ASE: mathutils.hh: improve fast_log2, add fast_log2_block

[swesterfeld]

• avoid using a union in fast_log2 (fix undefined behaviour)
• make code auto vectorizable by gcc
• add a block version for fast_log2 (should be auto vectorized)
• assume sign bit == 0, makex exponent extraction easier

[tim-janik]

@greptile

[greptile-apps]

Greptile Overview

Greptile Summary

This PR improves the fast_log2 function in ase/mathutils.hh by addressing undefined behavior and enhancing vectorization capabilities. The main changes replace union-based type punning with standards-compliant memcpy operations for float-to-int bit reinterpretation. The original implementation used a union to directly access float bits as integers, which violates C++ strict aliasing rules and constitutes undefined behavior, even though it worked in practice.

The refactored code maintains the same mathematical approach - extracting the IEEE 754 exponent and manipulating the mantissa - but uses memcpy for type conversion, which modern compilers optimize away while remaining standards-compliant. The implementation assumes positive input values (sign bit = 0), simplifying bit manipulation by avoiding sign checks. Additionally, a new fast_log2_block function processes arrays of values in a loop structure that compilers can auto-vectorize, improving performance for bulk audio processing operations common in DSP applications.

These changes integrate well with the codebase's audio synthesis engine (ASE), where fast logarithmic operations are crucial for real-time audio processing, parameter mapping, and frequency domain calculations.

Important Files Changed

Changed Files

Filename	Score	Overview
ase/mathutils.hh	5/5	Refactored fast_log2 to use memcpy instead of union type punning, fixing undefined behavior while maintaining performance and adding vectorizable block function

Confidence score: 5/5

• This PR is safe to merge with minimal risk as it fixes undefined behavior while preserving functionality
• Score reflects well-understood changes that replace problematic code with standards-compliant equivalents without altering the mathematical logic
• No files require special attention as the changes are localized to a single header with clear improvements

_{1 file reviewed, 2 comments}

_{Edit Code Review Agent Settings | Greptile}

[greptile-apps]

style: The comment 'rely on sign bit == 0' indicates this function assumes positive inputs only. Consider adding input validation or documenting this constraint in the function comment.

Prompt To Fix With AI

This is a comment left during a code review.
Path: ase/mathutils.hh
Line: 108:108

Comment:
**style:** The comment 'rely on sign bit == 0' indicates this function assumes positive inputs only. Consider adding input validation or documenting this constraint in the function comment.

How can I resolve this? If you propose a fix, please make it concise.

[greptile-apps]

style: Parameter should be const float *values since the input values are being read, not the pointer itself being modified.

Prompt To Fix With AI

This is a comment left during a code review.
Path: ase/mathutils.hh
Line: 133:133

Comment:
**style:** Parameter should be `const float *values` since the input values are being read, not the pointer itself being modified.

How can I resolve this? If you propose a fix, please make it concise.