Citation Context

...ofitable. Researchers have looked at the problem of exploiting parallelism through the use of data and event synchronization, where post and wait statements are used to synchronize between data items =-=[26]-=- or loop iterations [20]. In comparison, we do not attempt to extract fine-grain parallelism through synchronization. Instead, we try to expand the granularity of loop-level parallelism found by paral...

Citation Context

...ar88, ZY87, SY89]. Only a few works discuss compiler algorithms. The work in [MP87] deals mainly with single parallel loops. Recently, Tang, Yew and Zhu present an algorithm based on special counters =-=[TYZ90]-=-. Their paper considers a subset of our loop bounds and subscripts, but does not address if statements. They assume that each data word is associated with a counter, both of which may be accessed atom...

Citation Context

...ng methods for eliminating barrier synchronization, researchers investigated efficient use of data and event synchronization, where post and wait statements are used to synchronize between data items =-=[26]-=- or loop iterations [16]. Researchers compiling for fine-grain data-parallel languages sought to eliminate barriers following each expression evaluation [10, 21, 22]. Simple data dependence analysis c...

Citation Context

...ng methods for eliminating barrier synchronization, researchers investigated efficient use of data and event synchronization, where post and wait statements are used to synchronize between data items =-=[24]-=- or loop iterations [16]. Researchers compiling for finegrain data-parallel languages sought to eliminate barriers following each expression evaluation [10, 20, 21]. Simple data dependence analysis ca...

Citation Context

...ng methods for eliminating barrier synchronization, researchers investigated efficient use of data and event synchronization, where post and wait statements are used to synchronize between data items =-=[23]-=- or loop iterations [16]. Researchers compiling for fine-grain data-parallel languages sought to eliminate barriers following each expression evaluation [10, 20, 21]. Simple data dependence analysis c...

Citation Context

... instructions into DOACROSS loops [21]. Another example computes the access orders of each data element at compile time and then inserts data synchronization instructions to enforce those dependences =-=[32]-=-. Other research has taken aim at distributed shared-memory multiprocessors [31]. It uses di160 1 2 3 4 5 6 Core1 1.1 4.1 2.1 3.1 1.3 4.3 Core2 1.2 4.2 2.2 3.2 0 1 2 3 4 5 6 7 Core1 Core2 1.1 4.1 2.1...