Malte Ubl
def6f86a0e
Nearly every long-lived object becomes a cycle candidate once, soon after it is built: the constructing frame's release leaves it held only by the structure it was stored into. A nursery pass then walks it (about 75 ns per object), it survives, and it is re-buffered for the mature level, whose next pass walks it again. In tsc-ts single-threaded, nursery passes walked 6.0M objects on Excalidraw and 36.5M on Sentry and freed 7%; 99% of what they walked was promoted. Most of that walking goes to types whose objects never die in a young cycle: SymbolLinks, NodeLinks and Symbol were 2.0M of Excalidraw's candidates and the collector freed none of them, while young garbage comes from a few types (inference contexts and mappers, closures, some arrays). Types with a collector yield under 0.5% are 42-55% of what nursery passes walk. So the collector now keeps per-thread statistics per type and level. Each ScrCycType gains a `kind` index: emitted descriptors number themselves from 32 (FIRST_EMITTED_CYC_KIND / SCR_CYC_KIND_EMITTED), the runtime's own families have fixed kinds below that, 0 keeps no statistics. For the nursery and mature levels, a pass counts the candidates of each kind it walks and the objects of each kind it frees. After every 8192 walked candidates a kind bypasses that level for the next window if fewer than 1 in 128 of them were freed. A release that buffers an object at a level its type bypasses promotes it past the level first, and a nursery pass that re-buffers a survivor into a bypassed mature level moves it to the old level. One candidate in 16 stays put, so the yield keeps being measured and a type whose objects start dying young again is walked again after one window. Only the generation an object is judged in changes. Any assignment is sound (a restricted pass counts edges from older objects as external), a bypassing candidate stays buffered, and full passes walk everything they did before, so the old level's bounds hold unchanged. Cyclic garbage made only of objects of a bypassing type waits for the next pass at an older level. Programs that never complete a window (8192 walked candidates of one type) are collected exactly as before. The descriptor layout is part of the emitted-code/runtime ABI, so the program-object marker moves to scr_runtime_abi_v12. Pass log (instrumented build, single-threaded, one run each, with the preceding null-slot change): all collector passes Excalidraw 606 -> 313 ms (nursery 435 -> 153), TypeORM 629 -> 383 ms, tsc-ts self-check 377 -> 215 ms, Sentry 7.61 -> 5.75 s (nursery 2.92 -> 1.04 s, mature 1.17 -> 1.02 s; the full pass is unchanged at 3.6 s). tsc-ts (Linux sandbox, interleaved, medians; 10 rounds, Sentry 5 rounds pinned to 8 CPUs; both changes against int/best-v6): single-threaded Excalidraw -8.0%, TypeORM -4.2%, self-check -4.3%, Sentry -5.3%; --checkers 8 -12.2%, -0.8%, -2.3%, -1.6%. Peak RSS -0.8% to +0.6%. Diagnostics identical. Tests: test_cycle gains check_nursery_bypass, check_mature_bypass and check_rebuffer_past_mature (a test-only entry point, scr_cyc_pass_at under SCR_CYC_TESTING, runs passes at exact levels); each fails under the corresponding mutation (no bypass, a sticky decision, no sampling, no freed or walked counts, a bypassing candidate left unbuffered, survivors re-buffered into a bypassed level). Corpus 4690 builds a long-lived declaration graph large enough for both levels to bypass and lets objects of the same classes die in cycles; under --sanitize the exit RC audit proves they are reclaimed.
scriptc
Compile ordinary TypeScript and JavaScript to small, fast native executables or WASI WebAssembly modules — no Node, no V8, no JavaScript engine in the artifact.
$ cat fib.ts
function fib(n: number): number {
return n < 2 ? n : fib(n - 1) + fib(n - 2);
}
console.log(fib(30));
$ scriptc run fib.ts
832040
$ scriptc build fib.ts -o fib && ./fib
832040
Install
$ npm install -g scriptc
The installed compiler runs natively on supported macOS, Linux, and Windows hosts. Compilation, compile-time evaluation, and native execution do not require Node. --emit=ir|llvm|asm|obj uses the bundled TypeScript checker and LLVM helper without an external compiler, archiver, linker, or SDK. Executable builds additionally need a platform linker driver and SDK/sysroot; precompiled runtime packs supply the C runtime. Set SCRIPTC_LINKER to choose that driver. Runtime development with --sanitize additionally needs a C compiler. Node.js 24 or newer is needed for npm installation and scriptc run of WASI modules.
The native compiler includes its host runtime pack. Cross-compilation uses additional @scriptc/runtime-<target> packages installed in your project at the same version as scriptc --version. WASI builds use @scriptc/runtime-wasm32-wasi, for example. Run the compiler from that project or set SCRIPTC_RUNTIME_PACK to the pack directory. Linux GNU distributions require glibc 2.34 or newer.
Builds use a bounded persistent cache. Unchanged source can reuse the validated frontend result and LLVM program objects. Library identity getters occupy a separate LLVM module, so an identity change can reuse the large program object. Runtime objects come from the installed pack. Executable cache entries verify their native dependencies; FFI builds relink against current external inputs. Set SCRIPTC_NO_CACHE=1 to bypass the cache or SCRIPTC_CACHE_DIR to select its location. An existing POSIX override must already be private.
Commands
scriptc build <file.ts>— compile to a native executable or selected target artifactscriptc run <file.ts>— compile and runscriptc coverage <file.ts>— what compiles statically, and why the rest doesn't
scriptc build app.ts --emit=ir|llvm|asm|obj selects serialized typed IR, textual LLVM IR, target assembly, or a relocatable program object as the one primary artifact. --emit=exe is the default. Assembly/object emission uses the matching helper on supported macOS, Linux, and Windows hosts (and produces WASI artifacts when that target is selected). Objects retain undefined scr_* runtime references plus the scr_runtime_abi_v12 compatibility marker; they are not library archives. External consumption is experimental and requires the exact matching runtime. scriptc build app.ts --print=native-link-info -o app.o prints the versioned JSON target/runtime/link recipe without performing a link. --emit=asm|obj --sanitize is rejected until ASan pipeline parity is available.
For embedder-hosted modules that are not installed npm packages, coverage can map an exact bare specifier to a local declaration with repeatable --external-types <specifier=file.d.ts> options. This is analysis-only: the types unblock application measurement, while runtime module uses remain reported as blockers.
No annotations, no dialect, no special stdlib: the same TypeScript you run on Node, type-checked by the real TypeScript compiler. Programs outside the static tier can opt into --dynamic, which embeds a small JavaScript engine (~620KB) for the parts that can't be static; everything else fails the build with a specific error code and usually a rewrite hint.
WebAssembly is available as a production LLVM target: SCRIPTC_TARGET=wasm32-wasi scriptc build app.ts. It emits a WASI Preview 1 .wasm module, and scriptc run supplies a WASI host. The complete executable language tier—including async, generators, timers, and --dynamic—is supported. APIs needing capabilities WASI P1 does not expose (network sockets/fetch, child processes, OS signals, and filesystem watching), sanitizer builds, and native FFI are rejected with SC3002. With --lib --profile, the target produces a callable Wasm reactor with named host imports and per-instance memory. See WebAssembly Modules for the embedding contract.
Native code can be called through an explicit, link-time C ABI manifest: declare the function signature in TypeScript, bind it to a C symbol, and build with --ffi <manifest.json>. See the Native FFI guide.
Docs: scriptc.dev