Profile guided optimization (PGO) refers to compiler techniques that use data collected from actual program execution — branch outcomes, hot and cold code paths, call frequencies — to make optimization decisions such as code layout and basic block reordering. In this dataset, the claims most often centre on how that execution profile is captured, how sampling frequency and instrumentation overhead are balanced, and how stale profile data from an earlier program version is detected and corrected. It is a compiler and runtime discipline rather than a hardware one, which is reflected in the dominance of the G06F classification across 94.6% of the 185 records in scope.

The assignee ranking covers 56 companies, with a single leader holding 37 of the 185 records in scope, well ahead of fifth place at 9 records. The top 5 assignees together account for 61.1% of all records, and the top 10 extend that to 78.9%, so filing activity is concentrated among a small group rather than spread evenly. Beyond that group, the ranking thins quickly into a long tail of single- or low-digit filers.

Filings rose from 1 in 2021 to 3 in 2024, a 200% increase over that span and the most recent period that can be treated as filing-complete. That said, the field’s peak year so far was 2020 at 13 filings, so recent activity remains well below that high point. Counts for 2025 and 2026 will still rise as publications catch up, since publication typically lags filing by around 18 months — treat those most recent years as provisional, not as evidence of a slowdown.