Inner Loops: A Sourcebook for Fast 32-Bit Software Design by Rick Booth (1997)

Inner Loops focuses on squeezing maximum performance from 32-bit software by attacking the hot inner loops that dominate runtime. It presents low-level techniques rooted in processor microarchitecture - register allocation, instruction selection and ordering, addressing modes, memory alignment, cache behavior, pipelining and branch costs - and shows how those factors translate into measurable cycle and throughput differences. Coverage emphasizes practical interaction between C compilers and hand-written assembly, when to trust optimizing compilers and when to intervene, plus the trade-offs between maintainability and raw speed.

Concrete material includes a catalog of optimized inner-loop patterns and implementation recipes for common operations (data movement, string and memory primitives, arithmetic kernels), guidance on loop unrolling and strength reduction, and advice on exploiting available instruction set extensions and calling conventions. Performance measurement and profiling techniques are integral: how to identify true hotspots, construct realistic benchmarks, and validate that an optimization yields real gains across processor variants. The result is a hands-on sourcebook for developers who must write or tune performance-critical 32-bit code, with practical examples and rules of thumb that map hardware realities to better software design.