Many of ImageMagick's internal algorithms are threaded to take advantage of speed-ups offered by multicore processor chips and OpenMP, an API specification for parallel programming. If your compiler supports OpenMP directives (such as GCC or Visual Studio 2005), ImageMagick automatically includes support.

Verifying OpenMP Support

To verify if OpenMP is active in your ImageMagick build, look for the OpenMP feature string using the following command:

$ magick identify -version
Version: ImageMagick 7.0.10-50 2021-01-04 Q16 https://imagemagick.org
Copyright: © 1999-2021 ImageMagick Studio LLC
Features: OpenMP(4.5)

How Parallel Execution Works

With OpenMP enabled, most ImageMagick algorithms execute across all available system cores in parallel. ImageMagick typically divides the work so that each thread processes 64 rows of pixels. As rows are completed, OpenMP assigns additional chunks to each thread until the task is complete. For example, running a resize operation on a quad-core system utilizes 4 cores (or 8 if hyperthreading is enabled).

Performance Tip

You can further increase performance by reducing memory lock contention with the tcmalloc memory allocation library. To enable it, add --with-tcmalloc to the configure command-line when building ImageMagick.

The Perils of Parallel Execution and Benchmarking

Predicting behavior in a parallel environment can be complex. Performance depends on various factors including your compiler, OpenMP library version, processor type, core count, available memory, hyperthreading status, concurrent applications, and the specific image-processing algorithm used.

Because of these variables, benchmarking is the only reliable way to ensure optimal performance. ImageMagick supports progressive threading benchmarks, returning elapsed time and efficiency metrics across one or more threads. Below is an example benchmark measuring threads 1 through 8:

$ magick -bench 40 model.png -sharpen 0x1 null:
Performance[1]: 40i 0.712ips 1.000e 14.000u 0:14.040
Performance[2]: 40i 1.362ips 0.657e 14.550u 0:07.340
Performance[3]: 40i 2.033ips 0.741e 14.530u 0:04.920
Performance[4]: 40i 2.667ips 0.789e 14.590u 0:03.750
Performance[5]: 40i 3.236ips 0.820e 14.970u 0:03.090
Performance[6]: 40i 3.802ips 0.842e 15.280u 0:02.630
Performance[7]: 40i 4.274ips 0.857e 15.540u 0:02.340
Performance[8]: 40i 4.831ips 0.872e 15.680u 0:02.070

Higher values of IPS (iterations-per-second) indicate better performance. In the example above, 8 cores are optimal. However, certain workloads perform best restricted to a single thread (e.g., using -limit thread 1).

Disabling OpenMP

If you need to disable OpenMP entirely, add --disable-openmp to your configure script command line, then rebuild and reinstall ImageMagick.