How much faster is FastMath compared to Delphi's built-in math?

Benchmarks show speedups from 1.2x to over 90x, depending on the operation. For example, FastSinCos on Arm64 is up to 93.3x faster than RTL's SinCos for vectorized calculations.

Can I use FastMath for accurate scientific simulations?

No, FastMath is not recommended for scientific calculations due to its approximate functions and single-precision focus. It trades accuracy for speed, making it unsuitable where precision is critical.

How to disable SIMD optimizations in FastMath?

Compile with the FM_NOSIMD define to use plain Pascal versions instead of SIMD code. This is useful for compatibility or performance comparison, but will significantly reduce speed benefits.

FastMath or Delphi RTL math for game development?

FastMath is generally better for games due to its speed optimizations, especially for vector and matrix operations. However, if you need high precision or robust error handling, the RTL might be safer but slower.

Does FastMath support double-precision floats?

No, FastMath only operates on single-precision (Single) values. If your project requires double-precision arithmetic, you'll need to use the Delphi RTL or another library.

What's the best way to handle errors with FastMath?

Call DisableFloatingPointExceptions at startup to suppress exceptions; errors return values like NaN. You must check for these extreme values in your code to handle failures properly.

FastMath

NOASSERTIONPascal

A high-performance Delphi math library using SIMD assembly and approximate functions for multimedia and games.

GitHub

What is FastMath?

FastMath is a Delphi library optimized for high-performance mathematical computations, leveraging hand-tuned assembly code and SIMD instructions to significantly outperform standard Delphi RTL math functions. It is designed for performance-critical applications like games and multimedia, where calculation speed often outweighs the need for perfect accuracy or extensive error checking. The library provides vector and matrix types with overloaded operators, as well as approximate functions prefixed with 'Fast' for dramatic speed gains.

Target Audience

Delphi developers building performance-intensive applications such as real-time games, multimedia software, and animation tools where mathematical computation speed is critical. It is also suitable for developers needing SIMD-accelerated vector and matrix operations across x86 and Arm platforms.

Value Proposition

Developers choose FastMath for its substantial performance improvements—up to 90x faster for some operations—through SIMD optimization and approximate functions, while maintaining interoperability with Delphi's standard math types. Its cross-platform support and design philosophy prioritizing speed over exhaustive precision make it ideal for real-time interactive applications.

Overview

Fast Math Library for Delphi

Use Cases

Best For

Related Projects

Community-curated · Updated weekly · 100% open source

Found a gem we're missing?

Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.

Submit a project Star on GitHub

185 stars43 forks0 contributors

Developing real-time 3D games in Delphi requiring fast vector and matrix operations.
Building multimedia applications in Delphi that need high-speed mathematical computations for audio or video processing.
Optimizing animation and graphics rendering in Delphi projects where approximate trigonometric functions (like FastSin) are acceptable.
Creating cross-platform Delphi applications targeting both desktop (x86) and mobile (Arm) with consistent SIMD-accelerated math performance.
Integrating with existing Delphi RTL math types (like TPointF or TMatrix3D) while upgrading to faster, SIMD-optimized alternatives.
Performing linear algebra operations in Delphi, such as matrix inversions or multiplications, with up to 10x speed improvements over standard libraries.

Not Ideal For

Scientific or financial applications requiring high precision and rigorous error checking
Projects that depend on double-precision floating-point arithmetic
Development targeting the iOS simulator, as it lacks hardware SIMD acceleration and uses slower Pascal code
Systems where detailed floating-point exception handling is critical for debugging or safety

Pros & Cons

Pros

SIMD-Optimized Speed

Uses SSE2 and NEON instructions to process vectors in parallel, achieving up to 90x faster operations like FastSinCos, as shown in detailed benchmarks across platforms.

Cross-Platform Consistency

Delivers significant performance improvements on x86-32/64, Arm32, and Arm64 architectures, with benchmarks showing consistent speedups for vector and matrix operations.

Seamless RTL Integration

Types are memory-compatible with Delphi's standard math types like TPointF and TMatrix3D, allowing easy typecasting and conversion without data restructuring overhead.

Approximate Functions for Real-Time

Provides Fast-prefixed functions that trade minimal accuracy for massive speed gains, ideal for games and graphics where frame rate is a priority, as noted in the README.

Cons

Single-Precision Only

Lacks support for double-precision floating-point arithmetic, limiting its use in applications requiring higher numerical accuracy, as admitted in the README.

Accuracy Compromises

The approximate Fast-functions sacrifice precision for speed, making them unsuitable for scientific or financial calculations where exact results are essential.

SIMD Dependency

Optimal performance requires SSE2 or NEON support; while common, disabling it via FM_NOSIMD reduces benefits, and the iOS simulator uses unoptimized Pascal code.

Minimal Error Handling

Recommends disabling floating-point exceptions, which can mask errors by returning NaN or Infinity instead of raising exceptions, complicating debugging and error recovery.

Frequently Asked Questions

Home

Pascal

CAI Neural API

CAI NEURAL API - Pascal based deep learning neural network API optimized for AVX, AVX2 and AVX512 instruction sets plus OpenCL capable devices including AMD, Intel and NVIDIA.

Stars417

Forks206

Last commit2 days ago

ALGLIB

. [Delphi] [FPC] Cross-platform numerical analysis and data processing library. It supports several operating systems (Windows and POSIX, including Linux). ALGLIB features include: Data analysis (classification/regression, statistics); Optimization and nonlinear solvers; Interpolation and linear/nonlinear least-squares fitting; Linear algebra (direct algorithms, EVD/SVD), direct and iterative linear solvers; Fast Fourier Transform and many other algorithms. // Free edition is Delphi wrapper around generic C core licensed for Personal and Academic Use

. [Delphi] [FPC] Big Floating Point, Big Integers, Astronomical calcualtions

. [Delphi] Multi precision integer, rational, real, and complex arithmetic

Stars0

Forks0