Code

Event performance branch#4

Code

Open

sebres wants to merge 78 commits intocore-8-5-branchsebres/tcl:core-8-5-branchfrom sebres-8-5-event-perf-branchsebres/tcl:sebres-8-5-event-perf-branchCopy head branch name to clipboard

+12,334-17,896

Commits

Commits on Jul 3, 2017

• [performance] after-event list optimized (interp-assoc switched to doubly linked list, because requires handling from both ends of the list)

  sebres

  committed
  8ae594a

• timer resp. idle events optimized: better handling using doubly linked lists, prevents allocating memory twice for the "after" events (use memory inside timer/idle event for the "after" structure),…

  Show description for 7ea8eec
  sebres

  committed
  7ea8eec

• rewrite interpreter limit handling using new timer event handling (with delete callback)

  sebres

  committed
  59b7439

• after-id: introduced object of type "afterObjType" as self-referenced weak pointer to timer/idle event, used for fast access to the "after" event (cancel, info etc.);

  Show description for 303f179
  sebres

  committed
  303f179

• [bug/stable fix] don't execute TimerSetupProc directly (may be unwanted, because changes the blocking time, also if TCL_TIMER_EVENTS|TCL_IDLE_EVENTS not set), so let do that within Tcl_DoOneEvent c…

  Show description for f743ad6
  sebres

  committed
  f743ad6

• bug fix: wrong release of after-id tcl-object if it switch type (object leak)

  sebres

  committed
  ea640f9

• [performance] introduced additional queue for prompt timer events (after 0) that should be executed immediately (no time);

  Show description for daed4df
  sebres

  committed
  daed4df

• [performance] large performance increase by event servicing cycles (3x - 5x faster now);

  Show description for 616371f
  sebres

  committed
  616371f

• [performance] much better handling for timer events within Tcl_ServiceEvent using timer marker in the queue and direct call of TclServiceTimerEvents if marker reached (instead of continuous adding …

  Show description for 6cec57b
  sebres

  committed
  6cec57b

• command "vwait" extended with timeout argument (in ms), 0 could be used to process pending events only (without wait), negative value equivalent execution of "vwait" without timeout (infinite);

  Show description for 27b0e45
  sebres

  committed
  27b0e45

• [performance] do one event (update / event servicing) cycle optimized (introduced threshold to prevent sourcing resp. waiting for new events by no-wait).

  Show description for e3a3cb7
  sebres

  committed
  e3a3cb7

• interim commit: try to extend "vwait" with same options as "update"

  sebres

  committed
  3505de8

• [enhancement] extend "vwait" with same options as "update", new syntax "vwait ?options? ?timeout? varname".

  Show description for 70f5692
  sebres

  committed
  70f5692

• dynamic increase of timer resolution corresponding wait-time;

  Show description for d7b36b9
  sebres

  committed
  d7b36b9

• optimization of Tcl_LimitExceeded by internal usage (tclInt header)

  sebres

  committed
  74aaaff

• Use auto-reset event object (system automatically resets the event state to nonsignaled after wake-up), avoids unwanted reset if wake-up for some other reasons (timeout/aio/message).

  sebres

  committed
  5a5c7ea

• [win32] use timer resolution handling in Tcl_Sleep also;

  sebres

  committed
  91d77fe

• [win] fallback to replace C++ keyword "inline" with C keyword "__inline"

  Show description for 82ba340
  sebres

  committed
  82ba340

• code review + better usage of the waiting tolerance (fewer CPU-greedy now, avoid busy-wait if the rest of wait-time too small and can be neglected);

  Show description for 8f7576a
  sebres

  committed
  8f7576a

• resolved some warnings / fixed unix resp. x64 compilation

  sebres

  committed
  6315911

• after info, after cancel: compare interpreter of the timer-events by direct retrieving via internal representation (ignore foreign events), test cases extended.

  sebres

  committed
  50bbd06

• make timer test-case more precise and time-independent, ignores short tolerance (deviation by waiting);

  Show description for 754a8c1
  sebres

  committed
  754a8c1

• fix sporadic errors on some fast cpu/platforms (because bgerror executed in background and it is an idle-event, give enough time to process it (resp. wait until last idle event is done);

  sebres

  committed
  b9c33d0

• added performance test-cases to cover timer-events speed resp. event-driven tcl-handling

  Show description for 7455b3f
  sebres

  committed
  7455b3f

• [unix] optimized Tcl_WaitForEvent similar to windows changes (makes Tcl for *nix more "RTS" resp. NRT-capable):

  Show description for 6deb0b8
  sebres

  committed
  6deb0b8

• extended performance test-cases (test-nrt-capability): covering of brief wait-times and other RTS-near constructs.

  sebres

  committed
  5bc8e9f

• bug fix: prevent setting of negative block-time by too few initial wait-time, that may expire immediately (for example vwait 0.0001 test).

  sebres

  committed
  bfeb666

• chanio.test: optimize several tests cases running too long (shorten unwanted large sleeps)

  sebres

  committed
  9148f08

• extended performance test-cases (test-nrt-capability): RTS-near sleeps with very brief sleep-time.

  sebres

  committed
  81af4a8

• interim commit: try to fix time-drift backwards (too long offset after calibration? something else?)...

  sebres

  committed
  0b58606

• interim commit: try to fix time-drift backwards (calibration? something else?)...

  sebres

  committed
  e57cc9e

• interim commit: time-drift backwards fix...

  sebres

  committed
  206f8c7

• interim commit: time-drift backwards fix... (try to resolve using frequency factor)

  sebres

  committed
  6b0752e

• interim commit: time-drift backwards fix... (try to resolve using counter variance + frequency factor)

  sebres

  committed
  864762e

• interim commit: amend with optimization

  sebres

  committed
  17aaad4

• win: calibration cycle completely rewritten (no calibration thread needed, soft drifts within 250ms intervals, fewer discrepancy and fewer virtual time gradation, etc).

  Show description for 2a3c75f
  sebres

  committed
  2a3c75f

• call TclWinResetTimerResolution at end of sleep resp. wait for event (no calibration thread anymore)

  sebres

  committed
  e4df9d5

• interim commit: trying to resolve time-freezes with new facilities timeJump/timeJumpEpoch

  sebres

  committed
  45d875a

• interim commit: trying to resolve time-freezes with new facilities timeJump/timeJumpEpoch

  sebres

  committed
  9c04414

• don't use tolerance in vwait, because of dual usage, it causes canceling of wait before end-time, on small timeout values (like 0.5, etc.)

  sebres

  committed
  c581bda

• amend with review

  sebres

  committed
  9a0ec98

• revert dual lists (relative/absolute) back to single list (because of better handling, a bit faster, etc.)

  sebres

  committed
  9385efe

• interim commit: code review, rewrite tclTimer, etc.

  sebres

  committed
  d5f19fd

• code review, rewrite tclTimer, prolongation, etc.

  sebres

  committed
  655c496

• "after at" set factor to 1000000 (seconds), test cases fixed

  sebres

  committed
  8326183

• Merge branch 'fix-busy-prompt-timers' into event-perf-branch:

  Show description for 6e993d0
  sebres

  committed
  6e993d0

• interim commit: trying to pass test-cases - timer-3.*, chan-io-29.34, socket-2.12 (busy prompt timer, unexpected too early timer event, wrong timer-marker usage)

  sebres

  committed
  91f7185

• fixed timer-marker handling: timer should be always executed after queued event (of the same generation), it was marked (be sure it marked to immediate execution in corresponding checkProc only).

  Show description for 35e88e5
  sebres

  committed
  35e88e5

• don't cancel scheduled event as long as the event list is not bidirectional (too slow by large queue) - rewritten to cancel delayed (by execute it).

  sebres

  committed
  c780afc

• Introduced monotonic time as ultimate fix for time-jump issue (fixed for windows and unix now, TIP #302 fully implemented now);

  Show description for 6c801ea
  sebres

  committed
  6c801ea

• [unix] fixes conditional-wait: timeout is monotonic based;

  sebres

  committed
  1895544

• unix: implements wide-clicks on unix (1 wide-click == 0.001 microseconds (1 nanosecond)), so more precise now (e. g. by time measurement etc.);

  Show description for f9d02b4
  sebres

  committed
  f9d02b4

• fix check event source threshold (corresponds 100-ns ranges, if the wide-clicks supported);

  Show description for 2303ca1
  sebres

  committed
  2303ca1

• code review and small optimizations

  sebres

  committed
  7cf6ef1

• after at: added simple workaround for absolute timers/sleep ("after at real-time"): because we use monotonic time in all wait functions, so to avoid too long wait by the absolute timers (to be able…

  Show description for 590508d
  sebres

  committed
  590508d

• cleanup...

  sebres

  committed
  4af2721

• performance test cases extended

  sebres

  committed
  1041322

Commits on Jul 10, 2017

• merge resp. reintegrate sebres-8-5-event-perf-branch to 8.6

  sebres

  committed
  6e09c2f

• Stability fix: queue epoch to guarantee avoid broken queue, service level to avoid reset block time by nested event-cycles (if Tcl_SetTimer does not create it), etc.

  Show description for 96b79db
  sebres

  committed
  96b79db

• tclUnixNotfy.c: we should wait for notifier at least once in case of waitForFiles (regardless of the too short timeout), e. g. test case "chan-io-53.8", etc.

  sebres

  committed
  d1fba64

• reintegrate sebres-8-6-event-perf-branch back to 8.5

  sebres

  committed
  bb515d5

• avoid blocking wait if we've some retarded events (from last event-cycle) and max blocking time was not set outside an event source traversal.

  sebres

  committed
  d2ae8f3

• merge core-8-5-branch

  sebres

  committed
  606f029

Commits on Jul 13, 2017

• timer: fixed setup of the block-time to 0.0 by pending timer of new generation, example:

  Show description for a54af9a
  sebres

  committed
  a54af9a

• avoid busy wait if new short block-time will be set within service event-cycle (Tcl_DoOneEvent) produced in event-cycle without processing

  sebres

  committed
  78b9602

• merge (integrate) sebres-event-perf-fix-busy-wait

  sebres

  committed
  f4704c4

Commits on Mar 8, 2019

• merge sebres-8-5-timerate (prepare for consolidation with 8.5)

  sebres

  committed
  b4cbd5b

• merge 8.5 - timerate is part of Tcl now (since TIP#527 got merged), conflicts resolved, merge-points restored

  sebres

  committed
  67ef69d

• cpu-count (performance counter) recognition is rewritten (also resolves gcc warning: suggest parentheses around '&&' within '||')

  sebres

  committed
  d9b096a

• merge 8.5 (fixed performance test for short exec-time)

  sebres

  committed
  95719f8

• unix: monotonic time (performance counter/wide-clicks) - implements several fallbacks for platforms without monotonic time (older Darwin, MacOSX or xcode)

  sebres

  committed
  f5b24d3

Commits on Mar 9, 2019

• TclpScaleUTime fixed typo, only if native scaling used (tclScaleTimeProcPtr == NativeScaleTime) the relationship is 1:1 and nothing has to be done.

  sebres

  committed
  f0b40b1

• macosx: Tcl_Sleep -> TclpUSleep (provides microseconds sleep, fixed unresolved external "TclpUSleep" bug).

  sebres

  committed
  9fb5419

• non-native time scale fixed: do scale time only once (not in cycle, since all other times used in calculation already scaled).

  sebres

  committed
  2761fa9

• NRT relevant only: provides self-calibrating mechanism for TclpUSleep (the actual time of usleep/nanosleep may be longer, due to system latencies and possible limitations in the timer resolution of…

  Show description for fe57079
  sebres

  committed
  fe57079

Commits on May 22, 2019

• merge 8.5

  sebres

  committed
  61bcbc6

Commits on May 31, 2022

• correct comment for TclpGetUTimeMonotonic

  sebres

  committed
  2f096ff

• [windows] avoid overflow by calculation of perf-counter to 100ns as signed wide int (also fixes negative initial base for monotonic time and incorrect time distance by very large offsets)

  sebres

  committed
  ef4c1ac