Add UPLC → JVM bytecode demo built from eo optics by kryptt · Pull Request #132 · Constructive-Programming/eo

kryptt · 2026-10-13T11:45:03Z

What

A demo compiler from Untyped Plutus Core (UPLC, the language Cardano smart contracts run in) to JVM class files. It is built from lawful eo optics, all starting at the script as the ledger stores it: a prism, an iso and a prism.

CborScript ⇀ TypedProgram ≅ JvmClass ⇀ JvmClassScript
           toTyped        lowering   compiling

val script = CborScript.of(program)                 // Either[String, CborScript]; or CborScript.parse(bytes)
UplcJvm.toTyped                                     // CborScript ⇀ TypedProgram: typed Scala-like expression (.body.show prints it)
UplcJvm.toJvmClass                                  // CborScript ⇀ JvmClass: = toTyped andThen lowering
UplcJvm.toBytecode                                  // CborScript ⇀ JvmClassScript: = toJvmClass andThen compiling

UplcJvm.toBytecode.getOption(script) match          // the compiler
  case Some(compiled) =>
    UplcJvm.toBytecode.reverseGet(compiled) == script // the class hands its script back
    Exec.validate(compiled, scriptContext)            // Right(traces) | Left(reason), CIP-117
  case None => // an unknown Plutus Core version, or too large for a class file (UplcJvm.compile says which limit)

Why each arrow walks back. Every stage keeps the script it came from. A TypedProgram keeps its source. A class stores the script as Base64 string constants: a class file has no byte-array constants, and one string constant holds at most 65535 bytes. So reverseGet walks back exactly.
Two prisms around an iso. lowering is an iso: there and back is the identity in both directions, on everything the compiler produces. The other two are prisms, total on the way back:
- toTyped: the bytes can declare any Plutus Core version, and Scalus decodes them all. The ledger only takes the versions Plutus V3 has at protocol version 11, 1.0.0 and 1.1.0 (Semantics.versions), so a script of any other version has no typed program. Before this, a script declaring 2.0.0 or 7.3.9 parsed, compiled and ran as if it were 1.1.0.
- compiling: the JVM caps a method at 64 KB of bytecode and a class at 65 535 constants, so a class can be too large for a class file, and then there is none.
Prism, iso and prism compose to a prism, so toJvmClass and the compiler, toBytecode, are prisms too.
Sealed values.
- CborScript and JvmClassScript are opaque types over Scalus' ByteString. The two byte encodings can't be swapped by accident, and the laws can compare them with ==.
- JvmClass is an opaque type over ASM's ClassNode. Only the compiler and UplcJvm.decompile make one, so it always embeds its script.
- None of the three hands out what it holds: bytes and .tree are copies, and CborScript.parse checks a copy of the caller's array. A value cannot change after the compiler made or checked it, so the way back always works.
- Both entry points are total. CborScript.of returns Left for a program Scalus can't flat-encode, such as one with a free variable. CborScript.parse returns Left for bytes that don't decode.
- The demo's main code uses no unsafe* constructors.
Stack safety. Compiling cannot overflow the JVM stack, however deep the program. Decoding, typing and lowering are walks of Steps (Step.scala): plain calls while a walk is shallow, and from 128 calls deep the rest runs on the heap, from cats.Eval's loop. Every recursive call goes through Walk.nested. A program nested 100 000 deep compiles on a thread's default 1 MB stack.
- Trampolining every node with plain Eval made translation 2–2.7× slower. Trampolining only when deep costs 13–19%.
- A closure's captures are now one fold over its subtree's node ids (pre-order numbering makes a subtree a contiguous range), so FreeVars is gone.
Typer (toTyped). Works directly on Scalus' decoded Term tree. It recognises what UPLC compilers emit for constructs UPLC lacks:
- let, and if over delayed branches;
- saturated builtin calls;
- direct calls of let-bound functions and forced builtins;
- recursion tied by self-application, compiled to direct calls;
- the case (constr k args…) multi-argument call.
Every rewrite is exact under the CEK machine's evaluation order. An expression that can only fail is never followed by code, so the generated class has no unreachable instructions.
Hybrid typing. A static type is a Scalus DefaultUni. A builtin's signature is Scalus' own TypeScheme, instantiated at its argument types, so fstPair (unConstrData d) is an integer and the constructor-tag match on it becomes a JVM switch. Integers, booleans, bytestrings and strings stay unboxed wherever the typer can prove them.
Lowering (lowering). Writes bytecode straight into ASM's class tree (the JvmClass) with asm-commons' InstructionAdapter. It converts closures using a capture layout fixed by the source, so results can be "discharged" back to terms and compared with CEK results. A saturated builtin is a direct call of the scalus.uplc.builtin.Builtins function Scalus' CardanoBuiltins wraps.
Compiling (compiling). UplcJvm.compile is ASM's ClassWriter, which computes the frames, and returns Left for a class over a JVM limit (the one place ASM's MethodTooLargeException / ClassTooLargeException become a value). UplcJvm.decompile is its ClassReader. Both are public. A class file survives being read and written again byte for byte.
Runtime. List and pair builtins read Data lazily. Builtins used as values, and the rest, run Scalus' CEK meanings. A failure is a stackless MachineError, turned into Scalus' Result by Exec.run.
No optimisation of UPLC programs. The compiler translates; it does not optimise. There is no constant folding, inlining, dead-code elimination, CSE, specialisation or tail-call elimination: every operation runs, in the CEK machine's order. The speed-ups come from running as JVM bytecode under the JIT, with values unboxed and calls made directly where the types allow.

Correctness

Semantics. Pinned in Semantics.scala: Plutus V3, PV11, variant E, the configuration Scalus 1.3.0 runs the conformance suite under, and the Plutus Core versions it accepts (1.0.0, 1.1.0).
Test run. sbt demoUplcJvm/test runs 1067 tests: 1011 passed, 56 skipped, 0 failed.
Optic laws. cats-eo-laws' PrismTests check toTyped and toBytecode, over every validator, a closure-valued program, a script declaring Plutus Core 2.0.0, and a program too large for a class file. (JvmClass has no equality, so lowering is checked by the round trips rather than IsoTests.) Oracle.compare also checks the round trips on every program it runs.
Sealing and stack safety. OpticSpec checks that changing the arrays and trees a value hands out changes nothing, and compiles programs nested 100 000 deep (force (delay …), builtin calls, lambdas) on a 1 MB stack: each gives a class file, or the JVM limit its class exceeds — never a StackOverflowError.
Plutus conformance suite. Plutus 1.63.0.0, vendored. All 943 evaluation cases agree with the CEK machine. The 56 skips are parser-only tests and one bare free-variable program.
Validators. Owner + greeting, hash preimage, time lock, 2-of-3 multisig, Fibonacci and HTLC, each compiled with scalus-plugin. There are 20 accepting and rejecting V3 transactions, each checked compiled-then-applied and with the context baked in.
The JVM's verifier checks every generated class. Execution budgets are not metered.

Benchmarks (JMH, `demos/uplc-jvm-bench`, 3 forks)

Validation:
- 8–22× faster than the CEK machine on validators that mostly decode the script context.
- About 66× faster on the Fibonacci puzzle.
- 2.9–7.1× faster than this demo's first raw-ASM version.
- Rejections cost about what acceptances do.
Fibonacci scaling: 48× at n=10, 50× at n=100, 22× at n=1000, where large-integer arithmetic dominates.
Translation:
- 82–222 µs to translate, 0.39–0.79 ms with class loading (re-measured after the stack-safety change; a same-session A/B puts its cost at 13–19% of translate). That is about 15% more than the first raw-ASM emitter cost, and about 40% of the intermediate Cojen Maker version's cost.
- Decoding the script is a pure fold over Scalus' Term tree. It is 2–3× slower than writing into arrays, at a cost of a few microseconds per script.
- A script pays for its compilation after roughly 13–115 validations.

Size

Code lines (no comments or blanks) against the first raw-ASM version:

	before	now	change
main	1425	1105	−22.5%
whole demo	2218	1745	−21.3%

The prisms, sealing and stack safety added 115 lines of main code (the 37-line Step.scala, plus Lower and Typer written as walks of steps, the prisms and the copies) and 80 lines of tests (the prism laws, misses, copies and depth tests).

Build and CI

Three unpublished modules, all outside the root aggregate:
- demoUplcJvm depends on Scalus and asm-commons. Its tests use asm-util and cats-eo-laws.
- demoUplcJvmValidators holds the plugin-compiled validators and shared fixtures.
- demoUplcJvmBench is the JMH project.
ci.yml runs the demo tests and compiles the benchmarks on each JDK lane.
Scala 3 optimizer.
- demoUplcJvm compiles with -opt -opt-inline:**,!scala.**. It is unpublished, so it may inline from Scalus, ASM, cats and eo. Inlining from the Scala library is excluded because it made the short validations 12–27% slower: the inlined code makes Rt.force too big for the JIT to inline (seen with -XX:+PrintInlining). Every other scope measured neutral to ~5% faster. The generated classes are byte-identical either way.
- This touches the published library. scala3LibrarySettings adds -opt-inline:dev.constructive.eo.**, so every library module may inline code from eo's own packages, and never from cats, the JDK or the Scala library. This replaces core's former <sources> scope.
- The full root-aggregate test suite passes on JDK 25 with these flags.

Limitations

No optimisation of UPLC programs (above).
No budget metering.
One semantics: Plutus V3 at PV11. A script of another Plutus Core version has no typed program.
Methods are split at let boundaries only. A class that still exceeds a JVM limit has no class file: toBytecode.getOption is None.
Scalus' flat decoder and encoder, which reading and writing a script go through, recurse: on a 1 MB stack they overflow between 2 000 and 20 000 levels deep.
Generated code recurses on the JVM stack, so Exec uses a thread with a 512 MB stack.

🤖 Generated with Claude Code

https://claude.ai/code/session_01JG5LJsPA524HKbSMvjStrv

A compiler from Untyped Plutus Core (Cardano smart contracts) to JVM class files whose spine is a single polymorphic optic: Iso[Array[Byte], Array[Byte], TypedProgram, JvmProgram] S: flat UPLC T: class file A: typed expr B: JVM-shaped IR `get` decodes and types, `reverseGet` emits bytecode (ASM), and the compiler is `iso.modify(Lower.lower)`; `cborScript andThen iso` takes on-chain CBOR. - Typer (S ⇒ A): re-sugars let / if / saturated builtin calls / direct calls of let-bound functions and forced builtins / case-on-constr multi-argument calls, with rewrites exact under CEK evaluation order. Hybrid typing keeps provable integers, booleans, bytestrings and strings unboxed. - Lower (A ⇒ B): closure conversion with a capture layout fixed by the source, slot allocation, explicit boxing, typed fast paths, and let-body outlining to stay under the JVM's 64 KB method limit. - Runtime: closures, partial builtins, PV11 case-on-builtins, discharge back to UPLC terms; builtins delegate to Scalus except the five selectors. Correctness is differential against the Scalus 1.3.0 CEK machine (Plutus V3, PV11, variant E): the vendored Plutus conformance suite (plutus 1.63.0.0, 943/943 evaluation cases agree) and Scalus-plugin-compiled validators with accepting and rejecting V3 script contexts. Every generated class passes ASM's verifier. Results and errors are compared; budgets are not metered. The demo and its validator corpus live outside the root aggregate; CI runs `demoUplcJvm/test` as its own step and checks their formatting and scalafix. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JG5LJsPA524HKbSMvjStrv

github-actions · 2026-10-13T11:48:37Z

🚀 Cloudflare Pages preview for ccr-1ec1a317-uds4rn is live:

https://0927a1f3.cats-eo-docs.pages.dev

Branch alias: https://ccr-1ec1a317-uds4rn.cats-eo-docs.pages.dev

_{Built from commit 94a8ee0ecbd4717af683298b1f2d575f0301ffec · updated on every push.}

site/docs/demos/uplc-jvm.md (new "Demos" section in the site nav): - The idea: a compiler is a polymorphic Iso — S flat UPLC, A typed expression, B JVM code, T class file — and `iso.modify(Lower.lower)` is the compiler; `cborScript andThen iso` composes on-chain CBOR in front; the law that matters is the commuting square against the CEK machine. - Back and forth: an inline, theme-aware SVG of the round trip, then each stage on real output — UPLC text, A (`iso.get`), B (`Lower.lower`), the generated bytecode (`javap`), and a closure discharged back to a term, α-equal to the CEK machine's result — plus the Fibonacci validator read back as A. - Benchmarks: JMH tables for validation (2.5–3.6× on context-decoding validators, ~20× on the recursion-heavy one, 2–14× less allocation), Fibonacci scaling (10× → 25× → 18×), and translation cost with the break-even number of validations (~10–130). Supporting changes: - demos/uplc-jvm-bench: a JMH project (outside the root aggregate and the library's nightly sweep) — ValidatorBench, FibonacciBench, TranslateBench; setup checks both evaluators accept each transaction before timing. CI compiles it and checks its formatting and scalafix. - Validators module gains the shared off-chain fixtures (Parties, Contexts.spending, Scenarios); ValidatorSpec uses them. - JvmProgram.show renders B as pseudo-code; Demo becomes the stage-by-stage walkthrough the page quotes (timings moved to JMH). - README, CLAUDE.md, benchmarks.md cross-links; ci.yml regenerated. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JG5LJsPA524HKbSMvjStrv

Scalus already has types for most of what the demo redefined: - SourceProgram works on Scalus' own decoded Term tree, numbered by object identity, instead of a mirror Node enum. Constants are looked up by node id. - A static type is Option[DefaultUni]. Builtin signatures come from Scalus' TypeScheme (numTypeVars / arity), replacing Ty, Step, Signature and the hand-written PrimOp table. The typed fast paths are the builtins Rt defines a method for. - Evaluation failures are Scalus' MachineError (UplcFailure is gone), and Exec.run returns scalus.uplc.eval.Result (Exec.Outcome is gone). Emit now uses Cojen Maker instead of raw ASM. Code is written against variables and method names, and Cojen resolves overloads, allocates locals and computes stack maps. That removes descriptors, the reflective descriptor table, the B-level case slots and Rt.spread. The emitter shrinks from 424 to 228 lines, and the generated bytecode is the same shape. The embedded program is now Base64. Cojen's constant-pool writer fails on the non-ASCII strings the Latin-1 encoding produced. asm-util's verifier is dropped too: the JVM verifies every class when Exec.load defines it. Main sources go from 2047 to 1766 lines, and the build depends on cojen-maker instead of asm + asm-util. The suite is unchanged: 1052 tests, 996 passed, 56 skipped. Re-run JMH: validation speed is the same within noise (2.6-3.4x on context-decoding validators, ~20x on Fibonacci). Translation is ~3x slower than with ASM (160-585 us), so break-even moves to ~25-260 validations. The docs page, README and CLAUDE.md are updated. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JG5LJsPA524HKbSMvjStrv

github-actions · 2026-10-13T15:47:06Z

Benchmark A/B

Allocation (B/op) — authoritative

Benchmark	params	base	head	Δ
`JsoniterBench.jReadId`	`size=64`	48.0	72.0	+50.0%
`OrderJsoniterBench.eoReadStreet`	`size=512`	128.2	72.2	-43.7%
`JsoniterBench.jReadId`	`size=8`	72.0	48.0	-33.3%
`JsoniterBench.jReadId`	`size=512`	72.0	48.0	-33.3%
`JsoniterBench.jReadStreet`	`size=8`	128.0	152.0	+18.7%
`JsoniterBench.jReadStreet`	`size=512`	128.0	152.0	+18.7%
`OptionalBench.mModify_loyalty_empty`	`-`	0.0	0.0	+18.0%
`JsoniterBench.jReadStreet`	`size=64`	152.0	128.0	-15.8%
`AffineFoldBench.mGetOption_6`	`-`	0.0	0.0	+11.4%
`GetterBench.mGet_6`	`-`	0.0	0.0	-8.7%
`GetterBench.eoGet_0`	`-`	0.0	0.0	-6.5%
`OptionalBench.mModify_0_empty`	`-`	0.0	0.0	+6.2%
`AffineFoldBench.eoGetOption_0_asAffineFold`	`-`	0.0	0.0	-5.8%
`GetterBench.mGet_3`	`-`	0.0	0.0	+5.5%
`AffineFoldBench.eoGetOption_0`	`-`	0.0	0.0	-4.7%
`JsoniterBench.jMiss`	`size=64`	0.0	0.0	-4.1%
`OrderCirceBench.eoNamesIor`	`size=512`	671,197.2	695,823.9	+3.7%
`JsoniterBench.jMiss`	`size=512`	0.0	0.0	+3.6%
`AffineFoldBench.eoGetOption_0_empty`	`-`	0.0	0.0	-3.6%
`JsoniterBench.jMiss`	`size=8`	0.0	0.0	-3.3%

442 more benchmarks

Benchmark	params	base	head	Δ
`PowerSeriesNestedBench.eoModify_nested`	`size=4`	2,192.0	2,264.0	+3.3%
`OpticBuildBench.buildAndUse`	`-`	2,816.0	2,888.0	+2.6%
`PrismBench.eoGetOptionPresent`	`-`	0.0	0.0	-2.3%
`SchemesBench.handCata`	`-`	0.1	0.1	+2.2%
`OrderCirceBench.directNames`	`size=64`	76,156.5	77,717.2	+2.0%
`OrderCirceBench.hcursorNames`	`size=512`	601,633.2	613,950.6	+2.0%
`GetterBench.mGet_orderId`	`-`	0.0	0.0	-2.0%
`OrderCirceBench.eoNames`	`size=64`	77,738.7	79,298.6	+2.0%
`OrderCirceBench.hcursorNames`	`size=8`	10,768.1	10,552.1	-2.0%
`OrderCirceBench.eoNamesIor`	`size=8`	11,360.1	11,560.1	+1.8%
`GetterBench.eoGet_orderId`	`-`	0.0	0.0	+1.6%
`PowerSeriesNestedBench.eoModify_nested`	`size=16`	4,792.0	4,720.0	-1.5%
`GenericsBench.handPrismModifyMiss`	`-`	0.0	0.0	-1.5%
`AvroVulcanBench.encode_bridged`	`-`	1,240.0	1,224.0	-1.3%
`AvroVulcanBench.encode_vulcanRaw`	`-`	1,240.0	1,224.0	-1.3%
`MultiFocusCollectBench.eoCollectMap_constSum`	`-`	0.0	0.0	+1.3%
`AvroJsonBridgeBench.eoClickToJson`	`-`	3,952.0	4,000.0	+1.2%
`GenericsBench.genPrismGetMiss`	`-`	0.0	0.0	+1.2%
`PlatedBench.mUniverseJson`	`n=512`	699,936.4	691,743.5	-1.2%
`CapsBench.foldMapDirect`	`-`	0.0	0.0	+1.1%
`MultiFocusBench.naive_sumQty`	`size=4`	0.0	0.0	-1.0%
`PrismBench.eoGetOptionAbsent`	`-`	0.0	0.0	-0.9%
`GenericsBench.genLensGet`	`-`	0.0	0.0	-0.9%
`AffineFoldBench.mGetOption_0_empty`	`-`	0.0	0.0	+0.9%
`OrderCirceBench.directStreet`	`size=8`	2,728.0	2,752.0	+0.9%
`GenericsBench.handLensGet`	`-`	0.0	0.0	-0.7%
`JsoniterBench.cSumPrices`	`size=64`	156,496.8	157,517.3	+0.7%
`AffineFoldBench.eoGetOption_loyalty_empty`	`-`	0.0	0.0	+0.6%
`PowerSeriesNestedBench.monocle_nested`	`size=4`	5,536.0	5,568.0	+0.6%
`OptionalBench.eoModify_0_empty`	`-`	0.0	0.0	-0.6%
`GenericsBench.genPrismModifyMiss`	`-`	0.0	0.0	+0.5%
`PrismBench.eoGetRightAbsent`	`-`	0.0	0.0	-0.5%
`SchemesBench.handHylo`	`-`	0.1	0.1	-0.5%
`AffineFoldBench.mGetOption_3`	`-`	0.0	0.0	-0.5%
`PowerSeriesNestedBench.eoModify_nested`	`size=64`	14,592.5	14,664.5	+0.5%
`ZioDiBench.eoServiceGet`	`-`	0.0	0.0	-0.5%
`GenericsBench.handPrismGetMiss`	`-`	0.0	0.0	+0.5%
`CapsBench.foldMapDerivedPerCall`	`-`	0.0	0.0	+0.5%
`PrismBench.mGetOptionPresent`	`-`	0.0	0.0	-0.4%
`FoldBench.eoFoldMap`	`size=8`	0.0	0.0	-0.4%
`GetterBench.eoGet_6`	`-`	0.0	0.0	-0.4%
`CapsBench.foldMapCap`	`-`	0.0	0.0	-0.4%
`KyoDiBench.handMapGet`	`-`	0.0	0.0	-0.4%
`CapsBench.getDeepCap`	`-`	0.0	0.0	-0.4%
`CapsBench.foldMapDerivedHeld`	`-`	0.0	0.0	-0.4%
`CapsBench.getDerivedHeld`	`-`	0.0	0.0	-0.3%
`LensBench.mGet`	`-`	0.0	0.0	-0.3%
`OptionalBench.eoModify_loyalty_empty`	`-`	0.0	0.0	-0.3%
`CapsBench.getDirect`	`-`	0.0	0.0	-0.2%
`PrismBench.mGetRightAbsent`	`-`	0.0	0.0	-0.2%
`PrismBench.mGetOptionAbsent`	`-`	0.0	0.0	-0.2%
`OrderJsoniterBench.nativeSumPrices`	`size=8`	1,512.8	1,515.3	+0.2%
`KyoDiBench.eoMapGet`	`-`	0.0	0.0	-0.1%
`OrderJsoniterBench.nativeSumPrices`	`size=512`	86,346.1	86,472.2	+0.1%
`ZioDiBench.eoDrillGet`	`-`	0.0	0.0	-0.1%
`GetterBench.mGet_0`	`-`	0.0	0.0	+0.1%
`AffineFoldBench.mGetOption_0`	`-`	0.0	0.0	+0.1%
`OrderJsoniterBench.monocleSumPrices`	`size=8`	6,448.0	6,440.0	-0.1%
`JsoniterBench.cReplaceId`	`size=8`	20,643.3	20,667.3	+0.1%
`JsoniterBench.cModifyId`	`size=8`	20,691.3	20,715.3	+0.1%
`PowerSeriesBench.monocle_powerEach`	`size=64`	14,520.1	14,504.1	-0.1%
`CapsBench.getDerivedPerCall`	`-`	0.0	0.0	-0.1%
`FoldBench.mFoldMap`	`size=8`	0.0	0.0	+0.1%
`CapsBench.getCap`	`-`	0.0	0.0	+0.1%
`OrderAvroBench.naiveModifyNames`	`size=64`	27,912.2	27,936.2	+0.1%
`OrderCirceBench.monocleNames`	`size=8`	19,456.1	19,472.1	+0.1%
`AffineFoldBench.eoGetOption_loyalty`	`-`	0.0	0.0	+0.1%
`AffineFoldBench.mGetOption_loyalty_empty`	`-`	0.0	0.0	+0.1%
`LensBench.eoGet`	`-`	0.0	0.0	+0.1%
`OrderCirceBench.eoNamesIor`	`size=64`	87,490.1	87,552.0	+0.1%
`OrderAvroBench.monocleModifyNames`	`size=64`	39,841.1	39,865.0	+0.1%
`OrderCirceBench.eoStreet`	`size=64`	2,721.2	2,720.1	-0.0%
`OrderJsoniterBench.nativeSumPrices`	`size=64`	10,948.1	10,944.3	-0.0%
`PowerSeriesNestedBench.eoModify_nested`	`size=1024`	210,442.7	210,512.4	+0.0%
`CapsBench.getDeepDirect`	`-`	0.0	0.0	-0.0%
`OrderCirceBench.eoStreet`	`size=8`	2,720.8	2,720.0	-0.0%
`OrderCirceBench.monocleStreet`	`size=64`	113,812.1	113,844.0	+0.0%
`OrderCirceBench.eoStreetIor`	`size=8`	2,736.8	2,736.0	-0.0%
`AffineFoldBench.mGetOption_loyalty`	`-`	0.0	0.0	-0.0%
`GenericsBench.rawLensGet`	`-`	0.0	0.0	-0.0%
`PlatedBench.visitorUniverseJson`	`n=4096`	295,053.4	294,994.8	-0.0%
`OrderCirceBench.naiveNames`	`size=64`	120,808.5	120,832.4	+0.0%
`JsoniterBench.cModifyId`	`size=64`	136,271.4	136,295.6	+0.0%
`JsoniterBench.cReplaceId`	`size=64`	136,223.4	136,247.6	+0.0%
`GetterBench.eoGet_3`	`-`	0.0	0.0	-0.0%
`OrderJsoniterBench.nativeReadStreet`	`size=512`	454.3	454.3	-0.0%
`MultiFocusBench.eoModify_multiFocus`	`size=1024`	331,102.8	331,078.4	-0.0%
`JsoniterBench.cModifyId`	`size=512`	1,066,728.6	1,066,803.8	+0.0%
`OrderAvroBench.monocleModifyStreet`	`size=512`	169,078.1	169,084.1	+0.0%
`JsoniterBench.cReplaceId`	`size=512`	1,066,678.5	1,066,710.2	+0.0%
`OrderCirceBench.hcursorNames`	`size=64`	77,808.1	77,806.2	-0.0%
`OrderCirceBench.eoNames`	`size=512`	626,313.5	626,304.2	-0.0%
`PlatedBench.mUniverseJson`	`n=4096`	6,491,325.0	6,491,241.4	-0.0%
`PlatedBench.mUniverseExpr`	`n=4096`	4,689,262.3	4,689,205.1	-0.0%
`PowerSeriesBench.naive_powerEach`	`size=4096`	114,779.5	114,780.8	+0.0%
`PowerSeriesNestedBench.monocle_nested`	`size=1024`	1,118,799.3	1,118,812.4	+0.0%
`OrderJsoniterBench.nativeReadStreet`	`size=64`	424.5	424.5	+0.0%
`OrderAvroBench.eoModifyNames`	`size=512`	97,422.9	97,421.8	-0.0%
`MultiFocusBench.eoFold_powerEach`	`size=1024`	43,036.7	43,036.2	-0.0%
`PowerSeriesBench.monocle_powerEach`	`size=4096`	967,861.9	967,854.7	-0.0%
`OrderAvroBench.monocleReadStreet`	`size=512`	69,786.8	69,786.3	-0.0%
`OrderCirceBench.hcursorStreet`	`size=512`	3,032.7	3,032.7	-0.0%
`OrderCirceBench.eoStreetIor`	`size=512`	2,736.7	2,736.7	+0.0%
`OrderCirceBench.directNames`	`size=512`	601,636.3	601,633.3	-0.0%
`JsoniterBench.cReadStreet`	`size=64`	101,303.9	101,303.4	-0.0%
`MultiFocusBench.eoModify_powerEach`	`size=1024`	115,176.7	115,176.2	-0.0%
`PowerSeriesPrismBench.eoModify_sparse`	`size=2048`	104,681.2	104,680.7	-0.0%
`TraversalBench.mModify`	`size=512`	176,897.2	176,898.0	+0.0%
`PowerSeriesBench.monocle_powerEach`	`size=4`	1,176.0	1,176.0	+0.0%
`OrderJsoniterBench.monocleSumPrices`	`size=512`	1,178,381.5	1,178,376.9	-0.0%
`OrderJsoniterBench.eoModifyStreet`	`size=512`	42,029.9	42,029.7	-0.0%
`PowerSeriesBench.eoModify_powerEach`	`size=4096`	164,368.5	164,367.9	-0.0%
`OrderJsoniterBench.naiveSumPrices`	`size=512`	230,108.4	230,107.6	-0.0%
`OrderJsoniterBench.naiveModifyStreet`	`size=512`	333,548.1	333,546.9	-0.0%
`PlatedBench.eoTransformDeep`	`n=4096`	624,386.6	624,384.4	-0.0%
`MultiFocusBench.naive_sumQty`	`size=1024`	16,129.4	16,129.3	-0.0%
`OrderAvroBench.naiveReadStreet`	`size=512`	69,786.3	69,786.1	-0.0%
`OrderJsoniterBench.eoSumPrices`	`size=512`	63,684.1	63,684.0	-0.0%
`OrderCirceBench.naiveStreet`	`size=64`	113,806.7	113,807.0	+0.0%
`PowerSeriesBench.eoModify_powerEach`	`size=1024`	41,413.7	41,413.6	-0.0%
`OrderCirceBench.eoStreet`	`size=512`	2,720.7	2,720.7	-0.0%
`OrderAvroBench.naiveModifyStreet`	`size=512`	169,060.4	169,060.0	-0.0%
`JsoniterBench.jSumPrices`	`size=512`	63,630.3	63,630.4	+0.0%
`PowerSeriesPrismBench.monocle_sparse`	`size=2048`	523,139.1	523,137.9	-0.0%
`OrderJsoniterBench.monocleModifyStreet`	`size=512`	333,572.3	333,571.6	-0.0%
`OrderJsoniterBench.monocleReadStreet`	`size=512`	193,214.8	193,214.4	-0.0%
`PowerSeriesPrismBench.monocle_sparse`	`size=512`	166,691.9	166,691.5	-0.0%
`PlatedBench.eoTransformExpr`	`n=4096`	655,356.4	655,355.1	-0.0%
`OrderJsoniterBench.monocleSumPrices`	`size=64`	45,835.2	45,835.3	+0.0%
`OrderCirceBench.hcursorStreet`	`size=64`	3,032.1	3,032.1	-0.0%
`PowerSeriesNestedBench.eoModify_nested`	`size=256`	53,770.7	53,770.9	+0.0%
`PlatedBench.visitorTransformExpr`	`n=4096`	360,472.5	360,471.8	-0.0%
`JsoniterBench.cReadId`	`size=64`	101,303.9	101,303.7	-0.0%
`PlatedBench.mTransformExpr`	`n=4096`	753,724.5	753,723.3	-0.0%
`PlatedBench.eoUniverseJson`	`n=4096`	786,459.6	786,460.9	+0.0%
`PowerSeriesNestedBench.naive_nested`	`size=1024`	115,071.7	115,071.5	-0.0%
`OrderCirceBench.monocleNames`	`size=64`	132,761.6	132,761.4	-0.0%
`OrderAvroBench.eoModifyStreet`	`size=512`	328.0	328.0	-0.0%
`PowerSeriesNestedBench.naive_nested`	`size=256`	29,019.3	29,019.3	-0.0%
`PlatedBench.visitorUniverseDeep`	`n=4096`	196,707.9	196,707.6	-0.0%
`OrderCirceBench.directStreet`	`size=512`	2,728.7	2,728.7	+0.0%
`JsoniterBench.jModifyId`	`size=512`	41,921.7	41,921.7	-0.0%
`PowerSeriesBench.naive_powerEach`	`size=1024`	28,730.8	28,730.7	-0.0%
`PlatedBench.eoUniverseDeep`	`n=4096`	786,610.4	786,609.5	-0.0%
`OrderJsoniterBench.monocleModifyStreet`	`size=64`	30,081.6	30,081.5	-0.0%
`PlatedBench.visitorUniverseExpr`	`n=4096`	196,654.3	196,654.1	-0.0%
`JsoniterBench.cSumPrices`	`size=512`	1,240,754.0	1,240,752.7	-0.0%
`OrderJsoniterBench.naiveModifyStreet`	`size=64`	30,057.7	30,057.7	-0.0%
`OrderAvroBench.eoReadStreet`	`size=512`	88.0	88.0	-0.0%
`PowerSeriesPrismBench.eoModify_sparse`	`size=512`	24,785.5	24,785.5	-0.0%
`PlatedBench.mUniverseExpr`	`n=512`	475,028.5	475,028.9	+0.0%
`JsoniterBench.jReplaceId`	`size=512`	41,889.7	41,889.6	-0.0%
`OrderCirceBench.eoStreetIor`	`size=64`	2,736.1	2,736.1	+0.0%
`MultiFocusBench.naive_listMap`	`size=1024`	65,578.7	65,578.7	-0.0%
`OrderCirceBench.naiveStreet`	`size=512`	908,003.7	908,003.1	-0.0%
`OrderCirceBench.monocleNames`	`size=512`	1,121,753.7	1,121,754.4	+0.0%
`OrderJsoniterBench.eoSumPrices`	`size=8`	1,176.0	1,176.0	+0.0%
`PowerSeriesBench.monocle_powerEach`	`size=1024`	279,426.9	279,427.1	+0.0%
`OrderJsoniterBench.eoReadStreet`	`size=8`	72.0	72.0	-0.0%
`OrderAvroBench.monocleModifyNames`	`size=512`	382,754.5	382,754.3	-0.0%
`TraversalBench.eoFoldPrices`	`size=512`	12,312.3	12,312.3	-0.0%
`FoldBench.mFoldPrices`	`size=512`	12,312.4	12,312.4	-0.0%
`FoldBench.eoFoldMap`	`size=512`	7,936.2	7,936.2	-0.0%
`PlatedBench.eoUniverseExpr`	`n=4096`	786,416.8	786,417.2	+0.0%
`OrderAvroBench.naiveModifyNames`	`size=512`	226,292.5	226,292.6	+0.0%
`PowerSeriesPrismBench.naive_sparse`	`size=2048`	24,613.2	24,613.2	-0.0%
`OrderCirceBench.naiveNames`	`size=512`	965,242.9	965,243.3	+0.0%
`OrderCirceBench.directStreet`	`size=64`	2,728.1	2,728.1	-0.0%
`PowerSeriesPrismBench.eoModify_sparse`	`size=8`	496.0	496.0	-0.0%
`PlatedBench.eoUniverseJson`	`n=64`	12,168.1	12,168.1	-0.0%
`OrderCirceBench.monocleStreet`	`size=512`	908,004.1	908,003.8	-0.0%
`PowerSeriesPrismBench.naive_sparse`	`size=512`	6,176.4	6,176.4	-0.0%
`MultiFocusBench.eoFold_powerEach`	`size=256`	9,240.2	9,240.2	-0.0%
`PowerSeriesBench.eoModify_powerEach`	`size=256`	10,688.4	10,688.4	+0.0%
`PlatedBench.eoUniverseDeep`	`n=64`	12,360.0	12,360.0	-0.0%
`JsoniterBench.jSumPrices`	`size=64`	8,120.3	8,120.3	-0.0%
`AvroBytesBench.eoModifyPartner`	`-`	984.0	984.0	-0.0%
`SchemesBench.eoCata`	`-`	197,568.5	197,568.4	-0.0%
`AvroBytesBench.naiveReadPartner`	`-`	5,424.0	5,424.0	-0.0%
`PowerSeriesNestedBench.monocle_nested`	`size=64`	58,833.1	58,833.0	-0.0%
`PlatedBench.visitorUniverseExpr`	`n=512`	24,584.7	24,584.7	-0.0%
`PlatedBench.visitorTransformDeep`	`n=512`	20,496.5	20,496.5	-0.0%
`OrderJsoniterBench.nativeReadStreet`	`size=8`	424.0	424.0	+0.0%
`FoldBench.mFoldMap`	`size=512`	7,936.2	7,936.2	-0.0%
`PlatedBench.eoTransformDeep`	`n=512`	57,361.5	57,361.4	-0.0%
`PlatedBench.mTransformExpr`	`n=512`	69,585.4	69,585.4	-0.0%
`PlatedBench.visitorUniverseJson`	`n=512`	40,954.3	40,954.3	+0.0%
`OrderJsoniterBench.eoSumPrices`	`size=64`	8,121.1	8,121.1	+0.0%
`OrderJsoniterBench.eoModifyStreet`	`size=64`	5,432.1	5,432.1	-0.0%
`PowerSeriesPrismBench.eoModify_sparse`	`size=32`	1,360.0	1,360.0	-0.0%
`PlatedBench.visitorUniverseJson`	`n=64`	5,096.0	5,096.0	-0.0%
`AvroBytesBench.eoReadCountry`	`-`	520.0	520.0	-0.0%
`OrderJsoniterBench.naiveSumPrices`	`size=64`	29,337.0	29,337.0	+0.0%
`JsoniterBench.cReadStreet`	`size=512`	797,920.6	797,920.4	-0.0%
`PlatedBench.visitorUniverseDeep`	`n=512`	24,632.8	24,632.8	-0.0%
`OrderJsoniterBench.naiveModifyStreet`	`size=8`	4,640.1	4,640.1	-0.0%
`JsoniterBench.jSumPrices`	`size=8`	1,176.0	1,176.0	+0.0%
`PowerSeriesPrismBench.naive_sparse`	`size=128`	1,568.0	1,568.0	-0.0%
`TraversalBench.eoFoldNested`	`size=512`	20,200.8	20,200.8	-0.0%
`OrderJsoniterBench.naiveReadStreet`	`size=512`	193,214.6	193,214.6	-0.0%
`PlatedBench.eoTransformExpr`	`n=512`	81,825.7	81,825.7	-0.0%
`PlatedBench.eoUniverseDeep`	`n=512`	98,377.9	98,377.9	-0.0%
`PowerSeriesPrismBench.eoModify_sparse`	`size=128`	4,816.1	4,816.1	-0.0%
`PowerSeriesPrismBench.monocle_sparse`	`size=128`	27,808.2	27,808.2	-0.0%
`PowerSeriesNestedBench.monocle_nested`	`size=256`	430,194.6	430,194.7	+0.0%
`OrderJsoniterBench.monocleReadStreet`	`size=64`	24,704.9	24,704.9	-0.0%
`PowerSeriesNestedBench.monocle_nested`	`size=16`	15,776.1	15,776.1	-0.0%
`PlatedBench.visitorTransformExpr`	`n=512`	45,032.8	45,032.8	-0.0%
`PowerSeriesBench.eoModify_powerEach`	`size=64`	3,008.0	3,008.0	-0.0%
`AvroJsonBridgeBench.eoWideToJson`	`-`	1,472.0	1,472.0	-0.0%
`OrderAvroBench.monocleModifyStreet`	`size=64`	20,904.2	20,904.2	-0.0%
`AvroBytesBench.prunedReadCountry`	`-`	1,976.0	1,976.0	-0.0%
`OrderJsoniterBench.monocleModifyStreet`	`size=8`	4,664.1	4,664.1	-0.0%
`OrderJsoniterBench.eoModifyStreet`	`size=8`	1,072.0	1,072.0	-0.0%
`PlatedBench.mUniverseExpr`	`n=64`	45,424.6	45,424.6	+0.0%
`TraversalBench.mFoldPrices`	`size=8`	2,000.0	2,000.0	-0.0%
`OrderAvroBench.monocleReadStreet`	`size=64`	8,848.1	8,848.1	-0.0%
`TraversalBench.eoFoldNested`	`size=64`	2,680.0	2,680.0	-0.0%
`OrderJsoniterBench.naiveReadStreet`	`size=64`	24,704.9	24,704.9	-0.0%
`PowerSeriesBench.monocle_powerEach`	`size=256`	107,335.5	107,335.5	+0.0%
`MultiFocusBench.naive_sumQty`	`size=256`	3,840.0	3,840.0	-0.0%
`MultiFocusBench.naive_listMap`	`size=256`	16,424.2	16,424.2	-0.0%
`PlatedBench.visitorTransformDeep`	`n=4096`	163,888.0	163,888.0	+0.0%
`OrderAvroBench.naiveModifyNames`	`size=8`	3,752.0	3,752.0	-0.0%
`SchemesBench.eoHylo`	`-`	295,848.5	295,848.6	+0.0%
`AvroBytesBench.naiveModifyPartner`	`-`	8,696.0	8,696.0	-0.0%
`TraversalBench.mFoldPrices`	`size=512`	162,564.6	162,564.6	+0.0%
`JsoniterBench.jModifyId`	`size=64`	5,328.0	5,328.0	-0.0%
`PowerSeriesNestedBench.naive_nested`	`size=16`	2,136.0	2,136.0	-0.0%
`PowerSeriesBench.eoModify_powerEach`	`size=16`	1,088.0	1,088.0	-0.0%
`OrderAvroBench.naiveModifyStreet`	`size=64`	20,880.2	20,880.2	-0.0%
`MultiFocusBench.eoFold_powerEach`	`size=32`	920.0	920.0	-0.0%
`PowerSeriesBench.naive_powerEach`	`size=256`	7,224.2	7,224.2	-0.0%
`FoldBench.eoFoldPrices`	`size=512`	12,312.4	12,312.4	-0.0%
`AvroJsonBridgeBench.naiveClickToJson`	`-`	5,856.0	5,856.0	-0.0%
`OrderJsoniterBench.naiveReadStreet`	`size=8`	3,648.0	3,648.0	+0.0%
`AvroBytesBench.eoReadPartner`	`-`	480.0	480.0	-0.0%
`OrderCirceBench.eoNames`	`size=8`	10,944.1	10,944.1	-0.0%
`OrderAvroBench.monocleModifyStreet`	`size=8`	2,992.0	2,992.0	-0.0%
`PlatedBench.eoUniverseJson`	`n=512`	98,186.5	98,186.5	-0.0%
`OrderJsoniterBench.eoReadStreet`	`size=64`	128.0	128.0	+0.0%
`SchemesBench.eoAna`	`-`	589,712.8	589,712.7	-0.0%
`TraversalBench.eoModify`	`size=64`	4,640.0	4,640.0	-0.0%
`FoldBench.eoFoldPrices`	`size=8`	216.0	216.0	-0.0%
`AvroBytesBench.prunedReadPartner`	`-`	1,592.0	1,592.0	-0.0%
`AvroBytesBench.naivePassthroughPayload`	`-`	14,088.1	14,088.1	-0.0%
`OrderAvroBench.eoModifyNames`	`size=64`	12,616.1	12,616.1	+0.0%
`MultiFocusBench.eoModify_powerEach`	`size=256`	26,080.5	26,080.5	+0.0%
`AvroVulcanBench.fieldGet_native`	`-`	432.0	432.0	-0.0%
`OrderCirceBench.naiveStreet`	`size=8`	17,040.1	17,040.1	-0.0%
`PlatedBench.mUniverseJson`	`n=64`	73,208.7	73,208.7	-0.0%
`PowerSeriesPrismBench.monocle_sparse`	`size=8`	1,952.0	1,952.0	-0.0%
`SchemesBench.drosteHylo`	`-`	328,640.4	328,640.4	-0.0%
`ClickRecordBench.encode_derivedThroughPrism`	`-`	744.0	744.0	+0.0%
`AvroEncodeRouteBench.naive_freshPlumbing`	`-`	2,624.0	2,624.0	-0.0%
`PowerSeriesNestedBench.naive_nested`	`size=64`	7,512.2	7,512.2	-0.0%
`OrderAvroBench.eoReadStreet`	`size=64`	88.0	88.0	-0.0%
`OptionalBench.mModify_loyalty`	`-`	112.0	112.0	+0.0%
`OpticBuildBench.reuseUse`	`-`	2,720.0	2,720.0	+0.0%
`AvroBytesBench.naiveModifyCountry`	`-`	7,616.0	7,616.0	+0.0%
`MultiFocusBench.eoModify_multiFocus`	`size=32`	8,600.0	8,600.0	-0.0%
`OrderAvroBench.eoReadStreet`	`size=8`	88.0	88.0	-0.0%
`MultiFocusBench.eoModify_multiFocus`	`size=256`	76,112.7	76,112.7	-0.0%
`PowerSeriesPrismBench.monocle_sparse`	`size=32`	7,040.0	7,040.0	+0.0%
`OrderAvroBench.naiveReadStreet`	`size=64`	8,848.1	8,848.1	-0.0%
`OrderJsoniterBench.monocleReadStreet`	`size=8`	3,648.0	3,648.0	-0.0%
`PowerSeriesPrismBench.naive_sparse`	`size=32`	416.0	416.0	-0.0%
`AffineFoldBench.eoGetOption_6`	`-`	16.0	16.0	-0.0%
`AvroJsonBridgeBench.eoClickToAvro`	`-`	4,664.0	4,664.0	-0.0%
`OrderCirceBench.hcursorStreet`	`size=8`	3,032.0	3,032.0	-0.0%
`AvroBytesBench.eoModifyCountry`	`-`	928.0	928.0	-0.0%
`AvroDecodeReuseBench.freshDecodeRecord`	`-`	3,344.0	3,344.0	-0.0%
`SchemesBench.handAna`	`-`	163,816.2	163,816.2	-0.0%
`PowerSeriesNestedBench.naive_nested`	`size=4`	792.0	792.0	-0.0%
`OptionalBench.eoModify_6`	`-`	208.0	208.0	+0.0%
`FoldBench.mFoldPrices`	`size=64`	1,560.0	1,560.0	-0.0%
`AvroDecodeReuseBench.confluentRecordReader`	`-`	1,560.0	1,560.0	-0.0%
`OrderCirceBench.monocleStreet`	`size=8`	17,048.1	17,048.1	-0.0%
`MultiFocusBench.eoModify_multiFocus`	`size=4`	1,320.0	1,320.0	-0.0%
`AvroEncodeRouteBench.eo_encodeToAny`	`-`	312.0	312.0	-0.0%
`JsoniterBench.jModifyId`	`size=8`	992.0	992.0	-0.0%
`MultiFocusCollectBench.eoModify_multiFocusTuple3`	`-`	96.0	96.0	-0.0%
`JsoniterBench.cReadStreet`	`size=8`	15,568.0	15,568.0	+0.0%
`TraversalBench.eoModify`	`size=512`	36,897.3	36,897.3	+0.0%
`MultiFocusBench.eoModify_powerEach`	`size=32`	3,152.0	3,152.0	-0.0%
`PowerSeriesBench.eoModify_powerEach`	`size=4`	608.0	608.0	-0.0%
`FoldBench.mFoldMap`	`size=64`	768.0	768.0	-0.0%
`AvroVulcanBench.rootGet_native`	`-`	600.0	600.0	+0.0%
`TraversalBench.eoFoldPrices`	`size=64`	1,560.0	1,560.0	-0.0%
`TraversalBench.mFoldPrices`	`size=64`	15,408.1	15,408.1	-0.0%
`MultiFocusBench.naive_listMap`	`size=32`	2,088.0	2,088.0	-0.0%
`OrderAvroBench.naiveReadStreet`	`size=8`	1,208.0	1,208.0	-0.0%
`ClickRecordBench.encode_derived`	`-`	744.0	744.0	+0.0%
`OptionalBench.mModify_6`	`-`	496.0	496.0	+0.0%
`IsoBench.mReverseGet`	`-`	32.0	32.0	-0.0%
`FoldBench.eoFoldMap`	`size=64`	768.0	768.0	-0.0%
`MultiFocusCollectBench.eoModify_multiFocusTuple6`	`-`	248.0	248.0	-0.0%
`FoldBench.eoFoldPrices`	`size=64`	1,560.0	1,560.0	-0.0%
`AvroJsonBridgeBench.naiveWideToAvro`	`-`	3,504.0	3,504.0	-0.0%
`KyoDiBench.handMapDrillModify`	`-`	216.0	216.0	-0.0%
`MultiFocusBench.naive_sumQty`	`size=32`	256.0	256.0	-0.0%
`PlatedBench.eoUniverseExpr`	`n=64`	12,168.0	12,168.0	-0.0%
`CompositionBench.reuseLens1`	`-`	40.0	40.0	-0.0%
`MultiFocusCollectBench.naive_constSum`	`-`	16.0	16.0	+0.0%
`AvroEncodeRouteBench.handwritten_stream`	`-`	240.0	240.0	+0.0%
`JsoniterBench.jReplaceId`	`size=64`	5,296.0	5,296.0	-0.0%
`JsoniterBench.cReadId`	`size=512`	797,920.9	797,920.8	-0.0%
`PlatedBench.visitorTransformExpr`	`n=64`	5,608.0	5,608.0	-0.0%
`OrderAvroBench.naiveModifyStreet`	`size=8`	2,968.0	2,968.0	-0.0%
`OrderAvroBench.monocleReadStreet`	`size=8`	1,208.0	1,208.0	+0.0%
`OrderAvroBench.eoModifyStreet`	`size=8`	328.0	328.0	-0.0%
`JsoniterBench.jReplaceId`	`size=8`	960.0	960.0	-0.0%
`KyoDiBench.eoMapDrillModify`	`-`	216.0	216.0	-0.0%
`MultiFocusCollectBench.eoCollectList_listSum`	`-`	56.0	56.0	-0.0%
`GenericsBench.genLensModify`	`-`	24.0	24.0	-0.0%
`CompositionBench.buildLens6`	`-`	352.0	352.0	-0.0%
`MultiFocusCollectBench.eoCollectMap_zipMean`	`-`	880.0	880.0	-0.0%
`MultiFocusBench.naive_listMap`	`size=4`	296.0	296.0	-0.0%
`OptionalBench.eoModify_0`	`-`	112.0	112.0	-0.0%
`TraversalBench.eoModify`	`size=8`	608.0	608.0	-0.0%
`PrismBench.eoGetRightPresent`	`-`	16.0	16.0	-0.0%
`JsoniterBench.cSumPrices`	`size=8`	22,720.1	22,720.1	-0.0%
`OrderJsoniterBench.naiveSumPrices`	`size=8`	4,248.0	4,248.0	-0.0%
`JsoniterBench.cReadId`	`size=8`	15,568.0	15,568.0	-0.0%
`PlatedBench.eoTransformExpr`	`n=64`	10,144.0	10,144.0	-0.0%
`PlatedBench.visitorUniverseDeep`	`n=64`	3,128.0	3,128.0	-0.0%
`PlatedBench.mTransformExpr`	`n=64`	11,728.0	11,728.0	-0.0%
`GenericsBench.rawPrismGetHit`	`-`	16.0	16.0	-0.0%
`LensBench.eoModifyDeep`	`-`	152.0	152.0	-0.0%
`AvroBytesBench.naiveReadCountry`	`-`	4,256.0	4,256.0	+0.0%
`IsoBench.mGet`	`-`	32.0	32.0	-0.0%
`TraversalBench.mModify`	`size=64`	14,408.1	14,408.1	-0.0%
`AvroVulcanBench.fieldGet_bridged`	`-`	432.0	432.0	-0.0%
`GenericsBench.rawLensModify`	`-`	24.0	24.0	-0.0%
`PlatedBench.eoTransformDeep`	`n=64`	7,184.0	7,184.0	-0.0%
`OrderAvroBench.eoModifyStreet`	`size=64`	328.0	328.0	-0.0%
`AffineFoldBench.eoGetOption_0_asOptional`	`-`	16.0	16.0	-0.0%
`SetterBench.mModify_orderId`	`-`	40.0	40.0	-0.0%
`MultiFocusCollectBench.naive_tuple3Rewrite`	`-`	96.0	96.0	-0.0%
`SetterBench.eoModify_3`	`-`	72.0	72.0	-0.0%
`PrismBench.mGetRightPresent`	`-`	16.0	16.0	-0.0%
`PowerSeriesBench.naive_powerEach`	`size=64`	1,848.0	1,848.0	+0.0%
`PrismBench.mRightReverseGet`	`-`	16.0	16.0	-0.0%
`TraversalBench.mModify`	`size=8`	1,896.0	1,896.0	-0.0%
`OptionalBench.mReplace_0`	`-`	40.0	40.0	-0.0%
`MultiFocusBench.eoModify_powerEach`	`size=4`	800.0	800.0	-0.0%
`PowerSeriesBench.naive_powerEach`	`size=16`	504.0	504.0	-0.0%
`LensBench.mModifyDeep`	`-`	176.0	176.0	+0.0%
`ReviewBench.eoReverseGet_6`	`-`	120.0	120.0	-0.0%
`OrderAvroBench.eoModifyNames`	`size=8`	1,992.0	1,992.0	-0.0%
`ReviewBench.naiveBuild_6`	`-`	120.0	120.0	-0.0%
`KyoDiBench.handVarUpdate`	`-`	200.0	200.0	+0.0%
`SchemesBench.drosteCata`	`-`	164,824.3	164,824.2	-0.0%
`CompositionBench.buildLens3`	`-`	184.0	184.0	-0.0%
`AvroJsonBridgeBench.naiveClickToAvro`	`-`	3,928.0	3,928.0	-0.0%
`LensBench.mModify`	`-`	40.0	40.0	-0.0%
`AvroDecodeReuseBench.confluentRecordReaderFresh`	`-`	3,696.0	3,696.0	-0.0%
`PrismBench.mReverseGet`	`-`	16.0	16.0	-0.0%
`AvroBytesBench.eoSliceGraftPayload`	`-`	1,192.0	1,192.0	+0.0%
`AvroVulcanBench.decode_native`	`-`	48.0	48.0	-0.0%
`ClickRecordBench.encode_hand`	`-`	768.0	768.0	-0.0%
`GenericsBench.genPrismModifyHit`	`-`	24.0	24.0	-0.0%
`ReviewBench.naiveBuild_0`	`-`	24.0	24.0	-0.0%
`GenericsBench.genPrismGetHit`	`-`	16.0	16.0	-0.0%
`AvroBytesBench.eoGraftPayload`	`-`	720.0	720.0	-0.0%
`GenericsBench.handPrismModifyHit`	`-`	24.0	24.0	-0.0%
`CompositionBench.reuseLens6`	`-`	120.0	120.0	-0.0%
`OptionalBench.mModify_0`	`-`	112.0	112.0	-0.0%
`CapsBench.modifyDirect`	`-`	40.0	40.0	-0.0%
`CompositionBench.buildLens1`	`-`	72.0	72.0	-0.0%
`SchemesBench.drosteAna`	`-`	327,632.4	327,632.4	-0.0%
`GenericsBench.handLensModify`	`-`	24.0	24.0	+0.0%
`PrismBench.eoRightReverseGet`	`-`	16.0	16.0	-0.0%
`OptionalBench.eoModify_loyalty`	`-`	112.0	112.0	-0.0%
`AvroDecodeReuseBench.cachedDecodeRecord`	`-`	1,224.0	1,224.0	-0.0%
`IsoBench.eoGet`	`-`	32.0	32.0	-0.0%
`OrderCirceBench.directNames`	`size=8`	10,688.1	10,688.1	-0.0%
`ClickRecordBench.encode_vulcanFull`	`-`	28,408.0	28,408.0	-0.0%
`ZioDiBench.eoDrillModify`	`-`	472.0	472.0	+0.0%
`GenericsBench.rawPrismModifyHit`	`-`	24.0	24.0	-0.0%
`TraversalBench.eoFoldPrices`	`size=8`	216.0	216.0	-0.0%
`MultiFocusCollectBench.naive_zipMeanBroadcast`	`-`	1,176.0	1,176.0	-0.0%
`AvroEncodeRouteBench.eo_encodeValue`	`-`	568.0	568.0	+0.0%
`SetterBench.mModify_0`	`-`	24.0	24.0	+0.0%
`TraversalBench.eoFoldNested`	`size=8`	328.0	328.0	-0.0%
`SetterBench.mModify_6`	`-`	288.0	288.0	+0.0%
`CapsBench.modifyDeepDirect`	`-`	152.0	152.0	-0.0%
`CompositionBench.reuseLeaf`	`-`	24.0	24.0	-0.0%
`PlatedBench.visitorTransformDeep`	`n=64`	2,576.0	2,576.0	+0.0%
`OptionalBench.eoReplace_0`	`-`	40.0	40.0	-0.0%
`ClickRecordBench.encode_positional`	`-`	23,912.0	23,912.0	-0.0%
`AvroVulcanBench.decode_vulcanRaw`	`-`	832.0	832.0	-0.0%
`PowerSeriesBench.monocle_powerEach`	`size=16`	3,720.0	3,720.0	+0.0%
`FoldBench.mFoldPrices`	`size=8`	216.0	216.0	-0.0%
`PowerSeriesBench.naive_powerEach`	`size=4`	168.0	168.0	-0.0%
`CompositionBench.reuseLens3`	`-`	72.0	72.0	+0.0%
`ZioDiBench.handDrillModify`	`-`	592.0	592.0	+0.0%
`ReviewBench.eoReverseGet_0`	`-`	24.0	24.0	-0.0%
`OrderCirceBench.naiveNames`	`size=8`	17,808.1	17,808.1	+0.0%
`CompositionBench.reuseLensOptional3`	`-`	160.0	160.0	-0.0%
`PrismBench.eoReverseGet`	`-`	16.0	16.0	-0.0%
`CompositionBench.buildLensOptional3`	`-`	184.0	184.0	-0.0%
`LensBench.eoReplace`	`-`	40.0	40.0	-0.0%
`PowerSeriesPrismBench.naive_sparse`	`size=8`	128.0	128.0	-0.0%
`MultiFocusCollectBench.naive_tuple6Rewrite`	`-`	184.0	184.0	-0.0%
`AvroVulcanBench.encode_native`	`-`	56.0	56.0	+0.0%
`GenericsBench.handPrismGetHit`	`-`	16.0	16.0	+0.0%
`AvroJsonBridgeBench.eoWideToAvro`	`-`	2,040.0	2,040.0	-0.0%
`SetterBench.mModify_3`	`-`	168.0	168.0	-0.0%
`ZioDiBench.handServiceReplace`	`-`	520.0	520.0	-0.0%
`SetterBench.eoModify_orderId`	`-`	40.0	40.0	-0.0%
`OrderAvroBench.monocleModifyNames`	`size=8`	5,400.0	5,400.0	+0.0%
`PlatedBench.visitorUniverseExpr`	`n=64`	3,080.0	3,080.0	-0.0%
`OptionalBench.eoReplace_loyalty`	`-`	88.0	88.0	-0.0%
`OptionalBench.mModify_3`	`-`	304.0	304.0	-0.0%
`ZioDiBench.eoServiceReplace`	`-`	400.0	400.0	-0.0%
`SetterBench.eoModify_0`	`-`	24.0	24.0	-0.0%
`AvroVulcanBench.decode_bridged`	`-`	832.0	832.0	-0.0%
`ReviewBench.naiveBuild_3`	`-`	72.0	72.0	+0.0%
`AvroVulcanBench.rootGet_bridged`	`-`	1,424.0	1,424.0	-0.0%
`KyoDiBench.eoEnvFocus`	`-`	320.0	320.0	-0.0%
`MultiFocusBench.eoFold_powerEach`	`size=4`	328.0	328.0	-0.0%
`SetterBench.eoModify_6`	`-`	120.0	120.0	+0.0%
`CapsBench.modifyDeepCap`	`-`	176.0	176.0	+0.0%
`ReviewBench.eoReverseGet_3`	`-`	72.0	72.0	-0.0%
`LensBench.eoModify`	`-`	40.0	40.0	+0.0%
`OptionalBench.eoModify_3`	`-`	160.0	160.0	+0.0%
`PlatedBench.eoUniverseExpr`	`n=512`	98,185.8	98,185.8	+0.0%
`OpticBuildBench.build`	`-`	184.0	184.0	-0.0%
`CapsBench.modifyCap`	`-`	40.0	40.0	-0.0%
`AffineFoldBench.eoGetOption_3`	`-`	16.0	16.0	-0.0%
`CapsBench.modifyDerivedHeld`	`-`	40.0	40.0	+0.0%
`AvroJsonBridgeBench.naiveWideToJson`	`-`	4,376.0	4,376.0	+0.0%
`LensBench.mReplace`	`-`	40.0	40.0	-0.0%
`ZioDiBench.handServiceGet`	`-`	120.0	120.0	+0.0%
`ZioDiBench.handDrillGet`	`-`	120.0	120.0	-0.0%
`IsoBench.eoReverseGet`	`-`	32.0	32.0	-0.0%
`MultiFocusCollectBench.naive_listSum`	`-`	56.0	56.0	-0.0%
`OptionalBench.mReplace_loyalty`	`-`	88.0	88.0	-0.0%
`KyoDiBench.eoVarUpdateFocus`	`-`	216.0	216.0	+0.0%
`KyoDiBench.handEnvUse`	`-`	312.0	312.0	-0.0%
`CapsBench.modifyDerivedPerCall`	`-`	40.0	40.0	+0.0%

Timing (ns/op) — directional only, same-VM but shared runner

Benchmark	params	base	head	Δ
`PowerSeriesNestedBench.monocle_nested`	`size=4`	1,024.9	1,218.6	+18.9%
`OptionalBench.mModify_loyalty_empty`	`-`	1.6	1.9	+18.0%
`JsoniterBench.jReadId`	`size=64`	28.4	33.0	+16.2%
`PlatedBench.visitorUniverseJson`	`n=4096`	117,366.2	135,687.5	+15.6%
`JsoniterBench.jReadId`	`size=512`	32.7	27.8	-15.2%
`JsoniterBench.jReadId`	`size=8`	32.9	28.2	-14.4%
`SchemesBench.eoCata`	`-`	66,167.5	57,023.5	-13.8%
`AffineFoldBench.mGetOption_6`	`-`	18.7	20.8	+11.0%
`MultiFocusBench.eoFold_powerEach`	`size=1024`	17,991.9	16,164.3	-10.2%
`MultiFocusBench.eoModify_multiFocus`	`size=1024`	63,564.0	57,337.7	-9.8%
`GetterBench.mGet_6`	`-`	20.7	19.0	-8.3%
`SchemesBench.eoHylo`	`-`	75,582.7	80,727.0	+6.8%
`OptionalBench.mModify_0_empty`	`-`	1.3	1.4	+6.4%
`OrderCirceBench.eoStreet`	`size=8`	877.7	822.8	-6.3%
`GetterBench.eoGet_0`	`-`	0.8	0.8	-6.2%
`MultiFocusBench.eoModify_powerEach`	`size=1024`	33,685.3	31,597.8	-6.2%
`SchemesBench.eoAna`	`-`	107,696.7	101,027.4	-6.2%
`OrderCirceBench.hcursorStreet`	`size=64`	947.0	891.4	-5.9%
`AvroBytesBench.eoModifyPartner`	`-`	327.7	308.8	-5.8%
`OrderCirceBench.eoNames`	`size=512`	201,496.0	189,922.6	-5.7%
`FoldBench.eoFoldPrices`	`size=8`	31.3	29.5	-5.7%
`AffineFoldBench.eoGetOption_0_asAffineFold`	`-`	0.8	0.8	-5.7%
`AvroBytesBench.naiveReadPartner`	`-`	1,946.6	1,839.3	-5.5%
`GetterBench.mGet_3`	`-`	6.4	6.8	+5.4%
`SchemesBench.drosteHylo`	`-`	58,893.6	55,773.7	-5.3%
`AvroVulcanBench.fieldGet_native`	`-`	87.8	83.2	-5.2%
`AvroJsonBridgeBench.eoWideToJson`	`-`	536.5	508.6	-5.2%
`OptionalBench.mModify_loyalty`	`-`	16.8	17.7	+5.1%
`AvroBytesBench.eoReadCountry`	`-`	148.5	141.6	-4.7%
`AffineFoldBench.eoGetOption_0`	`-`	0.9	0.9	-4.5%
`OrderCirceBench.eoStreetIor`	`size=8`	843.2	880.4	+4.4%
`OrderAvroBench.naiveModifyNames`	`size=8`	949.2	907.7	-4.4%
`JsoniterBench.jMiss`	`size=64`	168.8	162.0	-4.0%
`OrderCirceBench.directStreet`	`size=8`	888.3	852.9	-4.0%
`OrderCirceBench.hcursorNames`	`size=512`	181,214.3	188,075.4	+3.8%
`PowerSeriesBench.eoModify_powerEach`	`size=16`	209.8	201.9	-3.7%
`MultiFocusBench.naive_sumQty`	`size=1024`	5,371.7	5,171.0	-3.7%
`MultiFocusCollectBench.eoModify_multiFocusTuple3`	`-`	13.6	13.1	-3.7%
`PowerSeriesPrismBench.monocle_sparse`	`size=8`	368.3	355.0	-3.6%
`MultiFocusBench.eoModify_multiFocus`	`size=4`	203.8	196.5	-3.6%
`JsoniterBench.jMiss`	`size=512`	162.3	168.1	+3.6%
`AffineFoldBench.eoGetOption_0_empty`	`-`	0.9	0.9	-3.6%
`OrderCirceBench.eoNamesIor`	`size=8`	3,474.2	3,595.7	+3.5%
`IsoBench.mReverseGet`	`-`	4.6	4.5	-3.4%
`PowerSeriesNestedBench.eoModify_nested`	`size=64`	4,268.5	4,122.5	-3.4%
`JsoniterBench.jMiss`	`size=8`	168.2	162.6	-3.3%
`SchemesBench.handAna`	`-`	31,256.0	30,222.5	-3.3%
`PowerSeriesNestedBench.monocle_nested`	`size=1024`	219,267.2	226,503.8	+3.3%
`PlatedBench.visitorUniverseJson`	`n=64`	1,850.9	1,789.9	-3.3%
`AvroEncodeRouteBench.naive_freshPlumbing`	`-`	766.0	740.9	-3.3%
`AvroEncodeRouteBench.eo_encodeToAny`	`-`	53.3	51.7	-3.1%
`JsoniterBench.jModifyId`	`size=512`	5,995.5	5,817.4	-3.0%
`KyoDiBench.handMapDrillModify`	`-`	35.6	34.5	-2.9%
`OrderAvroBench.monocleModifyStreet`	`size=8`	789.1	766.0	-2.9%
`MultiFocusCollectBench.eoModify_multiFocusTuple6`	`-`	36.1	35.0	-2.9%
`OrderCirceBench.hcursorStreet`	`size=512`	947.6	920.7	-2.8%
`OpticBuildBench.buildAndUse`	`-`	929.0	902.8	-2.8%
`OrderCirceBench.directNames`	`size=64`	24,667.0	25,356.5	+2.8%
`MultiFocusCollectBench.naive_constSum`	`-`	2.3	2.4	+2.8%
`AvroJsonBridgeBench.naiveClickToJson`	`-`	2,709.7	2,634.7	-2.8%
`OrderJsoniterBench.eoModifyStreet`	`size=512`	6,113.7	5,951.3	-2.7%
`JsoniterBench.jModifyId`	`size=8`	154.2	150.2	-2.6%
`OrderJsoniterBench.monocleModifyStreet`	`size=64`	16,949.7	16,507.6	-2.6%
`OpticBuildBench.reuseUse`	`-`	855.8	877.7	+2.6%
`AvroVulcanBench.encode_vulcanRaw`	`-`	198.6	193.5	-2.5%
`OrderJsoniterBench.eoReadStreet`	`size=512`	176.5	172.1	-2.5%
`PowerSeriesPrismBench.monocle_sparse`	`size=512`	31,666.8	30,878.0	-2.5%
`OrderJsoniterBench.eoSumPrices`	`size=8`	1,351.4	1,384.9	+2.5%
`PowerSeriesBench.eoModify_powerEach`	`size=64`	747.8	729.4	-2.5%
`JsoniterBench.jModifyId`	`size=64`	768.4	749.7	-2.4%
`PlatedBench.mUniverseJson`	`n=4096`	5,212,064.3	5,086,232.7	-2.4%
`TraversalBench.mModify`	`size=512`	31,646.8	32,406.9	+2.4%
`OrderCirceBench.eoStreetIor`	`size=64`	842.5	862.6	+2.4%
`PlatedBench.eoTransformDeep`	`n=4096`	124,485.6	121,509.0	-2.4%
`OrderCirceBench.hcursorNames`	`size=64`	24,732.8	24,142.8	-2.4%
`PowerSeriesBench.eoModify_powerEach`	`size=4`	100.4	98.0	-2.4%
`PowerSeriesNestedBench.monocle_nested`	`size=16`	2,771.1	2,706.4	-2.3%
`AvroVulcanBench.encode_bridged`	`-`	199.4	194.8	-2.3%
`PowerSeriesBench.monocle_powerEach`	`size=4096`	190,270.3	185,838.4	-2.3%
`OptionalBench.mModify_6`	`-`	91.6	93.8	+2.3%
`OrderJsoniterBench.eoModifyStreet`	`size=8`	272.2	266.0	-2.3%
`MultiFocusCollectBench.eoCollectMap_zipMean`	`-`	133.4	130.4	-2.3%
`PlatedBench.mUniverseJson`	`n=512`	383,001.5	374,354.8	-2.3%
`JsoniterBench.jReadStreet`	`size=8`	174.3	170.4	-2.3%
`OrderCirceBench.hcursorNames`	`size=8`	3,428.5	3,505.7	+2.3%
`TraversalBench.eoFoldNested`	`size=64`	589.7	576.6	-2.2%
`AvroVulcanBench.rootGet_native`	`-`	133.8	136.7	+2.2%
`SchemesBench.handCata`	`-`	14,431.3	14,742.6	+2.2%
`PowerSeriesNestedBench.eoModify_nested`	`size=1024`	57,575.7	56,341.3	-2.1%
`OrderCirceBench.hcursorStreet`	`size=8`	926.9	907.2	-2.1%
`OrderCirceBench.directNames`	`size=512`	185,031.0	181,230.9	-2.1%
`OrderCirceBench.eoStreetIor`	`size=512`	854.6	872.0	+2.0%
`GetterBench.mGet_orderId`	`-`	0.4	0.4	-2.0%
`GenericsBench.genLensModify`	`-`	4.0	4.0	-2.0%
`AvroBytesBench.naiveModifyPartner`	`-`	2,842.1	2,784.7	-2.0%
`OrderCirceBench.eoNames`	`size=8`	3,368.4	3,436.2	+2.0%
`JsoniterBench.cReplaceId`	`size=64`	42,114.9	41,273.7	-2.0%
`OrderCirceBench.naiveStreet`	`size=8`	2,984.2	2,924.8	-2.0%
`PowerSeriesNestedBench.eoModify_nested`	`size=16`	1,216.4	1,240.6	+2.0%
`PowerSeriesPrismBench.monocle_sparse`	`size=128`	4,496.4	4,407.1	-2.0%
`JsoniterBench.jReplaceId`	`size=512`	5,912.6	5,798.8	-1.9%
`JsoniterBench.jSumPrices`	`size=8`	1,289.3	1,313.8	+1.9%
`OrderAvroBench.eoModifyStreet`	`size=512`	100.9	99.0	-1.9%
`AvroDecodeReuseBench.confluentRecordReader`	`-`	352.7	346.1	-1.9%
`PowerSeriesBench.eoModify_powerEach`	`size=1024`	11,957.1	11,736.2	-1.8%
`AvroBytesBench.prunedReadCountry`	`-`	882.1	865.8	-1.8%
`JsoniterBench.cReadStreet`	`size=8`	3,351.1	3,412.9	+1.8%
`PowerSeriesBench.monocle_powerEach`	`size=64`	2,293.1	2,335.1	+1.8%
`KyoDiBench.eoMapDrillModify`	`-`	37.8	37.1	-1.8%
`PrismBench.eoGetOptionPresent`	`-`	0.8	0.8	-1.8%
`CompositionBench.buildLens6`	`-`	52.2	51.3	-1.8%
`TraversalBench.eoFoldPrices`	`size=512`	2,150.5	2,112.0	-1.8%
`PowerSeriesNestedBench.monocle_nested`	`size=64`	9,782.9	9,609.1	-1.8%
`OrderAvroBench.monocleModifyStreet`	`size=64`	5,031.7	4,943.3	-1.8%
`PowerSeriesPrismBench.monocle_sparse`	`size=32`	1,159.9	1,180.0	+1.7%
`OrderJsoniterBench.naiveModifyStreet`	`size=64`	16,736.0	16,447.5	-1.7%
`MultiFocusBench.naive_listMap`	`size=1024`	10,511.9	10,332.6	-1.7%
`OrderJsoniterBench.monocleSumPrices`	`size=8`	2,148.7	2,184.8	+1.7%
`MultiFocusBench.naive_listMap`	`size=4`	45.9	45.2	-1.7%
`PrismBench.eoGetRightPresent`	`-`	2.6	2.6	-1.6%
`MultiFocusBench.eoFold_powerEach`	`size=32`	323.6	318.3	-1.6%
`TraversalBench.eoModify`	`size=64`	898.0	883.5	-1.6%
`OrderCirceBench.eoNamesIor`	`size=512`	206,096.0	209,425.7	+1.6%
`PlatedBench.eoUniverseExpr`	`n=64`	2,033.2	2,000.4	-1.6%
`MultiFocusBench.eoModify_multiFocus`	`size=32`	1,573.2	1,548.1	-1.6%
`MultiFocusBench.eoFold_powerEach`	`size=256`	2,978.9	2,931.5	-1.6%
`FoldBench.eoFoldMap`	`size=512`	1,395.4	1,373.4	-1.6%
`PowerSeriesNestedBench.naive_nested`	`size=16`	375.5	369.6	-1.6%
`OrderAvroBench.naiveModifyStreet`	`size=512`	38,218.7	37,625.3	-1.6%
`OptionalBench.eoModify_0`	`-`	19.7	19.4	-1.5%
`PlatedBench.mUniverseExpr`	`n=4096`	4,931,665.0	4,857,872.5	-1.5%
`OrderJsoniterBench.naiveModifyStreet`	`size=8`	2,948.4	2,904.6	-1.5%
`AffineFoldBench.eoGetOption_0_asOptional`	`-`	2.7	2.6	-1.5%
`AvroBytesBench.eoReadPartner`	`-`	168.3	165.8	-1.5%
`OrderJsoniterBench.monocleSumPrices`	`size=512`	340,376.9	335,366.7	-1.5%
`GenericsBench.handPrismModifyMiss`	`-`	0.9	0.9	-1.5%
`JsoniterBench.jReplaceId`	`size=8`	147.6	145.5	-1.5%
`OrderAvroBench.monocleReadStreet`	`size=512`	19,881.7	19,593.3	-1.5%
`JsoniterBench.cReadStreet`	`size=64`	21,800.6	21,485.9	-1.4%
`MultiFocusCollectBench.naive_tuple3Rewrite`	`-`	12.8	12.6	-1.4%
`FoldBench.mFoldPrices`	`size=512`	2,588.0	2,551.0	-1.4%
`PowerSeriesPrismBench.eoModify_sparse`	`size=2048`	21,535.0	21,227.7	-1.4%
`OrderCirceBench.eoNames`	`size=64`	25,120.1	24,761.7	-1.4%
`AvroJsonBridgeBench.naiveWideToAvro`	`-`	1,080.0	1,064.7	-1.4%
`JsoniterBench.jReadStreet`	`size=64`	172.4	174.8	+1.4%
`AvroJsonBridgeBench.eoClickToAvro`	`-`	2,612.8	2,576.0	-1.4%
`MultiFocusBench.naive_listMap`	`size=32`	323.4	318.8	-1.4%
`ClickRecordBench.encode_derivedThroughPrism`	`-`	440.3	446.5	+1.4%
`PlatedBench.eoTransformExpr`	`n=4096`	126,991.9	125,208.2	-1.4%
`PrismBench.mGetRightPresent`	`-`	2.7	2.7	-1.4%
`PowerSeriesPrismBench.monocle_sparse`	`size=2048`	100,082.0	98,695.1	-1.4%
`PowerSeriesNestedBench.naive_nested`	`size=4`	133.7	131.8	-1.4%
`GenericsBench.rawLensModify`	`-`	3.5	3.5	-1.4%
`OrderAvroBench.eoReadStreet`	`size=8`	28.6	28.2	-1.4%
`PowerSeriesPrismBench.eoModify_sparse`	`size=512`	5,369.5	5,297.1	-1.3%
`AvroBytesBench.naivePassthroughPayload`	`-`	4,710.2	4,646.9	-1.3%
`PlatedBench.visitorTransformExpr`	`n=4096`	66,600.4	65,713.5	-1.3%
`GenericsBench.handLensModify`	`-`	3.5	3.6	+1.3%
`FoldBench.mFoldPrices`	`size=64`	219.6	216.7	-1.3%
`OrderJsoniterBench.eoModifyStreet`	`size=64`	879.5	867.9	-1.3%
`SetterBench.mModify_orderId`	`-`	5.4	5.3	-1.3%
`AvroVulcanBench.fieldGet_bridged`	`-`	83.1	82.0	-1.3%
`OptionalBench.mReplace_0`	`-`	5.5	5.4	-1.3%
`CompositionBench.reuseLens1`	`-`	11.8	11.7	-1.3%
`OrderAvroBench.monocleReadStreet`	`size=64`	2,520.9	2,488.7	-1.3%
`CompositionBench.buildLens1`	`-`	10.0	9.9	-1.3%
`PlatedBench.mTransformExpr`	`n=512`	13,524.3	13,352.3	-1.3%
`GenericsBench.rawPrismGetHit`	`-`	2.7	2.7	-1.3%
`PrismBench.mReverseGet`	`-`	2.5	2.5	-1.3%
`GetterBench.eoGet_orderId`	`-`	0.8	0.8	+1.3%
`OrderCirceBench.monocleStreet`	`size=8`	2,958.6	2,921.3	-1.3%
`PowerSeriesBench.naive_powerEach`	`size=1024`	5,837.9	5,764.7	-1.3%
`OrderAvroBench.monocleModifyStreet`	`size=512`	38,305.4	37,828.0	-1.2%
`LensBench.eoModifyDeep`	`-`	28.4	28.1	-1.2%
`PowerSeriesPrismBench.naive_sparse`	`size=128`	348.3	344.0	-1.2%
`IsoBench.mGet`	`-`	4.6	4.5	-1.2%
`OptionalBench.eoModify_6`	`-`	103.6	104.9	+1.2%
`PlatedBench.mUniverseExpr`	`n=512`	346,040.2	350,216.4	+1.2%
`PowerSeriesNestedBench.naive_nested`	`size=256`	5,270.2	5,207.1	-1.2%
`ReviewBench.naiveBuild_6`	`-`	16.6	16.4	-1.2%
`OrderAvroBench.naiveModifyStreet`	`size=64`	4,976.6	4,917.6	-1.2%
`AvroBytesBench.prunedReadPartner`	`-`	604.0	596.9	-1.2%
`OrderAvroBench.eoReadStreet`	`size=512`	28.5	28.2	-1.2%
`PowerSeriesPrismBench.naive_sparse`	`size=32`	89.2	88.1	-1.2%
`PlatedBench.visitorTransformDeep`	`n=512`	4,702.1	4,648.3	-1.1%
`PlatedBench.visitorUniverseExpr`	`n=512`	6,473.6	6,399.8	-1.1%
`SetterBench.eoModify_3`	`-`	11.8	11.7	-1.1%
`FoldBench.eoFoldPrices`	`size=64`	219.3	216.8	-1.1%
`OrderJsoniterBench.naiveSumPrices`	`size=512`	79,948.8	79,050.5	-1.1%
`PlatedBench.mTransformExpr`	`n=4096`	149,074.0	147,400.8	-1.1%
`MultiFocusBench.naive_listMap`	`size=256`	2,556.6	2,527.9	-1.1%
`ClickRecordBench.encode_hand`	`-`	125.3	123.9	-1.1%
`ClickRecordBench.encode_derived`	`-`	442.9	447.7	+1.1%
`OrderJsoniterBench.naiveReadStreet`	`size=8`	1,501.7	1,518.0	+1.1%
`PowerSeriesPrismBench.eoModify_sparse`	`size=8`	105.0	103.9	-1.1%
`PlatedBench.eoUniverseDeep`	`n=64`	2,076.3	2,054.0	-1.1%
`PowerSeriesPrismBench.eoModify_sparse`	`size=128`	1,249.4	1,236.0	-1.1%
`PlatedBench.eoTransformDeep`	`n=512`	13,814.5	13,667.4	-1.1%
`OrderJsoniterBench.eoReadStreet`	`size=8`	172.9	171.1	-1.1%
`JsoniterBench.cReadId`	`size=8`	3,412.3	3,376.7	-1.0%
`TraversalBench.eoModify`	`size=512`	7,803.3	7,884.4	+1.0%
`GenericsBench.genPrismModifyHit`	`-`	3.4	3.4	-1.0%
`GenericsBench.genPrismGetMiss`	`-`	1.0	1.0	+1.0%
`OrderAvroBench.monocleModifyNames`	`size=512`	80,103.0	79,280.9	-1.0%
`AvroJsonBridgeBench.eoClickToJson`	`-`	2,457.4	2,432.4	-1.0%
`LensBench.mModify`	`-`	5.6	5.5	-1.0%
`OrderJsoniterBench.monocleModifyStreet`	`size=8`	2,925.7	2,896.5	-1.0%
`PrismBench.mRightReverseGet`	`-`	2.5	2.5	-1.0%
`OrderJsoniterBench.naiveModifyStreet`	`size=512`	137,584.3	136,217.7	-1.0%
`OrderAvroBench.eoModifyNames`	`size=512`	28,939.7	28,653.3	-1.0%
`OrderJsoniterBench.nativeReadStreet`	`size=64`	4,322.7	4,365.4	+1.0%
`PrismBench.eoGetRightAbsent`	`-`	0.9	0.9	-1.0%
`OrderCirceBench.monocleNames`	`size=64`	22,328.5	22,108.7	-1.0%
`CapsBench.modifyDirect`	`-`	5.6	5.5	-1.0%
`MultiFocusBench.naive_sumQty`	`size=256`	659.3	652.8	-1.0%
`JsoniterBench.cReplaceId`	`size=512`	321,313.3	324,434.1	+1.0%
`FoldBench.mFoldMap`	`size=512`	1,376.8	1,363.5	-1.0%
`MultiFocusBench.naive_sumQty`	`size=32`	59.7	59.1	-1.0%
`OrderJsoniterBench.monocleSumPrices`	`size=64`	13,276.1	13,149.0	-1.0%
`OrderCirceBench.directStreet`	`size=64`	881.7	873.3	-1.0%
`JsoniterBench.jSumPrices`	`size=64`	9,880.4	9,786.1	-1.0%
`AvroDecodeReuseBench.freshDecodeRecord`	`-`	1,246.2	1,234.3	-1.0%
`CompositionBench.buildLens3`	`-`	27.2	26.9	-0.9%
`ReviewBench.eoReverseGet_6`	`-`	25.7	25.4	-0.9%
`OrderAvroBench.monocleReadStreet`	`size=8`	376.5	380.1	+0.9%
`JsoniterBench.cModifyId`	`size=8`	7,054.9	7,120.9	+0.9%
`OrderCirceBench.eoStreet`	`size=512`	849.1	841.2	-0.9%
`OptionalBench.mModify_0`	`-`	19.5	19.3	-0.9%
`SchemesBench.drosteAna`	`-`	55,107.2	54,597.2	-0.9%
`JsoniterBench.cModifyId`	`size=512`	326,594.3	323,632.9	-0.9%
`OptionalBench.eoModify_loyalty`	`-`	18.0	17.8	-0.9%
`GenericsBench.rawPrismModifyHit`	`-`	3.4	3.3	-0.9%
`AvroBytesBench.naiveModifyCountry`	`-`	2,452.9	2,474.9	+0.9%
`TraversalBench.eoModify`	`size=8`	103.9	103.0	-0.9%
`PlatedBench.eoUniverseDeep`	`n=4096`	146,206.7	144,916.1	-0.9%
`TraversalBench.eoFoldPrices`	`size=64`	219.1	217.1	-0.9%
`PlatedBench.eoUniverseDeep`	`n=512`	17,832.8	17,676.0	-0.9%
`TraversalBench.mModify`	`size=8`	331.4	328.5	-0.9%
`MultiFocusBench.eoModify_powerEach`	`size=32`	858.5	851.1	-0.9%
`JsoniterBench.cSumPrices`	`size=8`	5,344.1	5,298.3	-0.9%
`PlatedBench.eoUniverseJson`	`n=4096`	202,817.9	204,557.7	+0.9%
`JsoniterBench.cReplaceId`	`size=8`	7,056.4	6,996.0	-0.9%
`OrderAvroBench.naiveReadStreet`	`size=8`	383.7	380.5	-0.9%
`GenericsBench.handPrismGetHit`	`-`	2.4	2.5	+0.9%
`PlatedBench.eoTransformExpr`	`n=512`	15,942.9	15,807.8	-0.8%
`PowerSeriesPrismBench.eoModify_sparse`	`size=32`	326.4	323.6	-0.8%
`MultiFocusBench.eoModify_powerEach`	`size=4`	145.8	144.5	-0.8%
`OrderAvroBench.naiveModifyNames`	`size=512`	48,939.7	48,535.1	-0.8%
`JsoniterBench.cModifyId`	`size=64`	41,754.7	41,411.8	-0.8%
`JsoniterBench.cSumPrices`	`size=512`	283,705.4	281,379.8	-0.8%
`MultiFocusCollectBench.eoCollectMap_constSum`	`-`	1.6	1.6	+0.8%
`SetterBench.mModify_6`	`-`	53.5	54.0	+0.8%
`OrderAvroBench.naiveModifyStreet`	`size=8`	762.1	755.9	-0.8%
`PlatedBench.visitorTransformExpr`	`n=512`	7,815.9	7,753.5	-0.8%
`TraversalBench.mFoldPrices`	`size=64`	2,462.1	2,442.5	-0.8%
`TraversalBench.eoFoldNested`	`size=8`	52.6	52.2	-0.8%
`GenericsBench.genPrismGetHit`	`-`	2.5	2.4	-0.8%
`MultiFocusBench.naive_sumQty`	`size=4`	6.3	6.3	-0.8%
`OrderAvroBench.eoModifyNames`	`size=64`	3,441.5	3,468.2	+0.8%
`AvroBytesBench.eoModifyCountry`	`-`	267.7	265.7	-0.8%
`CapsBench.foldMapDirect`	`-`	19.1	19.3	+0.8%
`MultiFocusCollectBench.eoCollectList_listSum`	`-`	28.6	28.4	-0.8%
`AvroEncodeRouteBench.handwritten_stream`	`-`	157.4	158.6	+0.8%
`PrismBench.eoRightReverseGet`	`-`	2.6	2.6	-0.7%
`OrderJsoniterBench.nativeReadStreet`	`size=8`	791.8	797.7	+0.7%
`PowerSeriesNestedBench.eoModify_nested`	`size=4`	583.6	579.3	-0.7%
`PowerSeriesBench.eoModify_powerEach`	`size=256`	2,736.5	2,756.1	+0.7%
`PowerSeriesBench.monocle_powerEach`	`size=1024`	55,371.6	55,765.4	+0.7%
`JsoniterBench.jReadStreet`	`size=512`	174.7	173.5	-0.7%
`GenericsBench.genPrismModifyMiss`	`-`	1.0	1.0	+0.7%
`ReviewBench.naiveBuild_0`	`-`	3.5	3.4	-0.7%
`MultiFocusCollectBench.naive_zipMeanBroadcast`	`-`	223.0	221.5	-0.7%
`GenericsBench.handPrismModifyHit`	`-`	3.4	3.4	-0.7%
`JsoniterBench.jReplaceId`	`size=64`	745.4	740.3	-0.7%
`GenericsBench.handLensGet`	`-`	0.9	0.9	-0.7%
`TraversalBench.eoFoldPrices`	`size=8`	29.6	29.4	-0.7%
`AffineFoldBench.mGetOption_3`	`-`	7.1	7.0	-0.7%
`FoldBench.mFoldMap`	`size=64`	150.5	149.5	-0.7%
`OptionalBench.eoReplace_loyalty`	`-`	11.8	11.8	-0.7%
`JsoniterBench.jSumPrices`	`size=512`	78,292.5	78,812.6	+0.7%
`PlatedBench.visitorUniverseDeep`	`n=512`	7,329.4	7,281.0	-0.7%
`AvroVulcanBench.rootGet_bridged`	`-`	307.2	305.2	-0.6%
`OrderJsoniterBench.naiveSumPrices`	`size=64`	9,866.9	9,929.6	+0.6%
`ClickRecordBench.encode_vulcanFull`	`-`	5,613.2	5,577.8	-0.6%
`PrismBench.mGetOptionAbsent`	`-`	0.9	0.9	-0.6%
`OrderAvroBench.monocleModifyNames`	`size=64`	8,264.4	8,212.8	-0.6%
`OrderJsoniterBench.monocleModifyStreet`	`size=512`	137,874.5	137,013.6	-0.6%
`PlatedBench.visitorTransformExpr`	`n=64`	1,020.9	1,014.6	-0.6%
`PowerSeriesBench.monocle_powerEach`	`size=256`	19,758.7	19,881.0	+0.6%
`SchemesBench.drosteCata`	`-`	35,530.9	35,311.2	-0.6%
`PowerSeriesBench.eoModify_powerEach`	`size=4096`	48,391.2	48,096.3	-0.6%
`GenericsBench.genLensGet`	`-`	0.9	0.9	-0.6%
`MultiFocusBench.eoModify_multiFocus`	`size=256`	13,183.2	13,103.4	-0.6%
`OrderCirceBench.naiveStreet`	`size=512`	148,399.0	147,507.4	-0.6%
`AvroJsonBridgeBench.naiveClickToAvro`	`-`	1,439.3	1,430.7	-0.6%
`CompositionBench.reuseLeaf`	`-`	3.5	3.5	-0.6%
`OrderJsoniterBench.monocleReadStreet`	`size=8`	1,494.3	1,485.6	-0.6%
`FoldBench.eoFoldMap`	`size=64`	150.3	149.5	-0.6%
`AvroBytesBench.eoSliceGraftPayload`	`-`	236.7	238.0	+0.6%
`FoldBench.mFoldPrices`	`size=8`	29.7	29.5	-0.6%
`MultiFocusBench.eoFold_powerEach`	`size=4`	62.3	61.9	-0.6%
`PowerSeriesPrismBench.naive_sparse`	`size=512`	1,265.9	1,259.0	-0.5%
`OrderJsoniterBench.monocleReadStreet`	`size=512`	73,385.0	72,987.9	-0.5%
`PlatedBench.visitorUniverseDeep`	`n=4096`	60,254.1	59,929.1	-0.5%
`OrderAvroBench.eoModifyStreet`	`size=8`	99.5	99.0	-0.5%
`PlatedBench.visitorUniverseExpr`	`n=4096`	52,656.3	52,373.9	-0.5%
`SchemesBench.handHylo`	`-`	11,989.8	11,925.7	-0.5%
`AffineFoldBench.eoGetOption_6`	`-`	21.9	21.8	-0.5%
`PlatedBench.eoTransformExpr`	`n=64`	1,934.6	1,924.7	-0.5%
`IsoBench.eoGet`	`-`	4.6	4.5	-0.5%
`OrderCirceBench.directStreet`	`size=512`	850.1	854.4	+0.5%
`OrderAvroBench.naiveReadStreet`	`size=64`	2,504.2	2,491.8	-0.5%
`PowerSeriesNestedBench.eoModify_nested`	`size=256`	17,244.7	17,160.5	-0.5%
`TraversalBench.eoFoldNested`	`size=512`	5,067.9	5,043.3	-0.5%
`LensBench.eoReplace`	`-`	5.4	5.4	-0.5%
`OrderAvroBench.eoReadStreet`	`size=64`	28.3	28.2	-0.5%
`PrismBench.eoGetOptionAbsent`	`-`	0.8	0.8	-0.5%
`PlatedBench.eoTransformDeep`	`n=64`	1,444.0	1,437.1	-0.5%
`GenericsBench.handPrismGetMiss`	`-`	0.9	0.9	+0.5%
`ZioDiBench.handServiceReplace`	`-`	90.8	90.3	-0.5%
`CompositionBench.buildLensOptional3`	`-`	26.9	26.8	-0.5%
`TraversalBench.mModify`	`size=64`	2,260.8	2,250.4	-0.5%
`PowerSeriesBench.naive_powerEach`	`size=4`	27.2	27.0	-0.5%
`TraversalBench.mFoldPrices`	`size=8`	375.3	377.0	+0.5%
`PlatedBench.visitorUniverseJson`	`n=512`	21,901.3	22,000.3	+0.5%
`ZioDiBench.handDrillModify`	`-`	100.6	101.1	+0.5%
`PlatedBench.visitorUniverseDeep`	`n=64`	748.6	745.4	-0.4%
`AvroBytesBench.eoGraftPayload`	`-`	120.9	120.4	-0.4%
`ReviewBench.eoReverseGet_3`	`-`	14.8	14.7	-0.4%
`PowerSeriesNestedBench.naive_nested`	`size=1024`	21,398.6	21,306.4	-0.4%
`LensBench.mGet`	`-`	1.0	1.0	-0.4%
`OptionalBench.eoReplace_0`	`-`	5.7	5.7	-0.4%
`PrismBench.eoReverseGet`	`-`	2.6	2.6	-0.4%
`OrderJsoniterBench.monocleReadStreet`	`size=64`	9,269.6	9,230.7	-0.4%
`OptionalBench.eoModify_0_empty`	`-`	1.3	1.3	-0.4%
`PlatedBench.eoUniverseExpr`	`n=4096`	143,960.4	144,544.2	+0.4%
`LensBench.mModifyDeep`	`-`	28.2	28.4	+0.4%
`AvroJsonBridgeBench.naiveWideToJson`	`-`	1,646.6	1,640.0	-0.4%
`KyoDiBench.handMapGet`	`-`	1.6	1.6	-0.4%
`SetterBench.mModify_0`	`-`	3.4	3.4	+0.4%
`PowerSeriesBench.naive_powerEach`	`size=16`	91.0	90.6	-0.4%
`OptionalBench.eoModify_3`	`-`	44.0	43.8	-0.4%
`CapsBench.getDirect`	`-`	0.9	0.9	-0.4%
`PlatedBench.mTransformExpr`	`n=64`	2,295.7	2,286.8	-0.4%
`ZioDiBench.handDrillGet`	`-`	17.3	17.2	-0.4%
`OrderAvroBench.naiveReadStreet`	`size=512`	19,664.1	19,588.0	-0.4%
`AvroDecodeReuseBench.cachedDecodeRecord`	`-`	309.1	307.9	-0.4%
`CapsBench.foldMapDerivedHeld`	`-`	20.6	20.6	-0.4%
`CapsBench.getDeepCap`	`-`	1.3	1.3	-0.4%
`MultiFocusBench.eoModify_powerEach`	`size=256`	7,039.8	7,066.5	+0.4%
`AffineFoldBench.mGetOption_0_empty`	`-`	0.8	0.8	+0.4%
`ZioDiBench.eoDrillModify`	`-`	95.5	95.8	+0.4%
`OrderAvroBench.monocleModifyNames`	`size=8`	1,212.9	1,217.4	+0.4%
`AvroVulcanBench.decode_bridged`	`-`	156.1	155.5	-0.4%
`OrderAvroBench.eoModifyStreet`	`size=64`	99.3	99.0	-0.4%
`TraversalBench.mFoldPrices`	`size=512`	29,152.3	29,258.2	+0.4%
`PowerSeriesNestedBench.naive_nested`	`size=64`	1,323.1	1,318.3	-0.4%
`JsoniterBench.cReadStreet`	`size=512`	170,724.0	170,132.4	-0.3%
`OrderCirceBench.naiveNames`	`size=512`	159,045.7	159,594.1	+0.3%
`OrderJsoniterBench.nativeSumPrices`	`size=8`	992.7	996.1	+0.3%
`OrderAvroBench.naiveModifyNames`	`size=64`	6,276.0	6,254.8	-0.3%
`OrderJsoniterBench.nativeSumPrices`	`size=64`	6,665.1	6,642.9	-0.3%
`PowerSeriesBench.naive_powerEach`	`size=256`	1,342.9	1,338.5	-0.3%
`KyoDiBench.handVarUpdate`	`-`	66.8	67.1	+0.3%
`MultiFocusCollectBench.naive_tuple6Rewrite`	`-`	24.3	24.2	-0.3%
`GetterBench.eoGet_3`	`-`	11.8	11.7	-0.3%
`CapsBench.modifyDeepDirect`	`-`	28.2	28.1	-0.3%
`SetterBench.eoModify_6`	`-`	21.5	21.6	+0.3%
`OrderCirceBench.monocleStreet`	`size=512`	149,004.0	148,529.6	-0.3%
`PowerSeriesBench.monocle_powerEach`	`size=16`	630.4	632.4	+0.3%
`MultiFocusCollectBench.naive_listSum`	`-`	26.0	25.9	-0.3%
`PlatedBench.eoUniverseJson`	`n=512`	24,028.7	23,953.5	-0.3%
`CapsBench.getDerivedHeld`	`-`	1.9	1.9	-0.3%
`JsoniterBench.cReadId`	`size=64`	21,499.3	21,565.3	+0.3%
`AffineFoldBench.mGetOption_loyalty`	`-`	0.9	0.9	-0.3%
`CapsBench.foldMapDerivedPerCall`	`-`	19.3	19.4	+0.3%
`PrismBench.mGetOptionPresent`	`-`	0.9	0.9	-0.3%
`KyoDiBench.eoMapGet`	`-`	2.1	2.1	-0.3%
`OrderJsoniterBench.naiveReadStreet`	`size=64`	9,218.8	9,191.5	-0.3%
`GenericsBench.rawLensGet`	`-`	0.7	0.7	+0.3%
`OptionalBench.eoModify_loyalty_empty`	`-`	1.3	1.3	-0.3%
`AvroJsonBridgeBench.eoWideToAvro`	`-`	713.0	711.1	-0.3%
`GetterBench.eoGet_6`	`-`	22.8	22.7	-0.3%
`CapsBench.modifyDerivedHeld`	`-`	5.7	5.7	+0.3%
`OrderCirceBench.monocleStreet`	`size=64`	18,763.9	18,713.0	-0.3%
`LensBench.mReplace`	`-`	5.3	5.3	-0.3%
`FoldBench.eoFoldPrices`	`size=512`	2,594.5	2,587.7	-0.3%
`AvroVulcanBench.decode_native`	`-`	13.0	13.0	-0.3%
`SetterBench.eoModify_orderId`	`-`	5.3	5.3	-0.3%
`CompositionBench.reuseLensOptional3`	`-`	43.9	43.8	-0.3%
`OrderCirceBench.naiveNames`	`size=64`	20,074.6	20,023.4	-0.3%
`ClickRecordBench.encode_positional`	`-`	4,691.1	4,679.1	-0.3%
`JsoniterBench.cSumPrices`	`size=64`	35,115.9	35,028.0	-0.3%
`OrderAvroBench.eoModifyNames`	`size=8`	492.4	491.2	-0.2%
`OrderJsoniterBench.eoSumPrices`	`size=512`	79,098.2	78,901.5	-0.2%
`PowerSeriesBench.naive_powerEach`	`size=64`	342.3	343.2	+0.2%
`AvroDecodeReuseBench.confluentRecordReaderFresh`	`-`	1,311.4	1,308.3	-0.2%
`CapsBench.modifyDerivedPerCall`	`-`	5.6	5.6	+0.2%
`PrismBench.mGetRightAbsent`	`-`	1.0	1.0	-0.2%
`PowerSeriesPrismBench.naive_sparse`	`size=8`	23.0	22.9	-0.2%
`LensBench.eoGet`	`-`	0.9	0.9	-0.2%
`PowerSeriesPrismBench.naive_sparse`	`size=2048`	5,215.0	5,203.4	-0.2%
`OpticBuildBench.build`	`-`	41.6	41.5	-0.2%
`OrderCirceBench.monocleNames`	`size=512`	193,032.9	192,615.2	-0.2%
`ReviewBench.eoReverseGet_0`	`-`	3.5	3.5	-0.2%
`CapsBench.getDerivedPerCall`	`-`	1.3	1.3	+0.2%
`SetterBench.mModify_3`	`-`	27.4	27.4	-0.2%
`PowerSeriesBench.naive_powerEach`	`size=4096`	21,729.5	21,773.8	+0.2%
`OrderJsoniterBench.nativeReadStreet`	`size=512`	33,654.8	33,586.5	-0.2%
`OrderCirceBench.eoNamesIor`	`size=64`	27,872.5	27,927.0	+0.2%
`AffineFoldBench.eoGetOption_loyalty`	`-`	0.9	0.9	-0.2%
`CapsBench.modifyDeepCap`	`-`	28.8	28.9	+0.2%
`AvroVulcanBench.encode_native`	`-`	10.6	10.6	+0.2%
`PlatedBench.eoUniverseJson`	`n=64`	2,693.1	2,687.9	-0.2%
`ZioDiBench.eoServiceGet`	`-`	4.7	4.7	-0.2%
`PlatedBench.visitorUniverseExpr`	`n=64`	765.3	763.9	-0.2%
`JsoniterBench.cReadId`	`size=512`	171,438.8	171,746.8	+0.2%
`OrderJsoniterBench.eoSumPrices`	`size=64`	9,788.0	9,805.3	+0.2%
`KyoDiBench.handEnvUse`	`-`	89.6	89.4	-0.2%
`CompositionBench.reuseLens3`	`-`	32.7	32.8	+0.2%
`OrderJsoniterBench.naiveSumPrices`	`size=8`	1,567.6	1,565.0	-0.2%
`AffineFoldBench.eoGetOption_3`	`-`	10.6	10.6	-0.2%
`AvroEncodeRouteBench.eo_encodeValue`	`-`	519.5	520.3	+0.2%
`OrderCirceBench.naiveNames`	`size=8`	3,096.4	3,101.1	+0.2%
`CapsBench.modifyCap`	`-`	5.5	5.5	-0.1%
`ZioDiBench.eoDrillGet`	`-`	8.7	8.7	-0.1%
`PlatedBench.mUniverseJson`	`n=64`	34,801.7	34,850.2	+0.1%
`SetterBench.eoModify_0`	`-`	3.4	3.4	-0.1%
`KyoDiBench.eoEnvFocus`	`-`	91.5	91.3	-0.1%
`AffineFoldBench.eoGetOption_loyalty_empty`	`-`	0.9	0.9	+0.1%
`PowerSeriesNestedBench.monocle_nested`	`size=256`	83,320.8	83,432.5	+0.1%
`OrderCirceBench.monocleNames`	`size=8`	3,452.3	3,456.8	+0.1%
`ReviewBench.naiveBuild_3`	`-`	10.0	10.0	+0.1%
`OrderCirceBench.naiveStreet`	`size=64`	18,754.4	18,730.7	-0.1%
`OrderJsoniterBench.nativeSumPrices`	`size=512`	55,217.5	55,149.4	-0.1%
`OrderCirceBench.directNames`	`size=8`	3,307.4	3,303.4	-0.1%
`PlatedBench.mUniverseExpr`	`n=64`	30,488.8	30,523.6	+0.1%
`CapsBench.foldMapCap`	`-`	19.2	19.2	-0.1%
`FoldBench.mFoldMap`	`size=8`	19.0	19.0	+0.1%
`ZioDiBench.handServiceGet`	`-`	17.3	17.3	+0.1%
`CapsBench.getCap`	`-`	1.0	1.0	+0.1%
`IsoBench.eoReverseGet`	`-`	4.4	4.4	-0.1%
`AffineFoldBench.mGetOption_loyalty_empty`	`-`	0.9	0.9	-0.1%
`ZioDiBench.eoServiceReplace`	`-`	81.8	81.7	-0.1%
`AffineFoldBench.mGetOption_0`	`-`	0.8	0.8	-0.1%
`PlatedBench.visitorTransformDeep`	`n=4096`	43,970.7	43,996.9	+0.1%
`OptionalBench.mModify_3`	`-`	56.3	56.3	+0.1%
`OrderJsoniterBench.eoReadStreet`	`size=64`	176.6	176.7	+0.1%
`OrderJsoniterBench.naiveReadStreet`	`size=512`	73,175.2	73,137.7	-0.1%
`FoldBench.eoFoldMap`	`size=8`	19.1	19.1	-0.0%
`AvroVulcanBench.decode_vulcanRaw`	`-`	153.7	153.6	-0.0%
`LensBench.eoModify`	`-`	5.5	5.5	+0.0%
`PlatedBench.visitorTransformDeep`	`n=64`	520.2	520.0	-0.0%
`OptionalBench.mReplace_loyalty`	`-`	11.7	11.7	-0.0%
`CompositionBench.reuseLens6`	`-`	89.5	89.5	-0.0%
`CapsBench.getDeepDirect`	`-`	1.2	1.2	-0.0%
`OrderCirceBench.eoStreet`	`size=64`	856.7	856.5	-0.0%
`PowerSeriesBench.monocle_powerEach`	`size=4`	172.7	172.7	-0.0%
`GetterBench.mGet_0`	`-`	0.4	0.4	-0.0%
`PlatedBench.eoUniverseExpr`	`n=512`	17,424.7	17,426.3	+0.0%
`AvroBytesBench.naiveReadCountry`	`-`	1,417.5	1,417.5	+0.0%
`KyoDiBench.eoVarUpdateFocus`	`-`	74.7	74.7	-0.0%

_{base_sha: e996b12b897f2448ce09599f4b0d46345554e85b · head_sha: 94a8ee0ecbd4717af683298b1f2d575f0301ffec · jdk: temurin-25 · runner: ubuntu-22.04 · jmh_params: -i 3 -wi 2 -f 1 -t 1 -foe true -prof gc -rf json -jvmArgsAppend -XX:-ResizeTLAB -jvmArgsAppend -Xms1g -jvmArgsAppend -Xmx1g · profile: pr:-i3-wi2-f1-t1-gc-rt1g}

Informational only · head 94a8ee0ecbd4717af683298b1f2d575f0301ffec · benchmark run

Translation speed. The Cojen Maker version translated 2.3–3.9× slower than the original raw-ASM emitter. B is now ASM's own ClassNode, written directly by Lower with asm-commons' InstructionAdapter (no IR of our own, no library above ASM); reverseGet is ClassWriter. Strings ASM hashes are built once, switches use the last arm as default, Rt descriptors are precomputed. Translation is back to the raw-ASM emitter's cost (within noise; multisig ~20% slower), translate + load is level with it. Code size (code lines, no comments/blanks, vs the raw-ASM version): main 1425 → 991 (−30.5%), tests 534 → 330, validators 192 → 153, total 2218 → 1536 (−30.7%). JvmProgram, Emit and BuiltinCalls are gone. More Scalus reuse: - saturated builtins call scalus.uplc.builtin.Builtins directly on unboxed values; selectors, builtins used as values and everything else run Scalus' CEK meanings (BuiltinRuntime), with non-constant arguments swapped back by reference — no reflective twin; - chooseIndex runs the selector's own meaning on the branch indices; - builtin result types instantiate Scalus' TypeSchemes at the argument types, so `fstPair (unConstrData d)` is an integer and the tag match on it is a JVM switch. Validation speed (JMH, 3 forks): 8–21× the CEK machine on the context-decoding validators, 63× on Fibonacci; 2.9–7.1× faster than the raw-ASM version. Mechanisms: lazy Data lists (DataList / ConstrPair), direct Builtins calls, self-recursion as direct invokestatic, pure builtin lets computed where read, exact-arity applyAll, typed literals. Tests: ValidatorSpec is a loop over Scenarios.all (accepting and rejecting rows shared with the walkthrough and the benchmarks); CompilerSpec cases are one-liners; a vacuous `Rt.thunk` assertion now checks `Rt.closure`. dropList on a non-list now fails, as on the CEK machine. README, docs page (B as bytecode, new benchmark tables) and CLAUDE.md updated. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JG5LJsPA524HKbSMvjStrv

- Rt.fail throws a MachineError without a stack trace. Filling it cost ~1.3 µs — more than a whole validation — so rejecting transactions took 1.8–2.7 µs against 0.4–1.4 µs for accepting ones; now 0.5–1.6 µs. - Exec.run's one catch is a Try turned into Scalus' Result; a compiler bug (AssertionError) still escapes, and discharge runs outside it. - Measured alternative, not taken: a result value from every call needs a check after each one — validation unchanged, translation 20–35% slower. Rt's scaladoc and the docs page say so. - Expr renders constants with Scalus' own PrettyPrinter.prettyValue; builtin names use scalus.utils.Utils.lowerFirst. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JG5LJsPA524HKbSMvjStrv

The single Iso[Array[Byte], Array[Byte], TypedProgram, ClassNode] was two unrelated functions packaged as an optic: its S was flat bytes, its T class-file bytes, and nothing walked back. `cborScript andThen iso` made that visible: cborScript's reverseGet was the identity on class files, so the composed "iso" never returned to CBOR. Now every stage keeps the script it came from — TypedProgram its source, the class the script itself, in its constants — and each arrow is a lawful monomorphic iso: CborScript ≅ TypedProgram ≅ ClassNode ≅ JvmClassScript typed lowering writing UplcJvm.typed, bytecode (= typed andThen lowering) and classFile (= bytecode andThen writing) all start at the script; classFile.get is the compiler and classFile.reverseGet returns the script exactly. - CborScript and JvmClassScript are opaque types over Scalus' ByteString (value equality, so the laws compare with ==). CborScript.of(program) reuses Scalus' cborEncoded / fromCbor; CborScript.parse replaces UplcJvm.compile as the checked entry. - OpticSpec runs cats-eo-laws' IsoTests on typed and classFile; Oracle.compare checks the round trips on every program the suites run, all 943 conformance programs included. - writing (ClassWriter / ClassReader) is a byte-exact fixpoint because generated code no longer has unreachable instructions: a function that can only fail is never applied, and a call whose argument can only fail after nothing but values is that failure. (Dead code left constant-pool entries behind and reordered the pool on a rewrite.) - TypedProgram is equal by script and body; its constructor is private. - Thunk folds into Lam (arity 0); chooseList and trace run their CEK meanings; one duplicated CompilerSpec case removed. Code lines vs the raw-ASM version: main 1425 → 997 (−30.0%), total 2218 → 1549 (−30.2%). Translation unchanged within noise. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JG5LJsPA524HKbSMvjStrv

A doc comment ran one character past maxColumn; the local check had passed from scalafmt's incremental cache. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JG5LJsPA524HKbSMvjStrv

The middle of the chain is now `JvmClass`, an opaque type over ASM's ClassNode, so the optics read in the demo's own vocabulary: CborScript ≅ TypedProgram ≅ JvmClass ≅ JvmClassScript Only `Lower.lower` and `writing.reverseGet` make one, so every JvmClass embeds its script; `.tree` exposes ASM's tree for inspection. Docs, README, CLAUDE.md and the build comment follow. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JG5LJsPA524HKbSMvjStrv

…Bytecode Per review: - typed -> toTyped (CborScript ≅ TypedProgram) - writing -> compiling (JvmClass ≅ JvmClassScript) - bytecode -> toJvmClass (CborScript ≅ JvmClass) - classFile -> toBytecode (CborScript ≅ JvmClassScript) `compiling`'s two halves are now public as `UplcJvm.compile` and `UplcJvm.decompile`; the Oracle's class-file fixpoint check uses them directly. Docs, README, the walkthrough's labels and CLAUDE.md follow, and the last stage-letter labels ((A), (B), (T)) are gone from the comments and README. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JG5LJsPA524HKbSMvjStrv

Per review: - No unsafeFromArray (ByteString or IArray) or unsafeWrapArray left in the demo's main code. Scalus' unsafeFromArray skips a defensive copy; the scripts now pay that copy. - CborScript.of returns Either: Scalus' flat encoder throws on a program with a free variable, and the conformance suite used to wrap the call in a Try. Tests and benches unwrap it through one helper each. - Lower.direct no longer calls Option.get: an Option for-comprehension with cats' traverse replaces it. - chunks/fromChunks are now CborScript.asConstants/fromConstants, with the reason in the doc: a class carries its script, a class file has no byte-array constants, and one string constant holds at most 65535 bytes. - SourceProgram.decode is a right fold over a shared `subterms` instead of writes into arrays. Measured, the fold is 2-3x slower to decode and translation 2-9% slower; validation never decodes. children(n) is one unfold over subtree sizes. FreeVars walks `subterms` too but keeps its mutable set: immutable sets were 2.5x slower. - Docs: the compiler does no optimisation (no constant folding, inlining, dead-code elimination, CSE, specialisation or tail-call elimination); translation table re-measured. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JG5LJsPA524HKbSMvjStrv

- demoUplcJvm now compiles with `-opt -opt-inline:**,!scala.**`. It is unpublished, so it may inline from Scalus, ASM, cats and eo. Inlining from the Scala library is left out: measured with the demo's JMH benches, it made the short validations 12-27% slower (PrintInlining shows the inlined library code makes Rt.force too big for the JIT to inline). Everything else measures neutral to ~5% faster. The generated classes are byte-identical either way. - Published modules (scala3LibrarySettings) add `-opt-inline:dev.constructive.eo.**`: the inliner may copy code from eo's own packages only, never from cats, the JDK or the Scala library. This replaces core's `<sources>` scope and extends inlining to every library module. Full test suite green on JDK 25. - Docs: the demo's "no optimisation" note is about UPLC programs; the benchmark tables are re-measured with the new build. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JG5LJsPA524HKbSMvjStrv

…ealed - compiling and the composed toBytecode are prisms: a class can exceed the JVM's 64 KB method or 65 535-constant class limits, and then there is no class file. UplcJvm.compile returns the limit as a Left (the one place ASM's MethodTooLargeException / ClassTooLargeException become a value); toBytecode.getOption is the compiler. - CborScript, JvmClassScript and JvmClass hand out copies only (bytes, tree), so a value cannot change after the compiler made or checked it and reverseGet always walks back. - Compiling cannot overflow the JVM stack: decoding, typing and lowering are walks of Steps (Step.scala), plain calls while shallow and cats.Eval on the heap from 128 calls deep; every recursive call goes through Walk.nested. Trampolining every node instead doubled translation time; this costs 13-19% of it. A closure's captures are a fold over its subtree's ids, so FreeVars is gone. Programs nested 100 000 deep compile on a 1 MB stack. - OpticSpec: PrismTests for toBytecode (a too-large program among the scripts), the miss, the copies, and the 100 000-deep programs. - Docs: README, site page and CLAUDE.md describe two isos and a prism, sealing, stack safety and Scalus' still-recursive flat codec; the translation table is re-measured. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JG5LJsPA524HKbSMvjStrv

…ns miss Scalus decodes any declared Plutus Core version, and the compiler used to type every script as if it were 1.1.0: a script declaring 2.0.0 or 7.3.9 parsed, compiled and ran. The ledger only takes the versions Plutus V3 has at protocol version 11, 1.0.0 and 1.1.0 (Semantics.versions). - toTyped is a prism: getOption misses on a script of any other version; reverseGet stays the script the program was read from. toJvmClass and toBytecode compose to prisms; lowering stays an iso, typing the script a class embeds without the check (only a typed program is lowered). - SourceProgram keeps the decoded program's version. - OpticSpec runs PrismTests on toTyped, with a 2.0.0 script among the inputs, and checks the miss. Tests reach the stages through Oracle.typed, jvmClass and classFile. - Docs: the chain is CborScript ⇀ TypedProgram ≅ JvmClass ⇀ JvmClassScript. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JG5LJsPA524HKbSMvjStrv

CborScript ⇀ TypedProgram ≅ JvmClass ⇀ JvmClassScript: toTyped and compiling are prisms since the last two commits. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JG5LJsPA524HKbSMvjStrv

claude added 2 commits October 13, 2026 13:04

claude added 3 commits October 14, 2026 13:48

kryptt changed the title ~~Add UPLC → JVM bytecode demo built on one eo Iso~~ Add UPLC → JVM bytecode demo built from eo isos Oct 14, 2026

claude added 6 commits October 14, 2026 15:11

kryptt changed the title ~~Add UPLC → JVM bytecode demo built from eo isos~~ Add UPLC → JVM bytecode demo built from eo optics Oct 15, 2026

claude added 2 commits October 15, 2026 09:46

kryptt merged commit b9cf763 into main Oct 15, 2026
13 checks passed

kryptt mentioned this pull request Oct 25, 2026

release: v1.0.0 #151

Merged

Navigation Menu

Repository navigation

Add UPLC → JVM bytecode demo built from eo optics - #132

Add UPLC → JVM bytecode demo built from eo optics#132
kryptt merged 14 commits into
mainfrom
ccr-1ec1a317-uds4rn

kryptt commented Oct 13, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 13, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 13, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

kryptt commented Oct 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What

Correctness

Benchmarks (JMH, demos/uplc-jvm-bench, 3 forks)

Size

Build and CI

Limitations

Uh oh!

github-actions Bot commented Oct 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Oct 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Benchmark A/B

Allocation (B/op) — authoritative

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

kryptt commented Oct 13, 2026 •

edited

Loading

Benchmarks (JMH, `demos/uplc-jvm-bench`, 3 forks)

github-actions Bot commented Oct 13, 2026 •

edited

Loading

github-actions Bot commented Oct 13, 2026 •

edited

Loading