wcc: float arr[i]= spills X0 across index eval (#125)

Fix value-loss bug introduced as a #122 boundary in the float
arr[i]=v store: when the index sub-expr clobbers X0 (e.g.
`a[geti()]=1.5f32`), the value is lost. Mirror the scalar-deref
X0-spill template (cstage cgen.c:4187; line shifted from the brief's
stale :3859 cite by intervening #133/#135/#138 commits): for float
element only, replace PUSHQ AX (junk for floats — value is in X0)
with SUBQ $8,SP + MOVSS/MOVSD X0,(SP) before the idx/base eval;
mirror replace POPQ AX with MOVSS/MOVSD (SP),X0 + ADDQ $8,SP after.
Wwstage parallel. Non-float keeps PUSHQ/POPQ AX so the str/slice
3-word {ptr,len,cap} pop order at the end of the branch is preserved.
#122 trailing-store comment updated from "Deferred to #125" to a
positive cite.

Test 916: 5 rows — f64_call_index + f32_call_index canonical repros
(geti's body clobbers X0; pre-fix exit=2 from post-call residue,
post-fix exit=1 from the spilled 1.5) + f64_lit_index / _localvar /
_arith control rows for non-X0-clobbering index paths. f32_call_index
uses an int-arg call to dodge the sibling cs/ww f32-arg-push
MOVSD-vs-MOVSS divergence (#143, task #36 — orthogonal, filed).

Bootstrap NEUTRAL (zero current float arr[i]= callers in lib; only
[N]u8 byte-buffers like f64tos_buf). cs==ww byte-identical both
stages (990-997 + 916 inline cmp). Closes the #122 boundary-doc
loose end; completes the #122 family.
This commit is contained in:
2026-05-27 03:30:15 +09:00
parent 1de1b9a8a9
commit 61e6aab384
6 changed files with 362 additions and 29 deletions

View File

@@ -3951,7 +3951,20 @@ cgexpr(Cg *c, Node *n, Local *locals)
ins1(c, A_PUSHQ, areg(D_CX)); /* cap */
ins1(c, A_PUSHQ, areg(D_BX)); /* len */
}
ins1(c, A_PUSHQ, areg(D_AX));
/* Float element: spill X0 (not AX — AX is junk
* for floats) across the idx/base eval. A call-
* index (`a[geti()]=v`) clobbers X0 and would
* otherwise lose the value. Mirrors the *p=v
* float deref store at cgen.c:4187 (#125). */
int sp_isfloat = type_isfloat(esub);
int sp_mov = sp_isfloat
? (type_isf32(esub) ? A_MOVSS : A_MOVSD) : 0;
if (sp_isfloat) {
ins2(c, A_SUBQ, aimm(8), areg(D_SP));
ins2(c, sp_mov, areg(D_X0), amem(D_SP, 0));
} else {
ins1(c, A_PUSHQ, areg(D_AX));
}
cgexpr(c, n->lhs->rhs, locals); /* idx → AX */
if (esz > 1) {
ins2(c, A_MOVQ, aimm(esz), areg(D_CX));
@@ -3993,7 +4006,15 @@ cgexpr(Cg *c, Node *n, Local *locals)
}
ins1(c, A_POPQ, areg(D_AX)); /* scaled idx */
ins2(c, A_ADDQ, areg(D_AX), areg(D_BX));
ins1(c, A_POPQ, areg(D_AX)); /* value (ptr if str) */
/* Reload value: float reloads X0 from the spill
* slot; non-float pops AX. Twin of the value-spill
* site above (#125). */
if (sp_isfloat) {
ins2(c, sp_mov, amem(D_SP, 0), areg(D_X0));
ins2(c, A_ADDQ, aimm(8), areg(D_SP));
} else {
ins1(c, A_POPQ, areg(D_AX)); /* value (ptr if str) */
}
if (elem_is_str || elem_is_slice) {
/* str/slice: store ptr/len/cap (#1/Phase 3, #7). */
ins2(c, A_MOVQ, areg(D_AX), amem(D_BX, 0));
@@ -4008,12 +4029,10 @@ cgexpr(Cg *c, Node *n, Local *locals)
* CVTSD2SS narrowing only touches X0 — the AX path
* below would store the raw double low-bits (garbage
* for f32). Float-ness from esub, mirroring the read
* side at cgen.c:6423 (#122). X0 survives the index/
* base eval only for literal/local-var indices; a
* call-index (`a[geti()]=v`) clobbers X0 and loses the
* value — this path does not spill X0 across the eval,
* unlike the *p=v float deref store above. Deferred to
* #125. */
* side at cgen.c:6423 (#122). #125: the value-spill
* pair above keeps X0 live across the idx/base eval
* so this MOVSS/MOVSD is correct even on call-index
* shapes. */
if (type_isfloat(esub)) {
int mov = type_isf32(esub) ? A_MOVSS : A_MOVSD;
ins2(c, mov, areg(D_X0), amem(D_BX, 0));