lab/sx
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

ffi 0.4: focused FP-aggregate (HFA) baseline — FQuad + DQuad

Browse Source 
91/91 regression tests pass (+ffi-04-fp-struct).

Single-file regression net for the all-float / all-double aggregate
ABI path:

  FQuad — 16 B, 4×f32   (same slot as ffi-02's Vec4f)
  DQuad — 32 B, 4×f64   (UIEdgeInsets-shape — the f32-vs-f64 landmine)

Already nominally covered by ffi-02's Vec4f, but pinning it as a
focused single-file test means a future ABI rule change that breaks
the HFA path fails *this* test directly without a noisy drag-in from
the multi-shape baseline.

DQuad at 32 B straddles the AAPCS64 HFA limit (≤4 floats of same
type, total ≤64 B); it stays as a struct value passed through
v0..v3 rather than going indirect. The snapshot confirms the values
arrive intact.
This commit is contained in:
agra
2026-05-19 11:44:43 +03:00
parent 2463eea1d4
commit 736382d39c
5 changed files with 105 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

49
examples/ffi-04-fp-struct.sx Normal file
View File

@@ -0,0 +1,49 @@
// Phase 0 baseline (PLAN-FFI.md step 0.4): focused FP-aggregate (HFA)
// FFI test. All-float / all-double aggregates of ≤4 fields stay as
// struct values in LLVM and pass through the float register file
// (AAPCS64 v0..v3, SysV AMD64 xmm0..xmm7). Distinct from the int
// register-coercion paths (i64 / [2 x i64]).
//
// FQuad — 16 B, four f32 (same slot as ffi-02's Vec4f)
// DQuad — 32 B, four f64 (UIEdgeInsets-shape HFA — the
// f32-vs-f64 landmine from this session)
//
// Already nominally covered by ffi-02's Vec4f, but pinning it as a
// focused single-file test means a future ABI rule change that
// breaks the FP path fails *this* test directly without a noisy
// drag-in from the multi-shape baseline.
#import "modules/std.sx";
#import c {
#source "vendors/ffi_fp_struct/ffi_fp_struct.c";
};
FQuad :: struct { a: f32; b: f32; c: f32; d: f32; }
DQuad :: struct { a: f64; b: f64; c: f64; d: f64; }
ffi_fquad_make :: (a: f32, b: f32, c: f32, d: f32) -> FQuad #foreign;
ffi_fquad_reverse :: (v: FQuad) -> FQuad #foreign;
ffi_fquad_sum :: (v: FQuad) -> f32 #foreign;
ffi_dquad_make :: (a: f64, b: f64, c: f64, d: f64) -> DQuad #foreign;
ffi_dquad_reverse :: (v: DQuad) -> DQuad #foreign;
ffi_dquad_sum :: (v: DQuad) -> f64 #foreign;
main :: () -> s32 {
// ── FQuad (16 B, 4×f32 HFA) ────────────────────────────────────
f := ffi_fquad_make(1.0, 2.0, 3.0, 4.0);
print("fquad make = ({}, {}, {}, {})\n", f.a, f.b, f.c, f.d);
g := ffi_fquad_reverse(f);
print("fquad rev = ({}, {}, {}, {})\n", g.a, g.b, g.c, g.d);
print("fquad sum = {}\n", ffi_fquad_sum(f));
// ── DQuad (32 B, 4×f64 HFA — UIEdgeInsets-shape) ──────────────
d := ffi_dquad_make(1.5, 2.5, 3.5, 4.5);
print("dquad make = ({}, {}, {}, {})\n", d.a, d.b, d.c, d.d);
e := ffi_dquad_reverse(d);
print("dquad rev = ({}, {}, {}, {})\n", e.a, e.b, e.c, e.d);
print("dquad sum = {}\n", ffi_dquad_sum(d));
0;
}