G. Barrett. occam 3 Reference Manual. INMOS Ltd, draft edition, 1992.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....for holding the mobilespace of the recursive instance. This is initialised to MOSTNEGINT on PROC entry, along with the mobiles already required, and allocated at the point of the recursive instance. 5. 6 RESULT Parameters and Abbreviations Result parameters were suggested by Barrett for occam 3 [21]. These are reference parameters which explicitly return results, as opposed to standard reference parameters, whose input output behaviour is unknown. The undefined usage checker [22] expects reference parameters to have defined arguments before the call and to leave defined data before ....

Geoff Barrett. occam 3 Reference Manual. Technical report, Inmos Limited, March 1992. Available at: http://wotug.ukc.ac.uk/parallel/occam/documentation/.

....THING yes yes yes Table 1: Summary of formal vs. actual parameter combinations 7 Undefined Usage Checks A variable whose current data value (r value) is undefined should never be used in that state. This is the initial state of occam variables unless explicitly declared with INITIAL values [21]. We now have mobile assignment and communication that set their source variables back to this undefined state. Checking against the use of uninitialised variables has been an omission from previous occam compilers. In the process of this research, we have added an undefined usage checker to the ....

....exactly once for replicated IFs and ALTs, and possibly more times for a replicated PAR. Value parameters must not be passed undefined or unsure variables. Reference parameters can be, however, since the invoked process may use it for returning a result. The explicit RESULT qualifier of occam3[21] would raise quality here, as we could then enforce all non RESULT reference arguments to be defined. Additionally, we could ensure that RESULT parameters inside a PROC were left in a defined state when the PROC returns to the caller. A more difficult issue is tracking the state of individual ....

Geoff Barrett. occam 3 Reference Manual. Technical report, Inmos Limited, March 1992. Available at: http://wotug.ukc.ac.uk/parallel/occam/documentation/.

.... read.line switch (tcp) read.line switch (tcp) file.get (tcp) tcp) profile stats.get stats (tcp) cgi.get (tcp) tcp) cgi.get (tcp) file.get acceptor Figure 11: Network diagram for the occam web server After initialisation, all processes will be blocked on either channels or occam3[8] style shared channels (implemented via the semaphore abstract type [9] with the exception of the acceptor process, which will be blocked in the Linux kernel on a socket accept. The acceptor process consists of a loop, which accepts an incoming connection, time stamps it, then passes it on ....

Geoff Barrett. occam 3 Reference Manual. Technical report, Inmos Limited, March 1992. Available at: http://wotug.ukc.ac.uk/parallel/occam/documentation/.

....increasing the efficiency of the expression a(i) b(i 1) The ALIGN statement can be performed only once on an array at the beginning of execution. Arrays can be declared as DYNAMIC in which case REALIGN statements can be performed on them at any point in the execution. 3. 6 OCCAM 3 OCCAM 3 [ Barrett, 1992 ] is a message passing language based on the CSP model of computation. OCCAM 3 adds a number of extensions to the OCCAM 2 language [ Inmos, 1988 ] including: ffl Initialisation of variables. ffl User defined data types. ffl Records and unions. ffl Shared channels. ffl Libraries and modules. ....

....or static. A dynamic program is where processes can be created at any point in the execution, as opposed to static where the number of processes is fixed once execution starts. Since Opal is meant to be a language that can be efficiently implemented I chose to use the static mode as in OCCAM 3 [ Barrett, 1992 ] This simplifies the design of the compiler and improves runtime performance as processes are not being continuously created and killed, which would also complicate the bulk synchronisation. The BSP model has already been chosen as the basis for Opal. Gerbossiotis and Valiant, 1992 ] suggest ....

[Article contains additional citation context not shown here]

Geoff Barrett. Occam3 Reference Manual. Draft Copy, 1992.

....obviate, gains in processor speed. An indication of the problem is that multiple caches, interleaved memory and decoupled architectures are all aimed at ameliorating memory latency. Therefore, shared memory was added by us to the CSP model. Support for shared memory has also been added in occam 3 [ 28 ] intended for the T9000 series transputer. Similarly, the need to relax the occam specification has been recognized in [ 29 ] where semaphores, resources, events, and buckets are alternative synchronization mechanisms to the channel. For our template design, counting semaphores had several ....

G. Barrett. occam3 reference manual. Technical report, Inmos Ltd., 1000 Aztec Way, Bristol, UK, 1992. 26

.... for practical parallel computing (i.e. processes, structured and synchronised message passing, explicit non determinism, priorities, real time and hardware interfacing) In the period 1990 92 and in consultation with the occam user community (both academic and industrial) INMOS defined occam3[1] an upwards compatible extension to occam2. Apart from introducing a full set of user definable data structures (e.g. records and tagged unions) occam3 built on the experiences of occam2 users to provide explicit high level support for client server systems, remote procedure calls, ....

Geoff Barrett. occam3 reference manual. Technical report, INMOS Limited, Bristol, BS12 4SQ, ENGLAND, March 1992.

....compute resource. On the software side, we have paradigms such as data parallel and message passing. We have functional methods for process decomposition, client server principles, barrier synchronisation, farming and remote procedure cells. We have structured messaging systems (e.g. occam[1] and Linda[2] and unstructured ones (e.g. PVM[3] At present, the choice of parallel hardware platform we make (or is made for us) severely constrains our choice of parallel software paradigm. Our views on the statements at the beginning of this section will be coloured by the parallel ....

....usage checking rules, that already satisfies the good behaviour rules for parallel software. However, occam was designed primarily for embedded applications and currently lacks many features that would make it more acceptable for general purpose computing although this is being addressed [1, 12, 13]. Professor Hoare gave the following definition[14] for a generic transputer: transputer, n: a computing device in which communications are primitive operations (like addition) and in which parallelism is a primitive construct (like the conditional) This definition is clearly satisfied by the ....

Geoff Barrett. occam 3 reference manual (draft). INMOS Ltd, Bristol, BS12 4SQ, England, March 1992. See URL http://www.hensa.ac.uk/parallel/ occam/documentation/manual3.ps.Z.

....Message Passing Synchronous message passing systems allow parallel programs to be written as collections of sequential processes composed in parallel, and interconnected with synchronous communication channels. These systems include complete programming languages such as LCS [7] and Occam [6], and libraries such as PVM [14] and P4[8] Synchronous message passing systems are the simplest form of parallel programming for workstations. When implemented as libraries, they require no changes to the sequential languages that use them, and when implemented as languages, they require very few ....

....programming model, and either weakened or generalized them in some way. We have produced a new model, which should allow casual applications programmers to write parallel programs on available hardware while Selection: 1, 2, 3, 4, 5]selectfx : x mod 2 6= 0g = 1, 3, 5] Mapping: 1, 2, 3, 4] [5, 6, 7, 8] = 6, 8, 10, 12] Reduction: reduce [1, 2, 3, 4] 10 Figure 1: Examples of Data Parallel Operations (a) Conventional DP structure (b) A section from a composite tree structure Node Host Machine Parallel Memory Node Node Node Node Children Collection Figure 2: Comparison of data parallel ....

[Article contains additional citation context not shown here]

Geoff Barrett. Occam 3 Reference Manual. INMOS, 1992.

No context found.

G. Barrett. occam 3 Reference Manual. INMOS Ltd, draft edition, 1992.

No context found.

Geoff Barrett. occam 3 Reference Manual. Technical report, Inmos Limited, March 1992. Available at: http://wotug.ukc.ac.uk/parallel/occam/documentation/.

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