AtrekesLibHFTDocumentationBenchmarksComplete roster

Complete target and evidence inventory

Every target stays visible.

Codec and full-session status for all 35 registered targets across six language groups. Missing, excluded and non-parity results remain in the table.

Source revision 86263971fa08Source date 2026-10-03Four canonical flows

Read the evidence grade first

The target engine comparison by language defines every engine this campaign aims to measure. The separate coverage record says which results have passed their gates. Only S5 pass on all four canonical flows is eligible for a report-grade full-engine comparison. S5 diag, preflight, smoke, non-parity and not-run rows are visible context, not speed claims. A network floor is not a FIX engine.

C++ · C · Java · .NET · Rust · Go

This table is rendered from the same LibHFT source data as those two Markdown documents, at revision 86263971fa08. Download the complete machine-readable status →

C++

EngineKindLicenceCodec targetCodec statusFull targetFull statusNOS → ERMarket dataMD → order → ERTick → tradeNotes
tcp-ping-pong-cppfloor-n/an/an/anetwork floor————network floor: the RTT before any FIX processing
libhft-cpp (C++)libhftlibhftrequiredimplementedrequiredS5 pass 4/4S5 pass · canonicalS5 pass · canonicalS5 pass · canonicalS5 pass · canonical
quickfix_cppengineQuickFIX SL 1.0requiredimplementedrequiredS5 diag 4/4S5 diag · canonicalS5 diag · non-parityS5 diag · non-parityS5 diag · non-parity
nexusfixengineMITrequiredimplementedrequiredS5 diag 4/4S5 diag · canonicalS5 diag · canonicalS5 diag · canonicalS5 diag · canonicalpinned to untagged upstream snapshot ffa8b0a; v1.0.0 lacks the public engine headers
fix8engineLGPL-3.0requiredimplementedrequiredS5 diag 4/4S5 diag · canonicalS5 diag · canonicalS5 diag · canonicalS5 diag · canonicalD/8/depth-10 W full-message codec; generated benchmark dictionary; 3/3 mutation probes passed; public codec admission recorded 2026-10-02
llfixengineMITadmissionpartialrequiredblockedblockedblockedblockedblockedpublic encoder only: D/8/depth-10 W; cold field/framing checks passed; no receive parser; adapter ready; retained session, store/recovery and host evidence missing
hffixcodecBSD-2-Clause-stylerequiredimplementedn/acodec only————
fixppcodecMITrequiredimplementedn/acodec only————
robaho_cpp_fix_codeccodecnoneexcludedexcludedn/acodec only————no licence at 4a95e7c: Stage 2 adapter removed; archived timings withheld; no tested session endpoint
openfixenginenoneexcludedexcludedexcludedblockedblockedblockedblockedblockedno licence; Bazel-only; gnu++23
robaho_cpp_fix_engineengineunknownexcludedexcludedexcludedblockedblockedblockedblockedblockedupstream build fails; licence unknown

C

EngineKindLicenceCodec targetCodec statusFull targetFull statusNOS → ERMarket dataMD → order → ERTick → tradeNotes
libtradingengineBSD-2-ClauserequiredimplementedrequiredS5 diag 4/4S5 diag · canonicalS5 diag · non-parityS5 diag · non-parityS5 diag · non-paritypublic numeric-tag parser/unparser; depth-10 W requires the existing 128-field capacity patch; upstream long dormant

Java

EngineKindLicenceCodec targetCodec statusFull targetFull statusNOS → ERMarket dataMD → order → ERTick → tradeNotes
tcp-ping-pong-javafloor-n/an/an/anetwork floor————network floor: the RTT before any FIX processing
libhft-java-pure (Java Pure)libhftlibhftrequiredimplementedrequiredS5 pass 4/4S5 pass · canonicalS5 pass · canonicalS5 pass · canonicalS5 pass · canonical
libhft-java-jni (Java/JNI)libhftlibhftrequiredimplementedrequiredS5 pass 4/4S5 pass · canonicalS5 pass · canonicalS5 pass · canonicalS5 pass · canonical
philadelphiaengineApache-2.0requiredimplementedrequiredS5 pass 4/4S5 pass · canonicalS5 pass · canonicalS5 pass · canonicalS5 pass · canonical
philadelphia-fastengineApache-2.0n/an/arequiredS5 pass 4/4S5 pass · canonicalS5 pass · canonicalS5 pass · canonicalS5 pass · canonicala transport configuration of philadelphia, not a separate codec
crossfixenginecommercial, licensedexcludedexcludedconditionalS5 pass 4/4S5 pass · canonicalS5 pass · canonicalS5 pass · canonicalS5 pass · canonicalits obfuscated buffer API cannot be reset and reparsed fairly; CrossFIX 1.6; 2.0 is not headline-grade; licensed installation only
quickfixjengineQuickFIX SL 1.0requiredimplementedrequiredS5 diag 4/4S5 diag · canonicalS5 diag · canonicalS5 diag · canonicalS5 diag · canonicalholds 750 msg/s; misses the contract rate
artioengineApache-2.0requiredimplementedrequiredS5 diag 4/4S5 diag · canonicalS5 diag · canonicalS5 diag · canonicalS5 diag · canonicalsource-owned D/8/W dictionary generation; retained host smoke/parity required; next run pinned at upstream 0.182
falconengineBSD-2-Clauseadmissionpartialn/aS3 smoke 1/4S3 smoke · canonicalunsupportedunsupportedunsupportedD and 8 only: no V/W/X market-data types; no market-data messages, so N/A for the four flows

.NET

EngineKindLicenceCodec targetCodec statusFull targetFull statusNOS → ERMarket dataMD → order → ERTick → tradeNotes
tcp-ping-pong-netfloor-n/an/an/anetwork floor————network floor: the RTT before any FIX processing
libhft-dotnet-managed (.NET Pure)libhftlibhftrequiredimplementedrequiredS5 pass 4/4S5 pass · canonicalS5 pass · canonicalS5 pass · canonicalS5 pass · canonicalSCAN mlx5 DAC S5 pass 4/4; HP Solarflare repeat pending
libhft-dotnet-native (.NET Native)libhftlibhftrequiredimplementedrequiredS5 pass 4/4S5 pass · canonicalS5 pass · canonicalS5 pass · canonicalS5 pass · canonicalSCAN mlx5 DAC S5 pass 4/4; HP Solarflare repeat pending
quickfix_nengineQuickFIX SL 1.0requiredimplementedrequiredS5 diag 4/4S5 diag · canonicalS5 diag · canonicalS5 diag · canonicalS5 diag · canonicalSCAN mlx5 DAC S5 diagnostic 4/4: canonical wire, but report rate below 10k/s
fixantenna_netengineApache-2.0admissionimplementedadmissionS5 diag 4/4S5 diag · canonicalS5 diag · canonicalS5 diag · canonicalS5 diag · canonicalpublic RawFixUtil/Message codec; source pinned ff38f51; stage-2 parity passed; SCAN mlx5 DAC S5 diagnostic 4/4: canonical wire, but report rate below 10k/s

Rust

EngineKindLicenceCodec targetCodec statusFull targetFull statusNOS → ERMarket dataMD → order → ERTick → tradeNotes
tcp-ping-pong-rustfloor-n/an/an/anetwork floor————closed-loop TCP RTT before FIX processing; bench/rust/ping-pong
libhft-rust-pure (Rust)libhftlibhftrequiredimplementedrequiredS4 diag 4/4S4 diag · canonicalS4 diag · canonicalS4 diag · canonicalS4 diag · canonicalcanonical four-flow two-process localhost diagnostics; physical Stage-5 integration pending
libhft-rust-native (Rust over the C ABI)libhftlibhftparkedparkedparkedparked————parked 2026-09-11; parked 2026-09-11
dfxengineMPL-2.0requiredimplementedrequiredS5 diag 4/4S5 diag · canonicalS5 diag · non-parityS5 diag · non-parityS5 diag · non-parityderived from QuickFIX/n; market data needs the exact ten-level contract
fixer_rsengineApache-2.0requiredimplementedrequiredS5 diag 4/4S5 diag · canonicalS5 diag · non-parityS5 diag · non-parityS5 diag · non-paritya port of QuickFIX/Go; market data needs the exact ten-level contract
truefixengineApache-2.0 or MITrequiredimplementedadmissionS4 diag 4/4S4 diag · canonicalS4 diag · canonicalS4 diag · canonicalS4 diag · canonicalcanonical four-flow Stage 4 diagnostics with V/W dictionary overlay; mixed-flow rate fails and tick stability warns; Stage 5 pending
nanofixengineMITrequiredimplementedadmissionS4 diag 4/4S4 diag · canonicalS4 diag · canonicalS4 diag · canonicalS4 diag · canonicalcanonical four-flow two-process Stage 4 medium diagnostics pass with verified admin observer overlay; physical Stage 5 pending
ferrumfixcodecApache-2.0 or MITrequiredimplementedn/acodec only————through the pinned v0.7.0 shim

Go

EngineKindLicenceCodec targetCodec statusFull targetFull statusNOS → ERMarket dataMD → order → ERTick → tradeNotes
quickfix_goengineQuickFIX SL 1.0requiredimplementedrequiredS5 diag 4/4S5 diag · canonicalS5 diag · canonicalS5 diag · canonicalS5 diag · canonicalcross-runtime anchor: libhft has no Go flavour

Latest qualified codec report

The 3 October 2026 converted-values-v2 campaign admits 25 adapters, 320 isolated field-work probes and 554 operation rows. It replaces the historical parser rankings with complete checked converted-value work.

Per-language top-five CDFs and full percentile tables → · Full qualified source report →

Retained host reports

These are historical committed host reports, not the new codec ranking above. Earlier parse timings do not establish the corrected field-work contract. Each report carries its own workload, profile and validity notes; rows from different hosts, rates or stages must not be merged into one ranking.

Browse every committed benchmark file on Atrekes → · View the source tree at this revision →