colortwist

This is a little finger exercise in SIMD-optimization. The example we try to optimize is an image processing operation known as "color twist". It is a matrix-multication of an RGB-triple and a given matrix. So, if we have the source pixel values r, g, b, then the destination pixels R, G, B are calculated as

$$ R = t_{11} \cdot r + t_{12} \cdot g + t_{13} \cdot b + t_{14} $$

$$ G = t_{21} \cdot r + t_{22} \cdot g + t_{23} \cdot b + t_{24} $$

$$ B = t_{31} \cdot r + t_{32} \cdot g + t_{33} \cdot b + t_{34} $$

where

$$ \begin{equation} T = \begin{pmatrix} t_{11} & t_{12} & t_{13} & t_{14} \\ t_{21} & t_{22} & t_{23} & t_{24} \\ t_{31} & t_{32} & t_{33} & t_{34} \end{pmatrix} \end{equation} $$

is the color twist matrix.

There are SIMD-optimized versions leveraging AVX-instructions (for x86) and Neon-instructions (for ARM).

How fast is it?

Here are some numbers (1MB = 1,000,000 bytes):

Intel Core i7-8700K @ 3.70GHz, msvc 19.27.29112, x64 : 2048x2048 RGB48 bitmap

version	performance in MB/s
colorTwistRGB48_C	1226.24
colorTwistRGB48_IPP	1720.98
colorTwistRGB48_AVX	3291.02
colorTwistRGB48_AVX2	4297.74
colorTwistRGB48_AVX3	5175.55
colorTwistRGB24_C	318.01
colorTwistRGB24_IPP	892.93
colorTwistRGB24_AVX3	2966.55

Intel Core i7-8700K @ 3.70GHz, Intel icl 19.1.3.311, x64 : 2048x2048 RGB48 bitmap

version	performance in MB/s
colorTwistRGB48_C	1313.22
colorTwistRGB48_IPP	1809.26
colorTwistRGB48_AVX	3354.22
colorTwistRGB48_AVX2	4452.38
colorTwistRGB48_AVX3	5207.50
colorTwistRGB24_C	340.01
colorTwistRGB24_IPP	922.14
colorTwistRGB24_AVX3	3180.88

Intel Core i7-8700K @ 3.70GHz, gcc 10.2.0, x64 : 2048x2048 RGB48 bitmap

version	performance in MB/s
colorTwistRGB48_C	1453.32
colorTwistRGB48_AVX	3334.76
colorTwistRGB48_AVX2	4315.26
colorTwistRGB48_AVX3	5011.39
colorTwistRGB24_C	328.43
colorTwistRGB24_AVX3	2998.93

Raspberry Pi 4, gcc 8.3.0-6, 32bit : 2048x2048 RGB48 bitmap

version	performance in MB/s
colorTwistRGB48_C	366.99
colorTwistRGB48_NEON	410.42
colorTwistRGB48_NEON2	1009.3
colorTwistRGB24_C	96.14
colorTwistRGB24_NEON2	480.69

Raspberry Pi 4, gcc 10.3.0, 64bit : 2048x2048 RGB48 bitmap

version	performance in MB/s
colorTwistRGB48_C	439.89
colorTwistRGB48_NEON	417.84
colorTwistRGB48_NEON2	942.86
colorTwistRGB24_C	120.64
colorTwistRGB24_NEON2	462.34

Raspberry Pi 3b, gcc 8.3.0-6, 32bit : 2048x2048 RGB48 bitmap

version	performance in MB/s
colorTwistRGB48_C	208.69
colorTwistRGB48_NEON	187.42
colorTwistRGB48_NEON2	435.30
colorTwistRGB24_C	63.30
colorTwistRGB24_NEON2	285.971

AMD Ryzen 7 4700U, msvc 19.27.29112, x64 : 2048x2048 RGB48 bitmap

version	performance in MB/s
colorTwistRGB48_C	1202.02
colorTwistRGB48_IPP	1836.64
colorTwistRGB48_AVX	3810.53
colorTwistRGB48_AVX2	4970.00
colorTwistRGB48_AVX3	5572.38
colorTwistRGB24_C	297.13
colorTwistRGB24_IPP	920.24
colorTwistRGB24_AVX3	2548.42

AMD Ryzen 7 4700U, Intel icl 19.1.3.311, x64 : 2048x2048 RGB48 bitmap

version	performance in MB/s
colorTwistRGB48_C	1355.20
colorTwistRGB48_IPP	1862.04
colorTwistRGB48_AVX	3881.71
colorTwistRGB48_AVX2	5150.99
colorTwistRGB48_AVX3	5736.88
colorTwistRGB24_C	287.26
colorTwistRGB24_IPP	952.50
colorTwistRGB24_AVX3	2508.71

Intel Core m3-6Y30 @ 0.9GHz, msvc 19.27.29112, x64 : 2048x2048 RGB48 bitmap

version	performance in MB/s
colorTwistRGB48_C	650.32
colorTwistRGB48_IPP	935.62
colorTwistRGB48_AVX	1769.30
colorTwistRGB48_AVX2	2370.21
colorTwistRGB48_AVX3	2943.29
colorTwistRGB24_C	157.44
colorTwistRGB24_IPP	466.74
colorTwistRGB24_AVX3	1793.66

Samsung Galaxy Book Go (Snapdragon 7c Gen2 @ 2.55GHz), msvc 19.29.30133.0, ARM64 : 2048x2048 RGB48 bitmap

version	performance in MB/s
colorTwistRGB48_C	827.08
colorTwistRGB48_NEON	1252.92
colorTwistRGB48_NEON2	2758.96
colorTwistRGB24_C	222.86
colorTwistRGB24_NEON2	1201.04

Samsung Galaxy Book Go (Snapdragon 7c Gen2 @ 2.55GHz), msvc 19.29.30133.0, ARM : 2048x2048 RGB48 bitmap

version	performance in MB/s
colorTwistRGB48_C	583.61
colorTwistRGB48_NEON	1241.57
colorTwistRGB48_NEON2	2581.25
colorTwistRGB24_C	208.34
colorTwistRGB24_NEON2	1336.50

AMD Ryzen 5 1600X, gcc 7.5.0, Linux x64 : 2048x2048 RGB48 bitmap

version	performance in MB/s
colorTwistRGB48_C	1344.27
colorTwistRGB48_AVX	3186.97
colorTwistRGB48_AVX2	3445.82
colorTwistRGB48_AVX3	4008.93
colorTwistRGB24_C	281.88
colorTwistRGB24_AVX3	2014.70

Intel Core i9-12900K @ 3.20GHz, msvc 19.31.31106.2, x64 : 2048x2048 bitmap

version	performance in MB/s
colorTwistRGB48_C	1928.36
colorTwistRGB48_IPP	2203.86
colorTwistRGB48_AVX	4058.93
colorTwistRGB48_AVX2	8323.63
colorTwistRGB48_AVX3	9091.08
colorTwistRGB24_C	440.10
colorTwistRGB24_IPP	1116.93
colorTwistRGB24_AVX3	5441.54

MacBook Pro M2, clang-1400.0.29.202, arm64 : 2048x2048 bitmap

version	performance in MB/s
colorTwistRGB48_C	5301.62
colorTwistRGB48_NEON	7226.25
colorTwistRGB48_NEON2	9188.08
colorTwistRGB24_C	2863.86
colorTwistRGB24_NEON2	4584.36

Lenovo Yoga Slim 7x, MSVC 19.41.34123.0, arm64 : 2048x2048 bitmap

version	performance in MB/s
colorTwistRGB48_C	1733.64
colorTwistRGB48_NEON	10817.1
colorTwistRGB24_C	353.98
colorTwistRGB24_NEON2	5184.21

Building

For Windows-on-ARM, run the following:

mkdir build
cmake .. -G "Visual Studio 16 2019" -A ARM64
cmake  --build . --config Release

For a 32-bit build, use

mkdir build
cmake .. -G "Visual Studio 16 2019" -A ARM
cmake  --build . --config Release

For x86-Windows, use for x64-build:

mkdir build
cmake .. -G "Visual Studio 16 2019" -A x64
cmake  --build . --config Release

For x86-build, run

mkdir build
cmake .. -G "Visual Studio 16 2019" -A Win32
cmake  --build . --config Release