ConwayLife.com

• added [R]History support which extends LifeHistory features to any supported Life-like rule
  • implemented [R]History state 6 execution
• added AutoFit capability
  • added new button (top left) to toggle AutoFit
  • added hotkey shift and "f" to toggle AutoFit
  • notification displayed when AutoFit switched on or off
  • ZOOM and pan controls are locked when AutoFit is on
  • notification displayed when AutoFit on and ZOOM or pan change attempted via hotkeys, mouse wheel, or drag
• improved ZOOM
  • extended minimum ZOOM from 0.25x (-4x) to 0.0625x (-16x) and adjusted how zoom values are displayed
  • much faster rendering with ZOOM < 1
  • ZOOM via hotkeys is now centered around cell under mouse pointer
• UI updates
  • made the Zoom control always available at the top of the window
  • added Fit Zoom control
  • added toggle button for Grid Lines display
  • hide the AutoFit, Fit Zoom, Zoom and Grid Lines controls when help is displayed
  • moved timing information (hotkey "t") so it no longer overlaps Help control
  • increased the height of the standalone Viewer so the area without UI controls is square
  • background drag now scrolls help information and error list when displayed
• hotkey updates
  • shift and hotkey "." or "," will rotate left or right in 90 degree steps
  • shift and hotkey "[" or "]" will zoom in or out by a factor of 2
  • shift and hotkey "e" will set STEP to maximum (32)
  • shift and hotkey "d" will set STEP to minimum (1)
  • shift and hotkey "+" will set GPS to maximum (60)
  • shift and hotkey "-" will set GPS to minimum (1)
  • if help is open then hotkey "i" will jump to information section
• help information
  • display counts of cells in each state at pattern load
  • added colored boxes when showing colors
  • added count of color tiles
  • show when grid lines are switched on but hidden (due to non-zero ANGLE or ZOOM < 4)
  • documented the fact that "Esc" stops playback
  • [R]History rules now include state colors as well as Theme colors (since the Theme only defines states 0, 1 and 2)
• script commands
  • added support for HTML5 color names e.g. [[ COLOR ALIVE Red ]]
  • added [[ COLOR TEXT R G B ]] to set color for waypoint messages
• increased number and depth of STARS (hotkey "s") to support larger zoom out
• pattern decoder now accepts well know rule aliases (e.g. 2x2, Bugs, Gnarl, etc.)

• pattern decoders
  • corrected generations rule decoding
  • rules without B or S prefixes were being read in the wrong order
  • refuse to run [R]History patterns with an invalid number of states
  • failing to decode an invalid RLE pattern could then be marked valid by another decoder
  • made the pattern decoder less lenient on files that look like Life 1.05 but aren't
  • added support for "life" as a rule name
  • fixed a crash after failing to decode an invalid multi-state pattern
  • improved validation of rule names for RLE decoder
• memory usage and performance
  • removed duplicate grid clears during startup
  • removed duplicate initialization during standalone Viewer creation
  • improved snapshot memory usage after reset
  • snapshot memory now dynamically grows to reduce GC
  • improved tile usage during snapshot restore and reset
  • when step back disabled do not allocate snapshot 0
  • fixed an issue with tile clear failing which caused extra tiles to be processed
  • only allocate the [R]History overlay for [R]History rule patterns that contain states >= 2
  • only allocate the [R]History state 6 bitmaps when pattern contains state 6 cells
• [R]History pattern fit zoom failed after grid grows
• display a more useful message when step back attempted but step back disabled with [[ NOSTEPBACK ]]
• with multiple Viewers and more than one non-library rule the rule names were incorrectly displayed
• corrected the highlight on UI items that are selected
• UI library can now draw icons on simple buttons

gmc_nxtman wrote:Here is an example:

Code: Select all

x = 4, y = 5, rule = 125/36
2bo2$obo$o2bo$o!

The pattern explodes instead of running as a 2x2 crawler in B36/S125, because it runs in B125/S36.

Scorbie wrote:@Chris
I am currently using an old netbook. And the main bottleneck of the lifeviewer here and my smartphone seems to be updating the canvas.(The speed doesn't get faster with more gps but it does get faster with xN.)
Can the "updating pattern" be considerably faster or does it depend on the speed of my browser rendering HTML5 canvas?
Not to criticize, but this implementation renders it faster. Maybe it contains some optimizations?

rowett wrote:LifeViewer drawing time is constant (it's actually related to the size of the display window, which in the forum is constant). It doesn't matter how many cells are alive or changing. Nor does it matter (much) what zoom level, pan or viewing angle you have over the grid. The penalty for this design decision is that it's slower (than the implementation you reference) for small populations. The advantage is that the rendering (at least) scales incredibly well for complex patterns.

rowett wrote:LifeViewer is probably an order of magnitude faster at next generation computation (regardless of supported rule) than the implementation you reference.

rowett wrote:I've improved the rendering speed slightly (~20% on some devices) in an upcoming build in the common case where the viewing angle is 0 and the grid doesn't need clipping to be displayed. I have a backlog item to investigate further rendering optimizations.

Scorbie wrote:Can the "updating pattern" be considerably faster or does it depend on the speed of my browser rendering HTML5 canvas?

1. Original renderer here
2. New renderer here

Scorbie wrote:Okay, If I got it right, I am supposed to average the "Work" (in ms) over gen 0 to gen 700, right?

Scorbie wrote:Here are the results:
On my Android phone: 20~30ms -> 19~22ms
On my Netbook: About 30ms -> About 20~22ms (I tested with the default wm in my distro, Enlightenment, and Xmonad that I installed today They seem to be about the same.)
So I guess it's a 1.5x increase in speed? Nice

Scorbie wrote:My netbook shows the speed above... when it's okay, but often the speed plunges abruptly for an unknown reason, to about 2~3fps or work taking about 200~300ms.

Scorbie wrote:Pale Moon, 32bit, optimized build for Intel Atom CPU. (Basically a lighter Firefox fork, but it's still pretty heavy)

• step back function
  • step back function is now unlimited
    • step back buffer increased to 1024 generations
    • when buffer exhausted automatically compute the previous set of generations
• performance improvements
  • much faster rendering when ANGLE = 0 and/or clipping is not required
  • improved performance when computing previous generations for step back function (hotkey "b")
  • [R]History state 6 execution is now much faster
  • modest performance improvements to boundary clear
  • only clear boundary if pattern is near the edge of the maximum grid
• UI and Display updates
  • Layers now use dynamic ZOOM level of detail
  • better [R]History ZOOM out with priority state reordering
  • better handling for non-explosive rule patterns hitting the maximum grid boundary
  • cell coordinates under mouse cursor are displayed (when ANGLE = 0)
    • coordinates automatically display when paused and disappear during playback unless statistics are on
  • improved notification message when Pause other Viewers is used (hotkey "z")
  • area outside maximum grid size is now drawn (default grey)
  • Depth, Theme and Layers controls reordered to better match hotkey positions
  • smooth camera transitions
    • between POIs (hotkey "j")
    • when restoring saved View (hotkey "v")
    • when using Fit Zoom (hotkey "f")
    • when setting ZOOM to a multiple (hotkeys "1", "2", "3", "4", "6" and "8")
  • Fit Zoom and AutoFit controls disabled during Waypoint playback
• hotkey updates
  • shift and hotkey "2" now sets ZOOM to -2x (used to be hotkey "5" which is now unassigned)
  • shift and hotkey "4" now sets ZOOM to -4x
  • shift and hotkey "8" now sets ZOOM to -8x
  • shift and hotkey "6" now sets ZOOM to -6x
  • shift and hotkey "z" now pauses all Viewers (unlike hotkey "z" with pauses all other Viewers)
  • shift and hotkey "v" now saves the current camera position, hotkey "v" restores camera position
  • shift and hotkey "c" sets default Theme
  • hokey "j" jumps to the next defined point of interest, shift and hotkey "j" jumps to the previous point of interest
• script commands
  • new script command [[ POI ]] followed by one or more of [[ X ]], [[ Y ]], [[ ZOOM ]] and [[ ANGLE ]] defines a point of interest
    • optional message at each POI e.g. [[ "This is a glider" ]]
    • hotkey "j" / shift hotkey "j" to jump to next / previous point of interest
    • list of POI definitions added to help information
  • new script commands [[ XOFFSET <number> ]] and [[ YOFFSET <number> ]] to define initial pattern position on grid
  • new script command [[ SHOWGENSTATS ]] will display the generation statistics at pattern load (hotkey "g" to toggle)
  • new script command [[ SHOWTIMING ]] will display timing information at pattern load (hotkey "t" to toggle)
  • [[ AUTOFIT ]] command will now switch on AutoFit at pattern load when no Waypoints are specified
  • added custom state colours for [R]History
    • [[ COLOR <name> R G B ]] where <name> is "OFF", "ON", "HISTORY", "MARK1", "MARKOFF", "MARK2" or "KILL"
  • added [[ COLOR BOUNDARY R G B ]] to define custom boundary colour
  • [[ X ]] and [[ Y ]] script commands now allow arguments between +/- half maximum grid size rather than half current grid size
  • [[ ZOOM ]] command now accepts negative values -16 to -1 (as well as 0.0625 to 1)
• help information
  • list of POI definitions added to help information
  • [R]History state names added to colours in help information

• pattern decoders
  • fixed an issue with the RLE pattern decoder which would ignore invalid characters in rule names
  • some invalid [R]History style rules were being marked as LifeHistory
  • RLE pattern decoder is now case-insensitive for rule names
  • RLE pattern decoder can now deal with incorrect x and y values
  • Life 1.05 pattern decoder now ignores non-standard # commands
• general
  • hotkey ZOOM changes were resetting ANGLE to 0
  • fixed potential issue with initial grid size and large STEP sizes
  • population count at T=0 was incorrect in some rare cases
  • Fit Zoom was incorrect for [R]History patterns that had no live cells

• using the XOFFSET command to place a pattern at the right hand side of the grid
• the improved handling of patterns hitting the grid boundary
• how Layers can be used for effects
• some techniques for Waypoints

• grid boundary
  • patterns that hit the large grid boundary (8192x8192 cells) are now instantly deleted in their entirety
  • new script command [[ DELETERANGE <1..16> ]] that defines how near live cells have to be to each other to be considered part of a single entity for deletion (default 3 cells)
• performance improvements
  • drawing the starfield is now 25% faster in the general case and 4x faster at ANGLE=0
  • drawing the gridlines is now 3.5x faster

• RLE decoder was hanging on some invalid patterns
• removed support for LV85 format patterns

#N 10-engine Cordership
#O Dean Hickerson
#C A c/12 period 96 diagonal Cordership that uses 10 switch engines, which was the fewest possible known at the time of its discovery.
#C http://www.conwaylife.com/wiki/index.php?title=10-engine_Cordership
x = 88, y = 88, rule = 23/3
42bo45b$42bo45b$44bo5bo37b$43bo6bobo35b$42bo3bo2bo38b$43bo2bobob2o36b$
48bob2o36b$62b2o24b$62b2o24b7$70b2o16b$26b2o2bo39b2o16b$29bobo56b$28bo
59b2$30b2o56b2$31b2o55b$30bo47b2o8b$28b2ob2o45b2o8b$31b2o55b$16bo12bo
58b$16bo22b3o46b$18bo5bo13bo49b$17bo6bobo10bo4b2o44b$16bo3bo2bo12bo3bo
47b$17bo2bobob2o10bo2bo4bo41b2o$22bob2o10bo3bo3bo32bo8b2o$37b2obob3o
31bobo9b$40bo47b$41b4o31bo2bo8b$30b3o10b2o33b2o8b$29bo3bo45bo8b$28bo4b
o54b$27bo3bo56b$27bo2bob3o53b$27bo7bo52b$2o2bo24bo3bobo40b2o10b$3bobo
23bo3bob2o41bo9b$2bo28b3ob2o39b2o10b2$4b2o82b2$5b2o81b$4bo83b$2b2ob2o
52bobo6bobo17b$5b2o51bo9bobo17b$3bo55bo2bo6bo18b$61b3o24b5$51bo36b$50b
obo35b2$50bo2bo34b$7b2o43b2o34b$7b2o44bo34b5$50b2o36b$52bo35b$15b2o33b
2o36b$15b2o71b5$33bobo6bobo43b$32bo9bobo43b$23b2o8bo2bo6bo44b$23b2o10b
3o50b7$31b2o55b$31b2o!
#C [[ DELETERANGE 13 THEME 0 ZOOM 2 YOFFSET -4019 XOFFSET -4019 STOP 485 NOSTEPBACK ]]

Kiran wrote:How does it decide which cells are in the "same pattern" as those that hit the edge?

muzik wrote:

Kiran wrote:How does it decide which cells are in the "same pattern" as those that hit the edge?

Taking a guess here: it searches for matching known patterns in a database or something

But then, they always are when the Life grid is bounded, anyway.

Kiran wrote:I found a bug in the edge detection script.

• Alan Hensel's isotropic non-totalistic rules (e.g. B2ce3ai_S23) - see here
• 2-state totalistc Hex neighbourhood rules (e.g. B2/S34H)
• Any 2-state totalistic rule with a bounded grid definition (e.g. LifeHistory:K40*,20) - see here