situation. Instead of just sending a request message when processor p initiates the inc operation, processor p waits some time. Related ideas for solving the congestion problem are known for some time (e.g. exponential back off [GGMM88] or queue lock mechanisms [And90, MS91] Note that these solutions may help against congestion but do not have an improved operation time. When there are other initiations of the inc operation in the meantime, processor p sends a bulk request message that does not request 1 value for processor p but z values instead. Processor ....

Thomas Anderson. The performance of spin lock alternatives for shared memory multiprocessors. IEEE Transactions on Parallel and Distributed Systems, 1(1):6--16, January 1990.

....TLR does not have a serialized validation phase and exploits hardware techniques to provide transactional behavior. Lock based synchronization. Lock based synchronization has been extensively studied in literature. These techniques attempt to optimize the lock and data transfer operations [9, 3, 26, 16, 31]. The techniques are not lock free. These techniques suffer from locking overhead and serialization due to lock acquisitions. Martnez and Torrellas introduced Speculative Locks, allowing speculative threads to bypass a held lock and enter a critical section [24] At any time the lock is always ....

T. E. Anderson. The performance of spin lock alternatives for shared-memory multiprocessors. IEEE Transactions on Parallel and Distributed Systems, 1(1):6--16, Jan. 1990.

No context found.

T. Anderson. The performance of spin lock alternatives for shared-memory multiprocessors. IEEE Transactions on Parallel and Distributed Systems, 1(1):6-16, January 1990.

No context found.

T. Anderson. The performance of spin lock alternatives for shared-memory multiprocessors. IEEE Transactions on Parallel and Distributed Systems, 1(1):6-16, January 1990. 38

No context found.

T. Anderson, "The Performance of Spin Lock Alternatives for Shared-Memory Multiprocessors ", IEEE Transactions on Parallel and Distributed Systems, 1(1), 1990, pp. 6-16.

No context found.

Thomas E. Anderson. The performance of spin lock alternatives for shared-memory multiprocessors. IEEE Transactions on Parallel and Distributed Systems, 1(1):6--16, January 1990.

No context found.

T. E. Anderson. The Performance of Spin Lock Alternatives for Shared-Memory Multiprocessors. IEEE Transactions on Parallel and Distributed Systems, PDS-1(1):6--16, January 1990.

No context found.

Thomas E. Anderson. The performance of spin lock alternatives for sharedmemory multiprocessors. IEEE Transactions on Parallel and Distributed Systems, 1(1):6--16, January 1990.

No context found.

T. E. Anderson. The Performance of Spin Lock Alternatives for Shared-Memory Multiprocessors. IEEE Trans. on Parallel and Distributed Systems, 1(1):6--16, Jan. 1990.

No context found.

Anderson, T. The Performance of Spin Lock Alternatives for Shared-Memory Multiprocessors. Technical Report 89-04-03, University of Washington, 1989.

No context found.

T. E. Anderson. The performance of spin lock alternatives for shared-memory multiprocessors. IEEE Transactions on Parallel and Distributed Systems, 1(1):6--16, Jan. 1990.

No context found.

T. Anderson. The performance of spin lock alternatives for shared-memory multiprocessors. IEEE Trans. on Parallel and Distributed Systems, 1(1):6--16, Jan. 1990.

No context found.

T.E. Anderson, "The Performance of Spin Lock Alternatives for Shared-Memory Multiprocessors ", IEEE Transactions Parallel and Distributed Systems, Vol. 1, No. 1, pp. 6-16, January 1990.

No context found.

Anderson, T., "The performance of spin lock alternatives for shared-memory multiprocessors," IEEE Transactions on Parallel and Distributed Systems 1, Vol. 1, pp. 6--16, January 1990.

No context found.

T. Anderson. The performance of spin lock alternatives for shared-memory multiprocessors. IEEE Transactions on Parallel and Distributed Systems, 1(1):6--16, January 1990.

No context found.

T. E. Anderson. The performance of spin-lock alternatives for sharedmemory multiprocessors. IEEE Transactions on Parallel and Distributed Systems 1(1):6--16, January 1990.

No context found.

T. Anderson, "The performance of spin lock alternatives for shared-memory multiprocessors," IEEE Transactions on Parallel and Distributed Systems 1, vol. 1, no. 1, pp. 6--16, January 1990.

No context found.

T. Anderson. The performance of spin lock alternatives for shared-memory multiprocessors. IEEE Transactions on Parallel and Distributed Systems, 1(1):6-16, January 1990.

No context found.

T.E. Anderson. The Performance of Spin Lock Alterna- tives for Shared-Memory Multiprocessors. IEEE Trans- actions on Parallel and Distributed Systems, 1(1):6-16, January 1990.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC