Compare Engine Results

Use this reference to choose a comparison report and explain what agreed. A runtime implementation is the program that evaluates your Lispex source or compiled instructions.

Read a Lispex engine comparison

The bytecode guide provides the complete rule.lspx example. It prints refund with display and newline. With the Native CLI, run:

lispex compare-engines --receipt tree-vm.json rule.lspx

Native writes the report, prints this summary, and exits 0.

JSON

{
  "all_agree": true,
  "receipt": "tree-vm.json",
  "schema": "lispex.compare-engines-summary/v1",
  "semantic_mismatch_axes": []
}

Both engines recorded refund\n as stdout and completed successfully. all_agree applies to the checked semantic observations for this source. Open tree-vm.json to read the individual results and identities. Choose a new report path. An existing file causes exit 3 and remains unchanged.

Choose a record

Record format	Compared or recorded data	Where to start
`lispex.rust-engine-comparison/v1`	Lispex Interpreter and Lispex VM, using one derived Core IR and bytecode file	`compare-engines` in the bytecode guide
`csk.meaning-differential-report/v1`	Lispex Interpreter behavior and Meaning Environment v1 for the selected source	`meaning-diff` in the CLI reference
`csk.differential-receipt/v0`	Recorded cases from the historical Rust, LIL, and LIT corpus	Vouch receipts
`sicp-observation/v1`	One SICP execution's termination, value, stdout, diagnostics, warnings, and artifact identity	SICP guide

A corpus is a named set of programs used for comparison. A SICP observation records one execution. Its lispex-trace/v1 trace records execution steps. Use the relevant format to interpret the recorded fields.

Separate semantic results from resource usage

Start with completion status, then read the reported differences. Compare the source, input, executor identities, profiles, and output channels named in the report.

Read semantic differences and resource differences separately. The Lispex VM uses the bytecode cost model, and the Lispex Interpreter uses its own depth model. Lispex Interpreter and Lispex VM lists the resource observations each comparison shares.

Find the implementation

Implementation	How it runs	Recorded information
Lispex Interpreter	Tree interpreter in Native and WASM	Support results for 205 of 205 tracked primitive rows
Lispex VM	Runs verified bytecode derived from Core IR	Source, Core IR, bytecode, result, diagnostics, and resources
SICP	Explicit profile in Native and `@lispex/sicp` WASM	`sicp-observation/v1` and `lispex-trace/v1`

A primitive row names one operation in the capability registry. Its supported or unsupported result describes implementation coverage. The counts above describe those registry rows. A corpus count instead tells you how many named programs were run. An engine comparison records the results for the exact files and engines it identifies.

LIL and LIT completed their role in internal cross-validation. They were retired in Lispex 1.21 to reduce future maintenance costs. The LIL 205/205 and initial LIT 84/205 counts belong to historical procedure tables. Read them with the retirement and versioned records.

For individual operations, open the current feature catalog. For product and implementation differences, read Lispex Interpreter and Lispex VM.