D. V. James. Combinable locks: NullSwap efficiently delays cache-line transfers. In Sixth International Workshop on SCI-based Low-cost/Highperformance Computing, September 1996. Home/Search Document Not in Database Summary Related Articles Check |
This paper is cited in the following contexts:
Scintilla: Cluster Computing with SCI - Ibel, Schauser, Scheiman, Schmitt (Correct)
....is to build a queue in user memory using hardware synchronization primitives. Since many processes can concurrently enqueue data, we need some sort of atomic read write operation, such as test and set, to prevent write conflicts [Oma97] The SCI standard provides support for lock transactions [Jam96] 1 Given such a lock transaction, our scheme for sending a message could be: lock the remote queue, read the destination write pointer, increment the pointer by the message size, check for overflow, write 1 Unfortunately, lock transactions are currently not supported by our SBus 2 network ....
D. V. James. Combinable locks: NullSwap efficiently delays cache-line transfers. In Sixth International Workshop on SCI-based Low-cost/Highperformance Computing, September 1996.
High-Performance Cluster Computing Using SCI - Ibel, Schauser, Scheiman, Weis (1997) (14 citations) (Correct)
....is to build a queue in user memory using hardware synchronization primitives. Since many processes can concurrently enqueue data, we need some sort of atomic read write operation, such as test and set, to prevent write conflicts [Oma97] The SCI standard provides support for lock transactions [Jam96] 1 Given such a lock transaction, our scheme for sending a message could be: lock the remote queue, read the destination write pointer, increment the pointer by the message size, check for overflow, write the 1 Unfortunately, lock transactions are currently not supported by our SBus 2 ....
D. V. James. Combinable locks: NullSwap efficiently delays cache-line transfers. In Sixth International Workshop on SCI-based Low-cost/Highperformance Computing, September 1996.
High-Performance Cluster Computing Using Scalable.. - Ibel, Schauser.. (1997) (2 citations) (Correct)
....way to implement the queue like functionality is to build a queue in user memory. Since many processes can concurrently enqueue data, we need some sort of atomic read write operation, such as test and set, to prevent write conflicts [Oma97] The SCI standard provides support for lock transactions [Jam96] 1 Given such a lock transaction, our scheme for sending a message could be: lock the remote queue, read the destination write pointer, increment the pointer by the message size, check for overflow, write the pointer back to the receiver, put the message in the queue, 1 Unfortunately, lock ....
D. V. James. Combinable locks: NullSwap efficiently delays cache-line transfers. In Sixth International Workshop on SCI-based Low-cost/Highperformance Computing, September 1996.
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