Release-over-release — Fission's own trend, not a re-ranking
Every archived public/benchmark-history/<version>.json is Fission's own measured envelope at that release, keyed by toolchain.fission_version. Other decompilers are shown for stability context only — this page never re-ranks tools.
Semantic pass rate over time. Dot size shows the corpus size measured at that release — watch for it changing alongside the rate, since a bigger corpus is a harder bar to clear.
Dot size ∝ corpus size measured at that release (10–10 Fission rows). Hollow dots are smoke-profile runs; filled dots are official publications.
v0.1.4 → v0.1.6Same corpus subset, same oracle, two Fission releases — every number below is computed only over the (function, compiler variant) pairs measured in bothruns, so a bigger or smaller corpus in either release can't skew the comparison. Fission's own numbers are what matter here — other tools are shown as a stability reference, not a re-ranking.
⚠️ Only 10functions overlap between these two releases' measured corpora — one flipped result can move the percentage a lot. Treat this delta as a signal to look at the function list below, not a precise measurement.
| Decompiler | v0.1.4 semantic | v0.1.6 semantic | Δ | v0.1.4 perfect | v0.1.6 perfect | Δ | Rows compared |
|---|---|---|---|---|---|---|---|
fissionthis project | 95.0% | 100.0% | ▲ +5.0% | 9 | 10 | ▲ +1 | 10 |
ghidra | 19.3% | 19.3% | ±0.0% | 1 | 1 | ±0 | 10 |
fibonacci gcc -O0