Hojun-Cho/ww
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

w6c+wwstage: full-size aggregate copy for deref-rhs let-init (#265 fold-1)

Browse Source 
A `let c: T = *p` (T a struct or array, >8B) copied no full aggregate:
cstage dropped the init entirely (c read garbage); wwstage emitted only
the scalar `MOVQ AX,off(BP)` tail (first 8 bytes). Both wrong, differently
— converge BOTH stages on a size-driven slot-to-slot memcpy: cgexpr the
deref operand to the source address in AX, MOVQ AX,SI, then a MOVQ run
plus a sized MOVL/MOVW/MOVB tail over the #254 non-slot-padded ABI extent
(lu->size / structabisize for a struct, tinfo.size for an array). Mirror
arms in cgen.c N_LET and cgenstmt.ww cglet, byte-identical (rule-10).

Unblocks sha256's faithful `let copy = *h`. The by-value aggregate RETURN
ABI (array/struct return truncates to AX) is fold-2 (#267, deferred).

949 gains 6 full-readback rows (every member written distinct + summed,
so a truncated copy fails): struct{[4]u32} 16B, struct{[8]u32} 32B via
both *(&s) and *p (sha256 shape), bare [4]u32, and non-8-mult tails
([3]u32 12B → MOVL, [11]u8 11B → MOVW+MOVB). w6c+wwdump combined.ww regen
(#110). 61/61 949, test-unit 240, sizelint, smoke green.
This commit is contained in:
Hojun-Cho
2026-06-02 09:31:44 +09:00
parent 0afe4225cd
commit 4d3f8467a8
5 changed files with 331 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

72
selfhost/cmd/wwdump/main.combined.ww
View File

@@ -28858,6 +28858,78 @@ fn cglet(c: *cgen, n: *node) void = {
};
};
};
// #265 fold-1: aggregate deref-rhs let-init `let c: T = *p`
// (T a struct or array, >8B). cgexpr(rhs.lhs) leaves the
// SOURCE ADDRESS in AX (a `*p` ident loads the pointer value;
// `*(&s)` LEAQs the slot); memcpy N bytes slot→slot via SI — a
// MOVQ run plus a sized MOVL/MOVW/MOVB tail. N is the #254
// non-slot-padded ABI extent: structabisize for a struct (=
// cstage lu->size), tinfo.size for an array. Pre-fix wwstage
// copied only the first 8B (scalar tail below) and cstage
// dropped the copy entirely — both wrong; converge on the full
// copy (rule-10). Mirror of cstage cgen.c N_LET deref arm; the
// by-value RETURN ABI is fold-2 (#267).
if (rhs.kind == nkind.N_UN) { if (rhs.op == tkind.TK_STAR) {
let ncopy: i32 = 0;
let dsi: *structinfo = structlookupchain(c, tn);
if (dsi != nil) {
ncopy = structabisize(dsi);
} else {
let dti: *tinfo = nil;
if (tn != nil) { dti = tn.type_: *tinfo; };
for (dti != nil && dti.kind == tykind.TY_NAMED) {
dti = dti.under;
};
if (dti != nil) {
if (dti.kind == tykind.TY_ARRAY) {
ncopy = dti.size: i32;
};
};
};
if (ncopy > 8) {
cgexpr(c, rhs.lhs);
emitline("\tMOVQ\tAX, SI\n");
let k: i32 = 0;
for (k + 8 <= ncopy) {
emitline("\tMOVQ\t");
emitoff(k: i64);
emitline("(SI), AX\n");
emitline("\tMOVQ\tAX, ");
emitoff((off + k): i64);
emitline("(BP)\n");
k += 8;
};
if (k + 4 <= ncopy) {
emitline("\tMOVL\t");
emitoff(k: i64);
emitline("(SI), AX\n");
emitline("\tMOVL\tAX, ");
emitoff((off + k): i64);
emitline("(BP)\n");
k += 4;
};
if (k + 2 <= ncopy) {
emitline("\tMOVW\t");
emitoff(k: i64);
emitline("(SI), AX\n");
emitline("\tMOVW\tAX, ");
emitoff((off + k): i64);
emitline("(BP)\n");
k += 2;
};
if (k + 1 <= ncopy) {
emitline("\tMOVB\t");
emitoff(k: i64);
emitline("(SI), AX\n");
emitline("\tMOVB\tAX, ");
emitoff((off + k): i64);
emitline("(BP)\n");
k += 1;
};
c.lastwasreturn = 0;
return;
};
}; };
cgexpr(c, rhs);
// Float local: cgexpr leaves the value in X0. Spill via
// MOVSS (f32, 4B) or MOVSD (f64, 8B).