Same I3S payload. Faster than the Esri C++ reference. On every kernel.
TurboLEPCC is a memory-safe, SIMD-first encoder and decoder for the LEPCC format used by Esri's ArcGIS I3S point-cloud payloads. Byte-exact parity with Esri's C++ reference on every test case — and faster on every measured workload.
Drop-in for Esri liblepcc (Apache-2.0). SIMD backends: AVX2, NEON, WASM SIMD128, SVE2 (nightly), scalar fallback.
By the numbers
A complete win across the matrix.
Beats the Esri C++ reference on every kernel except two — both tied within 4%.
Two outliers are a macOS-clang optimization gap, not a TurboLEPCC regression.
Color encode (AVX2) on the compressible repeated-color path.
Faster than the MoCongxin port — the only other memory-safe port.
The full kernel matrix
Every subsystem. Both architectures. No cherry-picks.
LEPCC ships four subsystems behind one API — point geometry, LiDAR intensity, per-point color, and classification flags. Every kernel was benchmarked head-to-head against Esri's own C++ reference, on both Windows AVX2 and macOS NEON, across a real LiDAR scan and synthetic batches at multiple sizes.
Source: turbo-lepcc-rs criterion benchmarks vs the Esri C++ reference (liblepcc, Apache-2.0), median time across a small real LiDAR scan plus mid-size and large synthetic point batches. Whiskers show the spread across those workloads; a single bar means the workloads tied within 5%.
The memory-safe LEPCC field
The only memory-safe port that beats the Esri C++ reference.
There is exactly one other memory-safe LEPCC implementation in public — and it's roughly a generation behind. Log scale below: small differences are meaningful, big ones are decisive.
Decode time for a small real LiDAR point cloud (a public Esri reference scan), macOS aarch64 NEON. Log scale because MoCongxin's memory-safe port (the only other one that exists) sits roughly a generation behind. TurboLEPCC is the only memory-safe port that beats the Esri C++ reference.
Prove It
Race it yourself. Right here.
A real ~1.06M-point LiDAR scan decoded in your browser — TurboLEPCC against Esri's own libcc wasm on the same bytes — rendered georeferenced over satellite imagery, with BVC and LAZ/COPC lanes to compare. Drag either map; the cameras are linked.
What's in the box
Four LEPCC subsystems, one native API.
XYZ
Point geometry
Quantized 3D coordinates — the bulk of a LiDAR or photogrammetry point cloud.
Up to 2.36× faster than the Esri C++ reference
Intensity
LiDAR return intensity
Per-point 16-bit signal strength — used for surface classification.
Up to 4.10× faster decode (NEON)
RGB
Per-point color
Per-point 8-bit color triplets with cluster-aware compression.
Up to 9.74× faster encode (AVX2)
FlagBytes
Classification flags
Per-point classification and flag bytes — terrain, vegetation, building, etc.
Byte-exact parity with the Esri C++ reference
Correctness model
Bit-exact on the wire, four ways.
"Round-trip equivalence" is not enough. TurboLEPCC asserts byte-identical compatibility with Esri's C++ reference along four independent axes — every test case must satisfy all four.
Axis A
Cross-decoder agreement on Esri payloads
our_decode(esri_blob) == esri_decode(esri_blob)
Every Esri-encoded payload decodes to the same bytes through TurboLEPCC as through Esri's own C++. No drift.
Axis B
Cross-encoder agreement on the wire
our_encode(data) == esri_encode(data)
Encoded output is byte-identical to the Esri C++ reference. Downstream consumers of Esri-encoded data cannot detect a vendor swap.
Axis C
Forward round-trip
our_decode(our_encode(data)) == data
Lossless on every payload TurboLEPCC encodes.
Axis D
Backward round-trip
our_encode(our_decode(blob)) == blob
Re-encoded payloads are byte-identical to the original Esri blob. No silent transcoding.
Why switch
A library swap. Not a migration.
- - Byte-exact wire compatibility with Esri liblepcc on encode and decode — no consumer changes, no transcoding, no risk of silent drift.
- - Memory-safe, zero-unsafe — deployable in WASM (SIMD128 backend) for browser-side I3S point-cloud rendering with no C dependency.
- - Faster than the Esri C++ reference on every kernel measured — not a niche win, a system-wide one.
- - 34–60× faster than the only other memory-safe LEPCC port (decode-only, not benchmarked for encode).
- - All four subsystems behind one native API: point geometry, LiDAR intensity, per-point color, classification flags.
- - Patent inheritance from LEPCC via LERC (US 9,002,126) — same legal footing as Esri's reference.
liblepcc.
Just faster, safer.
Existing pipelines, existing data, existing tooling. Swap the implementation, keep everything else.
Esri, ArcGIS, I3S, and LERC are trademarks of Environmental Systems Research Institute, Inc., used here solely to identify the products and formats being compared. No affiliation, sponsorship, or endorsement is implied. LEPCC is published by Esri under the Apache 2.0 license, and the live demo runs Esri's unmodified reference decoder under that license; TurboLEPCC is an independent, clean-room implementation of the LEPCC format. Performance comparisons reflect our own measurements under the stated methodology; results vary by workload and hardware.