The Go side of the secp256k1 Metal anchor is built only under cgo && lux_cevm_native && darwin. Its C file carried no constraint at all, so every other build still saw a C source in a package it was compiling without cgo, and refused: "C source files not allowed when not using cgo or SWIG". That took out `go build ./...` for the whole repository, not just this package. A file and its other half now agree about when they exist. Under the tag the anchor is still compiled in — go list reports it in CFiles — and without it, neither half is. Co-authored-by: Hanzo Dev <[email protected]>
cevm — C++ EVM execution backend
The CGO bridge to the C++ EVM in ~/work/luxcpp/cevm. Selected at link time
via build tags from chains/evm/backend_{cgo,nocgo}.go.
This is not a standalone VM. It is the GPU-accelerated execution backend
for C-Chain when built with LUX_CGO=1. The pure-Go fallback (luxfi/geth)
is used when LUX_CGO=0.
Layout
chains/evm/
├── main.go # luxd plugin entry
├── backend_cgo.go # cgo backend selector (LUX_CGO=1)
├── backend_nocgo.go # pure-Go selector (LUX_CGO=0)
└── cevm/ # this package — C++ FFI shim
├── cevm.go # Go API
├── cevm_cgo.go # cgo bindings (linked against luxcpp libs)
├── cevm_nocgo.go # no-op stubs for !cgo builds
└── plugin.go # plugin registration glue
Linking
cevm_cgo.go references ${SRCDIR}/../../../../luxcpp/... for headers and
libs. From ~/work/lux/chains/evm/cevm/ that resolves to ~/work/luxcpp/.
For module-cache builds, see accel's fetch-luxcpp.sh pattern (TODO: port
the same approach here).
Performance
When CGO is enabled and luxcpp libs are available:
- ~3–5× the throughput of pure-Go EVM
- SIMD opcode dispatch (AVX2/NEON)
- GPU batch operations via Block-STM (CUDA/Metal)
Provenance
Folded into chains/evm from the previously-standalone luxfi/cevm repo on
2026-04-30. cevm is internal-only — no operator daemon. C-Chain runs in luxd.
See ~/work/lux/chains/PLUGGABLE.md for the canonicalization pattern that
governs the rest of the chain VMs.