Description

The GR32_Blend unit serves as the primary low-level color compositing engine in Graphics32. It provides high-performance routines for alpha blending, associative alpha merging, linear interpolation, color channel algebra, brightness adjustments, and alpha premultiplication/unpremultiplication.

Key capabilities provided by GR32_Blend include:

• Alpha Blending: Mixes a foreground color $F$ onto an opaque background color $B$ using foreground alpha $F_a$:$$Z=F_a\cdot F_{rgb}+(1-F_a)\cdot B_{rgb}$$
• Associative Alpha Merging: Combines foreground $F$ and semi-transparent background $B$ using Bruce Wallace's associative compositing formula:$$A\text{ over }(B\text{ over }C)=(A\text{ over }B)\text{ over }C$$
• Linear Combination (Combine / Lerp): Performs linear interpolation between two 32-bit colors given a weight $W\in [0,255]$:$$Z=W\cdot X+(1-W)\cdot Y$$
• Color Algebra & Blend Modes: Includes standard color blend operations such as Add, Subtract, Divide, Modulate, Max, Min, Difference, Average, Exclusion, Scale, Screen, Dodge, Burn, BlendColorAdd, and BlendColorModulate.
• Alpha Premultiplication & Unpremultiplication: Conversions between straight ARGB and premultiplied ARGB color formats (Premultiply32, Unpremultiply32, PremultiplyMem, UnpremultiplyMem).
• Optimized Lookup Tables: Precalculated $256\times 256$ lookup tables (MulDiv255Table, DivMul255Table) and SIMD alignment tables (alpha_ptr, bias_ptr) to eliminate runtime division.
• Dynamic Hardware Bindings: Function delegates (BlendReg, BlendMem, BlendMems, etc.) dynamically bound to pure Pascal, x86/x64 assembly, or SSE2 vector implementations at runtime based on CPU feature detection.

---

See also

• Alpha Composition