/* Emit or eliminate collected PHIs. */
static void loop_emit_phi(jit_State *J, IRRef1 *subst, IRRef1 *phi, IRRef nphi)
{
int pass2 = 0;
IRRef i, nslots;
IRRef invar = J->chain[IR_LOOP];
/* Pass #1: mark redundant and potentially redundant PHIs. */
for (i = 0; i < nphi; i++) {
IRRef lref = phi[i];
IRRef rref = subst[lref];
if (lref == rref || rref == REF_DROP) { /* Invariants are redundant. */
irt_setmark(IR(lref)->t);
} else if (!(IR(rref)->op1 == lref || IR(rref)->op2 == lref)) {
/* Quick check for simple recurrences failed, need pass2. */
irt_setmark(IR(lref)->t);
pass2 = 1;
}
}
/* Pass #2: traverse variant part and clear marks of non-redundant PHIs. */
if (pass2) {
for (i = J->cur.nins-1; i > invar; i--) {
IRIns *ir = IR(i);
if (!irref_isk(ir->op1)) irt_clearmark(IR(ir->op1)->t);
if (!irref_isk(ir->op2)) irt_clearmark(IR(ir->op2)->t);
}
}
/* Pass #3: add PHIs for variant slots without a corresponding SLOAD. */
nslots = J->baseslot+J->maxslot;
for (i = 1; i < nslots; i++) {
IRRef ref = tref_ref(J->slot[i]);
while (!irref_isk(ref) && ref != subst[ref]) {
IRIns *ir = IR(ref);
irt_clearmark(ir->t); /* Unmark potential uses, too. */
if (irt_isphi(ir->t) || irt_ispri(ir->t))
break;
irt_setphi(ir->t);
if (nphi >= LJ_MAX_PHI)
lj_trace_err(J, LJ_TRERR_PHIOV);
phi[nphi++] = (IRRef1)ref;
ref = subst[ref];
if (ref > invar)
break;
}
}
/* Pass #4: emit PHI instructions or eliminate PHIs. */
for (i = 0; i < nphi; i++) {
IRRef lref = phi[i];
IRIns *ir = IR(lref);
if (!irt_ismarked(ir->t)) { /* Emit PHI if not marked. */
IRRef rref = subst[lref];
if (rref > invar)
irt_setphi(IR(rref)->t);
emitir_raw(IRT(IR_PHI, irt_type(ir->t)), lref, rref);
} else { /* Otherwise eliminate PHI. */
irt_clearmark(ir->t);
irt_clearphi(ir->t);
}
}
}
/* Unroll loop. */
static void loop_unroll(jit_State *J)
{
IRRef1 phi[LJ_MAX_PHI];
uint32_t nphi = 0;
IRRef1 *subst;
SnapNo onsnap;
SnapShot *osnap, *loopsnap;
SnapEntry *loopmap, *psentinel;
IRRef ins, invar;
/* Use temp buffer for substitution table.
** Only non-constant refs in [REF_BIAS,invar) are valid indexes.
** Caveat: don't call into the VM or run the GC or the buffer may be gone.
*/
invar = J->cur.nins;
subst = (IRRef1 *)lj_str_needbuf(J->L, &G(J->L)->tmpbuf,
(invar-REF_BIAS)*sizeof(IRRef1)) - REF_BIAS;
subst[REF_BASE] = REF_BASE;
/* LOOP separates the pre-roll from the loop body. */
emitir_raw(IRTG(IR_LOOP, IRT_NIL), 0, 0);
/* Grow snapshot buffer and map for copy-substituted snapshots.
** Need up to twice the number of snapshots minus #0 and loop snapshot.
** Need up to twice the number of entries plus fallback substitutions
** from the loop snapshot entries for each new snapshot.
** Caveat: both calls may reallocate J->cur.snap and J->cur.snapmap!
*/
onsnap = J->cur.nsnap;
lj_snap_grow_buf(J, 2*onsnap-2);
lj_snap_grow_map(J, J->cur.nsnapmap*2+(onsnap-2)*J->cur.snap[onsnap-1].nent);
/* The loop snapshot is used for fallback substitutions. */
loopsnap = &J->cur.snap[onsnap-1];
loopmap = &J->cur.snapmap[loopsnap->mapofs];
/* The PC of snapshot #0 and the loop snapshot must match. */
psentinel = &loopmap[loopsnap->nent];
lua_assert(*psentinel == J->cur.snapmap[J->cur.snap[0].nent]);
*psentinel = SNAP(255, 0, 0); /* Replace PC with temporary sentinel. */
/* Start substitution with snapshot #1 (#0 is empty for root traces). */
osnap = &J->cur.snap[1];
/* Copy and substitute all recorded instructions and snapshots. */
for (ins = REF_FIRST; ins < invar; ins++) {
IRIns *ir;
IRRef op1, op2;
if (ins >= osnap->ref) /* Instruction belongs to next snapshot? */
loop_subst_snap(J, osnap++, loopmap, subst); /* Copy-substitute it. */
/* Substitute instruction operands. */
ir = IR(ins);
op1 = ir->op1;
if (!irref_isk(op1)) op1 = subst[op1];
op2 = ir->op2;
if (!irref_isk(op2)) op2 = subst[op2];
if (irm_kind(lj_ir_mode[ir->o]) == IRM_N &&
op1 == ir->op1 && op2 == ir->op2) { /* Regular invariant ins? */
subst[ins] = (IRRef1)ins; /* Shortcut. */
} else {
/* Re-emit substituted instruction to the FOLD/CSE/etc. pipeline. */
IRType1 t = ir->t; /* Get this first, since emitir may invalidate ir. */
IRRef ref = tref_ref(emitir(ir->ot & ~IRT_ISPHI, op1, op2));
subst[ins] = (IRRef1)ref;
if (ref != ins) {
IRIns *irr = IR(ref);
if (ref < invar) { /* Loop-carried dependency? */
/* Potential PHI? */
if (!irref_isk(ref) && !irt_isphi(irr->t) && !irt_ispri(irr->t)) {
irt_setphi(irr->t);
if (nphi >= LJ_MAX_PHI)
lj_trace_err(J, LJ_TRERR_PHIOV);
phi[nphi++] = (IRRef1)ref;
}
/* Check all loop-carried dependencies for type instability. */
if (!irt_sametype(t, irr->t)) {
if (irt_isinteger(t) && irt_isinteger(irr->t))
continue;
else if (irt_isnum(t) && irt_isinteger(irr->t)) /* Fix int->num. */
ref = tref_ref(emitir(IRTN(IR_CONV), ref, IRCONV_NUM_INT));
else if (irt_isnum(irr->t) && irt_isinteger(t)) /* Fix num->int. */
ref = tref_ref(emitir(IRTGI(IR_CONV), ref,
IRCONV_INT_NUM|IRCONV_CHECK));
else
lj_trace_err(J, LJ_TRERR_TYPEINS);
subst[ins] = (IRRef1)ref;
irr = IR(ref);
goto phiconv;
}
} else if (ref != REF_DROP && irr->o == IR_CONV &&
ref > invar && irr->op1 < invar) {
/* May need an extra PHI for a CONV. */
ref = irr->op1;
irr = IR(ref);
phiconv:
if (ref < invar && !irref_isk(ref) && !irt_isphi(irr->t)) {
irt_setphi(irr->t);
if (nphi >= LJ_MAX_PHI)
lj_trace_err(J, LJ_TRERR_PHIOV);
phi[nphi++] = (IRRef1)ref;
}
}
}
}
}
if (!irt_isguard(J->guardemit)) /* Drop redundant snapshot. */
J->cur.nsnapmap = (uint16_t)J->cur.snap[--J->cur.nsnap].mapofs;
lua_assert(J->cur.nsnapmap <= J->sizesnapmap);
*psentinel = J->cur.snapmap[J->cur.snap[0].nent]; /* Restore PC. */
loop_emit_phi(J, subst, phi, nphi, onsnap);
}
/* Emit or eliminate collected PHIs. */
static void loop_emit_phi(jit_State *J, IRRef1 *subst, IRRef1 *phi, IRRef nphi,
SnapNo onsnap)
{
int passx = 0;
IRRef i, j, nslots;
IRRef invar = J->chain[IR_LOOP];
/* Pass #1: mark redundant and potentially redundant PHIs. */
for (i = 0, j = 0; i < nphi; i++) {
IRRef lref = phi[i];
IRRef rref = subst[lref];
if (lref == rref || rref == REF_DROP) { /* Invariants are redundant. */
irt_clearphi(IR(lref)->t);
} else {
phi[j++] = (IRRef1)lref;
if (!(IR(rref)->op1 == lref || IR(rref)->op2 == lref)) {
/* Quick check for simple recurrences failed, need pass2. */
irt_setmark(IR(lref)->t);
passx = 1;
}
}
}
nphi = j;
/* Pass #2: traverse variant part and clear marks of non-redundant PHIs. */
if (passx) {
SnapNo s;
for (i = J->cur.nins-1; i > invar; i--) {
IRIns *ir = IR(i);
if (!irref_isk(ir->op2)) irt_clearmark(IR(ir->op2)->t);
if (!irref_isk(ir->op1)) {
irt_clearmark(IR(ir->op1)->t);
if (ir->op1 < invar &&
ir->o >= IR_CALLN && ir->o <= IR_CARG) { /* ORDER IR */
ir = IR(ir->op1);
while (ir->o == IR_CARG) {
if (!irref_isk(ir->op2)) irt_clearmark(IR(ir->op2)->t);
if (irref_isk(ir->op1)) break;
ir = IR(ir->op1);
irt_clearmark(ir->t);
}
}
}
}
for (s = J->cur.nsnap-1; s >= onsnap; s--) {
SnapShot *snap = &J->cur.snap[s];
SnapEntry *map = &J->cur.snapmap[snap->mapofs];
MSize n, nent = snap->nent;
for (n = 0; n < nent; n++) {
IRRef ref = snap_ref(map[n]);
if (!irref_isk(ref)) irt_clearmark(IR(ref)->t);
}
}
}
/* Pass #3: add PHIs for variant slots without a corresponding SLOAD. */
nslots = J->baseslot+J->maxslot;
for (i = 1; i < nslots; i++) {
IRRef ref = tref_ref(J->slot[i]);
while (!irref_isk(ref) && ref != subst[ref]) {
IRIns *ir = IR(ref);
irt_clearmark(ir->t); /* Unmark potential uses, too. */
if (irt_isphi(ir->t) || irt_ispri(ir->t))
break;
irt_setphi(ir->t);
if (nphi >= LJ_MAX_PHI)
lj_trace_err(J, LJ_TRERR_PHIOV);
phi[nphi++] = (IRRef1)ref;
ref = subst[ref];
if (ref > invar)
break;
}
}
/* Pass #4: propagate non-redundant PHIs. */
while (passx) {
passx = 0;
for (i = 0; i < nphi; i++) {
IRRef lref = phi[i];
IRIns *ir = IR(lref);
if (!irt_ismarked(ir->t)) { /* Propagate only from unmarked PHIs. */
IRIns *irr = IR(subst[lref]);
if (irt_ismarked(irr->t)) { /* Right ref points to other PHI? */
irt_clearmark(irr->t); /* Mark that PHI as non-redundant. */
passx = 1; /* Retry. */
}
}
}
}
/* Pass #5: emit PHI instructions or eliminate PHIs. */
for (i = 0; i < nphi; i++) {
IRRef lref = phi[i];
IRIns *ir = IR(lref);
if (!irt_ismarked(ir->t)) { /* Emit PHI if not marked. */
IRRef rref = subst[lref];
if (rref > invar)
irt_setphi(IR(rref)->t);
emitir_raw(IRT(IR_PHI, irt_type(ir->t)), lref, rref);
} else { /* Otherwise eliminate PHI. */
irt_clearmark(ir->t);
irt_clearphi(ir->t);
}
}
}
/* Transform the old IR to the new IR. */
static void split_ir(jit_State *J)
{
IRRef nins = J->cur.nins, nk = J->cur.nk;
MSize irlen = nins - nk;
MSize need = (irlen+1)*(sizeof(IRIns) + sizeof(IRRef1));
IRIns *oir = (IRIns *)lj_str_needbuf(J->L, &G(J->L)->tmpbuf, need);
IRRef1 *hisubst;
IRRef ref, snref;
SnapShot *snap;
/* Copy old IR to buffer. */
memcpy(oir, IR(nk), irlen*sizeof(IRIns));
/* Bias hiword substitution table and old IR. Loword kept in field prev. */
hisubst = (IRRef1 *)&oir[irlen] - nk;
oir -= nk;
/* Remove all IR instructions, but retain IR constants. */
J->cur.nins = REF_FIRST;
J->loopref = 0;
/* Process constants and fixed references. */
for (ref = nk; ref <= REF_BASE; ref++) {
IRIns *ir = &oir[ref];
if ((LJ_SOFTFP && ir->o == IR_KNUM) || ir->o == IR_KINT64) {
/* Split up 64 bit constant. */
TValue tv = *ir_k64(ir);
ir->prev = lj_ir_kint(J, (int32_t)tv.u32.lo);
hisubst[ref] = lj_ir_kint(J, (int32_t)tv.u32.hi);
} else {
ir->prev = ref; /* Identity substitution for loword. */
hisubst[ref] = 0;
}
}
/* Process old IR instructions. */
snap = J->cur.snap;
snref = snap->ref;
for (ref = REF_FIRST; ref < nins; ref++) {
IRIns *ir = &oir[ref];
IRRef nref = lj_ir_nextins(J);
IRIns *nir = IR(nref);
IRRef hi = 0;
if (ref >= snref) {
snap->ref = nref;
split_subst_snap(J, snap++, oir);
snref = snap < &J->cur.snap[J->cur.nsnap] ? snap->ref : ~(IRRef)0;
}
/* Copy-substitute old instruction to new instruction. */
nir->op1 = ir->op1 < nk ? ir->op1 : oir[ir->op1].prev;
nir->op2 = ir->op2 < nk ? ir->op2 : oir[ir->op2].prev;
ir->prev = nref; /* Loword substitution. */
nir->o = ir->o;
nir->t.irt = ir->t.irt & ~(IRT_MARK|IRT_ISPHI);
hisubst[ref] = 0;
/* Split 64 bit instructions. */
#if LJ_SOFTFP
if (irt_isnum(ir->t)) {
nir->t.irt = IRT_INT | (nir->t.irt & IRT_GUARD); /* Turn into INT op. */
/* Note: hi ref = lo ref + 1! Required for SNAP_SOFTFPNUM logic. */
switch (ir->o) {
case IR_ADD:
hi = split_call_ll(J, hisubst, oir, ir, IRCALL_softfp_add);
break;
case IR_SUB:
hi = split_call_ll(J, hisubst, oir, ir, IRCALL_softfp_sub);
break;
case IR_MUL:
hi = split_call_ll(J, hisubst, oir, ir, IRCALL_softfp_mul);
break;
case IR_DIV:
hi = split_call_ll(J, hisubst, oir, ir, IRCALL_softfp_div);
break;
case IR_POW:
hi = split_call_li(J, hisubst, oir, ir, IRCALL_lj_vm_powi);
break;
case IR_FPMATH:
/* Try to rejoin pow from EXP2, MUL and LOG2. */
if (nir->op2 == IRFPM_EXP2 && nir->op1 > J->loopref) {
IRIns *irp = IR(nir->op1);
if (irp->o == IR_CALLN && irp->op2 == IRCALL_softfp_mul) {
IRIns *irm4 = IR(irp->op1);
IRIns *irm3 = IR(irm4->op1);
IRIns *irm12 = IR(irm3->op1);
IRIns *irl1 = IR(irm12->op1);
if (irm12->op1 > J->loopref && irl1->o == IR_CALLN &&
irl1->op2 == IRCALL_lj_vm_log2) {
IRRef tmp = irl1->op1; /* Recycle first two args from LOG2. */
IRRef arg3 = irm3->op2, arg4 = irm4->op2;
J->cur.nins--;
tmp = split_emit(J, IRT(IR_CARG, IRT_NIL), tmp, arg3);
tmp = split_emit(J, IRT(IR_CARG, IRT_NIL), tmp, arg4);
ir->prev = tmp = split_emit(J, IRTI(IR_CALLN), tmp, IRCALL_pow);
hi = split_emit(J, IRT(IR_HIOP, IRT_SOFTFP), tmp, tmp);
break;
}
}
}
hi = split_call_l(J, hisubst, oir, ir, IRCALL_lj_vm_floor + ir->op2);
break;
case IR_ATAN2:
hi = split_call_ll(J, hisubst, oir, ir, IRCALL_atan2);
break;
case IR_LDEXP:
hi = split_call_li(J, hisubst, oir, ir, IRCALL_ldexp);
break;
case IR_NEG: case IR_ABS:
nir->o = IR_CONV; /* Pass through loword. */
nir->op2 = (IRT_INT << 5) | IRT_INT;
hi = split_emit(J, IRT(ir->o == IR_NEG ? IR_BXOR : IR_BAND, IRT_SOFTFP),
hisubst[ir->op1], hisubst[ir->op2]);
break;
case IR_SLOAD:
if ((nir->op2 & IRSLOAD_CONVERT)) { /* Convert from int to number. */
nir->op2 &= ~IRSLOAD_CONVERT;
ir->prev = nref = split_emit(J, IRTI(IR_CALLN), nref,
IRCALL_softfp_i2d);
hi = split_emit(J, IRT(IR_HIOP, IRT_SOFTFP), nref, nref);
break;
}
/* fallthrough */
case IR_ALOAD: case IR_HLOAD: case IR_ULOAD: case IR_VLOAD:
case IR_STRTO:
hi = split_emit(J, IRT(IR_HIOP, IRT_SOFTFP), nref, nref);
break;
case IR_XLOAD: {
IRIns inslo = *nir; /* Save/undo the emit of the lo XLOAD. */
J->cur.nins--;
hi = split_ptr(J, oir, ir->op1); /* Insert the hiref ADD. */
nref = lj_ir_nextins(J);
nir = IR(nref);
*nir = inslo; /* Re-emit lo XLOAD immediately before hi XLOAD. */
hi = split_emit(J, IRT(IR_XLOAD, IRT_SOFTFP), hi, ir->op2);
#if LJ_LE
ir->prev = nref;
#else
ir->prev = hi; hi = nref;
#endif
break;
}
case IR_ASTORE: case IR_HSTORE: case IR_USTORE: case IR_XSTORE:
split_emit(J, IRT(IR_HIOP, IRT_SOFTFP), nir->op1, hisubst[ir->op2]);
break;
case IR_CONV: { /* Conversion to number. Others handled below. */
IRType st = (IRType)(ir->op2 & IRCONV_SRCMASK);
UNUSED(st);
#if LJ_32 && LJ_HASFFI
if (st == IRT_I64 || st == IRT_U64) {
hi = split_call_l(J, hisubst, oir, ir,
st == IRT_I64 ? IRCALL_fp64_l2d : IRCALL_fp64_ul2d);
break;
}
#endif
lua_assert(st == IRT_INT ||
(LJ_32 && LJ_HASFFI && (st == IRT_U32 || st == IRT_FLOAT)));
nir->o = IR_CALLN;
#if LJ_32 && LJ_HASFFI
nir->op2 = st == IRT_INT ? IRCALL_softfp_i2d :
st == IRT_FLOAT ? IRCALL_softfp_f2d :
IRCALL_softfp_ui2d;
#else
nir->op2 = IRCALL_softfp_i2d;
#endif
hi = split_emit(J, IRT(IR_HIOP, IRT_SOFTFP), nref, nref);
break;
}
case IR_CALLN:
case IR_CALLL:
case IR_CALLS:
case IR_CALLXS:
goto split_call;
case IR_PHI:
if (nir->op1 == nir->op2)
J->cur.nins--; /* Drop useless PHIs. */
if (hisubst[ir->op1] != hisubst[ir->op2])
split_emit(J, IRT(IR_PHI, IRT_SOFTFP),
hisubst[ir->op1], hisubst[ir->op2]);
break;
case IR_HIOP:
J->cur.nins--; /* Drop joining HIOP. */
ir->prev = nir->op1;
hi = nir->op2;
break;
default:
lua_assert(ir->o <= IR_NE || ir->o == IR_MIN || ir->o == IR_MAX);
hi = split_emit(J, IRTG(IR_HIOP, IRT_SOFTFP),
hisubst[ir->op1], hisubst[ir->op2]);
break;
}
} else
#endif
#if LJ_32 && LJ_HASFFI
if (irt_isint64(ir->t)) {
IRRef hiref = hisubst[ir->op1];
nir->t.irt = IRT_INT | (nir->t.irt & IRT_GUARD); /* Turn into INT op. */
switch (ir->o) {
case IR_ADD:
case IR_SUB:
/* Use plain op for hiword if loword cannot produce a carry/borrow. */
if (irref_isk(nir->op2) && IR(nir->op2)->i == 0) {
ir->prev = nir->op1; /* Pass through loword. */
nir->op1 = hiref; nir->op2 = hisubst[ir->op2];
hi = nref;
break;
}
/* fallthrough */
case IR_NEG:
hi = split_emit(J, IRTI(IR_HIOP), hiref, hisubst[ir->op2]);
break;
case IR_MUL:
hi = split_call_ll(J, hisubst, oir, ir, IRCALL_lj_carith_mul64);
break;
case IR_DIV:
hi = split_call_ll(J, hisubst, oir, ir,
irt_isi64(ir->t) ? IRCALL_lj_carith_divi64 :
IRCALL_lj_carith_divu64);
break;
case IR_MOD:
hi = split_call_ll(J, hisubst, oir, ir,
irt_isi64(ir->t) ? IRCALL_lj_carith_modi64 :
IRCALL_lj_carith_modu64);
break;
case IR_POW:
hi = split_call_ll(J, hisubst, oir, ir,
irt_isi64(ir->t) ? IRCALL_lj_carith_powi64 :
IRCALL_lj_carith_powu64);
break;
case IR_FLOAD:
lua_assert(ir->op2 == IRFL_CDATA_INT64);
hi = split_emit(J, IRTI(IR_FLOAD), nir->op1, IRFL_CDATA_INT64_4);
#if LJ_BE
ir->prev = hi; hi = nref;
#endif
break;
case IR_XLOAD:
hi = split_emit(J, IRTI(IR_XLOAD), split_ptr(J, oir, ir->op1), ir->op2);
#if LJ_BE
ir->prev = hi; hi = nref;
#endif
break;
case IR_XSTORE:
split_emit(J, IRTI(IR_HIOP), nir->op1, hisubst[ir->op2]);
break;
case IR_CONV: { /* Conversion to 64 bit integer. Others handled below. */
IRType st = (IRType)(ir->op2 & IRCONV_SRCMASK);
#if LJ_SOFTFP
if (st == IRT_NUM) { /* NUM to 64 bit int conv. */
hi = split_call_l(J, hisubst, oir, ir,
irt_isi64(ir->t) ? IRCALL_fp64_d2l : IRCALL_fp64_d2ul);
} else if (st == IRT_FLOAT) { /* FLOAT to 64 bit int conv. */
nir->o = IR_CALLN;
nir->op2 = irt_isi64(ir->t) ? IRCALL_fp64_f2l : IRCALL_fp64_f2ul;
hi = split_emit(J, IRTI(IR_HIOP), nref, nref);
}
#else
if (st == IRT_NUM || st == IRT_FLOAT) { /* FP to 64 bit int conv. */
hi = split_emit(J, IRTI(IR_HIOP), nir->op1, nref);
}
#endif
else if (st == IRT_I64 || st == IRT_U64) { /* 64/64 bit cast. */
/* Drop cast, since assembler doesn't care. But fwd both parts. */
hi = hiref;
goto fwdlo;
} else if ((ir->op2 & IRCONV_SEXT)) { /* Sign-extend to 64 bit. */
IRRef k31 = lj_ir_kint(J, 31);
nir = IR(nref); /* May have been reallocated. */
ir->prev = nir->op1; /* Pass through loword. */
nir->o = IR_BSAR; /* hi = bsar(lo, 31). */
nir->op2 = k31;
hi = nref;
} else { /* Zero-extend to 64 bit. */
hi = lj_ir_kint(J, 0);
goto fwdlo;
}
break;
}
case IR_CALLXS:
goto split_call;
case IR_PHI: {
IRRef hiref2;
if ((irref_isk(nir->op1) && irref_isk(nir->op2)) ||
nir->op1 == nir->op2)
J->cur.nins--; /* Drop useless PHIs. */
hiref2 = hisubst[ir->op2];
if (!((irref_isk(hiref) && irref_isk(hiref2)) || hiref == hiref2))
split_emit(J, IRTI(IR_PHI), hiref, hiref2);
break;
}
case IR_HIOP:
J->cur.nins--; /* Drop joining HIOP. */
ir->prev = nir->op1;
hi = nir->op2;
break;
default:
lua_assert(ir->o <= IR_NE); /* Comparisons. */
split_emit(J, IRTGI(IR_HIOP), hiref, hisubst[ir->op2]);
break;
}
} else
#endif
#if LJ_SOFTFP
if (ir->o == IR_SLOAD) {
if ((nir->op2 & IRSLOAD_CONVERT)) { /* Convert from number to int. */
nir->op2 &= ~IRSLOAD_CONVERT;
if (!(nir->op2 & IRSLOAD_TYPECHECK))
nir->t.irt = IRT_INT; /* Drop guard. */
split_emit(J, IRT(IR_HIOP, IRT_SOFTFP), nref, nref);
ir->prev = split_num2int(J, nref, nref+1, irt_isguard(ir->t));
}
} else if (ir->o == IR_TOBIT) {
IRRef tmp, op1 = ir->op1;
J->cur.nins--;
#if LJ_LE
tmp = split_emit(J, IRT(IR_CARG, IRT_NIL), oir[op1].prev, hisubst[op1]);
#else
tmp = split_emit(J, IRT(IR_CARG, IRT_NIL), hisubst[op1], oir[op1].prev);
#endif
ir->prev = split_emit(J, IRTI(IR_CALLN), tmp, IRCALL_lj_vm_tobit);
} else if (ir->o == IR_TOSTR) {
if (hisubst[ir->op1]) {
if (irref_isk(ir->op1))
nir->op1 = ir->op1;
else
split_emit(J, IRT(IR_HIOP, IRT_NIL), hisubst[ir->op1], nref);
}
} else if (ir->o == IR_HREF || ir->o == IR_NEWREF) {
if (irref_isk(ir->op2) && hisubst[ir->op2])
nir->op2 = ir->op2;
} else
#endif
if (ir->o == IR_CONV) { /* See above, too. */
IRType st = (IRType)(ir->op2 & IRCONV_SRCMASK);
#if LJ_32 && LJ_HASFFI
if (st == IRT_I64 || st == IRT_U64) { /* Conversion from 64 bit int. */
#if LJ_SOFTFP
if (irt_isfloat(ir->t)) {
split_call_l(J, hisubst, oir, ir,
st == IRT_I64 ? IRCALL_fp64_l2f : IRCALL_fp64_ul2f);
J->cur.nins--; /* Drop unused HIOP. */
}
#else
if (irt_isfp(ir->t)) { /* 64 bit integer to FP conversion. */
ir->prev = split_emit(J, IRT(IR_HIOP, irt_type(ir->t)),
hisubst[ir->op1], nref);
}
#endif
else { /* Truncate to lower 32 bits. */
fwdlo:
ir->prev = nir->op1; /* Forward loword. */
/* Replace with NOP to avoid messing up the snapshot logic. */
nir->ot = IRT(IR_NOP, IRT_NIL);
nir->op1 = nir->op2 = 0;
}
}
#endif
#if LJ_SOFTFP && LJ_32 && LJ_HASFFI
else if (irt_isfloat(ir->t)) {
if (st == IRT_NUM) {
split_call_l(J, hisubst, oir, ir, IRCALL_softfp_d2f);
J->cur.nins--; /* Drop unused HIOP. */
} else {
nir->o = IR_CALLN;
nir->op2 = st == IRT_INT ? IRCALL_softfp_i2f : IRCALL_softfp_ui2f;
}
} else if (st == IRT_FLOAT) {
nir->o = IR_CALLN;
nir->op2 = irt_isint(ir->t) ? IRCALL_softfp_f2i : IRCALL_softfp_f2ui;
} else
#endif
#if LJ_SOFTFP
if (st == IRT_NUM || (LJ_32 && LJ_HASFFI && st == IRT_FLOAT)) {
if (irt_isguard(ir->t)) {
lua_assert(st == IRT_NUM && irt_isint(ir->t));
J->cur.nins--;
ir->prev = split_num2int(J, nir->op1, hisubst[ir->op1], 1);
} else {
split_call_l(J, hisubst, oir, ir,
#if LJ_32 && LJ_HASFFI
st == IRT_NUM ?
(irt_isint(ir->t) ? IRCALL_softfp_d2i : IRCALL_softfp_d2ui) :
(irt_isint(ir->t) ? IRCALL_softfp_f2i : IRCALL_softfp_f2ui)
#else
IRCALL_softfp_d2i
#endif
);
J->cur.nins--; /* Drop unused HIOP. */
}
}
#endif
} else if (ir->o == IR_CALLXS) {
IRRef hiref;
split_call:
hiref = hisubst[ir->op1];
if (hiref) {
IROpT ot = nir->ot;
IRRef op2 = nir->op2;
nir->ot = IRT(IR_CARG, IRT_NIL);
#if LJ_LE
nir->op2 = hiref;
#else
nir->op2 = nir->op1; nir->op1 = hiref;
#endif
ir->prev = nref = split_emit(J, ot, nref, op2);
}
if (LJ_SOFTFP ? irt_is64(ir->t) : irt_isint64(ir->t))
hi = split_emit(J,
IRT(IR_HIOP, (LJ_SOFTFP && irt_isnum(ir->t)) ? IRT_SOFTFP : IRT_INT),
nref, nref);
} else if (ir->o == IR_CARG) {
IRRef hiref = hisubst[ir->op1];
if (hiref) {
IRRef op2 = nir->op2;
#if LJ_LE
nir->op2 = hiref;
#else
nir->op2 = nir->op1; nir->op1 = hiref;
#endif
ir->prev = nref = split_emit(J, IRT(IR_CARG, IRT_NIL), nref, op2);
nir = IR(nref);
}
hiref = hisubst[ir->op2];
if (hiref) {
#if !LJ_TARGET_X86
int carg = 0;
IRIns *cir;
for (cir = IR(nir->op1); cir->o == IR_CARG; cir = IR(cir->op1))
carg++;
if ((carg & 1) == 0) { /* Align 64 bit arguments. */
IRRef op2 = nir->op2;
nir->op2 = REF_NIL;
nref = split_emit(J, IRT(IR_CARG, IRT_NIL), nref, op2);
nir = IR(nref);
}
#endif
#if LJ_BE
{ IRRef tmp = nir->op2; nir->op2 = hiref; hiref = tmp; }
#endif
ir->prev = split_emit(J, IRT(IR_CARG, IRT_NIL), nref, hiref);
}
} else if (ir->o == IR_CNEWI) {
if (hisubst[ir->op2])
split_emit(J, IRT(IR_HIOP, IRT_NIL), nref, hisubst[ir->op2]);
} else if (ir->o == IR_LOOP) {
J->loopref = nref; /* Needed by assembler. */
}
hisubst[ref] = hi; /* Store hiword substitution. */
}
if (snref == nins) { /* Substitution for last snapshot. */
snap->ref = J->cur.nins;
split_subst_snap(J, snap, oir);
}
/* Add PHI marks. */
for (ref = J->cur.nins-1; ref >= REF_FIRST; ref--) {
IRIns *ir = IR(ref);
if (ir->o != IR_PHI) break;
if (!irref_isk(ir->op1)) irt_setphi(IR(ir->op1)->t);
if (ir->op2 > J->loopref) irt_setphi(IR(ir->op2)->t);
}
}
/* Transform the old IR to the new IR. */
static void split_ir(jit_State *J)
{
IRRef nins = J->cur.nins, nk = J->cur.nk;
MSize irlen = nins - nk;
MSize need = (irlen+1)*(sizeof(IRIns) + sizeof(IRRef1));
IRIns *oir = (IRIns *)lj_str_needbuf(J->L, &G(J->L)->tmpbuf, need);
IRRef1 *hisubst;
IRRef ref;
/* Copy old IR to buffer. */
memcpy(oir, IR(nk), irlen*sizeof(IRIns));
/* Bias hiword substitution table and old IR. Loword kept in field prev. */
hisubst = (IRRef1 *)&oir[irlen] - nk;
oir -= nk;
/* Remove all IR instructions, but retain IR constants. */
J->cur.nins = REF_FIRST;
/* Process constants and fixed references. */
for (ref = nk; ref <= REF_BASE; ref++) {
IRIns *ir = &oir[ref];
if (ir->o == IR_KINT64) { /* Split up 64 bit constant. */
TValue tv = *ir_k64(ir);
ir->prev = lj_ir_kint(J, (int32_t)tv.u32.lo);
hisubst[ref] = lj_ir_kint(J, (int32_t)tv.u32.hi);
} else {
ir->prev = ref; /* Identity substitution for loword. */
hisubst[ref] = 0;
}
}
/* Process old IR instructions. */
for (ref = REF_FIRST; ref < nins; ref++) {
IRIns *ir = &oir[ref];
IRRef nref = lj_ir_nextins(J);
IRIns *nir = IR(nref);
IRRef hi = 0;
/* Copy-substitute old instruction to new instruction. */
nir->op1 = ir->op1 < nk ? ir->op1 : oir[ir->op1].prev;
nir->op2 = ir->op2 < nk ? ir->op2 : oir[ir->op2].prev;
ir->prev = nref; /* Loword substitution. */
nir->o = ir->o;
nir->t.irt = ir->t.irt & ~(IRT_MARK|IRT_ISPHI);
hisubst[ref] = 0;
/* Split 64 bit instructions. */
if (irt_isint64(ir->t)) {
IRRef hiref = hisubst[ir->op1];
nir->t.irt = IRT_INT | (nir->t.irt & IRT_GUARD); /* Turn into INT op. */
switch (ir->o) {
case IR_ADD:
case IR_SUB:
/* Use plain op for hiword if loword cannot produce a carry/borrow. */
if (irref_isk(nir->op2) && IR(nir->op2)->i == 0) {
ir->prev = nir->op1; /* Pass through loword. */
nir->op1 = hiref; nir->op2 = hisubst[ir->op2];
hi = nref;
break;
}
/* fallthrough */
case IR_NEG:
hi = split_emit(J, IRTI(IR_HIOP), hiref, hisubst[ir->op2]);
break;
case IR_MUL:
hi = split_call64(J, hisubst, oir, ir, IRCALL_lj_carith_mul64);
break;
case IR_DIV:
hi = split_call64(J, hisubst, oir, ir,
irt_isi64(ir->t) ? IRCALL_lj_carith_divi64 :
IRCALL_lj_carith_divu64);
break;
case IR_MOD:
hi = split_call64(J, hisubst, oir, ir,
irt_isi64(ir->t) ? IRCALL_lj_carith_modi64 :
IRCALL_lj_carith_modu64);
break;
case IR_POW:
hi = split_call64(J, hisubst, oir, ir,
irt_isi64(ir->t) ? IRCALL_lj_carith_powi64 :
IRCALL_lj_carith_powu64);
break;
case IR_FLOAD:
lua_assert(ir->op2 == IRFL_CDATA_INT64);
hi = split_emit(J, IRTI(IR_FLOAD), nir->op1, IRFL_CDATA_INT64HI);
#if LJ_BE
ir->prev = hi; hi = nref;
#endif
break;
case IR_XLOAD:
hi = split_emit(J, IRTI(IR_XLOAD), split_ptr(J, nir->op1), ir->op2);
#if LJ_BE
ir->prev = hi; hi = nref;
#endif
break;
case IR_XSTORE:
#if LJ_LE
hiref = hisubst[ir->op2];
#else
hiref = nir->op2; nir->op2 = hisubst[ir->op2];
#endif
split_emit(J, IRTI(IR_XSTORE), split_ptr(J, nir->op1), hiref);
break;
case IR_CONV: { /* Conversion to 64 bit integer. Others handled below. */
IRType st = (IRType)(ir->op2 & IRCONV_SRCMASK);
if (st == IRT_NUM || st == IRT_FLOAT) { /* FP to 64 bit int conv. */
hi = split_emit(J, IRTI(IR_HIOP), nir->op1, nref);
} else if (st == IRT_I64 || st == IRT_U64) { /* 64/64 bit cast. */
/* Drop cast, since assembler doesn't care. */
goto fwdlo;
} else if ((ir->op2 & IRCONV_SEXT)) { /* Sign-extend to 64 bit. */
IRRef k31 = lj_ir_kint(J, 31);
nir = IR(nref); /* May have been reallocated. */
ir->prev = nir->op1; /* Pass through loword. */
nir->o = IR_BSAR; /* hi = bsar(lo, 31). */
nir->op2 = k31;
hi = nref;
} else { /* Zero-extend to 64 bit. */
hi = lj_ir_kint(J, 0);
goto fwdlo;
}
break;
}
case IR_PHI: {
IRRef hiref2;
if ((irref_isk(nir->op1) && irref_isk(nir->op2)) ||
nir->op1 == nir->op2)
J->cur.nins--; /* Drop useless PHIs. */
hiref2 = hisubst[ir->op2];
if (!((irref_isk(hiref) && irref_isk(hiref2)) || hiref == hiref2))
split_emit(J, IRTI(IR_PHI), hiref, hiref2);
break;
}
default:
lua_assert(ir->o <= IR_NE); /* Comparisons. */
split_emit(J, IRTGI(IR_HIOP), hiref, hisubst[ir->op2]);
break;
}
} else if (ir->o == IR_CONV) { /* See above, too. */
IRType st = (IRType)(ir->op2 & IRCONV_SRCMASK);
if (st == IRT_I64 || st == IRT_U64) { /* Conversion from 64 bit int. */
if (irt_isfp(ir->t)) { /* 64 bit integer to FP conversion. */
ir->prev = split_emit(J, IRT(IR_HIOP, irt_type(ir->t)),
hisubst[ir->op1], nref);
} else { /* Truncate to lower 32 bits. */
fwdlo:
ir->prev = nir->op1; /* Forward loword. */
/* Replace with NOP to avoid messing up the snapshot logic. */
nir->ot = IRT(IR_NOP, IRT_NIL);
nir->op1 = nir->op2 = 0;
}
}
} else if (ir->o == IR_CNEWI) {
if (hisubst[ir->op2])
split_emit(J, IRT(IR_HIOP, IRT_NIL), nref, hisubst[ir->op2]);
} else if (ir->o == IR_LOOP) {
J->loopref = nref; /* Needed by assembler. */
}
hisubst[ref] = hi; /* Store hiword substitution. */
}
/* Add PHI marks. */
for (ref = J->cur.nins-1; ref >= REF_FIRST; ref--) {
IRIns *ir = IR(ref);
if (ir->o != IR_PHI) break;
if (!irref_isk(ir->op1)) irt_setphi(IR(ir->op1)->t);
if (ir->op2 > J->loopref) irt_setphi(IR(ir->op2)->t);
}
/* Substitute snapshot maps. */
oir[nins].prev = J->cur.nins; /* Substitution for last snapshot. */
{
SnapNo i, nsnap = J->cur.nsnap;
for (i = 0; i < nsnap; i++) {
SnapShot *snap = &J->cur.snap[i];
SnapEntry *map = &J->cur.snapmap[snap->mapofs];
MSize n, nent = snap->nent;
snap->ref = oir[snap->ref].prev;
for (n = 0; n < nent; n++) {
SnapEntry sn = map[n];
map[n] = ((sn & 0xffff0000) | oir[snap_ref(sn)].prev);
}
}
}
}