Understanding how to squeeze performance from very long instruction word (VLIW) systems can be as much about strategy as hardware.
This book explains compiler-driven techniques that unlock fine‑grained parallelism and turn complex architectures into practical, high‑throughput machines.
Readers will discover methods to map serial code to VLIW pipelines, balancing local and global scheduling, region transformations, and software pipelining. The discussion covers trace scheduling, percolation scheduling, region scheduling, and the evolution of VLIW ideas from horizontal microcode to modern compilers.
- How VLIW hardware and compilers collaborate to control resources and timing.
- Strategies for local and global code compaction, and when each is effective.
- Techniques like trace scheduling, percolation scheduling, and region scheduling.
- Foundations of software pipelining and its advantages over loop unrolling.
Ideal for readers of computer architecture, compiler design, and parallel processing who want to understand how to generate efficient code for VLIW machines.