1 files changed, 2074 insertions, 0 deletions

diff --git a/src/lj_asm_ppc.h b/src/lj_asm_ppc.h
new file mode 100644
index 00000000..72e4c956
--- /dev/null
+++ b/src/lj_asm_ppc.h
@@ -0,0 +1,2074 @@
+/*
+** PPC IR assembler (SSA IR -> machine code).
+** Copyright (C) 2005-2011 Mike Pall. See Copyright Notice in luajit.h
+*/
+
+/* -- Register allocator extensions --------------------------------------- */
+
+/* Allocate a register with a hint. */
+static Reg ra_hintalloc(ASMState *as, IRRef ref, Reg hint, RegSet allow)
+{
+  Reg r = IR(ref)->r;
+  if (ra_noreg(r)) {
+    if (!ra_hashint(r) && !iscrossref(as, ref))
+      ra_sethint(IR(ref)->r, hint);  /* Propagate register hint. */
+    r = ra_allocref(as, ref, allow);
+  }
+  ra_noweak(as, r);
+  return r;
+}
+
+/* Allocate two source registers for three-operand instructions. */
+static Reg ra_alloc2(ASMState *as, IRIns *ir, RegSet allow)
+{
+  IRIns *irl = IR(ir->op1), *irr = IR(ir->op2);
+  Reg left = irl->r, right = irr->r;
+  if (ra_hasreg(left)) {
+    ra_noweak(as, left);
+    if (ra_noreg(right))
+      right = ra_allocref(as, ir->op2, rset_exclude(allow, left));
+    else
+      ra_noweak(as, right);
+  } else if (ra_hasreg(right)) {
+    ra_noweak(as, right);
+    left = ra_allocref(as, ir->op1, rset_exclude(allow, right));
+  } else if (ra_hashint(right)) {
+    right = ra_allocref(as, ir->op2, allow);
+    left = ra_alloc1(as, ir->op1, rset_exclude(allow, right));
+  } else {
+    left = ra_allocref(as, ir->op1, allow);
+    right = ra_alloc1(as, ir->op2, rset_exclude(allow, left));
+  }
+  return left | (right << 8);
+}
+
+/* -- Guard handling ------------------------------------------------------ */
+
+/* Setup exit stubs after the end of each trace. */
+static void asm_exitstub_setup(ASMState *as, ExitNo nexits)
+{
+  ExitNo i;
+  MCode *mxp = as->mctop;
+  /* 1: mflr r0; bl ->vm_exit_handler; li r0, traceno; bl <1; bl <1; ... */
+  for (i = nexits-1; (int32_t)i >= 0; i--)
+    *--mxp = PPCI_BL|(((-3-i)&0x00ffffffu)<<2);
+  *--mxp = PPCI_LI|PPCF_T(RID_TMP)|as->T->traceno;  /* Read by exit handler. */
+  mxp--;
+  *mxp = PPCI_BL|((((MCode *)(void *)lj_vm_exit_handler-mxp)&0x00ffffffu)<<2);
+  *--mxp = PPCI_MFLR|PPCF_T(RID_TMP);
+  as->mctop = mxp;
+}
+
+static MCode *asm_exitstub_addr(ASMState *as, ExitNo exitno)
+{
+  /* Keep this in-sync with exitstub_trace_addr(). */
+  return as->mctop + exitno + 3;
+}
+
+/* Emit conditional branch to exit for guard. */
+static void asm_guardcc(ASMState *as, PPCCC cc)
+{
+  MCode *target = asm_exitstub_addr(as, as->snapno);
+  MCode *p = as->mcp;
+  if (LJ_UNLIKELY(p == as->invmcp)) {
+    as->loopinv = 1;
+    *p = PPCI_B | (((target-p) & 0x00ffffffu) << 2);
+    emit_condbranch(as, PPCI_BC, cc^4, p);
+    return;
+  }
+  emit_condbranch(as, PPCI_BC, cc, target);
+}
+
+/* -- Operand fusion ------------------------------------------------------ */
+
+/* Limit linear search to this distance. Avoids O(n^2) behavior. */
+#define CONFLICT_SEARCH_LIM     31
+
+/* Check if there's no conflicting instruction between curins and ref. */
+static int noconflict(ASMState *as, IRRef ref, IROp conflict)
+{
+  IRIns *ir = as->ir;
+  IRRef i = as->curins;
+  if (i > ref + CONFLICT_SEARCH_LIM)
+    return 0;  /* Give up, ref is too far away. */
+  while (--i > ref)
+    if (ir[i].o == conflict)
+      return 0;  /* Conflict found. */
+  return 1;  /* Ok, no conflict. */
+}
+
+/* Fuse the array base of colocated arrays. */
+static int32_t asm_fuseabase(ASMState *as, IRRef ref)
+{
+  IRIns *ir = IR(ref);
+  if (ir->o == IR_TNEW && ir->op1 <= LJ_MAX_COLOSIZE &&
+      !neverfuse(as) && noconflict(as, ref, IR_NEWREF))
+    return (int32_t)sizeof(GCtab);
+  return 0;
+}
+
+/* Indicates load/store indexed is ok. */
+#define AHUREF_LSX      ((int32_t)0x80000000)
+
+/* Fuse array/hash/upvalue reference into register+offset operand. */
+static Reg asm_fuseahuref(ASMState *as, IRRef ref, int32_t *ofsp, RegSet allow)
+{
+  IRIns *ir = IR(ref);
+  if (ra_noreg(ir->r)) {
+    if (ir->o == IR_AREF) {
+      if (mayfuse(as, ref)) {
+        if (irref_isk(ir->op2)) {
+          IRRef tab = IR(ir->op1)->op1;
+          int32_t ofs = asm_fuseabase(as, tab);
+          IRRef refa = ofs ? tab : ir->op1;
+          ofs += 8*IR(ir->op2)->i;
+          if (checki16(ofs)) {
+            *ofsp = ofs;
+            return ra_alloc1(as, refa, allow);
+          }
+        }
+        if (*ofsp == AHUREF_LSX) {
+          Reg base = ra_alloc1(as, ir->op1, allow);
+          Reg idx = ra_alloc1(as, ir->op2, rset_exclude(RSET_GPR, base));
+          return base | (idx << 8);
+        }
+      }
+    } else if (ir->o == IR_HREFK) {
+      if (mayfuse(as, ref)) {
+        int32_t ofs = (int32_t)(IR(ir->op2)->op2 * sizeof(Node));
+        if (checki16(ofs)) {
+          *ofsp = ofs;
+          return ra_alloc1(as, ir->op1, allow);
+        }
+      }
+    } else if (ir->o == IR_UREFC) {
+      if (irref_isk(ir->op1)) {
+        GCfunc *fn = ir_kfunc(IR(ir->op1));
+        int32_t ofs = i32ptr(&gcref(fn->l.uvptr[(ir->op2 >> 8)])->uv.tv);
+        int32_t jgl = (intptr_t)J2G(as->J);
+        if ((uint32_t)(ofs-jgl) < 65536) {
+          *ofsp = ofs-jgl-32768;
+          return RID_JGL;
+        } else {
+          *ofsp = (int16_t)ofs;
+          return ra_allock(as, ofs-(int16_t)ofs, allow);
+        }
+      }
+    }
+  }
+  *ofsp = 0;
+  return ra_alloc1(as, ref, allow);
+}
+
+/* Fuse XLOAD/XSTORE reference into load/store operand. */
+static void asm_fusexref(ASMState *as, PPCIns pi, Reg rt, IRRef ref,
+                         RegSet allow)
+{
+  IRIns *ir = IR(ref);
+  int32_t ofs = 0;
+  Reg base;
+  if (ra_noreg(ir->r) && mayfuse(as, ref)) {
+    if (ir->o == IR_ADD) {
+      if (irref_isk(ir->op2) && (ofs = IR(ir->op2)->i, checki16(ofs))) {
+        ref = ir->op1;
+      } else {
+        Reg right, left = ra_alloc2(as, ir, allow);
+        right = (left >> 8); left &= 255;
+        emit_fab(as, PPCI_LWZX | ((pi >> 20) & 0x780), rt, left, right);
+        return;
+      }
+    } else if (ir->o == IR_STRREF) {
+      ofs = (int32_t)sizeof(GCstr);
+      if (irref_isk(ir->op2)) {
+        ofs += IR(ir->op2)->i;
+        ref = ir->op1;
+      } else if (irref_isk(ir->op1)) {
+        ofs += IR(ir->op1)->i;
+        ref = ir->op2;
+      } else {
+        /* NYI: Fuse ADD with constant. */
+        Reg right, left = ra_alloc2(as, ir, allow);
+        right = (left >> 8); left &= 255;
+        emit_fai(as, pi, rt, rt, ofs);
+        emit_tab(as, PPCI_ADD, rt, left, right);
+        return;
+      }
+      if (!checki16(ofs)) {
+        Reg left = ra_alloc1(as, ref, allow);
+        Reg right = ra_allock(as, ofs, rset_exclude(allow, left));
+        emit_fab(as, PPCI_LWZX | ((pi >> 20) & 0x780), rt, left, right);
+        return;
+      }
+    }
+  }
+  base = ra_alloc1(as, ref, allow);
+  emit_fai(as, pi, rt, base, ofs);
+}
+
+/* Fuse to multiply-add/sub instruction. */
+static int asm_fusemadd(ASMState *as, IRIns *ir, PPCIns pi, PPCIns pir)
+{
+  IRRef lref = ir->op1, rref = ir->op2;
+  IRIns *irm;
+  if (lref != rref &&
+      ((mayfuse(as, lref) && (irm = IR(lref), irm->o == IR_MUL) &&
+        ra_noreg(irm->r)) ||
+       (mayfuse(as, rref) && (irm = IR(rref), irm->o == IR_MUL) &&
+        (rref = lref, pi = pir, ra_noreg(irm->r))))) {
+    Reg dest = ra_dest(as, ir, RSET_FPR);
+    Reg add = ra_alloc1(as, rref, RSET_FPR);
+    Reg right, left = ra_alloc2(as, irm, rset_exclude(RSET_FPR, add));
+    right = (left >> 8); left &= 255;
+    emit_facb(as, pi, dest, left, right, add);
+    return 1;
+  }
+  return 0;
+}
+
+/* -- Calls --------------------------------------------------------------- */
+
+/* Generate a call to a C function. */
+static void asm_gencall(ASMState *as, const CCallInfo *ci, IRRef *args)
+{
+  uint32_t n, nargs = CCI_NARGS(ci);
+  int32_t ofs = 8;
+  Reg gpr = REGARG_FIRSTGPR, fpr = REGARG_FIRSTFPR;
+  if ((void *)ci->func)
+    emit_call(as, (void *)ci->func);
+  for (n = 0; n < nargs; n++) {  /* Setup args. */
+    IRRef ref = args[n];
+    if (ref) {
+      IRIns *ir = IR(ref);
+      if (irt_isfp(ir->t)) {
+        if (fpr <= REGARG_LASTFPR) {
+          lua_assert(rset_test(as->freeset, fpr));  /* Already evicted. */
+          ra_leftov(as, fpr, ref);
+          fpr++;
+        } else {
+          Reg r = ra_alloc1(as, ref, RSET_FPR);
+          if (irt_isnum(ir->t)) ofs = (ofs + 4) & ~4;
+          emit_spstore(as, ir, r, ofs);
+          ofs += irt_isnum(ir->t) ? 8 : 4;
+        }
+      } else {
+        if (gpr <= REGARG_LASTGPR) {
+          lua_assert(rset_test(as->freeset, gpr));  /* Already evicted. */
+          ra_leftov(as, gpr, ref);
+          gpr++;
+        } else {
+          Reg r = ra_alloc1(as, ref, RSET_GPR);
+          emit_spstore(as, ir, r, ofs);
+          ofs += 4;
+        }
+      }
+    } else {
+      if (gpr <= REGARG_LASTGPR)
+        gpr++;
+      else
+        ofs += 4;
+    }
+  }
+}
+
+/* Setup result reg/sp for call. Evict scratch regs. */
+static void asm_setupresult(ASMState *as, IRIns *ir, const CCallInfo *ci)
+{
+  RegSet drop = RSET_SCRATCH;
+  int hiop = ((ir+1)->o == IR_HIOP);
+  if ((ci->flags & CCI_NOFPRCLOBBER))
+    drop &= ~RSET_FPR;
+  if (ra_hasreg(ir->r))
+    rset_clear(drop, ir->r);  /* Dest reg handled below. */
+  if (hiop && ra_hasreg((ir+1)->r))
+    rset_clear(drop, (ir+1)->r);  /* Dest reg handled below. */
+  ra_evictset(as, drop);  /* Evictions must be performed first. */
+  if (ra_used(ir)) {
+    lua_assert(!irt_ispri(ir->t));
+    if (irt_isfp(ir->t)) {
+      if ((ci->flags & CCI_CASTU64)) {
+        /* Use spill slot or temp slots. */
+        int32_t ofs = ir->s ? sps_scale(ir->s) : SPOFS_TMP;
+        Reg dest = ir->r;
+        if (ra_hasreg(dest)) {
+          ra_free(as, dest);
+          ra_modified(as, dest);
+          emit_fai(as, PPCI_LFD, dest, RID_SP, ofs);
+        }
+        emit_tai(as, PPCI_STW, RID_RETHI, RID_SP, ofs);
+        emit_tai(as, PPCI_STW, RID_RETLO, RID_SP, ofs+4);
+      } else {
+        ra_destreg(as, ir, RID_FPRET);
+      }
+    } else if (hiop) {
+      ra_destpair(as, ir);
+    } else {
+      ra_destreg(as, ir, RID_RET);
+    }
+  }
+}
+
+static void asm_call(ASMState *as, IRIns *ir)
+{
+  IRRef args[CCI_NARGS_MAX];
+  const CCallInfo *ci = &lj_ir_callinfo[ir->op2];
+  asm_collectargs(as, ir, ci, args);
+  asm_setupresult(as, ir, ci);
+  asm_gencall(as, ci, args);
+}
+
+static void asm_callx(ASMState *as, IRIns *ir)
+{
+  IRRef args[CCI_NARGS_MAX];
+  CCallInfo ci;
+  ci.flags = asm_callx_flags(as, ir);
+  asm_collectargs(as, ir, &ci, args);
+  asm_setupresult(as, ir, &ci);
+  if (irref_isk(ir->op2)) {  /* Call to constant address. */
+    ci.func = (ASMFunction)(void *)(IR(ir->op2)->i);
+  } else {  /* Need a non-argument register for indirect calls. */
+    RegSet allow = RSET_GPR & ~RSET_RANGE(RID_R0, REGARG_LASTGPR+1);
+    Reg freg = ra_alloc1(as, ir->op2, allow);
+    *--as->mcp = PPCI_BCTRL;
+    *--as->mcp = PPCI_MTCTR | PPCF_T(freg);
+    ci.func = (ASMFunction)(void *)0;
+  }
+  asm_gencall(as, &ci, args);
+}
+
+static void asm_callid(ASMState *as, IRIns *ir, IRCallID id)
+{
+  const CCallInfo *ci = &lj_ir_callinfo[id];
+  IRRef args[2];
+  args[0] = ir->op1;
+  args[1] = ir->op2;
+  asm_setupresult(as, ir, ci);
+  asm_gencall(as, ci, args);
+}
+
+/* -- Returns ------------------------------------------------------------- */
+
+/* Return to lower frame. Guard that it goes to the right spot. */
+static void asm_retf(ASMState *as, IRIns *ir)
+{
+  Reg base = ra_alloc1(as, REF_BASE, RSET_GPR);
+  void *pc = ir_kptr(IR(ir->op2));
+  int32_t delta = 1+bc_a(*((const BCIns *)pc - 1));
+  as->topslot -= (BCReg)delta;
+  if ((int32_t)as->topslot < 0) as->topslot = 0;
+  emit_setgl(as, base, jit_base);
+  emit_addptr(as, base, -8*delta);
+  asm_guardcc(as, CC_NE);
+  emit_ab(as, PPCI_CMPW, RID_TMP,
+          ra_allock(as, i32ptr(pc), rset_exclude(RSET_GPR, base)));
+  emit_tai(as, PPCI_LWZ, RID_TMP, base, -8);
+}
+
+/* -- Type conversions ---------------------------------------------------- */
+
+static void asm_tointg(ASMState *as, IRIns *ir, Reg left)
+{
+  RegSet allow = RSET_FPR;
+  Reg tmp = ra_scratch(as, rset_clear(allow, left));
+  Reg fbias = ra_scratch(as, rset_clear(allow, tmp));
+  Reg dest = ra_dest(as, ir, RSET_GPR);
+  Reg hibias = ra_allock(as, 0x43300000, rset_exclude(RSET_GPR, dest));
+  asm_guardcc(as, CC_NE);
+  emit_fab(as, PPCI_FCMPU, 0, tmp, left);
+  emit_fab(as, PPCI_FSUB, tmp, tmp, fbias);
+  emit_fai(as, PPCI_LFD, tmp, RID_SP, SPOFS_TMP);
+  emit_tai(as, PPCI_STW, RID_TMP, RID_SP, SPOFS_TMPLO);
+  emit_tai(as, PPCI_STW, hibias, RID_SP, SPOFS_TMPHI);
+  emit_asi(as, PPCI_XORIS, RID_TMP, dest, 0x8000);
+  emit_tai(as, PPCI_LWZ, dest, RID_SP, SPOFS_TMPLO);
+  emit_lsptr(as, PPCI_LFS, (fbias & 31),
+             (void *)lj_ir_k64_find(as->J, U64x(59800004,59800000)),
+             RSET_GPR);
+  emit_fai(as, PPCI_STFD, tmp, RID_SP, SPOFS_TMP);
+  emit_fb(as, PPCI_FCTIWZ, tmp, left);
+}
+
+static void asm_tobit(ASMState *as, IRIns *ir)
+{
+  RegSet allow = RSET_FPR;
+  Reg dest = ra_dest(as, ir, RSET_GPR);
+  Reg left = ra_alloc1(as, ir->op1, allow);
+  Reg right = ra_alloc1(as, ir->op2, rset_clear(allow, left));
+  Reg tmp = ra_scratch(as, rset_clear(allow, right));
+  emit_tai(as, PPCI_LWZ, dest, RID_SP, SPOFS_TMPLO);
+  emit_fai(as, PPCI_STFD, tmp, RID_SP, SPOFS_TMP);
+  emit_fab(as, PPCI_FADD, tmp, left, right);
+}
+
+static void asm_conv(ASMState *as, IRIns *ir)
+{
+  IRType st = (IRType)(ir->op2 & IRCONV_SRCMASK);
+  int stfp = (st == IRT_NUM || st == IRT_FLOAT);
+  IRRef lref = ir->op1;
+  lua_assert(irt_type(ir->t) != st);
+  lua_assert(!(irt_isint64(ir->t) ||
+               (st == IRT_I64 || st == IRT_U64))); /* Handled by SPLIT. */
+  if (irt_isfp(ir->t)) {
+    Reg dest = ra_dest(as, ir, RSET_FPR);
+    if (stfp) {  /* FP to FP conversion. */
+      if (st == IRT_NUM)  /* double -> float conversion. */
+        emit_fb(as, PPCI_FRSP, dest, ra_alloc1(as, lref, RSET_FPR));
+      else  /* float -> double conversion is a no-op on PPC. */
+        ra_leftov(as, dest, lref);  /* Do nothing, but may need to move regs. */
+    } else {  /* Integer to FP conversion. */
+      /* IRT_INT: Flip hibit, bias with 2^52, subtract 2^52+2^31. */
+      /* IRT_U32: Bias with 2^52, subtract 2^52. */
+      RegSet allow = RSET_GPR;
+      Reg left = ra_alloc1(as, lref, allow);
+      Reg hibias = ra_allock(as, 0x43300000, rset_clear(allow, left));
+      Reg fbias = ra_scratch(as, rset_exclude(RSET_FPR, dest));
+      const float *kbias;
+      if (irt_isfloat(ir->t)) emit_fb(as, PPCI_FRSP, dest, dest);
+      emit_fab(as, PPCI_FSUB, dest, dest, fbias);
+      emit_fai(as, PPCI_LFD, dest, RID_SP, SPOFS_TMP);
+      kbias = (const float *)lj_ir_k64_find(as->J, U64x(59800004,59800000));
+      if (st == IRT_U32) kbias++;
+      emit_lsptr(as, PPCI_LFS, (fbias & 31), (void *)kbias,
+                 rset_clear(allow, hibias));
+      emit_tai(as, PPCI_STW, st == IRT_U32 ? left : RID_TMP,
+               RID_SP, SPOFS_TMPLO);
+      emit_tai(as, PPCI_STW, hibias, RID_SP, SPOFS_TMPHI);
+      if (st != IRT_U32) emit_asi(as, PPCI_XORIS, RID_TMP, left, 0x8000);
+    }
+  } else if (stfp) {  /* FP to integer conversion. */
+    if (irt_isguard(ir->t)) {
+      /* Checked conversions are only supported from number to int. */
+      lua_assert(irt_isint(ir->t) && st == IRT_NUM);
+      asm_tointg(as, ir, ra_alloc1(as, lref, RSET_FPR));
+    } else {
+      Reg dest = ra_dest(as, ir, RSET_GPR);
+      Reg left = ra_alloc1(as, lref, RSET_FPR);
+      Reg tmp = ra_scratch(as, rset_exclude(RSET_FPR, left));
+      if (irt_isu32(ir->t)) {
+        /* Convert both x and x-2^31 to int and merge results. */
+        Reg tmpi = ra_scratch(as, rset_exclude(RSET_GPR, dest));
+        emit_asb(as, PPCI_OR, dest, dest, tmpi);  /* Select with mask idiom. */
+        emit_asb(as, PPCI_AND, tmpi, tmpi, RID_TMP);
+        emit_asb(as, PPCI_ANDC, dest, dest, RID_TMP);
+        emit_tai(as, PPCI_LWZ, tmpi, RID_SP, SPOFS_TMPLO);  /* tmp = (int)(x) */
+        emit_tai(as, PPCI_ADDIS, dest, dest, 0x8000);  /* dest += 2^31 */
+        emit_asb(as, PPCI_SRAWI, RID_TMP, dest, 31);  /* mask = -(dest < 0) */
+        emit_fai(as, PPCI_STFD, tmp, RID_SP, SPOFS_TMP);
+        emit_tai(as, PPCI_LWZ, dest,
+                 RID_SP, SPOFS_TMPLO);  /* dest = (int)(x-2^31) */
+        emit_fb(as, PPCI_FCTIWZ, tmp, left);
+        emit_fai(as, PPCI_STFD, tmp, RID_SP, SPOFS_TMP);
+        emit_fb(as, PPCI_FCTIWZ, tmp, tmp);
+        emit_fab(as, PPCI_FSUB, tmp, left, tmp);
+        emit_lsptr(as, PPCI_LFS, (tmp & 31),
+                   (void *)lj_ir_k64_find(as->J, U64x(4f000000,00000000)),
+                   RSET_GPR);
+      } else {
+        emit_tai(as, PPCI_LWZ, dest, RID_SP, SPOFS_TMPLO);
+        emit_fai(as, PPCI_STFD, tmp, RID_SP, SPOFS_TMP);
+        emit_fb(as, PPCI_FCTIWZ, tmp, left);
+      }
+    }
+  } else {
+    Reg dest = ra_dest(as, ir, RSET_GPR);
+    if (st >= IRT_I8 && st <= IRT_U16) {  /* Extend to 32 bit integer. */
+      Reg left = ra_alloc1(as, ir->op1, RSET_GPR);
+      lua_assert(irt_isint(ir->t) || irt_isu32(ir->t));
+      if ((ir->op2 & IRCONV_SEXT))
+        emit_as(as, st == IRT_I8 ? PPCI_EXTSB : PPCI_EXTSH, dest, left);
+      else
+        emit_rot(as, PPCI_RLWINM, dest, left, 0, st == IRT_U8 ? 24 : 16, 31);
+    } else {  /* 32/64 bit integer conversions. */
+      /* Only need to handle 32/32 bit no-op (cast) on 32 bit archs. */
+      ra_leftov(as, dest, lref);  /* Do nothing, but may need to move regs. */
+    }
+  }
+}
+
+#if LJ_HASFFI
+static void asm_conv64(ASMState *as, IRIns *ir)
+{
+  IRType st = (IRType)((ir-1)->op2 & IRCONV_SRCMASK);
+  IRType dt = (((ir-1)->op2 & IRCONV_DSTMASK) >> IRCONV_DSH);
+  IRCallID id;
+  const CCallInfo *ci;
+  IRRef args[2];
+  args[0] = ir->op1;
+  args[1] = (ir-1)->op1;
+  if (st == IRT_NUM || st == IRT_FLOAT) {
+    id = IRCALL_fp64_d2l + ((st == IRT_FLOAT) ? 2 : 0) + (dt - IRT_I64);
+    ir--;
+  } else {
+    id = IRCALL_fp64_l2d + ((dt == IRT_FLOAT) ? 2 : 0) + (st - IRT_I64);
+  }
+  ci = &lj_ir_callinfo[id];
+  asm_setupresult(as, ir, ci);
+  asm_gencall(as, ci, args);
+}
+#endif
+
+static void asm_strto(ASMState *as, IRIns *ir)
+{
+  const CCallInfo *ci = &lj_ir_callinfo[IRCALL_lj_str_tonum];
+  IRRef args[2];
+  int32_t ofs;
+  RegSet drop = RSET_SCRATCH;
+  if (ra_hasreg(ir->r)) rset_set(drop, ir->r);  /* Spill dest reg (if any). */
+  ra_evictset(as, drop);
+  asm_guardcc(as, CC_EQ);
+  emit_ai(as, PPCI_CMPWI, RID_RET, 0);  /* Test return status. */
+  args[0] = ir->op1;      /* GCstr *str */
+  args[1] = ASMREF_TMP1;  /* TValue *n  */
+  asm_gencall(as, ci, args);
+  /* Store the result to the spill slot or temp slots. */
+  ofs = ir->s ? sps_scale(ir->s) : SPOFS_TMP;
+  emit_tai(as, PPCI_ADDI, ra_releasetmp(as, ASMREF_TMP1), RID_SP, ofs);
+}
+
+/* Get pointer to TValue. */
+static void asm_tvptr(ASMState *as, Reg dest, IRRef ref)
+{
+  IRIns *ir = IR(ref);
+  if (irt_isnum(ir->t)) {
+    if (irref_isk(ref))  /* Use the number constant itself as a TValue. */
+      ra_allockreg(as, i32ptr(ir_knum(ir)), dest);
+    else  /* Otherwise force a spill and use the spill slot. */
+      emit_tai(as, PPCI_ADDI, dest, RID_SP, ra_spill(as, ir));
+  } else {
+    /* Otherwise use g->tmptv to hold the TValue. */
+    RegSet allow = rset_exclude(RSET_GPR, dest);
+    Reg type;
+    emit_tai(as, PPCI_ADDI, dest, RID_JGL, offsetof(global_State, tmptv)-32768);
+    if (!irt_ispri(ir->t)) {
+      Reg src = ra_alloc1(as, ref, allow);
+      emit_setgl(as, src, tmptv.gcr);
+    }
+    type = ra_allock(as, irt_toitype(ir->t), allow);
+    emit_setgl(as, type, tmptv.it);
+  }
+}
+
+static void asm_tostr(ASMState *as, IRIns *ir)
+{
+  IRRef args[2];
+  args[0] = ASMREF_L;
+  as->gcsteps++;
+  if (irt_isnum(IR(ir->op1)->t) || (ir+1)->o == IR_HIOP) {
+    const CCallInfo *ci = &lj_ir_callinfo[IRCALL_lj_str_fromnum];
+    args[1] = ASMREF_TMP1;  /* const lua_Number * */
+    asm_setupresult(as, ir, ci);  /* GCstr * */
+    asm_gencall(as, ci, args);
+    asm_tvptr(as, ra_releasetmp(as, ASMREF_TMP1), ir->op1);
+  } else {
+    const CCallInfo *ci = &lj_ir_callinfo[IRCALL_lj_str_fromint];
+    args[1] = ir->op1;  /* int32_t k */
+    asm_setupresult(as, ir, ci);  /* GCstr * */
+    asm_gencall(as, ci, args);
+  }
+}
+
+/* -- Memory references --------------------------------------------------- */
+
+static void asm_aref(ASMState *as, IRIns *ir)
+{
+  Reg dest = ra_dest(as, ir, RSET_GPR);
+  Reg idx, base;
+  if (irref_isk(ir->op2)) {
+    IRRef tab = IR(ir->op1)->op1;
+    int32_t ofs = asm_fuseabase(as, tab);
+    IRRef refa = ofs ? tab : ir->op1;
+    ofs += 8*IR(ir->op2)->i;
+    if (checki16(ofs)) {
+      base = ra_alloc1(as, refa, RSET_GPR);
+      emit_tai(as, PPCI_ADDI, dest, base, ofs);
+      return;
+    }
+  }
+  base = ra_alloc1(as, ir->op1, RSET_GPR);
+  idx = ra_alloc1(as, ir->op2, rset_exclude(RSET_GPR, base));
+  emit_tab(as, PPCI_ADD, dest, RID_TMP, base);
+  emit_slwi(as, RID_TMP, idx, 3);
+}
+
+/* Inlined hash lookup. Specialized for key type and for const keys.
+** The equivalent C code is:
+**   Node *n = hashkey(t, key);
+**   do {
+**     if (lj_obj_equal(&n->key, key)) return &n->val;
+**   } while ((n = nextnode(n)));
+**   return niltv(L);
+*/
+static void asm_href(ASMState *as, IRIns *ir, IROp merge)
+{
+  RegSet allow = RSET_GPR;
+  int destused = ra_used(ir);
+  Reg dest = ra_dest(as, ir, allow);
+  Reg tab = ra_alloc1(as, ir->op1, rset_clear(allow, dest));
+  Reg key = RID_NONE, tmp1 = RID_TMP, tmp2;
+  Reg tisnum = RID_NONE, tmpnum = RID_NONE;
+  IRRef refkey = ir->op2;
+  IRIns *irkey = IR(refkey);
+  IRType1 kt = irkey->t;
+  uint32_t khash;
+  MCLabel l_end, l_loop, l_next;
+
+  rset_clear(allow, tab);
+  if (irt_isnum(kt)) {
+    key = ra_alloc1(as, refkey, RSET_FPR);
+    tmpnum = ra_scratch(as, rset_exclude(RSET_FPR, key));
+    tisnum = ra_allock(as, (int32_t)LJ_TISNUM, allow);
+    rset_clear(allow, tisnum);
+  } else if (!irt_ispri(kt)) {
+    key = ra_alloc1(as, refkey, allow);
+    rset_clear(allow, key);
+  }
+  tmp2 = ra_scratch(as, allow);
+  rset_clear(allow, tmp2);
+
+  /* Key not found in chain: jump to exit (if merged) or load niltv. */
+  l_end = emit_label(as);
+  as->invmcp = NULL;
+  if (merge == IR_NE)
+    asm_guardcc(as, CC_EQ);
+  else if (destused)
+    emit_loada(as, dest, niltvg(J2G(as->J)));
+
+  /* Follow hash chain until the end. */
+  l_loop = --as->mcp;
+  emit_ai(as, PPCI_CMPWI, dest, 0);
+  emit_tai(as, PPCI_LWZ, dest, dest, (int32_t)offsetof(Node, next));
+  l_next = emit_label(as);
+
+  /* Type and value comparison. */
+  if (merge == IR_EQ)
+    asm_guardcc(as, CC_EQ);
+  else
+    emit_condbranch(as, PPCI_BC|PPCF_Y, CC_EQ, l_end);
+  if (irt_isnum(kt)) {
+    emit_fab(as, PPCI_FCMPU, 0, tmpnum, key);
+    emit_condbranch(as, PPCI_BC, CC_GE, l_next);
+    emit_ab(as, PPCI_CMPLW, tmp1, tisnum);
+    emit_fai(as, PPCI_LFD, tmpnum, dest, (int32_t)offsetof(Node, key.n));
+  } else {
+    if (!irt_ispri(kt)) {
+      emit_ab(as, PPCI_CMPW, tmp2, key);
+      emit_condbranch(as, PPCI_BC, CC_NE, l_next);
+    }
+    emit_ai(as, PPCI_CMPWI, tmp1, irt_toitype(irkey->t));
+    if (!irt_ispri(kt))
+      emit_tai(as, PPCI_LWZ, tmp2, dest, (int32_t)offsetof(Node, key.gcr));
+  }
+  emit_tai(as, PPCI_LWZ, tmp1, dest, (int32_t)offsetof(Node, key.it));
+  *l_loop = PPCI_BC | PPCF_Y | PPCF_CC(CC_NE) |
+            (((char *)as->mcp-(char *)l_loop) & 0xffffu);
+
+  /* Load main position relative to tab->node into dest. */
+  khash = irref_isk(refkey) ? ir_khash(irkey) : 1;
+  if (khash == 0) {
+    emit_tai(as, PPCI_LWZ, dest, tab, (int32_t)offsetof(GCtab, node));
+  } else {
+    Reg tmphash = tmp1;
+    if (irref_isk(refkey))
+      tmphash = ra_allock(as, khash, allow);
+    emit_tab(as, PPCI_ADD, dest, dest, tmp1);
+    emit_tai(as, PPCI_MULLI, tmp1, tmp1, sizeof(Node));
+    emit_asb(as, PPCI_AND, tmp1, tmp2, tmphash);
+    emit_tai(as, PPCI_LWZ, dest, tab, (int32_t)offsetof(GCtab, node));
+    emit_tai(as, PPCI_LWZ, tmp2, tab, (int32_t)offsetof(GCtab, hmask));
+    if (irref_isk(refkey)) {
+      /* Nothing to do. */
+    } else if (irt_isstr(kt)) {
+      emit_tai(as, PPCI_LWZ, tmp1, key, (int32_t)offsetof(GCstr, hash));
+    } else {  /* Must match with hash*() in lj_tab.c. */
+      emit_tab(as, PPCI_SUBF, tmp1, tmp2, tmp1);
+      emit_rotlwi(as, tmp2, tmp2, HASH_ROT3);
+      emit_asb(as, PPCI_XOR, tmp1, tmp1, tmp2);
+      emit_rotlwi(as, tmp1, tmp1, (HASH_ROT2+HASH_ROT1)&31);
+      emit_tab(as, PPCI_SUBF, tmp2, dest, tmp2);
+      if (irt_isnum(kt)) {
+        int32_t ofs = ra_spill(as, irkey);
+        emit_asb(as, PPCI_XOR, tmp2, tmp2, tmp1);
+        emit_rotlwi(as, dest, tmp1, HASH_ROT1);
+        emit_tab(as, PPCI_ADD, tmp1, tmp1, tmp1);
+        emit_tai(as, PPCI_LWZ, tmp2, RID_SP, ofs+4);
+        emit_tai(as, PPCI_LWZ, tmp1, RID_SP, ofs);
+      } else {
+        emit_asb(as, PPCI_XOR, tmp2, key, tmp1);
+        emit_rotlwi(as, dest, tmp1, HASH_ROT1);
+        emit_tai(as, PPCI_ADDI, tmp1, tmp2, HASH_BIAS);
+        emit_tai(as, PPCI_ADDIS, tmp2, key, (HASH_BIAS + 32768)>>16);
+      }
+    }
+  }
+}
+
+static void asm_hrefk(ASMState *as, IRIns *ir)
+{
+  IRIns *kslot = IR(ir->op2);
+  IRIns *irkey = IR(kslot->op1);
+  int32_t ofs = (int32_t)(kslot->op2 * sizeof(Node));
+  int32_t kofs = ofs + (int32_t)offsetof(Node, key);
+  Reg dest = (ra_used(ir)||ofs > 65535) ? ra_dest(as, ir, RSET_GPR) : RID_NONE;
+  Reg node = ra_alloc1(as, ir->op1, RSET_GPR);
+  Reg key = RID_NONE, type = RID_TMP, idx = node;
+  RegSet allow = rset_exclude(RSET_GPR, node);
+  lua_assert(ofs % sizeof(Node) == 0);
+  if (ofs > 65535) {
+    idx = dest;
+    rset_clear(allow, dest);
+    kofs = (int32_t)offsetof(Node, key);
+  } else if (ra_hasreg(dest)) {
+    emit_tai(as, PPCI_ADDI, dest, node, ofs);
+  }
+  asm_guardcc(as, CC_NE);
+  if (!irt_ispri(irkey->t)) {
+    key = ra_scratch(as, allow);
+    rset_clear(allow, key);
+  }
+  rset_clear(allow, type);
+  if (irt_isnum(irkey->t)) {
+    emit_cmpi(as, key, (int32_t)ir_knum(irkey)->u32.lo);
+    asm_guardcc(as, CC_NE);
+    emit_cmpi(as, type, (int32_t)ir_knum(irkey)->u32.hi);
+  } else {
+    if (ra_hasreg(key)) {
+      emit_cmpi(as, key, irkey->i);  /* May use RID_TMP, i.e. type. */
+      asm_guardcc(as, CC_NE);
+    }
+    emit_ai(as, PPCI_CMPWI, type, irt_toitype(irkey->t));
+  }
+  if (ra_hasreg(key)) emit_tai(as, PPCI_LWZ, key, idx, kofs+4);
+  emit_tai(as, PPCI_LWZ, type, idx, kofs);
+  if (ofs > 65535) {
+    emit_tai(as, PPCI_ADDIS, dest, dest, (ofs + 32768) >> 16);
+    emit_tai(as, PPCI_ADDI, dest, node, ofs);
+  }
+}
+
+static void asm_newref(ASMState *as, IRIns *ir)
+{
+  const CCallInfo *ci = &lj_ir_callinfo[IRCALL_lj_tab_newkey];
+  IRRef args[3];
+  args[0] = ASMREF_L;     /* lua_State *L */
+  args[1] = ir->op1;      /* GCtab *t     */
+  args[2] = ASMREF_TMP1;  /* cTValue *key */
+  asm_setupresult(as, ir, ci);  /* TValue * */
+  asm_gencall(as, ci, args);
+  asm_tvptr(as, ra_releasetmp(as, ASMREF_TMP1), ir->op2);
+}
+
+static void asm_uref(ASMState *as, IRIns *ir)
+{
+  /* NYI: Check that UREFO is still open and not aliasing a slot. */
+  Reg dest = ra_dest(as, ir, RSET_GPR);
+  if (irref_isk(ir->op1)) {
+    GCfunc *fn = ir_kfunc(IR(ir->op1));
+    MRef *v = &gcref(fn->l.uvptr[(ir->op2 >> 8)])->uv.v;
+    emit_lsptr(as, PPCI_LWZ, dest, v, RSET_GPR);
+  } else {
+    Reg uv = ra_scratch(as, RSET_GPR);
+    Reg func = ra_alloc1(as, ir->op1, RSET_GPR);
+    if (ir->o == IR_UREFC) {
+      asm_guardcc(as, CC_NE);
+      emit_ai(as, PPCI_CMPWI, RID_TMP, 1);
+      emit_tai(as, PPCI_ADDI, dest, uv, (int32_t)offsetof(GCupval, tv));
+      emit_tai(as, PPCI_LBZ, RID_TMP, uv, (int32_t)offsetof(GCupval, closed));
+    } else {
+      emit_tai(as, PPCI_LWZ, dest, uv, (int32_t)offsetof(GCupval, v));
+    }
+    emit_tai(as, PPCI_LWZ, uv, func,
+             (int32_t)offsetof(GCfuncL, uvptr) + 4*(int32_t)(ir->op2 >> 8));
+  }
+}
+
+static void asm_fref(ASMState *as, IRIns *ir)
+{
+  UNUSED(as); UNUSED(ir);
+  lua_assert(!ra_used(ir));
+}
+
+static void asm_strref(ASMState *as, IRIns *ir)
+{
+  Reg dest = ra_dest(as, ir, RSET_GPR);
+  IRRef ref = ir->op2, refk = ir->op1;
+  int32_t ofs = (int32_t)sizeof(GCstr);
+  Reg r;
+  if (irref_isk(ref)) {
+    IRRef tmp = refk; refk = ref; ref = tmp;
+  } else if (!irref_isk(refk)) {
+    Reg right, left = ra_alloc1(as, ir->op1, RSET_GPR);
+    IRIns *irr = IR(ir->op2);
+    if (ra_hasreg(irr->r)) {
+      ra_noweak(as, irr->r);
+      right = irr->r;
+    } else if (mayfuse(as, irr->op2) &&
+               irr->o == IR_ADD && irref_isk(irr->op2) &&
+               checki16(ofs + IR(irr->op2)->i)) {
+      ofs += IR(irr->op2)->i;
+      right = ra_alloc1(as, irr->op1, rset_exclude(RSET_GPR, left));
+    } else {
+      right = ra_allocref(as, ir->op2, rset_exclude(RSET_GPR, left));
+    }
+    emit_tai(as, PPCI_ADDI, dest, dest, ofs);
+    emit_tab(as, PPCI_ADD, dest, left, right);
+    return;
+  }
+  r = ra_alloc1(as, ref, RSET_GPR);
+  ofs += IR(refk)->i;
+  if (checki16(ofs))
+    emit_tai(as, PPCI_ADDI, dest, r, ofs);
+  else
+    emit_tab(as, PPCI_ADD, dest, r,
+             ra_allock(as, ofs, rset_exclude(RSET_GPR, r)));
+}
+
+/* -- Loads and stores ---------------------------------------------------- */
+
+static PPCIns asm_fxloadins(IRIns *ir)
+{
+  switch (irt_type(ir->t)) {
+  case IRT_I8: return PPCI_LBZ;  /* Needs sign-extension. */
+  case IRT_U8: return PPCI_LBZ;
+  case IRT_I16: return PPCI_LHA;
+  case IRT_U16: return PPCI_LHZ;
+  case IRT_NUM: return PPCI_LFD;
+  case IRT_FLOAT: return PPCI_LFS;
+  default: return PPCI_LWZ;
+  }
+}
+
+static PPCIns asm_fxstoreins(IRIns *ir)
+{
+  switch (irt_type(ir->t)) {
+  case IRT_I8: case IRT_U8: return PPCI_STB;
+  case IRT_I16: case IRT_U16: return PPCI_STH;
+  case IRT_NUM: return PPCI_STFD;
+  case IRT_FLOAT: return PPCI_STFS;
+  default: return PPCI_STW;
+  }
+}
+
+static void asm_fload(ASMState *as, IRIns *ir)
+{
+  Reg dest = ra_dest(as, ir, RSET_GPR);
+  Reg idx = ra_alloc1(as, ir->op1, RSET_GPR);
+  PPCIns pi = asm_fxloadins(ir);
+  int32_t ofs;
+  if (ir->op2 == IRFL_TAB_ARRAY) {
+    ofs = asm_fuseabase(as, ir->op1);
+    if (ofs) {  /* Turn the t->array load into an add for colocated arrays. */
+      emit_tai(as, PPCI_ADDI, dest, idx, ofs);
+      return;
+    }
+  }
+  ofs = field_ofs[ir->op2];
+  lua_assert(!irt_isi8(ir->t));
+  emit_tai(as, pi, dest, idx, ofs);
+}
+
+static void asm_fstore(ASMState *as, IRIns *ir)
+{
+  Reg src = ra_alloc1(as, ir->op2, RSET_GPR);
+  IRIns *irf = IR(ir->op1);
+  Reg idx = ra_alloc1(as, irf->op1, rset_exclude(RSET_GPR, src));
+  int32_t ofs = field_ofs[irf->op2];
+  PPCIns pi = asm_fxstoreins(ir);
+  emit_tai(as, pi, src, idx, ofs);
+}
+
+static void asm_xload(ASMState *as, IRIns *ir)
+{
+  Reg dest = ra_dest(as, ir, irt_isfp(ir->t) ? RSET_FPR : RSET_GPR);
+  lua_assert(!(ir->op2 & IRXLOAD_UNALIGNED));
+  if (irt_isi8(ir->t))
+    emit_as(as, PPCI_EXTSB, dest, dest);
+  asm_fusexref(as, asm_fxloadins(ir), dest, ir->op1, RSET_GPR);
+}
+
+static void asm_xstore(ASMState *as, IRIns *ir)
+{
+  // NYI: fuse with bswap to stwbrx.
+  Reg src = ra_alloc1(as, ir->op2, irt_isfp(ir->t) ? RSET_FPR : RSET_GPR);
+  asm_fusexref(as, asm_fxstoreins(ir), src, ir->op1,
+               rset_exclude(RSET_GPR, src));
+}
+
+static void asm_ahuvload(ASMState *as, IRIns *ir)
+{
+  IRType1 t = ir->t;
+  Reg dest = RID_NONE, type = RID_TMP, tmp = RID_TMP, idx;
+  RegSet allow = RSET_GPR;
+  int32_t ofs = AHUREF_LSX;
+  if (ra_used(ir)) {
+    lua_assert(irt_isnum(t) || irt_isint(t) || irt_isaddr(t));
+    if (!irt_isnum(t)) ofs = 0;
+    dest = ra_dest(as, ir, irt_isnum(t) ? RSET_FPR : RSET_GPR);
+    rset_clear(allow, dest);
+  }
+  idx = asm_fuseahuref(as, ir->op1, &ofs, allow);
+  if (irt_isnum(t)) {
+    Reg tisnum = ra_allock(as, (int32_t)LJ_TISNUM, rset_exclude(allow, idx));
+    asm_guardcc(as, CC_GE);
+    emit_ab(as, PPCI_CMPLW, type, tisnum);
+    if (ra_hasreg(dest)) {
+      if (ofs == AHUREF_LSX) {
+        tmp = ra_scratch(as, rset_exclude(rset_exclude(RSET_GPR,
+                                                       (idx&255)), (idx>>8)));
+        emit_fab(as, PPCI_LFDX, dest, (idx&255), tmp);
+      } else {
+        emit_fai(as, PPCI_LFD, dest, idx, ofs);
+      }
+    }
+  } else {
+    asm_guardcc(as, CC_NE);
+    emit_ai(as, PPCI_CMPWI, type, irt_toitype(t));
+    if (ra_hasreg(dest)) emit_tai(as, PPCI_LWZ, dest, idx, ofs+4);
+  }
+  if (ofs == AHUREF_LSX) {
+    emit_tab(as, PPCI_LWZX, type, (idx&255), tmp);
+    emit_slwi(as, tmp, (idx>>8), 3);
+  } else {
+    emit_tai(as, PPCI_LWZ, type, idx, ofs);
+  }
+}
+
+static void asm_ahustore(ASMState *as, IRIns *ir)
+{
+  RegSet allow = RSET_GPR;
+  Reg idx, src = RID_NONE, type = RID_NONE;
+  int32_t ofs = AHUREF_LSX;
+  if (irt_isnum(ir->t)) {
+    src = ra_alloc1(as, ir->op2, RSET_FPR);
+  } else {
+    if (!irt_ispri(ir->t)) {
+      src = ra_alloc1(as, ir->op2, allow);
+      rset_clear(allow, src);
+      ofs = 0;
+    }
+    type = ra_allock(as, (int32_t)irt_toitype(ir->t), allow);
+    rset_clear(allow, type);
+  }
+  idx = asm_fuseahuref(as, ir->op1, &ofs, allow);
+  if (irt_isnum(ir->t)) {
+    if (ofs == AHUREF_LSX) {
+      emit_fab(as, PPCI_STFDX, src, (idx&255), RID_TMP);
+      emit_slwi(as, RID_TMP, (idx>>8), 3);
+    } else {
+      emit_fai(as, PPCI_STFD, src, idx, ofs);
+    }
+  } else {
+    if (ra_hasreg(src))
+      emit_tai(as, PPCI_STW, src, idx, ofs+4);
+    if (ofs == AHUREF_LSX) {
+      emit_tab(as, PPCI_STWX, type, (idx&255), RID_TMP);
+      emit_slwi(as, RID_TMP, (idx>>8), 3);
+    } else {
+      emit_tai(as, PPCI_STW, type, idx, ofs);
+    }
+  }
+}
+
+static void asm_sload(ASMState *as, IRIns *ir)
+{
+  int32_t ofs = 8*((int32_t)ir->op1-1) + ((ir->op2 & IRSLOAD_FRAME) ? 0 : 4);
+  IRType1 t = ir->t;
+  Reg dest = RID_NONE, type = RID_NONE, base;
+  RegSet allow = RSET_GPR;
+  lua_assert(!(ir->op2 & IRSLOAD_PARENT));  /* Handled by asm_head_side(). */
+  lua_assert(irt_isguard(t) || !(ir->op2 & IRSLOAD_TYPECHECK));
+  lua_assert(LJ_DUALNUM ||
+             !irt_isint(t) || (ir->op2 & (IRSLOAD_CONVERT|IRSLOAD_FRAME)));
+  if ((ir->op2 & IRSLOAD_CONVERT) && irt_isguard(t) && irt_isint(t)) {
+    dest = ra_scratch(as, RSET_FPR);
+    asm_tointg(as, ir, dest);
+    t.irt = IRT_NUM;  /* Continue with a regular number type check. */
+  } else if (ra_used(ir)) {
+    lua_assert(irt_isnum(t) || irt_isint(t) || irt_isaddr(t));
+    dest = ra_dest(as, ir, irt_isnum(t) ? RSET_FPR : RSET_GPR);
+    rset_clear(allow, dest);
+    base = ra_alloc1(as, REF_BASE, allow);
+    rset_clear(allow, base);
+    if ((ir->op2 & IRSLOAD_CONVERT)) {
+      if (irt_isint(t)) {
+        emit_tai(as, PPCI_LWZ, dest, RID_SP, SPOFS_TMPLO);
+        dest = ra_scratch(as, RSET_FPR);
+        emit_fai(as, PPCI_STFD, dest, RID_SP, SPOFS_TMP);
+        emit_fb(as, PPCI_FCTIWZ, dest, dest);
+        t.irt = IRT_NUM;  /* Check for original type. */
+      } else {
+        Reg tmp = ra_scratch(as, allow);
+        Reg hibias = ra_allock(as, 0x43300000, rset_clear(allow, tmp));
+        Reg fbias = ra_scratch(as, rset_exclude(RSET_FPR, dest));
+        emit_fab(as, PPCI_FSUB, dest, dest, fbias);
+        emit_fai(as, PPCI_LFD, dest, RID_SP, SPOFS_TMP);
+        emit_lsptr(as, PPCI_LFS, (fbias & 31),
+                   (void *)lj_ir_k64_find(as->J, U64x(59800004,59800000)),
+                   rset_clear(allow, hibias));
+        emit_tai(as, PPCI_STW, tmp, RID_SP, SPOFS_TMPLO);
+        emit_tai(as, PPCI_STW, hibias, RID_SP, SPOFS_TMPHI);
+        emit_asi(as, PPCI_XORIS, tmp, tmp, 0x8000);
+        dest = tmp;
+        t.irt = IRT_INT;  /* Check for original type. */
+      }
+    }
+    goto dotypecheck;
+  }
+  base = ra_alloc1(as, REF_BASE, allow);
+  rset_clear(allow, base);
+dotypecheck:
+  if (irt_isnum(t)) {
+    if ((ir->op2 & IRSLOAD_TYPECHECK)) {
+      Reg tisnum = ra_allock(as, (int32_t)LJ_TISNUM, allow);
+      asm_guardcc(as, CC_GE);
+      emit_ab(as, PPCI_CMPLW, RID_TMP, tisnum);
+      type = RID_TMP;
+    }
+    if (ra_hasreg(dest)) emit_fai(as, PPCI_LFD, dest, base, ofs-4);
+  } else {
+    if ((ir->op2 & IRSLOAD_TYPECHECK)) {
+      asm_guardcc(as, CC_NE);
+      emit_ai(as, PPCI_CMPWI, RID_TMP, irt_toitype(t));
+      type = RID_TMP;
+    }
+    if (ra_hasreg(dest)) emit_tai(as, PPCI_LWZ, dest, base, ofs);
+  }
+  if (ra_hasreg(type)) emit_tai(as, PPCI_LWZ, type, base, ofs-4);
+}
+
+/* -- Allocations --------------------------------------------------------- */
+
+#if LJ_HASFFI
+static void asm_cnew(ASMState *as, IRIns *ir)
+{
+  CTState *cts = ctype_ctsG(J2G(as->J));
+  CTypeID typeid = (CTypeID)IR(ir->op1)->i;
+  CTSize sz = (ir->o == IR_CNEWI || ir->op2 == REF_NIL) ?
+              lj_ctype_size(cts, typeid) : (CTSize)IR(ir->op2)->i;
+  const CCallInfo *ci = &lj_ir_callinfo[IRCALL_lj_mem_newgco];
+  IRRef args[2];
+  RegSet allow = (RSET_GPR & ~RSET_SCRATCH);
+  RegSet drop = RSET_SCRATCH;
+  lua_assert(sz != CTSIZE_INVALID);
+
+  args[0] = ASMREF_L;     /* lua_State *L */
+  args[1] = ASMREF_TMP1;  /* MSize size   */
+  as->gcsteps++;
+
+  if (ra_hasreg(ir->r))
+    rset_clear(drop, ir->r);  /* Dest reg handled below. */
+  ra_evictset(as, drop);
+  if (ra_used(ir))
+    ra_destreg(as, ir, RID_RET);  /* GCcdata * */
+
+  /* Initialize immutable cdata object. */
+  if (ir->o == IR_CNEWI) {
+    int32_t ofs = sizeof(GCcdata);
+    lua_assert(sz == 4 || sz == 8);
+    if (sz == 8) {
+      ofs += 4;
+      lua_assert((ir+1)->o == IR_HIOP);
+    }
+    for (;;) {
+      Reg r = ra_alloc1(as, ir->op2, allow);
+      emit_tai(as, PPCI_STW, r, RID_RET, ofs);
+      rset_clear(allow, r);
+      if (ofs == sizeof(GCcdata)) break;
+      ofs -= 4; ir++;
+    }
+  }
+  /* Initialize gct and typeid. lj_mem_newgco() already sets marked. */
+  emit_tai(as, PPCI_STB, RID_RET+1, RID_RET, offsetof(GCcdata, gct));
+  emit_tai(as, PPCI_STH, RID_TMP, RID_RET, offsetof(GCcdata, typeid));
+  emit_ti(as, PPCI_LI, RID_RET+1, ~LJ_TCDATA);
+  emit_ti(as, PPCI_LI, RID_TMP, typeid);  /* Lower 16 bit used. Sign-ext ok. */
+  asm_gencall(as, ci, args);
+  ra_allockreg(as, (int32_t)(sz+sizeof(GCcdata)),
+               ra_releasetmp(as, ASMREF_TMP1));
+}
+#else
+#define asm_cnew(as, ir)        ((void)0)
+#endif
+
+/* -- Write barriers ------------------------------------------------------ */
+
+static void asm_tbar(ASMState *as, IRIns *ir)
+{
+  Reg tab = ra_alloc1(as, ir->op1, RSET_GPR);
+  Reg mark = ra_scratch(as, rset_exclude(RSET_GPR, tab));
+  Reg link = RID_TMP;
+  MCLabel l_end = emit_label(as);
+  emit_tai(as, PPCI_STW, link, tab, (int32_t)offsetof(GCtab, gclist));
+  emit_tai(as, PPCI_STB, mark, tab, (int32_t)offsetof(GCtab, marked));
+  emit_setgl(as, tab, gc.grayagain);
+  lua_assert(LJ_GC_BLACK == 0x04);
+  emit_rot(as, PPCI_RLWINM, mark, mark, 0, 30, 28);  /* Clear black bit. */
+  emit_getgl(as, link, gc.grayagain);
+  emit_condbranch(as, PPCI_BC|PPCF_Y, CC_EQ, l_end);
+  emit_asi(as, PPCI_ANDIDOT, RID_TMP, mark, LJ_GC_BLACK);
+  emit_tai(as, PPCI_LBZ, mark, tab, (int32_t)offsetof(GCtab, marked));
+}
+
+static void asm_obar(ASMState *as, IRIns *ir)
+{
+  const CCallInfo *ci = &lj_ir_callinfo[IRCALL_lj_gc_barrieruv];
+  IRRef args[2];
+  MCLabel l_end;
+  Reg obj, val, tmp;
+  /* No need for other object barriers (yet). */
+  lua_assert(IR(ir->op1)->o == IR_UREFC);
+  ra_evictset(as, RSET_SCRATCH);
+  l_end = emit_label(as);
+  args[0] = ASMREF_TMP1;  /* global_State *g */
+  args[1] = ir->op1;      /* TValue *tv      */
+  asm_gencall(as, ci, args);
+  emit_tai(as, PPCI_ADDI, ra_releasetmp(as, ASMREF_TMP1), RID_JGL, -32768);
+  obj = IR(ir->op1)->r;
+  tmp = ra_scratch(as, rset_exclude(RSET_GPR, obj));
+  emit_condbranch(as, PPCI_BC|PPCF_Y, CC_EQ, l_end);
+  emit_asi(as, PPCI_ANDIDOT, tmp, tmp, LJ_GC_BLACK);
+  emit_condbranch(as, PPCI_BC, CC_EQ, l_end);
+  emit_asi(as, PPCI_ANDIDOT, RID_TMP, RID_TMP, LJ_GC_WHITES);
+  val = ra_alloc1(as, ir->op2, rset_exclude(RSET_GPR, obj));
+  emit_tai(as, PPCI_LBZ, tmp, obj,
+           (int32_t)offsetof(GCupval, marked)-(int32_t)offsetof(GCupval, tv));
+  emit_tai(as, PPCI_LBZ, RID_TMP, val, (int32_t)offsetof(GChead, marked));
+}
+
+/* -- Arithmetic and logic operations ------------------------------------- */
+
+static void asm_fparith(ASMState *as, IRIns *ir, PPCIns pi)
+{
+  Reg dest = ra_dest(as, ir, RSET_FPR);
+  Reg right, left = ra_alloc2(as, ir, RSET_FPR);
+  right = (left >> 8); left &= 255;
+  if (pi == PPCI_FMUL)
+    emit_fac(as, pi, dest, left, right);
+  else
+    emit_fab(as, pi, dest, left, right);
+}
+
+static void asm_fpunary(ASMState *as, IRIns *ir, PPCIns pi)
+{
+  Reg dest = ra_dest(as, ir, RSET_FPR);
+  Reg left = ra_hintalloc(as, ir->op1, dest, RSET_FPR);
+  emit_fb(as, pi, dest, left);
+}
+
+static int asm_fpjoin_pow(ASMState *as, IRIns *ir)
+{
+  IRIns *irp = IR(ir->op1);
+  if (irp == ir-1 && irp->o == IR_MUL && !ra_used(irp)) {
+    IRIns *irpp = IR(irp->op1);
+    if (irpp == ir-2 && irpp->o == IR_FPMATH &&
+        irpp->op2 == IRFPM_LOG2 && !ra_used(irpp)) {
+      const CCallInfo *ci = &lj_ir_callinfo[IRCALL_pow];
+      IRRef args[2];
+      args[0] = irpp->op1;
+      args[1] = irp->op2;
+      asm_setupresult(as, ir, ci);
+      asm_gencall(as, ci, args);
+      return 1;
+    }
+  }
+  return 0;
+}
+
+static void asm_add(ASMState *as, IRIns *ir)
+{
+  if (irt_isnum(ir->t)) {
+    if (!asm_fusemadd(as, ir, PPCI_FMADD, PPCI_FMADD))
+      asm_fparith(as, ir, PPCI_FADD);
+  } else {
+    Reg dest = ra_dest(as, ir, RSET_GPR);
+    Reg right, left = ra_hintalloc(as, ir->op1, dest, RSET_GPR);
+    PPCIns pi;
+    if (irref_isk(ir->op2)) {
+      int32_t k = IR(ir->op2)->i;
+      if (checki16(k)) {
+        pi = PPCI_ADDI;
+        /* May fail due to spills/restores above, but simplifies the logic. */
+        if (as->flagmcp == as->mcp) {
+          as->flagmcp = NULL;
+          as->mcp++;
+          pi = PPCI_ADDICDOT;
+        }
+        emit_tai(as, pi, dest, left, k);
+        return;
+      } else if ((k & 0xffff) == 0) {
+        emit_tai(as, PPCI_ADDIS, dest, left, (k >> 16));
+        return;
+      } else if (!as->sectref) {
+        emit_tai(as, PPCI_ADDIS, dest, dest, (k + 32768) >> 16);
+        emit_tai(as, PPCI_ADDI, dest, left, k);
+        return;
+      }
+    }
+    pi = PPCI_ADD;
+    /* May fail due to spills/restores above, but simplifies the logic. */
+    if (as->flagmcp == as->mcp) {
+      as->flagmcp = NULL;
+      as->mcp++;
+      pi |= PPCF_DOT;
+    }
+    right = ra_alloc1(as, ir->op2, rset_exclude(RSET_GPR, left));
+    emit_tab(as, pi, dest, left, right);
+  }
+}
+
+static void asm_sub(ASMState *as, IRIns *ir)
+{
+  if (irt_isnum(ir->t)) {
+    if (!asm_fusemadd(as, ir, PPCI_FMSUB, PPCI_FNMSUB))
+      asm_fparith(as, ir, PPCI_FSUB);
+  } else {
+    PPCIns pi = PPCI_SUBF;
+    Reg dest = ra_dest(as, ir, RSET_GPR);
+    Reg left, right;
+    if (irref_isk(ir->op1)) {
+      int32_t k = IR(ir->op1)->i;
+      if (checki16(k)) {
+        right = ra_alloc1(as, ir->op2, RSET_GPR);
+        emit_tai(as, PPCI_SUBFIC, dest, right, k);
+        return;
+      }
+    }
+    /* May fail due to spills/restores above, but simplifies the logic. */
+    if (as->flagmcp == as->mcp) {
+      as->flagmcp = NULL;
+      as->mcp++;
+      pi |= PPCF_DOT;
+    }
+    left = ra_hintalloc(as, ir->op1, dest, RSET_GPR);
+    right = ra_alloc1(as, ir->op2, rset_exclude(RSET_GPR, left));
+    emit_tab(as, pi, dest, right, left);  /* Subtract right _from_ left. */
+  }
+}
+
+static void asm_mul(ASMState *as, IRIns *ir)
+{
+  if (irt_isnum(ir->t)) {
+    asm_fparith(as, ir, PPCI_FMUL);
+  } else {
+    PPCIns pi = PPCI_MULLW;
+    Reg dest = ra_dest(as, ir, RSET_GPR);
+    Reg right, left = ra_hintalloc(as, ir->op1, dest, RSET_GPR);
+    if (irref_isk(ir->op2)) {
+      int32_t k = IR(ir->op2)->i;
+      if (checki16(k)) {
+        emit_tai(as, PPCI_MULLI, dest, left, k);
+        return;
+      }
+    }
+    /* May fail due to spills/restores above, but simplifies the logic. */
+    if (as->flagmcp == as->mcp) {
+      as->flagmcp = NULL;
+      as->mcp++;
+      pi |= PPCF_DOT;
+    }
+    right = ra_alloc1(as, ir->op2, rset_exclude(RSET_GPR, left));
+    emit_tab(as, pi, dest, left, right);
+  }
+}
+
+static void asm_neg(ASMState *as, IRIns *ir)
+{
+  if (irt_isnum(ir->t)) {
+    asm_fpunary(as, ir, PPCI_FNEG);
+  } else {
+    Reg dest, left;
+    PPCIns pi = PPCI_NEG;
+    if (as->flagmcp == as->mcp) {
+      as->flagmcp = NULL;
+      as->mcp++;
+      pi |= PPCF_DOT;
+    }
+    dest = ra_dest(as, ir, RSET_GPR);
+    left = ra_hintalloc(as, ir->op1, dest, RSET_GPR);
+    emit_tab(as, pi, dest, left, 0);
+  }
+}
+
+static void asm_arithov(ASMState *as, IRIns *ir, PPCIns pi)
+{
+  Reg dest, left, right;
+  if (as->flagmcp == as->mcp) {
+    as->flagmcp = NULL;
+    as->mcp++;
+  }
+  asm_guardcc(as, CC_SO);
+  dest = ra_dest(as, ir, RSET_GPR);
+  left = ra_alloc2(as, ir, RSET_GPR);
+  right = (left >> 8); left &= 255;
+  if (pi == PPCI_SUBFO) { Reg tmp = left; left = right; right = tmp; }
+  emit_tab(as, pi|PPCF_DOT, dest, left, right);
+}
+
+#if LJ_HASFFI
+static void asm_add64(ASMState *as, IRIns *ir)
+{
+  Reg dest = ra_dest(as, ir, RSET_GPR);
+  Reg right, left = ra_alloc1(as, ir->op1, RSET_GPR);
+  PPCIns pi = PPCI_ADDE;
+  if (irref_isk(ir->op2)) {
+    int32_t k = IR(ir->op2)->i;
+    if (k == 0)
+      pi = PPCI_ADDZE;
+    else if (k == -1)
+      pi = PPCI_ADDME;
+    else
+      goto needright;
+    right = 0;
+  } else {
+  needright:
+    right = ra_alloc1(as, ir->op2, rset_exclude(RSET_GPR, left));
+  }
+  emit_tab(as, pi, dest, left, right);
+  ir--;
+  dest = ra_dest(as, ir, RSET_GPR);
+  left = ra_alloc1(as, ir->op1, RSET_GPR);
+  if (irref_isk(ir->op2)) {
+    int32_t k = IR(ir->op2)->i;
+    if (checki16(k)) {
+      emit_tai(as, PPCI_ADDIC, dest, left, k);
+      return;
+    }
+  }
+  right = ra_alloc1(as, ir->op2, rset_exclude(RSET_GPR, left));
+  emit_tab(as, PPCI_ADDC, dest, left, right);
+}
+
+static void asm_sub64(ASMState *as, IRIns *ir)
+{
+  Reg dest = ra_dest(as, ir, RSET_GPR);
+  Reg left, right = ra_alloc1(as, ir->op2, RSET_GPR);
+  PPCIns pi = PPCI_SUBFE;
+  if (irref_isk(ir->op1)) {
+    int32_t k = IR(ir->op1)->i;
+    if (k == 0)
+      pi = PPCI_SUBFZE;
+    else if (k == -1)
+      pi = PPCI_SUBFME;
+    else
+      goto needleft;
+    left = 0;
+  } else {
+  needleft:
+    left = ra_alloc1(as, ir->op1, rset_exclude(RSET_GPR, right));
+  }
+  emit_tab(as, pi, dest, right, left);  /* Subtract right _from_ left. */
+  ir--;
+  dest = ra_dest(as, ir, RSET_GPR);
+  right = ra_alloc1(as, ir->op2, RSET_GPR);
+  if (irref_isk(ir->op1)) {
+    int32_t k = IR(ir->op1)->i;
+    if (checki16(k)) {
+      emit_tai(as, PPCI_SUBFIC, dest, right, k);
+      return;
+    }
+  }
+  left = ra_alloc1(as, ir->op1, rset_exclude(RSET_GPR, right));
+  emit_tab(as, PPCI_SUBFC, dest, right, left);
+}
+
+static void asm_neg64(ASMState *as, IRIns *ir)
+{
+  Reg dest = ra_dest(as, ir, RSET_GPR);
+  Reg left = ra_alloc1(as, ir->op1, RSET_GPR);
+  emit_tab(as, PPCI_SUBFZE, dest, left, 0);
+  ir--;
+  dest = ra_dest(as, ir, RSET_GPR);
+  left = ra_alloc1(as, ir->op1, RSET_GPR);
+  emit_tai(as, PPCI_SUBFIC, dest, left, 0);
+}
+#endif
+
+static void asm_bitnot(ASMState *as, IRIns *ir)
+{
+  Reg dest, left, right;
+  PPCIns pi = PPCI_NOR;
+  if (as->flagmcp == as->mcp) {
+    as->flagmcp = NULL;
+    as->mcp++;
+    pi |= PPCF_DOT;
+  }
+  dest = ra_dest(as, ir, RSET_GPR);
+  if (mayfuse(as, ir->op1)) {
+    IRIns *irl = IR(ir->op1);
+    if (irl->o == IR_BAND)
+      pi ^= (PPCI_NOR ^ PPCI_NAND);
+    else if (irl->o == IR_BXOR)
+      pi ^= (PPCI_NOR ^ PPCI_EQV);
+    else if (irl->o != IR_BOR)
+      goto nofuse;
+    left = ra_hintalloc(as, irl->op1, dest, RSET_GPR);
+    right = ra_alloc1(as, irl->op2, rset_exclude(RSET_GPR, left));
+  } else {
+nofuse:
+    left = right = ra_hintalloc(as, ir->op1, dest, RSET_GPR);
+  }
+  emit_asb(as, pi, dest, left, right);
+}
+
+static void asm_bitswap(ASMState *as, IRIns *ir)
+{
+  // NYI: fuse with XLOAD to lwbrx.
+  Reg dest = ra_dest(as, ir, RSET_GPR);
+  Reg left = ra_alloc1(as, ir->op1, RSET_GPR);
+  Reg tmp = dest;
+  if (tmp == left) {
+    tmp = RID_TMP;
+    emit_mr(as, dest, RID_TMP);
+  }
+  emit_rot(as, PPCI_RLWIMI, tmp, left, 24, 16, 23);
+  emit_rot(as, PPCI_RLWIMI, tmp, left, 24, 0, 7);
+  emit_rotlwi(as, tmp, left, 8);
+}
+
+static void asm_bitop(ASMState *as, IRIns *ir, PPCIns pi, PPCIns pik)
+{
+  Reg dest = ra_dest(as, ir, RSET_GPR);
+  Reg right, left = ra_hintalloc(as, ir->op1, dest, RSET_GPR);
+  if (irref_isk(ir->op2)) {
+    int32_t k = IR(ir->op2)->i;
+    Reg tmp = left;
+    if ((checku16(k) || (k & 0xffff) == 0) || (tmp = dest, !as->sectref)) {
+      if (!checku16(k)) {
+        emit_asi(as, pik ^ (PPCI_ORI ^ PPCI_ORIS), dest, tmp, (k >> 16));
+        if ((k & 0xffff) == 0) return;
+      }
+      emit_asi(as, pik, dest, left, k);
+      return;
+    }
+  }
+  /* May fail due to spills/restores above, but simplifies the logic. */
+  if (as->flagmcp == as->mcp) {
+    as->flagmcp = NULL;
+    as->mcp++;
+    pi |= PPCF_DOT;
+  }
+  right = ra_alloc1(as, ir->op2, rset_exclude(RSET_GPR, left));
+  emit_asb(as, pi, dest, left, right);
+}
+
+static void asm_bitand(ASMState *as, IRIns *ir)
+{
+  Reg dest, left, right;
+  PPCIns dot = 0;
+  IRRef op2;
+  if (as->flagmcp == as->mcp) {
+    as->flagmcp = NULL;
+    as->mcp++;
+    dot = PPCF_DOT;
+  }
+  dest = ra_dest(as, ir, RSET_GPR);
+  left = ra_hintalloc(as, ir->op1, dest, RSET_GPR);
+  if (irref_isk(ir->op2)) {
+    int32_t k = IR(ir->op2)->i;
+    if (k) {
+      // NYI: fuse with shifts/rotates.
+      uint32_t s1 = lj_ffs((uint32_t)k);
+      uint32_t k1 = ((uint32_t)k >> s1);
+      if ((k1 & (k1+1)) == 0) {
+        emit_rot(as, PPCI_RLWINM|dot, dest, left, 0,
+                 31-lj_fls((uint32_t)k), 31-s1);
+        return;
+      }
+      if (~(uint32_t)k) {
+        uint32_t s2 = lj_ffs(~(uint32_t)k);
+        uint32_t k2 = (~(uint32_t)k >> s2);
+        if ((k2 & (k2+1)) == 0) {
+          emit_rot(as, PPCI_RLWINM|dot, dest, left, 0,
+                   32-s2, 30-lj_fls(~(uint32_t)k));
+          return;
+        }
+      }
+    }
+    if (checku16(k)) {
+      emit_asi(as, PPCI_ANDIDOT, dest, left, k);
+      return;
+    } else if ((k & 0xffff) == 0) {
+      emit_asi(as, PPCI_ANDISDOT, dest, left, (k >> 16));
+      return;
+    }
+  }
+  op2 = ir->op2;
+  if (mayfuse(as, op2) && IR(op2)->o == IR_BNOT) {
+    dot ^= (PPCI_AND ^ PPCI_ANDC);
+    op2 = IR(op2)->op1;
+  }
+  right = ra_alloc1(as, op2, rset_exclude(RSET_GPR, left));
+  emit_asb(as, PPCI_AND ^ dot, dest, left, right);
+}
+
+static void asm_bitshift(ASMState *as, IRIns *ir, PPCIns pi, PPCIns pik)
+{
+  // NYI: fuse with IR_BAND.
+  Reg dest, left;
+  Reg dot = 0;
+  if (as->flagmcp == as->mcp) {
+    as->flagmcp = NULL;
+    as->mcp++;
+    dot = PPCF_DOT;
+  }
+  dest = ra_dest(as, ir, RSET_GPR);
+  left = ra_alloc1(as, ir->op1, RSET_GPR);
+  if (irref_isk(ir->op2)) {  /* Constant shifts. */
+    int32_t shift = (IR(ir->op2)->i & 31);
+    if (pik == 0)  /* SLWI */
+      emit_rot(as, PPCI_RLWINM|dot, dest, left, shift, 0, 31-shift);
+    else if (pik == 1)  /* SRWI */
+      emit_rot(as, PPCI_RLWINM|dot, dest, left, (32-shift)&31, shift, 31);
+    else
+      emit_asb(as, pik|dot, dest, left, shift);
+  } else {
+    Reg right = ra_alloc1(as, ir->op2, rset_exclude(RSET_GPR, left));
+    emit_asb(as, pi|dot, dest, left, right);
+  }
+}
+
+static void asm_min_max(ASMState *as, IRIns *ir, int ismax)
+{
+  if (irt_isnum(ir->t)) {
+    Reg dest = ra_dest(as, ir, RSET_FPR);
+    Reg tmp = dest;
+    Reg right, left = ra_alloc2(as, ir, RSET_FPR);
+    right = (left >> 8); left &= 255;
+    if (tmp == left || tmp == right)
+      tmp = ra_scratch(as, rset_exclude(rset_exclude(rset_exclude(RSET_FPR,
+                                        dest), left), right));
+    emit_facb(as, PPCI_FSEL, dest, tmp,
+              ismax ? left : right, ismax ? right : left);
+    emit_fab(as, PPCI_FSUB, tmp, left, right);
+  } else {
+    Reg dest = ra_dest(as, ir, RSET_GPR);
+    Reg tmp1 = RID_TMP, tmp2 = dest;
+    Reg right, left = ra_alloc2(as, ir, RSET_GPR);
+    right = (left >> 8); left &= 255;
+    if (tmp2 == left || tmp2 == right)
+      tmp2 = ra_scratch(as, rset_exclude(rset_exclude(rset_exclude(RSET_GPR,
+                                         dest), left), right));
+    emit_tab(as, PPCI_ADD, dest, tmp2, right);
+    emit_asb(as, ismax ? PPCI_ANDC : PPCI_AND, tmp2, tmp2, tmp1);
+    emit_tab(as, PPCI_SUBFE, tmp1, tmp1, tmp1);
+    emit_tab(as, PPCI_SUBFC, tmp2, tmp2, tmp1);
+    emit_asi(as, PPCI_XORIS, tmp2, right, 0x8000);
+    emit_asi(as, PPCI_XORIS, tmp1, left, 0x8000);
+  }
+}
+
+/* -- Comparisons --------------------------------------------------------- */
+
+#define CC_UNSIGNED     0x08    /* Unsigned integer comparison. */
+#define CC_TWO          0x80    /* Check two flags for FP comparison. */
+
+/* Map of comparisons to flags. ORDER IR. */
+static const uint8_t asm_compmap[IR_ABC+1] = {
+  /* op     int cc                 FP cc */
+  /* LT  */ CC_GE               + (CC_GE<<4),
+  /* GE  */ CC_LT               + (CC_LE<<4) + CC_TWO,
+  /* LE  */ CC_GT               + (CC_GE<<4) + CC_TWO,
+  /* GT  */ CC_LE               + (CC_LE<<4),
+  /* ULT */ CC_GE + CC_UNSIGNED + (CC_GT<<4) + CC_TWO,
+  /* UGE */ CC_LT + CC_UNSIGNED + (CC_LT<<4),
+  /* ULE */ CC_GT + CC_UNSIGNED + (CC_GT<<4),
+  /* UGT */ CC_LE + CC_UNSIGNED + (CC_LT<<4) + CC_TWO,
+  /* EQ  */ CC_NE               + (CC_NE<<4),
+  /* NE  */ CC_EQ               + (CC_EQ<<4),
+  /* ABC */ CC_LE + CC_UNSIGNED + (CC_LT<<4) + CC_TWO  /* Same as UGT. */
+};
+
+static void asm_intcomp_(ASMState *as, IRRef lref, IRRef rref, Reg cr, PPCCC cc)
+{
+  Reg right, left = ra_alloc1(as, lref, RSET_GPR);
+  if (irref_isk(rref)) {
+    int32_t k = IR(rref)->i;
+    if ((cc & CC_UNSIGNED) == 0) {  /* Signed comparison with constant. */
+      if (checki16(k)) {
+        emit_tai(as, PPCI_CMPWI, cr, left, k);
+        /* Signed comparison with zero and referencing previous ins? */
+        if (k == 0 && lref == as->curins-1)
+          as->flagmcp = as->mcp;  /* Allow elimination of the compare. */
+        return;
+      } else if ((cc & 3) == (CC_EQ & 3)) {  /* Use CMPLWI for EQ or NE. */
+        if (checku16(k)) {
+          emit_tai(as, PPCI_CMPLWI, cr, left, k);
+          return;
+        } else if (!as->sectref && ra_noreg(IR(rref)->r)) {
+          emit_tai(as, PPCI_CMPLWI, cr, RID_TMP, k);
+          emit_asi(as, PPCI_XORIS, RID_TMP, left, (k >> 16));
+          return;
+        }
+      }
+    } else {  /* Unsigned comparison with constant. */
+      if (checku16(k)) {
+        emit_tai(as, PPCI_CMPLWI, cr, left, k);
+        return;
+      }
+    }
+  }
+  right = ra_alloc1(as, rref, rset_exclude(RSET_GPR, left));
+  emit_tab(as, (cc & CC_UNSIGNED) ? PPCI_CMPLW : PPCI_CMPW, cr, left, right);
+}
+
+static void asm_comp(ASMState *as, IRIns *ir)
+{
+  PPCCC cc = asm_compmap[ir->o];
+  if (irt_isnum(ir->t)) {
+    Reg right, left = ra_alloc2(as, ir, RSET_FPR);
+    right = (left >> 8); left &= 255;
+    asm_guardcc(as, (cc >> 4));
+    if ((cc & CC_TWO))
+      emit_tab(as, PPCI_CROR, ((cc>>4)&3), ((cc>>4)&3), (CC_EQ&3));
+    emit_fab(as, PPCI_FCMPU, 0, left, right);
+  } else {
+    IRRef lref = ir->op1, rref = ir->op2;
+    if (irref_isk(lref) && !irref_isk(rref)) {
+      /* Swap constants to the right (only for ABC). */
+      IRRef tmp = lref; lref = rref; rref = tmp;
+      if ((cc & 2) == 0) cc ^= 1;  /* LT <-> GT, LE <-> GE */
+    }
+    asm_guardcc(as, cc);
+    asm_intcomp_(as, lref, rref, 0, cc);
+  }
+}
+
+#if LJ_HASFFI
+/* 64 bit integer comparisons. */
+static void asm_comp64(ASMState *as, IRIns *ir)
+{
+  PPCCC cc = asm_compmap[(ir-1)->o];
+  if ((cc&3) == (CC_EQ&3)) {
+    asm_guardcc(as, cc);
+    emit_tab(as, (cc&4) ? PPCI_CRAND : PPCI_CROR,
+             (CC_EQ&3), (CC_EQ&3), 4+(CC_EQ&3));
+  } else {
+    asm_guardcc(as, CC_EQ);
+    emit_tab(as, PPCI_CROR, (CC_EQ&3), (CC_EQ&3), ((cc^~(cc>>2))&1));
+    emit_tab(as, (cc&4) ? PPCI_CRAND : PPCI_CRANDC,
+             (CC_EQ&3), (CC_EQ&3), 4+(cc&3));
+  }
+  /* Loword comparison sets cr1 and is unsigned, except for equality. */
+  asm_intcomp_(as, (ir-1)->op1, (ir-1)->op2, 4,
+               cc | ((cc&3) == (CC_EQ&3) ? 0 : CC_UNSIGNED));
+  /* Hiword comparison sets cr0. */
+  asm_intcomp_(as, ir->op1, ir->op2, 0, cc);
+  as->flagmcp = NULL;  /* Doesn't work here. */
+}
+#endif
+
+/* -- Support for 64 bit ops in 32 bit mode ------------------------------- */
+
+/* Hiword op of a split 64 bit op. Previous op must be the loword op. */
+static void asm_hiop(ASMState *as, IRIns *ir)
+{
+#if LJ_HASFFI
+  /* HIOP is marked as a store because it needs its own DCE logic. */
+  int uselo = ra_used(ir-1), usehi = ra_used(ir);  /* Loword/hiword used? */
+  if (LJ_UNLIKELY(!(as->flags & JIT_F_OPT_DCE))) uselo = usehi = 1;
+  if ((ir-1)->o == IR_CONV) {  /* Conversions to/from 64 bit. */
+    as->curins--;  /* Always skip the CONV. */
+    if (usehi || uselo)
+      asm_conv64(as, ir);
+    return;
+  } else if ((ir-1)->o <= IR_NE) {  /* 64 bit integer comparisons. ORDER IR. */
+    as->curins--;  /* Always skip the loword comparison. */
+    asm_comp64(as, ir);
+    return;
+  }
+  if (!usehi) return;  /* Skip unused hiword op for all remaining ops. */
+  switch ((ir-1)->o) {
+  case IR_ADD: as->curins--; asm_add64(as, ir); break;
+  case IR_SUB: as->curins--; asm_sub64(as, ir); break;
+  case IR_NEG: as->curins--; asm_neg64(as, ir); break;
+  case IR_CALLN:
+  case IR_CALLXS:
+    if (!uselo)
+      ra_allocref(as, ir->op1, RID2RSET(RID_RETLO));  /* Mark lo op as used. */
+    break;
+  case IR_CNEWI:
+    /* Nothing to do here. Handled by lo op itself. */
+    break;
+  default: lua_assert(0); break;
+  }
+#else
+  UNUSED(as); UNUSED(ir); lua_assert(0);  /* Unused without FFI. */
+#endif
+}
+
+/* -- Stack handling ------------------------------------------------------ */
+
+/* Check Lua stack size for overflow. Use exit handler as fallback. */
+static void asm_stack_check(ASMState *as, BCReg topslot,
+                            IRIns *irp, RegSet allow, ExitNo exitno)
+{
+  /* Try to get an unused temp. register, otherwise spill/restore RID_RET*. */
+  Reg tmp, pbase = irp ? (ra_hasreg(irp->r) ? irp->r : RID_TMP) : RID_BASE;
+  rset_clear(allow, pbase);
+  tmp = allow ? rset_pickbot(allow) :
+                (pbase == RID_RETHI ? RID_RETLO : RID_RETHI);
+  emit_condbranch(as, PPCI_BC, CC_LT, asm_exitstub_addr(as, exitno));
+  if (allow == RSET_EMPTY)  /* Restore temp. register. */
+    emit_tai(as, PPCI_LWZ, tmp, RID_SP, SPOFS_TMPW);
+  else
+    ra_modified(as, tmp);
+  emit_ai(as, PPCI_CMPLWI, RID_TMP, (int32_t)(8*topslot));
+  emit_tab(as, PPCI_SUBF, RID_TMP, pbase, tmp);
+  emit_tai(as, PPCI_LWZ, tmp, tmp, offsetof(lua_State, maxstack));
+  if (pbase == RID_TMP)
+    emit_getgl(as, RID_TMP, jit_base);
+  emit_getgl(as, tmp, jit_L);
+  if (allow == RSET_EMPTY)  /* Spill temp. register. */
+    emit_tai(as, PPCI_STW, tmp, RID_SP, SPOFS_TMPW);
+}
+
+/* Restore Lua stack from on-trace state. */
+static void asm_stack_restore(ASMState *as, SnapShot *snap)
+{
+  SnapEntry *map = &as->T->snapmap[snap->mapofs];
+  MSize n, nent = snap->nent;
+  SnapEntry *flinks = map + nent + snap->depth;
+  /* Store the value of all modified slots to the Lua stack. */
+  for (n = 0; n < nent; n++) {
+    SnapEntry sn = map[n];
+    BCReg s = snap_slot(sn);
+    int32_t ofs = 8*((int32_t)s-1);
+    IRRef ref = snap_ref(sn);
+    IRIns *ir = IR(ref);
+    if ((sn & SNAP_NORESTORE))
+      continue;
+    if (irt_isnum(ir->t)) {
+      Reg src = ra_alloc1(as, ref, RSET_FPR);
+      emit_fai(as, PPCI_STFD, src, RID_BASE, ofs);
+    } else {
+      Reg type;
+      RegSet allow = rset_exclude(RSET_GPR, RID_BASE);
+      lua_assert(irt_ispri(ir->t) || irt_isaddr(ir->t) || irt_isinteger(ir->t));
+      if (!irt_ispri(ir->t)) {
+        Reg src = ra_alloc1(as, ref, allow);
+        rset_clear(allow, src);
+        emit_tai(as, PPCI_STW, src, RID_BASE, ofs+4);
+      }
+      if ((sn & (SNAP_CONT|SNAP_FRAME))) {
+        if (s == 0) continue;  /* Do not overwrite link to previous frame. */
+        type = ra_allock(as, (int32_t)(*flinks--), allow);
+      } else {
+        type = ra_allock(as, (int32_t)irt_toitype(ir->t), allow);
+      }
+      emit_tai(as, PPCI_STW, type, RID_BASE, ofs);
+    }
+    checkmclim(as);
+  }
+  lua_assert(map + nent == flinks);
+}
+
+/* -- GC handling --------------------------------------------------------- */
+
+/* Check GC threshold and do one or more GC steps. */
+static void asm_gc_check(ASMState *as)
+{
+  const CCallInfo *ci = &lj_ir_callinfo[IRCALL_lj_gc_step_jit];
+  IRRef args[2];
+  MCLabel l_end;
+  Reg tmp;
+  ra_evictset(as, RSET_SCRATCH);
+  l_end = emit_label(as);
+  /* Exit trace if in GCSatomic or GCSfinalize. Avoids syncing GC objects. */
+  asm_guardcc(as, CC_NE);  /* Assumes asm_snap_prep() already done. */
+  emit_ai(as, PPCI_CMPWI, RID_RET, 0);
+  args[0] = ASMREF_TMP1;  /* global_State *g */
+  args[1] = ASMREF_TMP2;  /* MSize steps     */
+  asm_gencall(as, ci, args);
+  emit_tai(as, PPCI_ADDI, ra_releasetmp(as, ASMREF_TMP1), RID_JGL, -32768);
+  tmp = ra_releasetmp(as, ASMREF_TMP2);
+  emit_loadi(as, tmp, (int32_t)as->gcsteps);
+  /* Jump around GC step if GC total < GC threshold. */
+  emit_condbranch(as, PPCI_BC|PPCF_Y, CC_LT, l_end);
+  emit_ab(as, PPCI_CMPLW, RID_TMP, tmp);
+  emit_getgl(as, tmp, gc.threshold);
+  emit_getgl(as, RID_TMP, gc.total);
+  as->gcsteps = 0;
+  checkmclim(as);
+}
+
+/* -- Loop handling ------------------------------------------------------- */
+
+/* Fixup the loop branch. */
+static void asm_loop_fixup(ASMState *as)
+{
+  MCode *p = as->mctop;
+  MCode *target = as->mcp;
+  if (as->loopinv) {  /* Inverted loop branch? */
+    /* asm_guardcc already inverted the cond branch and patched the final b. */
+    p[-2] = (p[-2] & (0xffff0000u & ~PPCF_Y)) | (((target-p+2) & 0x3fffu) << 2);
+  } else {
+    p[-1] = PPCI_B|(((target-p+1)&0x00ffffffu)<<2);
+  }
+}
+
+/* -- Head of trace ------------------------------------------------------- */
+
+/* Coalesce BASE register for a root trace. */
+static void asm_head_root_base(ASMState *as)
+{
+  IRIns *ir = IR(REF_BASE);
+  Reg r = ir->r;
+  if (ra_hasreg(r)) {
+    ra_free(as, r);
+    if (rset_test(as->modset, r))
+      ir->r = RID_INIT;  /* No inheritance for modified BASE register. */
+    if (r != RID_BASE)
+      emit_mr(as, r, RID_BASE);
+  }
+}
+
+/* Coalesce BASE register for a side trace. */
+static RegSet asm_head_side_base(ASMState *as, IRIns *irp, RegSet allow)
+{
+  IRIns *ir = IR(REF_BASE);
+  Reg r = ir->r;
+  if (ra_hasreg(r)) {
+    ra_free(as, r);
+    if (rset_test(as->modset, r))
+      ir->r = RID_INIT;  /* No inheritance for modified BASE register. */
+    if (irp->r == r) {
+      rset_clear(allow, r);  /* Mark same BASE register as coalesced. */
+    } else if (ra_hasreg(irp->r) && rset_test(as->freeset, irp->r)) {
+      rset_clear(allow, irp->r);
+      emit_mr(as, r, irp->r);  /* Move from coalesced parent reg. */
+    } else {
+      emit_getgl(as, r, jit_base);  /* Otherwise reload BASE. */
+    }
+  }
+  return allow;
+}
+
+/* -- Tail of trace ------------------------------------------------------- */
+
+/* Fixup the tail code. */
+static void asm_tail_fixup(ASMState *as, TraceNo lnk)
+{
+  MCode *p = as->mctop;
+  MCode *target;
+  int32_t spadj = as->T->spadjust;
+  if (spadj == 0) {
+    *--p = PPCI_NOP;
+    *--p = PPCI_NOP;
+    as->mctop = p;
+  } else {
+    /* Patch stack adjustment. */
+    lua_assert(checki16(CFRAME_SIZE+spadj));
+    p[-3] = PPCI_ADDI | PPCF_T(RID_TMP) | PPCF_A(RID_SP) | (CFRAME_SIZE+spadj);
+    p[-2] = PPCI_STWU | PPCF_T(RID_TMP) | PPCF_A(RID_SP) | spadj;
+  }
+  /* Patch exit branch. */
+  target = lnk ? traceref(as->J, lnk)->mcode : (MCode *)lj_vm_exit_interp;
+  p[-1] = PPCI_B|(((target-p+1)&0x00ffffffu)<<2);
+}
+
+/* Prepare tail of code. */
+static void asm_tail_prep(ASMState *as)
+{
+  MCode *p = as->mctop - 1;  /* Leave room for exit branch. */
+  if (as->loopref) {
+    as->invmcp = as->mcp = p;
+  } else {
+    as->mcp = p-2;  /* Leave room for stack pointer adjustment. */
+    as->invmcp = NULL;
+  }
+}
+
+/* -- Instruction dispatch ------------------------------------------------ */
+
+/* Assemble a single instruction. */
+static void asm_ir(ASMState *as, IRIns *ir)
+{
+  switch ((IROp)ir->o) {
+  /* Miscellaneous ops. */
+  case IR_LOOP: asm_loop(as); break;
+  case IR_NOP: case IR_XBAR: lua_assert(!ra_used(ir)); break;
+  case IR_USE:
+    ra_alloc1(as, ir->op1, irt_isfp(ir->t) ? RSET_FPR : RSET_GPR); break;
+  case IR_PHI: asm_phi(as, ir); break;
+  case IR_HIOP: asm_hiop(as, ir); break;
+
+  /* Guarded assertions. */
+  case IR_EQ: case IR_NE:
+    if ((ir-1)->o == IR_HREF && ir->op1 == as->curins-1) {
+      as->curins--;
+      asm_href(as, ir-1, (IROp)ir->o);
+      break;
+    }
+    /* fallthrough */
+  case IR_LT: case IR_GE: case IR_LE: case IR_GT:
+  case IR_ULT: case IR_UGE: case IR_ULE: case IR_UGT:
+  case IR_ABC:
+    asm_comp(as, ir);
+    break;
+
+  case IR_RETF: asm_retf(as, ir); break;
+
+  /* Bit ops. */
+  case IR_BNOT: asm_bitnot(as, ir); break;
+  case IR_BSWAP: asm_bitswap(as, ir); break;
+
+  case IR_BAND: asm_bitand(as, ir); break;
+  case IR_BOR:  asm_bitop(as, ir, PPCI_OR, PPCI_ORI); break;
+  case IR_BXOR: asm_bitop(as, ir, PPCI_XOR, PPCI_XORI); break;
+
+  case IR_BSHL: asm_bitshift(as, ir, PPCI_SLW, 0); break;
+  case IR_BSHR: asm_bitshift(as, ir, PPCI_SRW, 1); break;
+  case IR_BSAR: asm_bitshift(as, ir, PPCI_SRAW, PPCI_SRAWI); break;
+  case IR_BROL: asm_bitshift(as, ir, PPCI_RLWNM|PPCF_MB(0)|PPCF_ME(31),
+                             PPCI_RLWINM|PPCF_MB(0)|PPCF_ME(31)); break;
+  case IR_BROR: lua_assert(0); break;
+
+  /* Arithmetic ops. */
+  case IR_ADD: asm_add(as, ir); break;
+  case IR_SUB: asm_sub(as, ir); break;
+  case IR_MUL: asm_mul(as, ir); break;
+  case IR_DIV: asm_fparith(as, ir, PPCI_FDIV); break;
+  case IR_MOD: asm_callid(as, ir, IRCALL_lj_vm_modi); break;
+  case IR_POW: asm_callid(as, ir, IRCALL_lj_vm_powi); break;
+  case IR_NEG: asm_neg(as, ir); break;
+
+  case IR_ABS: asm_fpunary(as, ir, PPCI_FABS); break;
+  case IR_ATAN2: asm_callid(as, ir, IRCALL_atan2); break;
+  case IR_LDEXP: asm_callid(as, ir, IRCALL_ldexp); break;
+  case IR_MIN: asm_min_max(as, ir, 0); break;
+  case IR_MAX: asm_min_max(as, ir, 1); break;
+  case IR_FPMATH:
+    if (ir->op2 == IRFPM_EXP2 && asm_fpjoin_pow(as, ir))
+      break;
+    asm_callid(as, ir, IRCALL_lj_vm_floor + ir->op2);
+    break;
+
+  /* Overflow-checking arithmetic ops. */
+  case IR_ADDOV: asm_arithov(as, ir, PPCI_ADDO); break;
+  case IR_SUBOV: asm_arithov(as, ir, PPCI_SUBFO); break;
+  case IR_MULOV: asm_arithov(as, ir, PPCI_MULLWO); break;
+
+  /* Memory references. */
+  case IR_AREF: asm_aref(as, ir); break;
+  case IR_HREF: asm_href(as, ir, 0); break;
+  case IR_HREFK: asm_hrefk(as, ir); break;
+  case IR_NEWREF: asm_newref(as, ir); break;
+  case IR_UREFO: case IR_UREFC: asm_uref(as, ir); break;
+  case IR_FREF: asm_fref(as, ir); break;
+  case IR_STRREF: asm_strref(as, ir); break;
+
+  /* Loads and stores. */
+  case IR_ALOAD: case IR_HLOAD: case IR_ULOAD: case IR_VLOAD:
+    asm_ahuvload(as, ir);
+    break;
+  case IR_FLOAD: asm_fload(as, ir); break;
+  case IR_XLOAD: asm_xload(as, ir); break;
+  case IR_SLOAD: asm_sload(as, ir); break;
+
+  case IR_ASTORE: case IR_HSTORE: case IR_USTORE: asm_ahustore(as, ir); break;
+  case IR_FSTORE: asm_fstore(as, ir); break;
+  case IR_XSTORE: asm_xstore(as, ir); break;
+
+  /* Allocations. */
+  case IR_SNEW: case IR_XSNEW: asm_snew(as, ir); break;
+  case IR_TNEW: asm_tnew(as, ir); break;
+  case IR_TDUP: asm_tdup(as, ir); break;
+  case IR_CNEW: case IR_CNEWI: asm_cnew(as, ir); break;
+
+  /* Write barriers. */
+  case IR_TBAR: asm_tbar(as, ir); break;
+  case IR_OBAR: asm_obar(as, ir); break;
+
+  /* Type conversions. */
+  case IR_CONV: asm_conv(as, ir); break;
+  case IR_TOBIT: asm_tobit(as, ir); break;
+  case IR_TOSTR: asm_tostr(as, ir); break;
+  case IR_STRTO: asm_strto(as, ir); break;
+
+  /* Calls. */
+  case IR_CALLN: case IR_CALLL: case IR_CALLS: asm_call(as, ir); break;
+  case IR_CALLXS: asm_callx(as, ir); break;
+  case IR_CARG: break;
+
+  default:
+    setintV(&as->J->errinfo, ir->o);
+    lj_trace_err_info(as->J, LJ_TRERR_NYIIR);
+    break;
+  }
+}
+
+/* -- Trace setup --------------------------------------------------------- */
+
+/* Ensure there are enough stack slots for call arguments. */
+static Reg asm_setup_call_slots(ASMState *as, IRIns *ir, const CCallInfo *ci)
+{
+  IRRef args[CCI_NARGS_MAX];
+  uint32_t i, nargs = (int)CCI_NARGS(ci);
+  int nslots = 2, ngpr = REGARG_NUMGPR, nfpr = REGARG_NUMFPR;
+  asm_collectargs(as, ir, ci, args);
+  for (i = 0; i < nargs; i++)
+    if (args[i] && irt_isfp(IR(args[i])->t)) {
+      if (nfpr > 0) nfpr--; else nslots = (nslots+3) & ~1;
+    } else {
+      if (ngpr > 0) ngpr--; else nslots++;
+    }
+  if (nslots > as->evenspill)  /* Leave room for args in stack slots. */
+    as->evenspill = nslots;
+  return irt_isfp(ir->t) ? REGSP_HINT(RID_FPRET) : REGSP_HINT(RID_RET);
+}
+
+static void asm_setup_target(ASMState *as)
+{
+  asm_exitstub_setup(as, as->T->nsnap + (as->parent ? 1 : 0));
+}
+
+/* -- Trace patching ------------------------------------------------------ */
+
+/* Patch exit jumps of existing machine code to a new target. */
+void lj_asm_patchexit(jit_State *J, GCtrace *T, ExitNo exitno, MCode *target)
+{
+  MCode *p = T->mcode;
+  MCode *pe = (MCode *)((char *)p + T->szmcode);
+  MCode *px = exitstub_trace_addr(T, exitno);
+  MCode *cstart = NULL;
+  MCode *mcarea = lj_mcode_patch(J, p, 0);
+  int clearso = 0;
+  for (; p < pe; p++) {
+    /* Look for exitstub branch, try to replace with branch to target. */
+    uint32_t ins = *p;
+    if ((ins & 0xfc000000u) == 0x40000000u &&
+        ((ins ^ ((char *)px-(char *)p)) & 0xffffu) == 0) {
+      ptrdiff_t delta = (char *)target - (char *)p;
+      if (((ins >> 16) & 3) == (CC_SO&3)) {
+        clearso = sizeof(MCode);
+        delta -= sizeof(MCode);
+      }
+      /* Many, but not all short-range branches can be patched directly. */
+      if (((delta + 0x8000) >> 16) == 0) {
+        *p = (ins & 0xffdf0000u) | ((uint32_t)delta & 0xffffu) |
+             ((delta & 0x8000) * (PPCF_Y/0x8000));
+        if (!cstart) cstart = p;
+      }
+    } else if ((ins & 0xfc000000u) == PPCI_B &&
+               ((ins ^ ((char *)px-(char *)p)) & 0x03ffffffu) == 0) {
+      ptrdiff_t delta = (char *)target - (char *)p;
+      lua_assert(((delta + 0x02000000) >> 26) == 0);
+      *p = PPCI_B | ((uint32_t)delta & 0x03ffffffu);
+      if (!cstart) cstart = p;
+    }
+  }
+  {  /* Always patch long-range branch in exit stub itself. */
+    ptrdiff_t delta = (char *)target - (char *)px - clearso;
+    lua_assert(((delta + 0x02000000) >> 26) == 0);
+    *px = PPCI_B | ((uint32_t)delta & 0x03ffffffu);
+  }
+  if (!cstart) cstart = px;
+  asm_cache_flush(cstart, px+1);
+  if (clearso) {  /* Extend the current trace. Ugly workaround. */
+    MCode *pp = J->cur.mcode;
+    J->cur.szmcode += sizeof(MCode);
+    *--pp = PPCI_MCRXR;  /* Clear SO flag. */
+    J->cur.mcode = pp;
+    asm_cache_flush(pp, pp+1);
+  }
+  lj_mcode_patch(J, mcarea, 1);
+}
+