Add fast API trampolines for AArch64 and x86_64.
IsJitMemorySupported() maps an RW page, writes a ret instruction
(0xD65F03C0 AArch64 / 0xC3 x86_64), and mprotects it to RX. A
successful RX transition is treated as the support signal. The page is
deliberately not executed: this check may run during normal operation
(when an FFI function is first created), and executing freshly written
code from a capability probe could SIGSEGV/SIGKILL the process on
systems that block executable memory. The real trampoline emitter
performs the same mprotect at creation time and falls back to libffi
when it is rejected.
The result is computed once via std::call_once and cached for the
lifetime of the process, so concurrent callers never observe a
provisional value.
Wired into CreateFastFFIMetadata() for early nullptr bail-out when
JIT memory is unavailable.
Windows stub returns false (no trampolines yet on this branch).
IsFastCallEligible() validates at parse time whether a signature can
use the fast-call path, covering:
- Return and argument type eligibility (numeric, pointer; no structs)
- Argument count cap (8, matching V8 fast-call limit)
- Per-ABI register pressure limits that mirror the trampoline emitters
(AArch64 and x86_64 SysV). Platforms without an emitter, including
Win64, are reported ineligible.
- Buffer and float args cannot coexist, and buffer args additionally
consume an extra GP register slot on both supported ABIs.
The arg/arg-name lengths are checked before the per-arg loop so a
malformed signature cannot index out of bounds.
Returns nullptr from CreateFastFFIMetadata() for ineligible signatures,
falling back to libffi.
Co-authored-by: Bryan English <bryan@bryanenglish.com>
Signed-off-by: Paolo Insogna <paolo@cowtech.it>
Signed-off-by: Bryan English <bryan@bryanenglish.com>
Assisted-By: OpenAI:GPT-5.5 <openai/gpt-5.5>
PR-URL: https://github.com/nodejs/node/pull/63068
Reviewed-By: Matteo Collina <matteo.collina@gmail.com>
Reviewed-By: Bryan English <bryan@bryanenglish.com>
Adds an ArrayBuffer-based invocation path for FFI functions whose
signatures are composed entirely of numeric types (i8..i64, u8..u64,
f32, f64, bool, char) and/or pointer types. The JS wrapper packs
arguments directly into a per-function AB via primordial DataView
setters and the C++ invoker (`InvokeFunctionSB`) reads them without
going through V8's `FunctionCallbackInfo`. Results are returned the same
way.
Pointer arguments use runtime dispatch: BigInt, null, and undefined take
the fast path, while Buffer, ArrayBuffer, ArrayBufferView, and String
fall back transparently to the classic `InvokeFunction` path via a
stashed `_invokeSlow` function. Signatures containing
non-numeric/non-pointer types also bypass the fast path.
The fast path is disabled on big-endian platforms.
Callers do not opt in, and the fast path is transparent in every way
users should rely on. One observable change: function wrappers returned
by `library.getFunction`, `library.getFunctions`, and `library.functions`
now have `.length` equal to the declared parameter count rather than
`0`. Code that relied on the previous value will need to be updated.
Adds microbenchmarks covering the common FFI call shapes so future
changes to the invoker can be evaluated:
- add-i32.js: 2-arg integer
- add-f64.js: 2-arg float
- many-args.js: 6-arg integer
- pointer-bigint.js: 1-arg pointer (BigInt)
- sum-buffer.js: pointer + length (Buffer)
A `common.js` helper resolves the fixture-library path from
`test/ffi/fixture_library` without pulling in the test harness, and
throws a clear message if the fixture hasn't been built yet.
Also adds `sum_6_i32` to the fixture library for the many-args case.
Signed-off-by: Bryan English <bryan@bryanenglish.com>
PR-URL: https://github.com/nodejs/node/pull/62918
Reviewed-By: Anna Henningsen <anna@addaleax.net>
Reviewed-By: Stephen Belanger <admin@stephenbelanger.com>
Reviewed-By: Paolo Insogna <paolo@cowtech.it>
Yilong Li