From Newsgroup: comp.compilers

This Google paper describes an AI approach to invent new compiler optimizations.

Abstract

Modern compilers rely on hand-crafted heuristics to guide optimization
passes. These human-designed rules often struggle to adapt to the
complexity of modern software and hardware and lead to high maintenance
burden. To address this challenge, we present Magellan, an agentic
framework that evolves the compiler pass itself by synthesizing executable
C++ decision logic. Magellan couples an LLM coding agent with evolutionary search and autotuning in a closed loop of generation, evaluation on user-provided macro-benchmarks, and refinement, producing compact
heuristics that integrate directly into existing compilers. Across several production optimization tasks, Magellan discovers policies that match or surpass expert baselines. In LLVM function inlining, Magellan synthesizes
new heuristics that outperform decades of manual engineering for both binary-size reduction and end-to-end performance. In register allocation,
it learns a concise priority rule for live-range processing that matches intricate human-designed policies on a large-scale workload. We also
report preliminary results on XLA problems, demonstrating portability
beyond LLVM with reduced engineering effort.

https://arxiv.org/abs/2601.21096

Regards,
John Levine, johnl@taugh.com, Taughannock Networks, Trumansburg NY
Please consider the environment before reading this e-mail. https://jl.ly
--- Synchronet 3.21b-Linux NewsLink 1.2

From Newsgroup: comp.compilers

John,

A paper with "novel" in the title is a major red flag.

This Google paper describes an AI approach to invent new compiler optimizations.

No they don't. They use an LLM to select the tuning parameters
for a well established optimization, function inlining.

surpass expert baselines. In LLVM function inlining, Magellan synthesizes
new heuristics that outperform decades of manual engineering for both binary-size reduction and end-to-end performance.

"... the continued Gemini-3-Pro run achieves consistent
positive speedups beyond 0%, ultimately surpassing the hand-
tuned baseline by 0.61%."

Figure 3/4 suggests a much bigger improvement, until the reader
realises that the comparison is not against human generated
rules. Results given to two decimal places and no error bars!

In register allocation,
it learns a concise priority rule for live-range processing that matches intricate human-designed policies on a large-scale workload.

This sentence in the abstract goes undiscussed in the paper, which
only looks at inlining.
--- Synchronet 3.21b-Linux NewsLink 1.2