src/Pure/Concurrent/future.ML
Tue, 14 May 2013 20:46:09 +0200 wenzelm more uniform Markup.print_real;
Tue, 09 Apr 2013 15:59:02 +0200 wenzelm clarified protocol_message undefinedness;
Mon, 08 Apr 2013 15:44:09 +0200 wenzelm discontinued odd magic number, which was once used for performance measurements;
Tue, 05 Mar 2013 11:37:01 +0100 wenzelm removed unused Future.flat, while leaving its influence of Future.map (see bcd6b1aa4db5);
Sun, 03 Mar 2013 17:34:42 +0100 wenzelm more uniform Future.map: always internalize failure;
Tue, 26 Feb 2013 13:38:34 +0100 wenzelm disallow shutdown from worker, which would lead to deadlock since the scheduler cannot terminate;
Tue, 26 Feb 2013 13:05:48 +0100 wenzelm signal work_available should be sufficient to initiate daisy-chained workers, and lead to separate broadcast work_finished eventually -- NB: broadcasting all worker threads tends to burn parallel CPU cycles;
Tue, 26 Feb 2013 12:50:52 +0100 wenzelm less eventful shutdown: merely wait for scheduler to terminate;
Tue, 26 Feb 2013 12:46:47 +0100 wenzelm tuned messages;
Thu, 24 Jan 2013 17:31:12 +0100 wenzelm report status more frequently on demand;
Sat, 19 Jan 2013 00:00:29 +0100 wenzelm tuned signature;
Fri, 18 Jan 2013 17:51:50 +0100 wenzelm more systematic task statistics;
Fri, 18 Jan 2013 16:20:09 +0100 wenzelm added "tasks_proof" statistics, via slighly odd global reference Future.forked_proofs (NB: Future.report_status is intertwined with scheduler thread);
Thu, 17 Jan 2013 12:04:52 +0100 wenzelm clarified Future.error_msg: slightly more robust id check, actually suppress displaced messages;
Wed, 16 Jan 2013 21:39:43 +0100 wenzelm proper runtime position (cf. fe4714886d92 and Toplevel.error_msg) -- to make error messages actually appear in the document;
Wed, 16 Jan 2013 20:41:29 +0100 wenzelm identify future results more carefully, to avoid odd duplication of error messages, notably from forked goals;
Wed, 16 Jan 2013 16:26:36 +0100 wenzelm more explicit treatment of (optional) exception properties, notably for "serial" -- avoid conflict with startPosition = offset;
Wed, 16 Jan 2013 11:25:26 +0100 wenzelm tuned signature;
Wed, 02 Jan 2013 13:50:59 +0100 wenzelm inline ML statistics into build log;
Tue, 01 Jan 2013 21:55:46 +0100 wenzelm more robust report_status: tolerate ML_statistics even if ignored right now, e.g. in batch build;
Fri, 07 Dec 2012 23:11:01 +0100 wenzelm final report_status within SYNCHRONIZED part of scheduler loop: required for sanity of data;
Thu, 29 Nov 2012 10:45:25 +0100 wenzelm more uniform ML statistics;
Wed, 28 Nov 2012 17:18:53 +0100 wenzelm some support for ML runtime statistics;
Fri, 19 Oct 2012 12:08:13 +0200 wenzelm clarified Future.map (again): finished value is mapped in-place, which saves task structures and changes error behaviour slightly (tolerance against canceled group of old value etc.);
Thu, 18 Oct 2012 13:53:02 +0200 wenzelm more uniform group for map_future, which is relevant for cancel in worker_task vs. future_job -- prefer peer group despite 81d03a29980c;
Thu, 18 Oct 2012 12:47:30 +0200 wenzelm tuned comment;
Thu, 18 Oct 2012 12:00:27 +0200 wenzelm more official Future.terminate;
Wed, 17 Oct 2012 21:18:32 +0200 wenzelm more robust cancel_now: avoid shooting yourself in the foot;
Thu, 30 Aug 2012 15:26:37 +0200 wenzelm refined status of forked goals;
Wed, 11 Apr 2012 13:49:09 +0200 wenzelm more robust Future.fulfill wrt. duplicate assignment and interrupt;
Wed, 11 Apr 2012 12:15:56 +0200 wenzelm always signal after cancel_group: passive tasks may have become active;
Mon, 09 Apr 2012 17:22:23 +0200 wenzelm simplified Future.cancel/cancel_group (again) -- running threads only;
Mon, 28 Nov 2011 22:05:32 +0100 wenzelm separate module for concrete Isabelle markup;
Thu, 13 Oct 2011 22:50:35 +0200 wenzelm static dummy_task (again) to avoid a few extra allocations;
Tue, 23 Aug 2011 17:43:06 +0200 wenzelm tuned signature;
Tue, 23 Aug 2011 16:39:21 +0200 wenzelm discontinued slightly odd Future/Lazy.get_finished, which do not really fit into the execution model of Future.cancel/join_tasks (canceled tasks need to be dequeued and terminated explicitly);
Tue, 23 Aug 2011 15:48:41 +0200 wenzelm some support for toplevel printing wrt. editor perspective (still inactive);
Sun, 21 Aug 2011 13:23:29 +0200 wenzelm purely functional task_queue.ML -- moved actual interrupt_unsynchronized to future.ML;
Sat, 20 Aug 2011 15:52:29 +0200 wenzelm added Future.joins convenience;
Fri, 19 Aug 2011 17:39:37 +0200 wenzelm tuned signature (again);
Fri, 19 Aug 2011 16:13:26 +0200 wenzelm tuned signature -- treat structure Task_Queue as private to implementation;
Fri, 19 Aug 2011 15:56:26 +0200 wenzelm refined Future.cancel: explicit future allows to join actual cancellation;
Fri, 19 Aug 2011 14:01:20 +0200 wenzelm Future.promise: explicit abort operation (like uninterruptible future job);
Fri, 19 Aug 2011 12:51:14 +0200 wenzelm more focused use of Multithreading.interrupted: retain interrupts within task group boundary, without loss of information;
Fri, 19 Aug 2011 12:03:44 +0200 wenzelm clarified Future.cond_forks: more uniform handling of exceptional situations;
Thu, 18 Aug 2011 16:07:58 +0200 wenzelm tuned comments;
Wed, 17 Aug 2011 23:37:23 +0200 wenzelm identify parallel exceptions where they emerge first -- to achieve unique results within evaluation graph;
Wed, 17 Aug 2011 22:14:22 +0200 wenzelm more systematic handling of parallel exceptions;
Mon, 15 Aug 2011 19:27:55 +0200 wenzelm explicit check of finished evaluation;
Sat, 13 Aug 2011 00:34:54 +0200 wenzelm immediate fork of initial workers -- avoid 5 ticks (250ms) for adaptive scheme (a07558eb5029);
Wed, 10 Aug 2011 16:26:05 +0200 wenzelm tuned signature;
Wed, 10 Aug 2011 16:05:14 +0200 wenzelm future_job: explicit indication of interrupts;
Wed, 10 Aug 2011 14:28:55 +0200 wenzelm synchronized cancel and flushing of Multithreading.interrupted state, to ensure that interrupts stay within task boundaries;
Wed, 10 Aug 2011 14:04:45 +0200 wenzelm tuned source structure;
Sat, 23 Jul 2011 22:22:21 +0200 wenzelm make double-sure that interrupts are flushed before executing new work (cf. 22f8c2483bd2);
Sat, 23 Jul 2011 21:29:56 +0200 wenzelm more detailed tracing;
Mon, 11 Jul 2011 22:55:47 +0200 wenzelm tuned signature -- corresponding to Scala version;
Tue, 05 Jul 2011 11:16:37 +0200 wenzelm tuned signature;
Thu, 23 Jun 2011 23:12:00 +0200 wenzelm more robust join_results: join_work needs to be uninterruptible, otherwise the task being dequeued by join_next might be never executed/finished!
Sat, 26 Mar 2011 21:28:04 +0100 wenzelm added Future.cond_forks convenience;
Fri, 18 Feb 2011 16:11:58 +0100 wenzelm less verbose tracing;
Wed, 09 Feb 2011 15:48:43 +0100 wenzelm tuned scope of Multithreading.interrupted vs. Multithreading.with_attributes;
Tue, 08 Feb 2011 16:11:52 +0100 wenzelm explicit Multithreading.interrupted to ensure that interrupts stay within the boundaries of managed evaluation blocks;
Fri, 04 Feb 2011 20:40:25 +0100 wenzelm tuned signature;
Wed, 02 Feb 2011 20:32:50 +0100 wenzelm eliminated slightly odd abstract type Task_Queue.deps;
Wed, 02 Feb 2011 15:04:09 +0100 wenzelm maintain Task_Queue.group within Task_Queue.task;
Wed, 02 Feb 2011 13:38:09 +0100 wenzelm Future.join_results: discontinued post-hoc recording of dynamic dependencies;
Tue, 01 Feb 2011 22:24:28 +0100 wenzelm more informative task timing: some dependency tracking;
Tue, 01 Feb 2011 21:05:22 +0100 wenzelm refined task timing: joining vs. waiting;
Mon, 31 Jan 2011 23:02:53 +0100 wenzelm tuned signature;
Mon, 31 Jan 2011 22:57:01 +0100 wenzelm support named tasks, for improved tracing;
Mon, 31 Jan 2011 21:54:49 +0100 wenzelm more direct Future.bulk, which potentially reduces overhead for Par_List;
Mon, 31 Jan 2011 16:34:10 +0100 wenzelm added basic task timing;
Tue, 09 Nov 2010 21:13:06 +0100 wenzelm explicitly identify forked/joined tasks;
Tue, 02 Nov 2010 20:55:12 +0100 wenzelm simplified some time constants;
Fri, 10 Sep 2010 14:54:08 +0200 wenzelm Future.promise: more robust treatment of concurrent abort vs. fulfill (amending 047c96f41455);
Thu, 09 Sep 2010 17:20:27 +0200 wenzelm more abstract treatment of interrupts in structure Exn -- hardly ever need to mention Interrupt literally;
Sun, 08 Aug 2010 19:36:31 +0200 wenzelm explicitly distinguish Output.status (essential feedback) vs. Output.report (useful markup);
Thu, 22 Jul 2010 10:41:12 +0200 wenzelm tuned comments;
Wed, 21 Jul 2010 13:25:14 +0200 wenzelm clarified/exported Future.worker_subgroup, which is already the default for Future.fork;
Tue, 20 Jul 2010 17:35:42 +0200 wenzelm back to more strict dependencies, even for canceled groups (reverting parts of 02936e77a07c);
Tue, 20 Jul 2010 14:44:33 +0200 wenzelm eliminated old-style sys_error/SYS_ERROR in favour of exception Fail -- after careful checking that there is no overlap with existing handling of that;
Sun, 04 Jul 2010 21:01:22 +0200 wenzelm general Future.report -- also for Toplevel.async_state;
Fri, 02 Jul 2010 20:44:17 +0200 wenzelm tuned white space;
Mon, 31 May 2010 21:06:57 +0200 wenzelm modernized some structure names, keeping a few legacy aliases;
Sat, 29 May 2010 15:31:15 +0200 wenzelm future result: retain plain Interrupt for vacuous group exceptions;
Fri, 21 May 2010 21:28:31 +0200 wenzelm future_job: propagate current Position.thread_data to the forked job -- this is important to provide a default position, e.g. for parallelizied Goal.prove within a package (proper command transactions are wrapped via Toplevel.setmp_thread_position);
Sat, 06 Feb 2010 22:06:18 +0100 wenzelm result: Single_Assignment.var;
Wed, 06 Jan 2010 18:22:43 +0100 wenzelm more robust cancelation, notably of passive futures without scheduler running;
Wed, 06 Jan 2010 18:14:16 +0100 wenzelm eliminated cache, which complicates the code without making a real difference (NB: deque_towards often disrupts caching, and daisy-chaining of workers already reduces queue overhead);
Wed, 06 Jan 2010 15:07:56 +0100 wenzelm tasks of canceled groups are considered "ready" -- enables to purge the queue from tasks depending on unfinished promises (also improves general reactivity);
Tue, 05 Jan 2010 23:38:10 +0100 wenzelm added Future.promise/fulfill -- promised futures that are fulfilled by external means;
Thu, 05 Nov 2009 13:16:22 +0100 wenzelm scheduler: clarified interrupt attributes and handling;
Thu, 05 Nov 2009 13:01:11 +0100 wenzelm worker_next: plain signalling via work_available only, not scheduler_event;
Wed, 04 Nov 2009 21:22:35 +0100 wenzelm avoid broadcast work_available, use daisy-chained signal instead;
Wed, 04 Nov 2009 20:31:36 +0100 wenzelm worker_next: treat wait for work_available as Sleeping, not Waiting;
Wed, 04 Nov 2009 11:58:29 +0100 wenzelm worker activity: distinguish between waiting (formerly active) and sleeping;
Wed, 04 Nov 2009 11:37:06 +0100 wenzelm tuned;
Wed, 04 Nov 2009 11:30:22 +0100 wenzelm tuned thread data;
Wed, 04 Nov 2009 00:29:58 +0100 wenzelm worker_next: ensure that work_available is passed on before sleeping (was occasionally lost when worker configuration changed, causing scheduler deadlock);
Tue, 03 Nov 2009 19:52:09 +0100 wenzelm slightly leaner and more direct control of worker activity etc.;
Thu, 22 Oct 2009 15:21:01 +0200 wenzelm use Synchronized.assign to achieve actual immutable results;
Thu, 01 Oct 2009 16:27:13 +0200 wenzelm added Task_Queue.depend (again) -- light-weight version for transitive graph;
Tue, 29 Sep 2009 11:49:22 +0200 wenzelm explicit indication of Unsynchronized.ref;
Mon, 28 Sep 2009 12:09:18 +0200 wenzelm added fork_deps_pri;
Tue, 22 Sep 2009 20:25:31 +0200 wenzelm full reserve of worker threads -- for improved CPU utilization;
Sun, 20 Sep 2009 19:17:33 +0200 wenzelm tuned tracing;
Sun, 20 Sep 2009 18:37:55 +0200 wenzelm scheduler backdoor: 9999 means 1 worker;
Wed, 16 Sep 2009 22:46:10 +0200 wenzelm Synchronized.value does not require locking, since assigments are atomic;
Thu, 27 Aug 2009 17:00:03 +0200 wenzelm tuned tracing;
Sat, 01 Aug 2009 00:17:03 +0200 wenzelm future scheduler: uninterruptible cancelation;
Sat, 01 Aug 2009 00:09:45 +0200 wenzelm renamed Multithreading.regular_interrupts to Multithreading.public_interrupts;
Thu, 30 Jul 2009 23:37:53 +0200 wenzelm tuned tracing;
Thu, 30 Jul 2009 23:06:06 +0200 wenzelm added Multithreading.sync_wait, which turns enabled interrupts to sync ones, to ensure that wait will reaquire its lock when interrupted;
Tue, 28 Jul 2009 16:30:23 +0200 wenzelm eliminated separate Future.enabled -- let Future.join fail explicitly in critical section, instead of entering sequential mode silently;
Tue, 28 Jul 2009 15:10:15 +0200 wenzelm future result: Synchronized.var;
Tue, 28 Jul 2009 14:35:27 +0200 wenzelm Task_Queue.dequeue: explicit thread;
Tue, 28 Jul 2009 14:29:25 +0200 wenzelm more precise treatment of scheduler_event: continous pulse (50ms) instead of flooding, which was burning many CPU cycles in spare threads;
Tue, 28 Jul 2009 14:11:15 +0200 wenzelm interruptible_task: unified treatment of Multithreading.with_attributes (cf. 9f6461b1c9cc);
Tue, 28 Jul 2009 14:04:33 +0200 wenzelm misc tuning;
less more (0) -120 tip