cd5b958d19
Introduce the welded comptime `compiler` library (`#library "compiler"` +
`abi(.zig) extern compiler`), per design/comptime-compiler-api.md, and unify
`callconv(...)` into the new `abi(...)` annotation.
abi(...) replaces callconv(...):
- New ABI enum { default, c, zig, pure }; `abi(.c|.zig|.pure)` parses in the
postfix slot before extern/export (and standalone). `kw_callconv` -> `kw_abi`.
- Migrated 52 sx files, the call-convention-mismatch diagnostic, and docs
(readme/specs) from `callconv(.c)` to `abi(.c)`.
Phase 1 — welded compiler library (parse -> registry -> validation -> bridge):
- `abi(.zig) extern compiler` parses on fn decls (carries abi/extern_lib) and
struct decls (StructDecl.abi/extern_lib).
- `#library "compiler"` is the comptime-only internal surface — never dlopen'd.
- src/ir/compiler_lib.zig: the binding registry (the safety boundary). `Field`
welded to StructInfo.Field with layout baked from the real Zig type
(@offsetOf/@sizeOf); `findType`/`findFn`. Welded structs are layout-validated
at registration (field set + total size) as a header checked against the impl.
- Host-call bridge: a `fn abi(.zig) extern compiler` dispatches under the
comptime interp to its registered Zig handler (intern/text_of round-trip),
never dlsym. IR Function.compiler_welded; validated in declareFunction.
- Comptime-only enforcement: a runtime call to a welded fn is a clean
build-gating error (emitCall), not an undefined-symbol link failure.
Phase 2.1 — byte-layout weld foundation:
- Decision: full byte-layout weld (sx struct laid out byte-identically to the
bound Zig type). Registered StructInfo (first non-natural / Zig-reordered
layout). `computeWeldPlan` — pure offset-ordered element plan + padding +
sx-field->LLVM-element remap; unit-tested. Emit/interp wiring is the next
sub-step (2.2+, see current/CHECKPOINT-COMPILER-API.md).
Examples: 0625/0626 (welded struct + fn round-trip), 1183/1184/1185
(layout-mismatch, unexported-fn, runtime-call diagnostics).
36 lines
1.4 KiB
Plaintext
36 lines
1.4 KiB
Plaintext
// ASM stream — the round trip: sx → asm → sx. A global-asm routine (`_caller`)
|
|
// CALLS BACK into an sx function (`cb`) by its symbol, then returns. For the asm
|
|
// `bl _cb` to resolve, the sx callback needs EXTERNAL LINKAGE and a stable C
|
|
// symbol — that is exactly what `export` provides (it also implies the C ABI, so
|
|
// no hidden context parameter). `abi(.c)` alone is NOT enough: it sets the
|
|
// ABI but leaves the function `internal`, so it is dead-code-eliminated (nothing
|
|
// in the IR references it — the `bl` is opaque to the optimizer) and `_cb` is
|
|
// undefined at link. macOS gives `export "cb"` the symbol `_cb` (leading
|
|
// underscore), which the template references. aarch64-macos-pinned; runs under
|
|
// the JIT here (sx run), ir-only elsewhere.
|
|
|
|
// The sx callback — `export` gives it external linkage + the `_cb` symbol + C ABI.
|
|
cb :: (n: i64) -> i64 export "cb" {
|
|
return n + 1;
|
|
}
|
|
|
|
// A global-asm trampoline that calls back into `cb`. It saves/restores the link
|
|
// register (x30) around the `bl` — it was itself reached via `bl`, so the return
|
|
// address must survive the nested call.
|
|
asm {
|
|
#string ASM
|
|
.global _caller
|
|
_caller:
|
|
stp x29, x30, [sp, #-16]!
|
|
bl _cb // x0 = cb(x0) — back into sx
|
|
ldp x29, x30, [sp], #16
|
|
ret
|
|
ASM,
|
|
};
|
|
|
|
caller :: (n: i64) -> i64 extern;
|
|
|
|
main :: () -> i64 {
|
|
return caller(41); // sx main → asm caller → bl _cb → sx cb → 42
|
|
}
|