~marionnet-drivers/ocamlbricks/trunk : changes

~marionnet-drivers/ocamlbricks/trunk » Changes from revision 561

From Revision 561 to 542

Rev	Summary	Authors	Date
561	In module Channel, added tools (`reactive_image', `reactive_... In module Channel, added tools (`reactive_image', `reactive_minion') simply derived from the fundamental ones (`map', `product', `cover') to show the ability of channels to program in a reactive style, remaining compliant with the OCaml garbage collection.	Jean-Vincent Loddo	3 years ago
560	Whole library adapted to OCaml versions 4.04.2, 4.06.1, 4.07... Whole library adapted to OCaml versions 4.04.2, 4.06.1, 4.07.1, 4.08.1, 4.09.1, and 4.10.0. On the other hand, we are forced to abandon versions older than 4.04.2 because of the standard module Ephemeron, appeared with version 4.03, which is now used too extensively to preserve backward compatibility.	Jean-Vincent Loddo	3 years ago
559	Minor changes in several modules Minor changes in several modules	Jean-Vincent Loddo	3 years ago
558	Minor but several changes to render Ocamlbrics compatible wi... Minor but several changes to render Ocamlbrics compatible with OCaml 4.07. Basically, the main solved problem concerns the separation between the types `bytes' and `string', because since OCaml 4.06, the two types are not anymore synonyms.	Jean-Vincent Loddo	3 years ago
557	Added module Collector to manage message collection, i.e. to... Added module Collector to manage message collection, i.e. to manage values put on the channel by several sources (threads) asynchronously. Several (priority) "maintainers" depose a value on the channel and, from time to time, a (non priority) "writer" flush the container. In case of conflict with a flusher, maintainers have time to put everything they want before the flusher can act.	Jean-Vincent Loddo	3 years ago
556	Minor changes (modules' renaming) about algebraic channels Minor changes (modules' renaming) about algebraic channels	Jean-Vincent Loddo	3 years ago
555	Created directory CHANNEL/ to group all sources related to a... Created directory CHANNEL/ to group all sources related to algebraic channels. Currently, all files contained in CHANNEL/ have been moved from the directory CORTEX. This last directory is not far from being obsolete and useless.	Jean-Vincent Loddo	3 years ago
554	Sixth satisfactory version (until the next) of algebraic cha... Sixth satisfactory version (until the next) of algebraic channels, with support for ephemeral (not persistent) structures. Stabilizations and notifications are performed only for writers and maintainers. During the access protocol to a channel, if the channel contains a persistent structure (by default) and the internal resistance has not really changed we do nothing (no stabilization, neither notifications to super-structures). The same (micro-)optimization cannot be applied to ephemeral (not persistent) structure, because for this kind of structures we cannot easily detect changes provoked by the `update' function (the 3th and 4th arguments of the body are the same object). Finally the more accurate tool to access a channel has the following signature: val access : ?level:int -> ?ephemeral:unit -> ?enter:('a -> bool) -> ?notify:int list -> ?update:('a -> 'a) -> ?leave:('a -> bool) -> 'a t -> ('a FTCC.t -> 'a details -> 'a -> 'a -> 'b) -> (exn, 'b) Either.t	Jean-Vincent Loddo	3 years ago
553	Fifth satisfactory version of algebraic channels, with tools... Fifth satisfactory version of algebraic channels, with tools to better control the non-determinism (+). Now, when a method is applied to a disjunction (many "files" stored in the "directory"), we are able to acquire again exactly the same structure acquired the first time, and continue the work on it (module Channel.Control).	Jean-Vincent Loddo	3 years ago
552	Added Lazynel, the module of thread-safe deferred computatio... Added Lazynel, the module of thread-safe deferred computations implemented as algebraic channels (to be composable). Also added the module TS_lazy of (T)hread-(S)afe deferred computations not implemented by algebraic channels but with simple mutexes. Hence, the latter cannot be composed neither may be the subject of a waiting method (as on the contrary is possible with Lazynel.wait).	Jean-Vincent Loddo	3 years ago
551	Added a `fold' function to algebraic channels, similar to th... Added a `fold' function to algebraic channels, similar to the homonymous tool for lists, arrays and many other structures, but with the specificity of acquiring and releasing the good locks during the visit. This tool is useful when the channel is a disjunction, in order to compute something during the visit of all stored values. The access provided by `fold' (and the derived logical tools as `exists' and `for_all') is performed as "reader" (level 0) and without guards (no waiting may happen).	Jean-Vincent Loddo	3 years ago
550	Added module TS_memo, of (T)hread-(S)afe weak memoization, w... Added module TS_memo, of (T)hread-(S)afe weak memoization, with or without projections (for uncurried functions with several arguments). This implementation is based on algebraic channels and, precisely, on hublets (Hublet.t). Indeed, hublets allow the thread computing the image y of an argument x to broadcast the answer y, when it will be ready, to all other threads waiting on the same question (i.e. the question of having the image y of x by the function which is memoized).	Jean-Vincent Loddo	3 years ago
549	Added module Hublet, of named, 1-to-N, synchronous/asynchron... Added module Hublet, of named, 1-to-N, synchronous/asynchronous, 1-size value-passing (algebraic) channels. The purpose of this structure is to allow multicast communications, being at the same time composable as member of the algebra of channels.	Jean-Vincent Loddo	3 years ago
548	Added module Milner, of named, 1-to-1, synchronous/asynchron... Added module Milner, of named, 1-to-1, synchronous/asynchronous, 1-size value-passing (algebraic) channels. Being first-class citizen, they allow the communication of channels like in the pi-calculus. Strictly speaking, (1) Milner's channels as initially defined in CCS are not value-passing (here they would be of type unit), and (2) both in CCS and pi-calculus, the `send' method is supposed to be executed in the synchronous mode, while in this module `send' is executed by default in the asynchronous mode. Hence, to force the synchronous mode we have to precise the option ~sync:(). Note that Milner's theoretical models were initially synchronous in order to have a basic definition of program equivalence (bisimilarity), but were later extended (by Milner and others) to cover the asynchronous case.	Jean-Vincent Loddo	3 years ago
547	Forth satisfactory version of algebraic channels: just cosme... Forth satisfactory version of algebraic channels: just cosmetic changes and (a lot of) renaming.	Jean-Vincent Loddo	3 years ago
546	Third satisfactory (better optimized) version of algebraic c... Third satisfactory (better optimized) version of algebraic channels. Also, grouped several (but very basic) tests into the source tests/test_channel.ml.	Jean-Vincent Loddo	3 years ago
545	In module Table, added the possibility to change the definit... In module Table, added the possibility to change the definition of the equality even when the argument ?identifier is provided. In this way, when both are provided, the identifier affects no more than the definition of hashing (i.e. x -> Hashtbl.hash (identifier x)).	Jean-Vincent Loddo	3 years ago
544	In module Extreme_sharing, added optional parameter ?trace_f... In module Extreme_sharing, added optional parameter ?trace_faults to function `weakly_memoize'. Only for debugging purposes.	Jean-Vincent Loddo	3 years ago
543	In module Directory_tree, added the possibility to redefine ... In module Directory_tree, added the possibility to redefine operations on "files" making products (times) or extracting projections (split, fst, snd). By default these values are taken by the module (File: ZIPPED_FUNCTOR) argument of the functor MAKE. So, the new interface of multiplicative tools is: val times : ?file_zip :('x File.t * 'y File.t -> ('x * 'y) File.t) -> (*) ('a,'x) t -> ('a,'y) t -> ('a, 'x 'y) t val split : ?file_unzip :(('x * 'y) File.t -> 'x File.t * 'y File.t) -> (*) ('a, 'x 'y) t -> ('a,'x) t * ('a,'y) t val fst : ?file_fst :(('x * 'y) File.t -> 'x File.t) -> (*) ('a, 'x 'y) t -> ('a,'x) t val snd : ?file_snd :(('x * 'y) File.t -> 'y File.t) -> (*) ('a, 'x 'y) t -> ('a,'y) t	Jean-Vincent Loddo	3 years ago
542	Second satisfactory (quite optimized) version of algebraic c... Second satisfactory (quite optimized) version of algebraic channels	Jean-Vincent Loddo	3 years ago

Older »