On the Road to 4.1

[David_Tschumperle (G'MIC staff)]

This is the running changelog of the G’MIC software, until version 4.1 is released. It lists all new features and changes done since latest major version 4.0.

New features

G’MIC-Qt plug-in:

• [gmic-qt-407] New filter Rendering / Perspective Water Waves procedurally renders a choppy water surface, in perspective.

• [gmic-qt-407] New filter Rendering / Paint Background generates a painted-wall / oil-sky style background (cloudy pigment layers, impasto, splatters).

• [gmic-qt-406] New filter Patterns / Penrose Tiling generate Penrose tilings by Robinson-triangle deflation.

• [gmic-qt-406] New filter Deformations / Page Curl curls up a corner of the image, as on a sheet of paper.

• [gmic-qt-406] New filter Deformations / Glass Door simulates optical distortion as if your image is seen through a glass block door.

• [gmic-qt-406] New filter Rendering / Cosmic Nebula generates a procedural cosmic nebula.

• [gmic-qt-406] New filter Rendering / Chladni Figures, mathematically simulates complex sand shapes created by 2D wave interference.

• [gmic-qt-404] New filter Rendering / Variable-Density Hilbert Curve draws a single variable-density Hilbert Curve that reproduces an input image.

• [gmic-qt-404] Filter Arrays & Tiles / Montage has been recoded to use the newest arrange command, with a few additionnal improvements (support a larger number of input layers, better memory management, etc.).

Core functionalities:

• [gmic-core-408] Add new native commands sigmoid, logit and log1p.

G’MIC Standard Library:

• [stdlib-408] New command perspective draws an image into another one, as a 3D quadrilatere in perspective.

• [stdlib-404] New command arrange arranges selected images according to a specified layout, described as a string (see post: https://discuss.pixls.us/t/new-command-arrange-a-better-version-of-montage/).

• [stdlib-404] New command shape_peano creates a 2D image of the Peano curve.

Improvements / Modifications

G’MIC-Qt plug-in:

• [gmic-qt-407] Filter Deformations / Perspective has been rewritten from scratch to make it more interactive and easy to use.

Core functionalities:

• [stdlib-406] Comprehensive code review has been done for nn_lib , with many small bugs fixed and performance improvements.

• [math-core-408] Optimizations (x6) of vector-valued functions vmax(), vmin(), … when only two arguments are provided.

• [math-core-406] Optimizations done in parsing macro definitions. Avoid multiple string reallocations.

• [core-404] Interpreter now pre-counts number of arguments given to a native command, to speed up argument parsing.

• [core-404] When interpreter parses a string that is expected to be a single double value, it now uses std::strtod() rather than std::sscanf(), which is 5 to 10 times faster (it does not make a huge difference on execution time of G’MIC scripts though).

• [math-core-404] Function find() has an extra max_iterations parameters that allows to narrow search in an image or vector subset.

G’MIC Standard Library:

• [stdlib-406] Library nn_lib has been greatly improved after a full code review. Calculations of network modules have been verified and corrected when necessary.

• [stdlib-405] Command map now proposes an adaptive random colormap option.

• [stdlib-404] Command colormap now implements a multi-threaded algorithm to extract the full colormap of images.

Bug fix

• [stdlib-406] Library nn_lib: Fix calculus in backprop when using convolutional layers with boundary_conditions!=0.

• [stdlib-404] Commands display, print and plot: Fix error occurring when one of the selected images has an empty name.

• [math-core-407] Fix bug in double assignment expression as x = X[k] = y.

• [math-core-405] Fix loading PNM files from stdin.

• [math-core-404] Better catch exceptions occuring in parallelized evaluations.

• [math-core-404] Fix bug in math evaluator, related to optimization attempts for certain expressions that led to a wrong memory access in some cases.

• [build-system-406] Fix Makefile to allow parallel compilation with make -j.

[Reptorian]

"* Function 'find(A,B,_starting_index,_search_step,_max_iterations)' returns the index where sub-vector 'B' appears "\

I suppose that max iteration means number of times to search? And can there be max sublength parameter? Vectors are const size, but what if I just want to find exact matches up to N?

rep_uniq_args could benefit from that, which to be honest is pretty small edge case.

[David_Tschumperle (G'MIC staff)]

@Reptorian

I suppose that max iteration means number of times to search?

That’s it. I didn’t want to call it sequence_length because when search_step is different from 1 it could be tricky to understand what it does.

[David_Tschumperle (G'MIC staff)]

2026/08/19: Release of G’MIC 4.0.4.

[lylejk (Lyle Kroll)]

Let’s hope this doesn’t actually happen with the version with the same nomenclature, David. lol

So far appears to work great.

All rendered with a few G’MIC settings.

[David_Tschumperle (G'MIC staff)]

• 2027/09/04 : release of v.4.0.5

[David_Tschumperle (G'MIC staff)]

• 2027/10/14: Release of v.4.0.6.

[lylejk (Lyle Kroll)]

Tested and works for the presets most import to me, David (i.e, no way can I test them all; lol). Again, appreciate you and your team.

[David_Tschumperle (G'MIC staff)]

2026/10/28: Release of G’MIC v.4.0.7.