/* Narrowing of power operator or math.pow. */
TRef lj_opt_narrow_pow(jit_State *J, TRef rb, TRef rc, TValue *vc)
{
lua_Number n;
if (tvisstr(vc) && !lj_str_tonum(strV(vc), vc))
lj_trace_err(J, LJ_TRERR_BADTYPE);
n = numV(vc);
/* Limit narrowing for pow to small exponents (or for two constants). */
if ((tref_isk(rc) && tref_isint(rc) && tref_isk(rb)) ||
((J->flags & JIT_F_OPT_NARROW) &&
(numisint(n) && n >= -65536.0 && n <= 65536.0))) {
TRef tmp;
if (!tref_isinteger(rc)) {
if (tref_isstr(rc))
rc = emitir(IRTG(IR_STRTO, IRT_NUM), rc, 0);
rc = emitir(IRTGI(IR_TOINT), rc, IRTOINT_CHECK); /* Guarded TOINT! */
}
if (!tref_isk(rc)) { /* Range guard: -65536 <= i <= 65536 */
tmp = emitir(IRTI(IR_ADD), rc, lj_ir_kint(J, 65536-2147483647-1));
emitir(IRTGI(IR_LE), tmp, lj_ir_kint(J, 2*65536-2147483647-1));
}
return emitir(IRTN(IR_POWI), rb, rc);
}
/* FOLD covers most cases, but some are easier to do here. */
if (tref_isk(rb) && tvispone(ir_knum(IR(tref_ref(rb)))))
return rb; /* 1 ^ x ==> 1 */
rc = lj_ir_tonum(J, rc);
if (tref_isk(rc) && ir_knum(IR(tref_ref(rc)))->n == 0.5)
return emitir(IRTN(IR_FPMATH), rb, IRFPM_SQRT); /* x ^ 0.5 ==> sqrt(x) */
/* Split up b^c into exp2(c*log2(b)). Assembler may rejoin later. */
rb = emitir(IRTN(IR_FPMATH), rb, IRFPM_LOG2);
rc = emitir(IRTN(IR_MUL), rb, rc);
return emitir(IRTN(IR_FPMATH), rc, IRFPM_EXP2);
}
static void LJ_FASTCALL recff_math_random(jit_State *J, RecordFFData *rd)
{
GCudata *ud = udataV(&J->fn->c.upvalue[0]);
TRef tr, one;
lj_ir_kgc(J, obj2gco(ud), IRT_UDATA); /* Prevent collection. */
tr = lj_ir_call(J, IRCALL_lj_math_random_step, lj_ir_kptr(J, uddata(ud)));
one = lj_ir_knum_one(J);
tr = emitir(IRTN(IR_SUB), tr, one);
if (J->base[0]) {
TRef tr1 = lj_ir_tonum(J, J->base[0]);
if (J->base[1]) { /* d = floor(d*(r2-r1+1.0)) + r1 */
TRef tr2 = lj_ir_tonum(J, J->base[1]);
tr2 = emitir(IRTN(IR_SUB), tr2, tr1);
tr2 = emitir(IRTN(IR_ADD), tr2, one);
tr = emitir(IRTN(IR_MUL), tr, tr2);
tr = emitir(IRTN(IR_FPMATH), tr, IRFPM_FLOOR);
tr = emitir(IRTN(IR_ADD), tr, tr1);
} else { /* d = floor(d*r1) + 1.0 */
tr = emitir(IRTN(IR_MUL), tr, tr1);
tr = emitir(IRTN(IR_FPMATH), tr, IRFPM_FLOOR);
tr = emitir(IRTN(IR_ADD), tr, one);
}
}
J->base[0] = tr;
UNUSED(rd);
}
/* Record math.asin, math.acos, math.atan. */
static void LJ_FASTCALL recff_math_atrig(jit_State *J, RecordFFData *rd)
{
TRef y = lj_ir_tonum(J, J->base[0]);
TRef x = lj_ir_knum_one(J);
uint32_t ffid = rd->data;
if (ffid != FF_math_atan) {
TRef tmp = emitir(IRTN(IR_MUL), y, y);
tmp = emitir(IRTN(IR_SUB), x, tmp);
tmp = emitir(IRTN(IR_FPMATH), tmp, IRFPM_SQRT);
if (ffid == FF_math_asin) { x = tmp; } else { x = y; y = tmp; }
}
J->base[0] = emitir(IRTN(IR_ATAN2), y, x);
}
static void LJ_FASTCALL recff_math_modf(jit_State *J, RecordFFData *rd)
{
TRef tr = J->base[0];
if (tref_isinteger(tr)) {
J->base[0] = tr;
J->base[1] = lj_ir_kint(J, 0);
} else {
TRef trt;
tr = lj_ir_tonum(J, tr);
trt = emitir(IRTN(IR_FPMATH), tr, IRFPM_TRUNC);
J->base[0] = trt;
J->base[1] = emitir(IRTN(IR_SUB), tr, trt);
}
rd->nres = 2;
}
static void LJ_FASTCALL recff_math_degrad(jit_State *J, RecordFFData *rd)
{
TRef tr = lj_ir_tonum(J, J->base[0]);
TRef trm = lj_ir_knum(J, numV(&J->fn->c.upvalue[0]));
J->base[0] = emitir(IRTN(IR_MUL), tr, trm);
UNUSED(rd);
}
/* Record math.atan2. */
static void LJ_FASTCALL recff_math_atan2(jit_State *J, RecordFFData *rd)
{
TRef tr = lj_ir_tonum(J, J->base[0]);
TRef tr2 = lj_ir_tonum(J, J->base[1]);
J->base[0] = emitir(IRTN(IR_ATAN2), tr, tr2);
UNUSED(rd);
}
static void LJ_FASTCALL recff_math_minmax(jit_State *J, RecordFFData *rd)
{
TRef tr = lj_ir_tonumber(J, J->base[0]);
uint32_t op = rd->data;
BCReg i;
for (i = 1; J->base[i] != 0; i++) {
TRef tr2 = lj_ir_tonumber(J, J->base[i]);
IRType t = IRT_INT;
if (!(tref_isinteger(tr) && tref_isinteger(tr2))) {
if (tref_isinteger(tr)) tr = emitir(IRTN(IR_CONV), tr, IRCONV_NUM_INT);
if (tref_isinteger(tr2)) tr2 = emitir(IRTN(IR_CONV), tr2, IRCONV_NUM_INT);
t = IRT_NUM;
}
tr = emitir(IRT(op, t), tr, tr2);
}
J->base[0] = tr;
}
/* Narrowing of modulo operator. */
TRef lj_opt_narrow_mod(jit_State *J, TRef rb, TRef rc)
{
TRef tmp;
if ((J->flags & JIT_F_OPT_NARROW) &&
tref_isk(rc) && tref_isint(rc)) { /* Optimize x % k. */
int32_t k = IR(tref_ref(rc))->i;
if (k > 0 && (k & (k-1)) == 0) { /* i % 2^k ==> band(i, 2^k-1) */
if (tref_isinteger(rb))
return emitir(IRTI(IR_BAND), rb, lj_ir_kint(J, k-1));
}
}
/* b % c ==> b - floor(b/c)*c */
rb = lj_ir_tonum(J, rb);
rc = lj_ir_tonum(J, rc);
tmp = emitir(IRTN(IR_DIV), rb, rc);
tmp = emitir(IRTN(IR_FPMATH), tmp, IRFPM_FLOOR);
tmp = emitir(IRTN(IR_MUL), tmp, rc);
return emitir(IRTN(IR_SUB), rb, tmp);
}
/* Record math.ldexp. */
static void LJ_FASTCALL recff_math_ldexp(jit_State *J, RecordFFData *rd)
{
TRef tr = lj_ir_tonum(J, J->base[0]);
#if LJ_TARGET_X86ORX64
TRef tr2 = lj_ir_tonum(J, J->base[1]);
#else
TRef tr2 = lj_opt_narrow_toint(J, J->base[1]);
#endif
J->base[0] = emitir(IRTN(IR_LDEXP), tr, tr2);
UNUSED(rd);
}
/* Record math.log. */
static void LJ_FASTCALL recff_math_log(jit_State *J, RecordFFData *rd)
{
TRef tr = lj_ir_tonum(J, J->base[0]);
if (J->base[1]) {
#ifdef LUAJIT_NO_LOG2
uint32_t fpm = IRFPM_LOG;
#else
uint32_t fpm = IRFPM_LOG2;
#endif
TRef trb = lj_ir_tonum(J, J->base[1]);
tr = emitir(IRTN(IR_FPMATH), tr, fpm);
trb = emitir(IRTN(IR_FPMATH), trb, fpm);
trb = emitir(IRTN(IR_DIV), lj_ir_knum_one(J), trb);
tr = emitir(IRTN(IR_MUL), tr, trb);
} else {
tr = emitir(IRTN(IR_FPMATH), tr, IRFPM_LOG);
}
J->base[0] = tr;
UNUSED(rd);
}
/* Record rounding functions math.floor and math.ceil. */
static void LJ_FASTCALL recff_math_round(jit_State *J, RecordFFData *rd)
{
TRef tr = J->base[0];
if (!tref_isinteger(tr)) { /* Pass through integers unmodified. */
tr = emitir(IRTN(IR_FPMATH), lj_ir_tonum(J, tr), rd->data);
/* Result is integral (or NaN/Inf), but may not fit an int32_t. */
if (LJ_DUALNUM) { /* Try to narrow using a guarded conversion to int. */
lua_Number n = lj_vm_foldfpm(numberVnum(&rd->argv[0]), rd->data);
if (n == (lua_Number)lj_num2int(n))
tr = emitir(IRTGI(IR_CONV), tr, IRCONV_INT_NUM|IRCONV_CHECK);
}
J->base[0] = tr;
}
}
static void LJ_FASTCALL recff_math_htrig(jit_State *J, RecordFFData *rd)
{
TRef tr = lj_ir_tonum(J, J->base[0]);
J->base[0] = emitir(IRTN(IR_CALLN), tr, rd->data);
}
/* Record unary math.* functions, mapped to IR_FPMATH opcode. */
static void LJ_FASTCALL recff_math_unary(jit_State *J, RecordFFData *rd)
{
J->base[0] = emitir(IRTN(IR_FPMATH), lj_ir_tonum(J, J->base[0]), rd->data);
}
static void LJ_FASTCALL recff_math_abs(jit_State *J, RecordFFData *rd)
{
TRef tr = lj_ir_tonum(J, J->base[0]);
J->base[0] = emitir(IRTN(IR_ABS), tr, lj_ir_knum_abs(J));
UNUSED(rd);
}
/* 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);
}
/* Replay snapshot state to setup side trace. */
void lj_snap_replay(jit_State *J, GCtrace *T)
{
SnapShot *snap = &T->snap[J->exitno];
SnapEntry *map = &T->snapmap[snap->mapofs];
MSize n, nent = snap->nent;
BloomFilter seen = 0;
int pass23 = 0;
J->framedepth = 0;
/* Emit IR for slots inherited from parent snapshot. */
for (n = 0; n < nent; n++) {
SnapEntry sn = map[n];
BCReg s = snap_slot(sn);
IRRef ref = snap_ref(sn);
IRIns *ir = &T->ir[ref];
TRef tr;
/* The bloom filter avoids O(nent^2) overhead for de-duping slots. */
if (bloomtest(seen, ref) && (tr = snap_dedup(J, map, n, ref)) != 0)
goto setslot;
bloomset(seen, ref);
if (irref_isk(ref)) {
tr = snap_replay_const(J, ir);
} else if (!regsp_used(ir->prev)) {
pass23 = 1;
lua_assert(s != 0);
tr = s;
} else {
IRType t = irt_type(ir->t);
uint32_t mode = IRSLOAD_INHERIT|IRSLOAD_PARENT;
if (LJ_SOFTFP && (sn & SNAP_SOFTFPNUM)) t = IRT_NUM;
if (ir->o == IR_SLOAD) mode |= (ir->op2 & IRSLOAD_READONLY);
tr = emitir_raw(IRT(IR_SLOAD, t), s, mode);
}
setslot:
J->slot[s] = tr | (sn&(SNAP_CONT|SNAP_FRAME)); /* Same as TREF_* flags. */
J->framedepth += ((sn & (SNAP_CONT|SNAP_FRAME)) && s);
if ((sn & SNAP_FRAME))
J->baseslot = s+1;
}
if (pass23) {
IRIns *irlast = &T->ir[snap->ref];
pass23 = 0;
/* Emit dependent PVALs. */
for (n = 0; n < nent; n++) {
SnapEntry sn = map[n];
IRRef refp = snap_ref(sn);
IRIns *ir = &T->ir[refp];
if (regsp_reg(ir->r) == RID_SUNK) {
if (J->slot[snap_slot(sn)] != snap_slot(sn)) continue;
pass23 = 1;
lua_assert(ir->o == IR_TNEW || ir->o == IR_TDUP ||
ir->o == IR_CNEW || ir->o == IR_CNEWI);
if (ir->op1 >= T->nk) snap_pref(J, T, map, nent, seen, ir->op1);
if (ir->op2 >= T->nk) snap_pref(J, T, map, nent, seen, ir->op2);
if (LJ_HASFFI && ir->o == IR_CNEWI) {
if (LJ_32 && refp+1 < T->nins && (ir+1)->o == IR_HIOP)
snap_pref(J, T, map, nent, seen, (ir+1)->op2);
} else {
IRIns *irs;
for (irs = ir+1; irs < irlast; irs++)
if (irs->r == RID_SINK && snap_sunk_store(T, ir, irs)) {
if (snap_pref(J, T, map, nent, seen, irs->op2) == 0)
snap_pref(J, T, map, nent, seen, T->ir[irs->op2].op1);
else if ((LJ_SOFTFP || (LJ_32 && LJ_HASFFI)) &&
irs+1 < irlast && (irs+1)->o == IR_HIOP)
snap_pref(J, T, map, nent, seen, (irs+1)->op2);
}
}
} else if (!irref_isk(refp) && !regsp_used(ir->prev)) {
lua_assert(ir->o == IR_CONV && ir->op2 == IRCONV_NUM_INT);
J->slot[snap_slot(sn)] = snap_pref(J, T, map, nent, seen, ir->op1);
}
}
/* Replay sunk instructions. */
for (n = 0; pass23 && n < nent; n++) {
SnapEntry sn = map[n];
IRRef refp = snap_ref(sn);
IRIns *ir = &T->ir[refp];
if (regsp_reg(ir->r) == RID_SUNK) {
TRef op1, op2;
if (J->slot[snap_slot(sn)] != snap_slot(sn)) { /* De-dup allocs. */
J->slot[snap_slot(sn)] = J->slot[J->slot[snap_slot(sn)]];
continue;
}
op1 = ir->op1;
if (op1 >= T->nk) op1 = snap_pref(J, T, map, nent, seen, op1);
op2 = ir->op2;
if (op2 >= T->nk) op2 = snap_pref(J, T, map, nent, seen, op2);
if (LJ_HASFFI && ir->o == IR_CNEWI) {
if (LJ_32 && refp+1 < T->nins && (ir+1)->o == IR_HIOP) {
lj_needsplit(J); /* Emit joining HIOP. */
op2 = emitir_raw(IRT(IR_HIOP, IRT_I64), op2,
snap_pref(J, T, map, nent, seen, (ir+1)->op2));
}
J->slot[snap_slot(sn)] = emitir(ir->ot & ~(IRT_MARK|IRT_ISPHI), op1, op2);
} else {
IRIns *irs;
TRef tr = emitir(ir->ot, op1, op2);
J->slot[snap_slot(sn)] = tr;
for (irs = ir+1; irs < irlast; irs++)
if (irs->r == RID_SINK && snap_sunk_store(T, ir, irs)) {
IRIns *irr = &T->ir[irs->op1];
TRef val, key = irr->op2, tmp = tr;
if (irr->o != IR_FREF) {
IRIns *irk = &T->ir[key];
if (irr->o == IR_HREFK)
key = lj_ir_kslot(J, snap_replay_const(J, &T->ir[irk->op1]),
irk->op2);
else
key = snap_replay_const(J, irk);
if (irr->o == IR_HREFK || irr->o == IR_AREF) {
IRIns *irf = &T->ir[irr->op1];
tmp = emitir(irf->ot, tmp, irf->op2);
}
}
tmp = emitir(irr->ot, tmp, key);
val = snap_pref(J, T, map, nent, seen, irs->op2);
if (val == 0) {
IRIns *irc = &T->ir[irs->op2];
lua_assert(irc->o == IR_CONV && irc->op2 == IRCONV_NUM_INT);
val = snap_pref(J, T, map, nent, seen, irc->op1);
val = emitir(IRTN(IR_CONV), val, IRCONV_NUM_INT);
} else if ((LJ_SOFTFP || (LJ_32 && LJ_HASFFI)) &&
irs+1 < irlast && (irs+1)->o == IR_HIOP) {
IRType t = IRT_I64;
if (LJ_SOFTFP && irt_type((irs+1)->t) == IRT_SOFTFP)
t = IRT_NUM;
lj_needsplit(J);
if (irref_isk(irs->op2) && irref_isk((irs+1)->op2)) {
uint64_t k = (uint32_t)T->ir[irs->op2].i +
((uint64_t)T->ir[(irs+1)->op2].i << 32);
val = lj_ir_k64(J, t == IRT_I64 ? IR_KINT64 : IR_KNUM,
lj_ir_k64_find(J, k));
} else {
val = emitir_raw(IRT(IR_HIOP, t), val,
snap_pref(J, T, map, nent, seen, (irs+1)->op2));
}
tmp = emitir(IRT(irs->o, t), tmp, val);
continue;
}
tmp = emitir(irs->ot, tmp, val);
} else if (LJ_HASFFI && irs->o == IR_XBAR && ir->o == IR_CNEW) {
emitir(IRT(IR_XBAR, IRT_NIL), 0, 0);
}
}
}
}
}
J->base = J->slot + J->baseslot;
J->maxslot = snap->nslots - J->baseslot;
lj_snap_add(J);
if (pass23) /* Need explicit GC step _after_ initial snapshot. */
emitir_raw(IRTG(IR_GCSTEP, IRT_NIL), 0, 0);
}
