PSDP — same-language parallelization (pick a specific service)
Phase-Synchronized Deterministic Parallelism. An independent service that parallelizes sequential batches while keeping output bit-exact. Use it on modern code — including code that came out of a SlimeNENC language migration. The right-most column of the table takes you to trial use / enquiry.
Supported languages
Same-language → same-language (e.g. Java 8 → Java 8, C# → C#). Cross-version (e.g. Java 8 → 17) combines conversion and parallelization in one pass.
| Language | Version / variant | Parallel construct | Validation | Specific service / enquiry |
|---|---|---|---|---|
| Java 8 → 8 | Same-version maintenance (bytecode 52 → 52) | ForkJoin / ExecutorService | Byte-exact (501/501) | Trial / Enquiry |
| Java 17 → 17 | Same-version maintenance (bytecode 61 → 61) | ForkJoin / sealed-class aware | Byte-exact | Trial / Enquiry |
| Java 21 → 21 | Same-version maintenance (bytecode 65 → 65) | Virtual Threads / structured concurrency | Byte-exact | Trial / Enquiry |
| Java 8 → 17 | Cross-version migration + parallelization (52 → 61) | ForkJoin → sealed | Byte-exact | Trial / Enquiry |
| Java 17 → 21 | Cross-version migration + parallelization (61 → 65, mixed-bytecode forward compat in one process) | sealed → Virtual Threads | Byte-exact | Trial / Enquiry |
| C# | .NET 8 (top-level) | Parallel.For / Task | Byte-exact (double.ToString("F17")) | Trial / Enquiry |
| Rust | stable 2021 | rayon / std::thread | 5.9x parallel speedup verified | Trial / Enquiry |
| Go | 1.21+ | goroutine + channel | Byte-exact | Trial / Enquiry |
| Python | 3.10+ | multiprocessing / concurrent.futures | Byte-exact (subject to double-precision language constraints) | Trial / Enquiry |
| Node.js | Worker Threads | Worker / SharedArrayBuffer | Byte-exact | Trial / Enquiry |
| PHP | 8.x + GMP | parallel ext / pcntl | Byte-exact (GMP required) | Trial / Enquiry |
| Swift | 5.x | TaskGroup / async-let | Byte-exact | Trial / Enquiry |
| Kotlin | JVM (1.9+) / coroutines | Coroutines / structured concurrency | Byte-exact (coroutines recommended on Android) | Trial / Enquiry |
| FORTRAN | F2018 free-form | OpenMP / coarray | Byte-exact (F2018 25/25) | Trial / Enquiry |
| C | C11+ | pthreads + barrier | Byte-exact (10/10) | Trial / Enquiry |
| C++ | C++17+ | std::thread / std::async | Byte-exact (10/10) | Trial / Enquiry |
| Scala | Future + Await barrier | Future / Akka | Byte-exact (co-located on JVM) | Trial / Enquiry |
| Clojure | future + deref | future + atom | Byte-exact (primitive-hint pitfalls avoided) | Trial / Enquiry |
| Common Lisp | SBCL | bordeaux-threads | Byte-exact | Trial / Enquiry |
| Erlang | OTP 26+ | actor model (BEAM) | Byte-exact (first non-JVM paradigm) | Trial / Enquiry |
Legend for "Evidence"
Byte-exact means the sha256 of sequential and parallel outputs is identical down to the last byte (bit-exact). Parenthetical notes record the language-side prerequisites or the number of samples checked.
- (501/501), (F2018 25/25), (10/10) — sample counts verified by the test harness
- (double.ToString("F17")) — output convention that serializes .NET
doubleas a 17-digit fixed string to preserve bit-exactness in C# - (subject to double-precision language constraints) — languages like Python whose transcendental functions are not strictly IEEE-754; the runtime contract absorbs the affected ranges
- (GMP required) — PHP arbitrary-precision arithmetic depends on the GMP extension for bit-exactness
- (co-located on JVM) — Scala shares Java's
doublerepresentation on the JVM - (first non-JVM paradigm) — first bit-exact verification outside the JVM (BEAM / Erlang actor model)
- (primitive-hint pitfalls avoided) — Clojure boxing traps are avoided with explicit primitive type hints
- (coroutines recommended on Android) — Kotlin syntax recommendation for Android targets
- 5.9x parallel speedup verified — on top of bit-exactness, measured parallel speedup is 5.9x (Rust)
How to use the service
- Pick a specific service from the table above (e.g. Java parallelization).
- Sign up for an account (admin approval; select the languages you want).
- Run a trial (no feature restrictions).
- Contact us for production use. Preferred terms are available for existing SlimeNENC customers.
Pricing (Provisional, 2026-05-18 — example at 10,000 LOC)
| Path | LOC | Buy-in / LOC | Buy-in total | Runtime rate | Runtime / year | Support |
|---|---|---|---|---|---|---|
| SlimeNENC (converter) | 10,000 | US$ 9.95 | US$ 99,500 | — | — | Full |
| PSDP for SlimeNENC customer | 10,000 | US$ 0.00 | US$ 0 | 2%/year | US$ 1,990 | Full |
| PSDP direct (via SIer / partner) | 10,000 | US$ 1.00 | US$ 10,000 | 20%/year | US$ 2,000 | Full (via partner) |
| PSDP end-user direct (30% OFF) | 10,000 | US$ 0.70 | US$ 7,000 | 20%/year | US$ 1,400 | FAQ only |
Per-LOC annual runtime cost converges to roughly US$ 0.2 / LOC / year across the first three paths. The end-user direct path is 30% off both buy-in and runtime, but support is limited to the FAQ; we do not handle issues caused by the post-deployment runtime environment (OS, JVM, libraries, container, build pipeline). For hands-on integration please go through a SIer / system-integration partner instead — see /partners/.
What makes PSDP different
- Bit-exact results — output is identical before and after parallelization; meets financial audit / FDA / FINRA expectations.
- Phase synchronization — workers advance through phases deterministically, structurally removing race conditions.
- Language-independent core — 16+ languages share a single Slot IR; only the emitter targets each language's parallel construct.
- Cross-version Java — convert a Java 8 batch and parallelize it under Java 17 / 21 + Virtual Threads in one pass.
Related
- Legacy conversion (cross-language) — when combining with a language migration
- PSDP technical details — TCSVT paper / Algorithm 1