Logic synthesis and technology-mapping research with two separately evaluated projects:
| Project | Scope | Entry |
|---|---|---|
| eGraphMap / emapper | AIG-to-standard-cell mapping, currently single-output and delay/area lexicographic optimization | emapper/README.md |
| ePMapping | Post-mapping area recovery under the initial mapped delay bound | epmapping/README.md |
| eGraphMap paper | DAC draft | emapper_paper/README.md |
| ePMapping paper | DATE paper and talk | epmapping_paper/README.md |
| Experiments | Published compact records, failures, provenance and selected CEC artifacts | data/README.md |
| Dependencies | Pinned baseline sources and the fixed project mockturtle overlay | tools/BASELINES.md |
WORKSPACE.md gives the workspace layout. Current evidence explicitly separates incomplete experiments, search timeouts, verified incumbents and completed optimizations. Do not combine these two projects' objectives, budgets or results.
Initialize the pinned submodules; E-morphic is the local name of upstream E-Syn2:
git submodule update --init
cmake -S . -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build --target epmapping_check baseline_mapper --parallel 4
cargo build --release --bin bench_rs_egraph_mapCMake requires a C++17 compiler, Python interpreter/development module, Coin-OR CBC dependencies, and Rust/Cargo for the project bridge. Experiment scripts additionally use Yosys, ABC and the Python packages recorded in their source/build provenance.
Before consumers compile, CMake verifies/applies four fixed project headers from
tools/patches/mockturtle/overlay. The corresponding
patches are under tools/patches/mockturtle/. Identical installed files are not
rewritten; conflicting extra edits stop the build. These are named project
extensions, not an unmodified upstream baseline.
python3 scripts/maintenance/test_mockturtle_overlay.py
python3 scripts/experiments/tests/test_mockturtle_patch.py
python3 scripts/maintenance/verify_published_data.py --deepCurrent eGraphMap studies use one600-second mapping-related search window; independent CEC is excluded. Historical budgets remain attached to their original cohorts. Baseline source/algorithms are not optimized; historical E-Syn source patches are disclosed separately from the newly registered clean upstream source.
The compact data export is versioned under data/, with per-file and per-archive
SHA-256 manifests. Large raw experiments and build caches remain local; this is
not a151GB raw-data mirror. The latest legacy database is exported losslessly as
compressed parts; the older root Git LFS object retains its original history.
Shell configuration, credentials and editor state are not publication inputs.
The earlier root documentation remains in docs/history/README-through-20261023.md.