Mandel Machine

Mandel Machine is a highly efficient Mandelbrot set explorer developed by Botond Kósa.

Download area:

• Latest version for Windows x64: 1.3.15
• Previous versions & changelog
• Source code for version 1.3.10:
• Java source files
• C++/ASM source files

Main features:

• Up to 536 MP (~23000x23000) image size
• Up to 32x32 supersampling
• Magnification up to 10000 zooms (image size ~3.2E-3010)
• Max iterations up to 800 millions
• Intuitive mouse navigation:
• Zoom in and out with the mouse wheel
• Zoom in faster by double-clicking
• Click-and-drag to select area to zoom into
• Selection rectangle can be moved, resized and rotated
• Right click and drag to pan (experimental)
• Optimizations to speed up calculations:
  • Extensive use of SIMD instruction sets (SSE2, SSE3, AVX) with double data types
  • Inner loops implemented in assembly language
  • Arbitrary precision fixed point arithmetic using 64-bit integer data types, fully unrolled loops
  • Perturbation algorithm with series approximation
  • Tweaked Mariani/Silver algorithm to guess areas with monotonic iteration counts
  • Multi-core support (up to 32 threads)
  • Pixel grouping to fully saturate the execution units of modern CPUs (up to 16 pixels with SSE2, up to 32 with AVX-capable CPUs)
• Iteration counts histogram
• Coloring options:
• 7 palette presets
• Palette can be extracted from any JPG/PNG/BMP/GIF image
• 8 different iteration transfer functions
• Color density & offset adjustable
• Smooth/solid/inverted dwell bands
• Bump mapping to create 3D-look
• Automatic correction of glitches caused by perturbation algorithm
• Detailed statistics
• Full history of changes
• Save/load location and rendering attributes
• Save images as JPG or PNG files
• Copy image to clipboard

Features coming soon:

• Calculate distance estimates
• Palette editor
• Movie maker

• NEW: Zoom into pattern centroid with a single click (experimental)
• FIXED: Application crashes when calculating with series approximation turned off
• FIXED: Application crashes when calculating with iteration limit higher than 10 millions
• Known issues:
  • Pixel guessing is not working during glitch correction
  • Glitch correction is not working when calculating with scaled doubles

• NEW: Palette preset: Petal Colors by Alef
• NEW: Window location is saved on exit and restored on startup
• NEW: Reference calculation can be aborted
• IMPROVED: Magnification limit increased to 10000
• IMPROVED: Max. iteration limit increased to 800 millions
• IMPROVED: Progress is displayed during reference calculation
• FIXED: Successive palette changes are not saved in history
• FIXED: Iteration data is sometimes saved incorrectly
• Known issues:
  • Pixel guessing is not working during glitch correction
  • Glitch correction is not working when calculating with scaled doubles

• NEW: Palette preset: Grayscale Histogram Coloring
• FIXED: Application crashes when aborting during reference calculation or series approximation initialization
• Known issues:
  • Pixel guessing is not working during glitch correction
  • Glitch correction is not working when calculating with scaled doubles

• CHANGED: Default value of OOM diff between terms raised from 1 to 2 to avoid some glitches caused by overly aggressive series approximation
• IMPROVED: History is loaded 20% faster
• FIXED: Application freezes when trying to run the benchmark
• FIXED: Application crashes when increasing image size or supersampling with bump mapping turned on
• FIXED: Successive zoom-in actions by the mouse wheel are not saved in history
• FIXED: Trying to zoom out with the mouse wheel or touchpad may result in zooming in instead
• Known issues:
  • Pixel guessing is not working during glitch correction
  • Glitch correction is not working when calculating with scaled doubles

• IMPROVED: Normal map caching allows bump mapping to be performed 2x faster (RAM usage is 12 bytes/pixel; can be switched off under Rendering / Post processing)
• Known issues:
  • Pixel guessing is not working during glitch correction
  • Glitch correction is not working when calculating with scaled doubles

• NEW: Bump mapping effect (under Rendering / Post processing)
• IMPROVED: Color density adjustment does not wrap palette around when rendering with histogram coloring
• CHANGED: Default pixel guessing strategy is Fill: only interior areas of the Mandelbrot set are guessed (Interpolate causes artifacts in interpolated areas when bump mapping is enabled)
• FIXED: JPEG file is not truncated to new size when overwriting a larger file
• Known issues:
  • Pixel guessing is not working during glitch correction
  • Glitch correction is not working when calculating with scaled doubles

• NEW: Fractal images can be saved in JPEG format
• Known issues:
  • Pixel guessing is not working during glitch correction
  • Glitch correction is not working when calculating with scaled doubles

• NEW: Optimization called Local bailout check deferral can be switched off under Computation
• NEW: Palette presets from SeryZone: Strange Soft Colors, Dusk Colors
• IMPROVED: More accurate palette interpolation with no overshoot
• FIXED: Palette preset phaumann's Histogram Coloring corrected by SeryZone
• Known issues:
  • Pixel guessing is not working during glitch correction
  • Glitch correction is not working when calculating with scaled doubles

• NEW: Sensitivity of series approximation is adjustable in the Location toolbox: the number set in OOM diff between terms specifies the difference in orders of magnitude between the last term and other terms.
• IMPROVED: Series approximation is more aggressive by default, using an OOM diff between terms of 1 instead of 16
• CHANGED: Ultra Fractal default palette modified (contains more orange and purple hues) to imitate the original more accurately
• Known issues:
  • Pixel guessing is not working during glitch correction
  • Glitch correction is not working when calculating with scaled doubles

• NEW: Compressed iteration data (.mmit) files can be loaded in the absence of the corresponding fractal definition (.mmf) file
• NEW: Kalles Fraktaler Palette (.kfp) files can be loaded
• NEW: History of previous changes is loaded on startup - work can be continued even after a crash
• CHANGED: History is saved in a subfolder under the application folder instead of the TEMP folder. Deletion of saved history items on application exit can be toggled under Settings.
• IMPROVED: Parameters can be selectively loaded from .kfr and .kfp files
• IMPROVED: References are calculated up to 80% faster when the iteration limit is below 10 million, at slightly increased RAM usage
• FIXED: State is saved in history for every intermediate value when continuously adjusting color density/offset with arrow keys
• FIXED: Lossy compression of iteration data used in .mmit files may cause stripes in slowly changing gradients
• Known issues:
  • Pixel guessing is not working during glitch correction
  • Glitch correction is not working when calculating with scaled doubles

• NEW: Smooth (C¹ continuous) interpolation of color gradients can be toggled in Palette window
• IMPROVED: New widget allowing finer control over color density/offset
• IMPROVED: Color density range extended to -500; fractional values are also allowed
• IMPROVED: Coloring of KFR files is loaded more accurately
• FIXED: State is saved in history twice when adjusting color density/offset with arrow keys
• FIXED: KFR files with iteration division greater than 1 are loaded incorrectly
• Known issues:
  • Pixel guessing is not working during glitch correction

• FIXED: Application crashes when trying to extract palette from an image
• FIXED: Application may crash when computing a reference with 0 imaginary part
• FIXED: Series approximation breaks around (-2,0)
• FIXED: Application may hang when computation is aborted
• Known issues:
  • Pixel guessing is not working during glitch correction

• NEW: Iteration data can be saved in uncompressed format
• Known issues:
  • Series approximation breaks around (-2,0)
  • Application may crash when computing a reference with 0 imaginary part
  • Pixel guessing is not working during glitch correction

• NEW: Iteration data can be saved/loaded beside fractal definition
• NEW: Compression level for iteration data can be adjusted under Settings
• IMPROVED: Iteration data is saved in a more compressible format
• IMPROVED: Smooth scrolling of toolboxes/history
• IMPROVED: Zoom range in Image toolbox extended to 1/16x and 16x
• IMPROVED: Up to 2x faster rendering of the image when adjusting rendering attributes due to lossless compression of palette
• IMPROVED: More precise palette interpolation with less overshoot
• FIXED: Glitches may appear in the centroids of embedded Julias when using perturbation with reduced scale floats
• FIXED: Application hangs when trying to compute an image with huge dimensions and magnification close to the limit
• FIXED: Iteration color density is sometimes loaded incorrectly from .mmf files
• Known issues:
  • Series approximation breaks around (-2,0)
  • Application may crash when computing a reference with 0 imaginary part
  • Pixel guessing is not working during glitch correction

• IMPROVED: Iteration data in history is saved and loaded up to 10x faster
• FIXED: Saved iteration data may be corrupted when the iteration span is larger than one million
• Known issues:
  • Series approximation breaks around (-2,0)
  • Application may crash when computing a reference with 0 imaginary part
  • Glitches may appear in the center of embedded Julias when using perturbation with reduced scale floats

• NEW: Toolbox for adjusting application-wide settings. Previously saved settings are loaded on startup.
• NEW: Mouse navigation settings (zoom increase by scrolling/double click, auto-apply selection) can be adjusted in the Settings toolbox
• FIXED: Iteration data is not saved in history - a recalculation is needed if location or computation attributes are changed
• Known issues:
  • Series approximation breaks around (-2,0)
  • Application may crash when computing a reference with 0 imaginary part
  • Glitches may appear in the center of embedded Julias when using perturbation with reduced scale floats

• NEW: Supersampled images can be saved or copied to clipboard at different resolutions
• CHANGED: Memory available to the application is set to match the amount of free RAM in the system
• CHANGED: Perturbation algorithm automatically used when standard double precision calculation is not enough (beyond ~42 zooms)
• CHANGED: Histogram is displayed for the covered iteration range only
• FIXED: Application may crash when zooming into an empty region
• FIXED: Memory leak in history when using a palette that was extracted from an image
• FIXED: Image is always recalculated after loading rendering parameters from a file
• Known issues:
  • Series approximation breaks around (-2,0)
  • Application may crash when computing a reference with 0 imaginary part
• Features temporarily turned off in this beta:
  • Saving of iteration data in history - a recalculation is needed if location or computation attributes are changed

• FIXED: Application may crash when zooming in/out during the glitch correction phase
• FIXED: Palette is not saved correctly in .mmf files
• Known issues:
  • Series approximation breaks around (-2,0)
  • Application may crash when computing a reference with 0 imaginary part
• Features temporarily turned off in this beta:
  • Saving of iteration data in history - a recalculation is needed if location or computation attributes are changed

• IMPROVED: Wider pixel grouping (16-fold on SSE2, 32-fold on AVX-capable CPUs) when using perturbation with reduced scale floats
• IMPROVED: Scale reduction to floats enabled from 80-bit extended precision as well
• FIXED: 16-fold pixel grouping is not used during glitch correction
• FIXED: Glitch correction is not working when calculating with 80-bit extended precision
• Known issues:
  • Palette is not saved correctly in .mmf files
  • Series approximation breaks around (-2,0)
  • Application may crash when computing a reference with 0 imaginary part
  • Application may crash when zooming in/out during the glitch correction phase
• Features temporarily turned off in this beta:
  • Saving of iteration data in history - a recalculation is needed if location or computation attributes are changed

• IMPROVED: Glitch centroids are located up to 70% faster
• FIXED: Minibrots are identified as glitches and get filled with junk
• FIXED: Application may crash or turn perturbation off when zooming in/out with the mouse wheel
• FIXED: Application hangs when calculating perturbation algorithm with ASFloat data type
• Known issues:
  • Palette is not saved correctly in .mmf files
  • Glitch correction is not working when calculating with 80-bit extended precision
  • Series approximation breaks around (-2,0)
• Features temporarily turned off in this beta:
  • Saving of iteration data in history - a recalculation is needed if location or computation attributes are changed

• NEW: Glitch correction based on the algorithm from Paul G. Derbyshire (Pauldelbrot)
• NEW: Iterations skipped by series approximation can be limited
• NEW: Check for newer versions on startup
• NEW: GUI controls are disabled during calculation
• IMPROVED: Magnification limit increased to 8000
• FIXED: False GFLOPS values are reported under Statistics
• FIXED: Save image function produces corrupted PNGs
• Known issues:
  • Palette is not saved correctly in .mmf files
  • Glitch correction is not working when calculating with 80-bit extended precision
  • Series approximation breaks around (-2,0)
• Features temporarily turned off in this beta:
  • Saving of iteration data in history - a recalculation is needed if location or computation attributes are changed

• IMPROVED: Magnification limit increased to 6000
• IMPROVED: Per-pixel cost of series approximation reduced by 65% - large images are calculated up to 30% faster
• FIXED: Patterned glitches appear in deep Julias when the reference pixel if off-center
• Known issues:
  • Palette is not saved correctly in .mmf files
  • Application may hang when adjusting a parameter during calculation - always cancel the calculation first with the Esc key
• Features temporarily turned off in this beta:
  • Glitch correction
  • Saving of iteration data in history - a recalculation is needed if location or computation attributes are changed

• FIXED: Glitches appear due to reference orbit calculation with too few digits (Tick-tock location)
• FIXED: Some locations are not rendered at all (Verstoppertje location)
• FIXED: Application hangs when saving very large PNG images due to excessive memory usage
• Known issues:
  • Palette is not saved correctly in .mmf files
  • Application may hang when adjusting a parameter during calculation - always cancel the calculation first with the Esc key
• Features temporarily turned off in this beta:
  • Glitch correction
  • Saving of iteration data in history - a recalculation is needed if location or computation attributes are changed

• IMPROVED: Magnification limit increased to 5300
• IMPROVED: Up to 5x faster initialization of series approximation
• IMPROVED: Up to 2x faster calculation when using perturbation with ASFloat data type
• IMPROVED: Up to 1.25x faster calculation when using perturbation with 80-bit extended data type
• IMPROVED: Wider pixel grouping (8-fold on SSE2, 16-fold on AVX-capable CPUs) at zoom levels below 50 when not using perturbation
• CHANGED: Benchmark runs made longer to provide more repeatable results
• CHANGED: "Ignore small addends" option renamed to "Ignore delta₀ if negligible"
• FIXED: Series approximation sometimes breaks embedded Julias near the edges of the image
• FIXED: Minibrots get distorted when using perturbation with too low iteration limit
• FIXED: Application crashes on CPUs or operating systems with no AVX support
• Known issues:
  • Palette is not saved correctly in .mmf files
  • Application may hang when adjusting a parameter during calculation - always cancel the calculation first with the Esc key
• Features temporarily turned off in this beta:
  • Glitch correction
  • Saving of iteration data in history - a recalculation is needed if location or computation attributes are changed

• NEW: Statistics about data types used for perturbation calculation
• NEW: Automatic reduction of scale from double to float on AVX-capable CPUs
• IMPROVED: Palette can be extracted from gif images as well
• FIXED: Image gets corrupted when using series approximation with too few terms
• FIXED: Application hangs when using extended scale without ignoring small addends
• Known issues:
  • Palette is not saved correctly in .mmf files
  • Application may hang when adjusting a parameter during calculation - always cancel the calculation first with the Esc key
• Features temporarily turned off in this beta:
  • Glitch correction
  • Saving of iteration data in history - a recalculation is needed if location or computation attributes are changed

• NEW: Magnification limit increased to 3700, 80-bit extended data type used beyond 1900 zooms
• NEW: Automatic reduction of scale from extended to double or scaled double when ignoring small addends - enables the use of SIMD codepaths beyond 1900 zooms
• IMPROVED: Number of terms used for series approximation is adjustable (up to 130)
• IMPROVED: Faster calculation using double precision through wider pixel grouping (8-fold on SSE2, 16-fold on AVX-capable CPUs) when using perturbation
• IMPROVED: Faster saving of PNG images
• IMPROVED: Faster palette updates
• CHANGED: Shorter transitions when zooming in/out
• CHANGED: Iteration values are stored in 64 bits/pixel - memory usage increased by 50% when rendering very large images
• CHANGED: JRE is no longer bundled with the application - it automatically searches for a 64-bit JRE of version 1.7.0 or newer
• FIXED: Palette changes are not saved in history
• FIXED: Inertial scrolling causes the image to zoom in/out further than desired
• Known issues:
  • Palette is not saved correctly in .mmf files
  • Application may hang when adjusting a parameter during calculation - always cancel the calculation first with the Esc key
  • Application hangs when using extended scale without ignoring small addends
• Features temporarily turned off in this beta:
  • Glitch correction
  • Saving of iteration data in history - a recalculation is needed if location or computation attributes are changed

• NEW: Palette can be saved into .mmf files
• FIXED: No history item created when dragging the color density/offset sliders
• FIXED: When adjusting the color density/offset sliders, the image may be rendered with the old values

• FIXED: Palette images with high width/height ratio are displayed as a thin line

• IMPROVED: Palette can be extracted from bmp images as well

• NEW: Image size in megapixels shown in the Image box
• FIXED: Application crashes when trying to create an image that doesn't fit into the memory or is larger than 536 MP

• NEW: Automatic correction of flat blobs caused by the perturbation algorithm
• FIXED: Stack overflow can occur when rendering super dense areas with 1000+ magnification and pixel grouping of 2

• FIXED: Application sometimes hangs when using perturbation method with pixel guessing off
• FIXED: Bogus pixels with infinite or NaN iteration count appear in super dense areas

• CHANGED: New distribution format, standalone app without installer, embedded JRE does not have to be downloaded with each upgrade
• FIXED: Zooming in/out with mouse wheels doesn't work if computation is still running
• FIXED: Application crashes when running with the standard HotSpot JVM

• NEW: Correct blobs manually by right clicking inside the blob to add a new reference point (experimental)
• CHANGED: Initial reference point is the center pixel; hillclimbing algorithm for finding a reference with highest iteration count is no longer used
• FIXED: Image becomes distorted when approaching a minibrot using perturbation method with scaled double
• IMPROVED: Maximum number of working threads increased from 8 to 32
• IMPROVED: Palette size increased from 4K to 1M to support KFR files with a high number of key colors
• IMPROVED: Minimum color density extended from -300 to -400 to support KFR files with high iteration division
• IMPROVED: Fractal window can be iconified to eliminate repaint overhead when rendering with high supersampling

• FIXED: Small addends are never ignored when using perturbation method with scaled double and SSE2/SSE3 codepaths
• FIXED: Application crashes when using perturbation method with scaled double, SSE2/SSE3 codepaths and pixel grouping of at least 6
• IMPROVED: Detailed history labels when zooming in/out or panning

• FIXED: Application hangs when using perturbation method with such a low iteration limit that the whole image becomes one-colored
• FIXED: Computation cannot be aborted with ESC key during reference orbit calculation and series approx. initialization

• FIXED: Application hangs when increasing iteration limit with histogram coloring
• FIXED: Application hangs when using perturbation method after a benchmark run
• IMPROVED: Reduced memory usage when using perturbation method and high iteration limit (millions)

• Initial release