{ $Id$ Copyright (c) 1998-2002 by Florian Klaempfl Type checking and register allocation for add nodes This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. **************************************************************************** } unit nadd; {$i fpcdefs.inc} { define addstringopt} {$ifdef callparatemp} {$undef addstringopt} {$endif} interface uses node; type taddnode = class(tbinopnode) constructor create(tt : tnodetype;l,r : tnode);override; function pass_1 : tnode;override; function det_resulttype:tnode;override; {$ifdef state_tracking} function track_state_pass(exec_known:boolean):boolean;override; {$endif} protected { override the following if you want to implement } { parts explicitely in the code generator (JM) } function first_addstring: tnode; virtual; function first_addset: tnode; virtual; { only implements "muln" nodes, the rest always has to be done in } { the code generator for performance reasons (JM) } function first_add64bitint: tnode; virtual; {$ifdef cpufpemu} { This routine calls internal runtime library helpers for all floating point arithmetic in the case where the emulation switches is on. Otherwise returns nil, and everything must be done in the code generation phase. } function first_addfloat : tnode; virtual; {$endif cpufpemu} end; taddnodeclass = class of taddnode; var { caddnode is used to create nodes of the add type } { the virtual constructor allows to assign } { another class type to caddnode => processor } { specific node types can be created } caddnode : taddnodeclass; implementation uses globtype,systems, cutils,verbose,globals,widestr, symconst,symtype,symdef,symsym,symtable,defutil,defcmp, cginfo,cgbase, htypechk,pass_1, nbas,nmat,ncnv,ncon,nset,nopt,ncal,ninl, {$ifdef state_tracking} nstate, {$endif} cpubase,cpuinfo; {***************************************************************************** TADDNODE *****************************************************************************} {$ifdef fpc} {$maxfpuregisters 0} {$endif fpc} constructor taddnode.create(tt : tnodetype;l,r : tnode); begin inherited create(tt,l,r); end; function taddnode.det_resulttype:tnode; var hp,t : tnode; lt,rt : tnodetype; rd,ld : tdef; htype : ttype; ot : tnodetype; concatstrings : boolean; resultset : Tconstset; i : longint; b : boolean; s1,s2 : pchar; ws1,ws2 : pcompilerwidestring; l1,l2 : longint; rv,lv : tconstexprint; rvd,lvd : bestreal; resultrealtype : ttype; {$ifdef state_tracking} factval : Tnode; change : boolean; {$endif} begin result:=nil; { first do the two subtrees } resulttypepass(left); resulttypepass(right); { both left and right need to be valid } set_varstate(left,true); set_varstate(right,true); if codegenerror then exit; { convert array constructors to sets, because there is no other operator possible for array constructors } if is_array_constructor(left.resulttype.def) then begin arrayconstructor_to_set(left); resulttypepass(left); end; if is_array_constructor(right.resulttype.def) then begin arrayconstructor_to_set(right); resulttypepass(right); end; { allow operator overloading } hp:=self; if isbinaryoverloaded(hp) then begin result:=hp; exit; end; { Stop checking when an error was found in the operator checking } if codegenerror then begin result:=cerrornode.create; exit; end; { Kylix allows enum+ordconstn in an enum declaration (blocktype is bt_type), we need to do the conversion here before the constant folding } if (m_delphi in aktmodeswitches) and (blocktype=bt_type) then begin if (left.resulttype.def.deftype=enumdef) and (right.resulttype.def.deftype=orddef) then begin { insert explicit typecast to s32bit } left:=ctypeconvnode.create_explicit(left,s32bittype); resulttypepass(left); end else if (left.resulttype.def.deftype=orddef) and (right.resulttype.def.deftype=enumdef) then begin { insert explicit typecast to s32bit } right:=ctypeconvnode.create_explicit(right,s32bittype); resulttypepass(right); end; end; { is one a real float, then both need to be floats, this need to be done before the constant folding so constant operation on a float and int are also handled } resultrealtype:=pbestrealtype^; if (right.resulttype.def.deftype=floatdef) or (left.resulttype.def.deftype=floatdef) then begin { when both floattypes are already equal then use that floattype for results } if (right.resulttype.def.deftype=floatdef) and (left.resulttype.def.deftype=floatdef) and (tfloatdef(left.resulttype.def).typ=tfloatdef(right.resulttype.def).typ) then resultrealtype:=left.resulttype { when there is a currency type then use currency, but only when currency is defined as float } else if (s64currencytype.def.deftype=floatdef) and (is_currency(right.resulttype.def) or is_currency(left.resulttype.def)) then begin resultrealtype:=s64currencytype; inserttypeconv(right,resultrealtype); inserttypeconv(left,resultrealtype); end else begin inserttypeconv(right,resultrealtype); inserttypeconv(left,resultrealtype); end; end; { if one operand is a widechar or a widestring, both operands } { are converted to widestring. This must be done before constant } { folding to allow char+widechar etc. } if is_widestring(right.resulttype.def) or is_widestring(left.resulttype.def) or is_widechar(right.resulttype.def) or is_widechar(left.resulttype.def) then begin inserttypeconv(right,cwidestringtype); inserttypeconv(left,cwidestringtype); end; { load easier access variables } rd:=right.resulttype.def; ld:=left.resulttype.def; rt:=right.nodetype; lt:=left.nodetype; if (nodetype = slashn) and (((rt = ordconstn) and (tordconstnode(right).value = 0)) or ((rt = realconstn) and (trealconstnode(right).value_real = 0.0))) then begin Message(parser_e_division_by_zero); case rt of ordconstn: tordconstnode(right).value := 1; realconstn: trealconstnode(right).value_real := 1.0; end; end; { both are int constants } if ((is_constintnode(left) and is_constintnode(right)) or (is_constboolnode(left) and is_constboolnode(right) and (nodetype in [slashn,ltn,lten,gtn,gten,equaln,unequaln,andn,xorn,orn])) or (is_constenumnode(left) and is_constenumnode(right) and (nodetype in [equaln,unequaln,ltn,lten,gtn,gten]))) or { support pointer arithmetics on constants (JM) } ((lt = pointerconstn) and is_constintnode(right) and (nodetype in [addn,subn])) or (((lt = pointerconstn) or (lt = niln)) and ((rt = pointerconstn) or (rt = niln)) and (nodetype in [ltn,lten,gtn,gten,equaln,unequaln,subn])) then begin { when comparing/substracting pointers, make sure they are } { of the same type (JM) } if (lt = pointerconstn) and (rt = pointerconstn) then begin if not(cs_extsyntax in aktmoduleswitches) and not(nodetype in [equaln,unequaln]) then CGMessage(type_e_mismatch) else if (nodetype <> subn) and is_voidpointer(rd) then inserttypeconv(right,left.resulttype) else if (nodetype <> subn) and is_voidpointer(ld) then inserttypeconv(left,right.resulttype) else if not(equal_defs(ld,rd)) then CGMessage2(type_e_incompatible_types,ld.typename,rd.typename); end else if (ld.deftype=enumdef) and (rd.deftype=enumdef) then begin if not(equal_defs(ld,rd)) then inserttypeconv(right,left.resulttype); end else if (lt=ordconstn) and (rt=ordconstn) then begin { make left const type the biggest (u32bit is bigger than s32bit for or,and,xor) } if (rd.size>ld.size) or ((torddef(rd).typ=u32bit) and (torddef(ld).typ=s32bit) and (nodetype in [orn,andn,xorn])) then inserttypeconv(left,right.resulttype); end; { load values } case lt of ordconstn: lv:=tordconstnode(left).value; pointerconstn: lv:=tpointerconstnode(left).value; niln: lv:=0; else internalerror(2002080202); end; case rt of ordconstn: rv:=tordconstnode(right).value; pointerconstn: rv:=tpointerconstnode(right).value; niln: rv:=0; else internalerror(2002080203); end; if (lt = pointerconstn) and (rt <> pointerconstn) then rv := rv * tpointerdef(left.resulttype.def).pointertype.def.size; if (rt = pointerconstn) and (lt <> pointerconstn) then lv := lv * tpointerdef(right.resulttype.def).pointertype.def.size; case nodetype of addn : if (lt <> pointerconstn) then t := genintconstnode(lv+rv) else t := cpointerconstnode.create(lv+rv,left.resulttype); subn : if (lt <> pointerconstn) or (rt = pointerconstn) then t := genintconstnode(lv-rv) else t := cpointerconstnode.create(lv-rv,left.resulttype); muln : t:=genintconstnode(lv*rv); xorn : t:=cordconstnode.create(lv xor rv,left.resulttype,true); orn : t:=cordconstnode.create(lv or rv,left.resulttype,true); andn : t:=cordconstnode.create(lv and rv,left.resulttype,true); ltn : t:=cordconstnode.create(ord(lvrv),booltype,true); gten : t:=cordconstnode.create(ord(lv>=rv),booltype,true); equaln : t:=cordconstnode.create(ord(lv=rv),booltype,true); unequaln : t:=cordconstnode.create(ord(lv<>rv),booltype,true); slashn : begin { int/int becomes a real } rvd:=rv; lvd:=lv; t:=crealconstnode.create(lvd/rvd,resultrealtype); end; else begin CGMessage(type_e_mismatch); t:=cnothingnode.create; end; end; result:=t; exit; end; { both real constants ? } if (lt=realconstn) and (rt=realconstn) then begin lvd:=trealconstnode(left).value_real; rvd:=trealconstnode(right).value_real; case nodetype of addn : t:=crealconstnode.create(lvd+rvd,resultrealtype); subn : t:=crealconstnode.create(lvd-rvd,resultrealtype); muln : t:=crealconstnode.create(lvd*rvd,resultrealtype); starstarn, caretn : begin if lvd<0 then begin Message(parser_e_invalid_float_operation); t:=crealconstnode.create(0,resultrealtype); end else if lvd=0 then t:=crealconstnode.create(1.0,resultrealtype) else t:=crealconstnode.create(exp(ln(lvd)*rvd),resultrealtype); end; slashn : t:=crealconstnode.create(lvd/rvd,resultrealtype); ltn : t:=cordconstnode.create(ord(lvdrvd),booltype,true); gten : t:=cordconstnode.create(ord(lvd>=rvd),booltype,true); equaln : t:=cordconstnode.create(ord(lvd=rvd),booltype,true); unequaln : t:=cordconstnode.create(ord(lvd<>rvd),booltype,true); else begin CGMessage(type_e_mismatch); t:=cnothingnode.create; end; end; result:=t; exit; end; { first, we handle widestrings, so we can check later for } { stringconstn only } { widechars are converted above to widestrings too } { this isn't veryy efficient, but I don't think } { that it does matter that much (FK) } if (lt=stringconstn) and (rt=stringconstn) and (tstringconstnode(left).st_type=st_widestring) and (tstringconstnode(right).st_type=st_widestring) then begin initwidestring(ws1); initwidestring(ws2); copywidestring(pcompilerwidestring(tstringconstnode(left).value_str),ws1); copywidestring(pcompilerwidestring(tstringconstnode(right).value_str),ws2); case nodetype of addn : begin concatwidestrings(ws1,ws2); t:=cstringconstnode.createwstr(ws1); end; ltn : t:=cordconstnode.create(byte(comparewidestrings(ws1,ws2)<0),booltype,true); lten : t:=cordconstnode.create(byte(comparewidestrings(ws1,ws2)<=0),booltype,true); gtn : t:=cordconstnode.create(byte(comparewidestrings(ws1,ws2)>0),booltype,true); gten : t:=cordconstnode.create(byte(comparewidestrings(ws1,ws2)>=0),booltype,true); equaln : t:=cordconstnode.create(byte(comparewidestrings(ws1,ws2)=0),booltype,true); unequaln : t:=cordconstnode.create(byte(comparewidestrings(ws1,ws2)<>0),booltype,true); else begin CGMessage(type_e_mismatch); t:=cnothingnode.create; end; end; donewidestring(ws1); donewidestring(ws2); result:=t; exit; end; { concating strings ? } concatstrings:=false; s1:=nil; s2:=nil; if (lt=ordconstn) and (rt=ordconstn) and is_char(ld) and is_char(rd) then begin s1:=strpnew(char(byte(tordconstnode(left).value))); s2:=strpnew(char(byte(tordconstnode(right).value))); l1:=1; l2:=1; concatstrings:=true; end else if (lt=stringconstn) and (rt=ordconstn) and is_char(rd) then begin s1:=tstringconstnode(left).getpcharcopy; l1:=tstringconstnode(left).len; s2:=strpnew(char(byte(tordconstnode(right).value))); l2:=1; concatstrings:=true; end else if (lt=ordconstn) and (rt=stringconstn) and is_char(ld) then begin s1:=strpnew(char(byte(tordconstnode(left).value))); l1:=1; s2:=tstringconstnode(right).getpcharcopy; l2:=tstringconstnode(right).len; concatstrings:=true; end else if (lt=stringconstn) and (rt=stringconstn) then begin s1:=tstringconstnode(left).getpcharcopy; l1:=tstringconstnode(left).len; s2:=tstringconstnode(right).getpcharcopy; l2:=tstringconstnode(right).len; concatstrings:=true; end; if concatstrings then begin case nodetype of addn : t:=cstringconstnode.createpchar(concatansistrings(s1,s2,l1,l2),l1+l2); ltn : t:=cordconstnode.create(byte(compareansistrings(s1,s2,l1,l2)<0),booltype,true); lten : t:=cordconstnode.create(byte(compareansistrings(s1,s2,l1,l2)<=0),booltype,true); gtn : t:=cordconstnode.create(byte(compareansistrings(s1,s2,l1,l2)>0),booltype,true); gten : t:=cordconstnode.create(byte(compareansistrings(s1,s2,l1,l2)>=0),booltype,true); equaln : t:=cordconstnode.create(byte(compareansistrings(s1,s2,l1,l2)=0),booltype,true); unequaln : t:=cordconstnode.create(byte(compareansistrings(s1,s2,l1,l2)<>0),booltype,true); end; ansistringdispose(s1,l1); ansistringdispose(s2,l2); result:=t; exit; end; { set constant evaluation } if (right.nodetype=setconstn) and not assigned(tsetconstnode(right).left) and (left.nodetype=setconstn) and not assigned(tsetconstnode(left).left) then begin { check if size adjusting is needed, only for left to right as the other way is checked in the typeconv } if (tsetdef(right.resulttype.def).settype=smallset) and (tsetdef(left.resulttype.def).settype<>smallset) then tsetdef(right.resulttype.def).changesettype(normset); { check base types } inserttypeconv(left,right.resulttype); if codegenerror then begin { recover by only returning the left part } result:=left; left:=nil; exit; end; {$ifdef oldset} case nodetype of addn : begin for i:=0 to 31 do resultset[i]:=tsetconstnode(right).value_set^[i] or tsetconstnode(left).value_set^[i]; t:=csetconstnode.create(@resultset,left.resulttype); end; muln : begin for i:=0 to 31 do resultset[i]:=tsetconstnode(right).value_set^[i] and tsetconstnode(left).value_set^[i]; t:=csetconstnode.create(@resultset,left.resulttype); end; subn : begin for i:=0 to 31 do resultset[i]:=tsetconstnode(left).value_set^[i] and not(tsetconstnode(right).value_set^[i]); t:=csetconstnode.create(@resultset,left.resulttype); end; symdifn : begin for i:=0 to 31 do resultset[i]:=tsetconstnode(left).value_set^[i] xor tsetconstnode(right).value_set^[i]; t:=csetconstnode.create(@resultset,left.resulttype); end; unequaln : begin b:=true; for i:=0 to 31 do if tsetconstnode(right).value_set^[i]=tsetconstnode(left).value_set^[i] then begin b:=false; break; end; t:=cordconstnode.create(ord(b),booltype,true); end; equaln : begin b:=true; for i:=0 to 31 do if tsetconstnode(right).value_set^[i]<>tsetconstnode(left).value_set^[i] then begin b:=false; break; end; t:=cordconstnode.create(ord(b),booltype,true); end; lten : begin b := true; for i := 0 to 31 Do if (tsetconstnode(right).value_set^[i] And tsetconstnode(left).value_set^[i]) <> tsetconstnode(left).value_set^[i] Then begin b := false; break end; t := cordconstnode.create(ord(b),booltype,true); end; gten : begin b := true; for i := 0 to 31 Do If (tsetconstnode(left).value_set^[i] And tsetconstnode(right).value_set^[i]) <> tsetconstnode(right).value_set^[i] Then begin b := false; break end; t := cordconstnode.create(ord(b),booltype,true); end; end; {$else} case nodetype of addn : begin resultset:=tsetconstnode(right).value_set^ + tsetconstnode(left).value_set^; t:=csetconstnode.create(@resultset,left.resulttype); end; muln : begin resultset:=tsetconstnode(right).value_set^ * tsetconstnode(left).value_set^; t:=csetconstnode.create(@resultset,left.resulttype); end; subn : begin resultset:=tsetconstnode(left).value_set^ - tsetconstnode(right).value_set^; t:=csetconstnode.create(@resultset,left.resulttype); end; symdifn : begin resultset:=tsetconstnode(right).value_set^ >< tsetconstnode(left).value_set^; t:=csetconstnode.create(@resultset,left.resulttype); end; unequaln : begin b:=tsetconstnode(right).value_set^ <> tsetconstnode(left).value_set^; t:=cordconstnode.create(byte(b),booltype,true); end; equaln : begin b:=tsetconstnode(right).value_set^ = tsetconstnode(left).value_set^; t:=cordconstnode.create(byte(b),booltype,true); end; lten : begin b:=tsetconstnode(left).value_set^ <= tsetconstnode(right).value_set^; t:=cordconstnode.create(byte(b),booltype,true); end; gten : begin b:=tsetconstnode(left).value_set^ >= tsetconstnode(right).value_set^; t:=cordconstnode.create(byte(b),booltype,true); end; end; {$endif} result:=t; exit; end; { but an int/int gives real/real! } if nodetype=slashn then begin if (left.resulttype.def.deftype <> floatdef) and (right.resulttype.def.deftype <> floatdef) then CGMessage(type_h_use_div_for_int); inserttypeconv(right,resultrealtype); inserttypeconv(left,resultrealtype); end { if both are orddefs then check sub types } else if (ld.deftype=orddef) and (rd.deftype=orddef) then begin { optimize multiplacation by a power of 2 } if not(cs_check_overflow in aktlocalswitches) and (nodetype = muln) and (((left.nodetype = ordconstn) and ispowerof2(tordconstnode(left).value,i)) or ((right.nodetype = ordconstn) and ispowerof2(tordconstnode(right).value,i))) then begin if left.nodetype = ordconstn then begin tordconstnode(left).value := i; result := cshlshrnode.create(shln,right,left); end else begin tordconstnode(right).value := i; result := cshlshrnode.create(shln,left,right); end; left := nil; right := nil; exit; end; { 2 booleans? Make them equal to the largest boolean } if is_boolean(ld) and is_boolean(rd) then begin if torddef(left.resulttype.def).size>torddef(right.resulttype.def).size then begin right:=ctypeconvnode.create_explicit(right,left.resulttype); ttypeconvnode(right).convtype:=tc_bool_2_int; resulttypepass(right); end else if torddef(left.resulttype.def).size0); ot:=nodetype; left.free; left:=nil; right:=nil; if (not(b) and (ot=equaln)) or (b and (ot=unequaln)) then begin hp:=cnotnode.create(hp); end; result:=hp; exit; end; if (right.nodetype=ordconstn) then begin hp:=left; b:=(tordconstnode(right).value<>0); ot:=nodetype; right.free; right:=nil; left:=nil; if (not(b) and (ot=equaln)) or (b and (ot=unequaln)) then begin hp:=cnotnode.create(hp); end; result:=hp; exit; end; end; end; else CGMessage(type_e_mismatch); end; end { Both are chars? } else if is_char(rd) and is_char(ld) then begin if nodetype=addn then begin resulttype:=cshortstringtype; if not(is_constcharnode(left) and is_constcharnode(right)) then begin inserttypeconv(left,cshortstringtype); {$ifdef addstringopt} hp := genaddsstringcharoptnode(self); result := hp; exit; {$endif addstringopt} end; end; end { is there a currency type ? } else if ((torddef(rd).typ=scurrency) or (torddef(ld).typ=scurrency)) then begin if (torddef(ld).typ<>scurrency) then inserttypeconv(left,s64currencytype); if (torddef(rd).typ<>scurrency) then inserttypeconv(right,s64currencytype); end { is there a signed 64 bit type ? } else if ((torddef(rd).typ=s64bit) or (torddef(ld).typ=s64bit)) then begin if (torddef(ld).typ<>s64bit) then inserttypeconv(left,cs64bittype); if (torddef(rd).typ<>s64bit) then inserttypeconv(right,cs64bittype); end { is there a unsigned 64 bit type ? } else if ((torddef(rd).typ=u64bit) or (torddef(ld).typ=u64bit)) then begin if (torddef(ld).typ<>u64bit) then inserttypeconv(left,cu64bittype); if (torddef(rd).typ<>u64bit) then inserttypeconv(right,cu64bittype); end { is there a cardinal? } else if ((torddef(rd).typ=u32bit) or (torddef(ld).typ=u32bit)) then begin if is_signed(ld) and { then rd = u32bit } (nodetype in [addn,subn,muln]) and { convert positive constants to u32bit } not(is_constintnode(left) and (tordconstnode(left).value >= 0)) then begin { perform the operation in 64bit } CGMessage(type_w_mixed_signed_unsigned); inserttypeconv(left,cs64bittype); inserttypeconv(right,cs64bittype); end else begin { and,or,xor work on bit patterns and don't care about the sign } if nodetype in [andn,orn,xorn] then inserttypeconv_explicit(left,u32bittype) else begin if is_signed(ld) and not(is_constintnode(left) and (tordconstnode(left).value >= 0)) then CGMessage(type_w_mixed_signed_unsigned2); inserttypeconv(left,u32bittype); end; if is_signed(rd) and { then ld = u32bit } (nodetype in [addn,subn,muln]) and { convert positive constants to u32bit } not(is_constintnode(right) and (tordconstnode(right).value >= 0)) then begin { perform the operation in 64bit } CGMessage(type_w_mixed_signed_unsigned); inserttypeconv(left,cs64bittype); inserttypeconv(right,cs64bittype); end else begin { and,or,xor work on bit patterns and don't care about the sign } if nodetype in [andn,orn,xorn] then inserttypeconv_explicit(left,u32bittype) else begin if is_signed(rd) and not(is_constintnode(right) and (tordconstnode(right).value >= 0)) then CGMessage(type_w_mixed_signed_unsigned2); inserttypeconv(right,u32bittype); end; end; end; end { generic ord conversion is s32bit } else begin { if the left or right value is smaller than the normal type s32bittype and is unsigned, and the other value is a constant < 0, the result will always be false/true for equal / unequal nodes. } if ( { left : unsigned ordinal var, right : < 0 constant } ( ((is_signed(ld)=false) and (is_constintnode(left) =false)) and ((is_constintnode(right)) and (tordconstnode(right).value < 0)) ) or { right : unsigned ordinal var, left : < 0 constant } ( ((is_signed(rd)=false) and (is_constintnode(right) =false)) and ((is_constintnode(left)) and (tordconstnode(left).value < 0)) ) ) then begin if nodetype = equaln then CGMessage(type_w_signed_unsigned_always_false) else if nodetype = unequaln then CGMessage(type_w_signed_unsigned_always_true) else if (is_constintnode(left) and (nodetype in [ltn,lten])) or (is_constintnode(right) and (nodetype in [gtn,gten])) then CGMessage(type_w_signed_unsigned_always_true) else if (is_constintnode(right) and (nodetype in [ltn,lten])) or (is_constintnode(left) and (nodetype in [gtn,gten])) then CGMessage(type_w_signed_unsigned_always_false); end; inserttypeconv(right,s32bittype); inserttypeconv(left,s32bittype); end; end { if both are floatdefs, conversion is already done before constant folding } else if (ld.deftype=floatdef) then begin { already converted } end { left side a setdef, must be before string processing, else array constructor can be seen as array of char (PFV) } else if (ld.deftype=setdef) then begin { trying to add a set element? } if (nodetype=addn) and (rd.deftype<>setdef) then begin if (rt=setelementn) then begin if not(equal_defs(tsetdef(ld).elementtype.def,rd)) then CGMessage(type_e_set_element_are_not_comp); end else CGMessage(type_e_mismatch) end else begin if not(nodetype in [addn,subn,symdifn,muln,equaln,unequaln,lten,gten]) then CGMessage(type_e_set_operation_unknown); { right def must be a also be set } if (rd.deftype<>setdef) or not(equal_defs(rd,ld)) then CGMessage(type_e_set_element_are_not_comp); end; { ranges require normsets } if (tsetdef(ld).settype=smallset) and (rt=setelementn) and assigned(tsetelementnode(right).right) then begin { generate a temporary normset def, it'll be destroyed when the symtable is unloaded } htype.setdef(tsetdef.create(tsetdef(ld).elementtype,255)); inserttypeconv(left,htype); end; { if the right side is also a setdef then the settype must be the same as the left setdef } if (rd.deftype=setdef) and (tsetdef(ld).settype<>tsetdef(rd).settype) then begin { when right is a normset we need to typecast both to normsets } if (tsetdef(rd).settype=normset) then inserttypeconv(left,right.resulttype) else inserttypeconv(right,left.resulttype); end; end { compare pchar to char arrays by addresses like BP/Delphi } else if ((is_pchar(ld) or (lt=niln)) and is_chararray(rd)) or ((is_pchar(rd) or (rt=niln)) and is_chararray(ld)) then begin if is_chararray(rd) then inserttypeconv(right,charpointertype) else inserttypeconv(left,charpointertype); end { pointer comparision and subtraction } else if (rd.deftype=pointerdef) and (ld.deftype=pointerdef) then begin case nodetype of equaln,unequaln : begin if is_voidpointer(right.resulttype.def) then inserttypeconv(right,left.resulttype) else if is_voidpointer(left.resulttype.def) then inserttypeconv(left,right.resulttype) else if not(equal_defs(ld,rd)) then CGMessage2(type_e_incompatible_types,ld.typename,rd.typename); end; ltn,lten,gtn,gten: begin if (cs_extsyntax in aktmoduleswitches) then begin if is_voidpointer(right.resulttype.def) then inserttypeconv(right,left.resulttype) else if is_voidpointer(left.resulttype.def) then inserttypeconv(left,right.resulttype) else if not(equal_defs(ld,rd)) then CGMessage2(type_e_incompatible_types,ld.typename,rd.typename); end else CGMessage(type_e_mismatch); end; subn: begin if (cs_extsyntax in aktmoduleswitches) then begin if is_voidpointer(right.resulttype.def) then inserttypeconv(right,left.resulttype) else if is_voidpointer(left.resulttype.def) then inserttypeconv(left,right.resulttype) else if not(equal_defs(ld,rd)) then CGMessage2(type_e_incompatible_types,ld.typename,rd.typename); end else CGMessage(type_e_mismatch); resulttype:=s32bittype; exit; end; addn: begin if (cs_extsyntax in aktmoduleswitches) then begin if is_voidpointer(right.resulttype.def) then inserttypeconv(right,left.resulttype) else if is_voidpointer(left.resulttype.def) then inserttypeconv(left,right.resulttype) else if not(equal_defs(ld,rd)) then CGMessage2(type_e_incompatible_types,ld.typename,rd.typename); end else CGMessage(type_e_mismatch); resulttype:=s32bittype; exit; end; else CGMessage(type_e_mismatch); end; end { is one of the operands a string?, chararrays are also handled as strings (after conversion), also take care of chararray+chararray and chararray+char. Note: Must be done after pointerdef+pointerdef has been checked, else pchar is converted to string } else if (rd.deftype=stringdef) or (ld.deftype=stringdef) or ((is_pchar(rd) or is_chararray(rd) or is_char(rd)) and (is_pchar(ld) or is_chararray(ld) or is_char(ld))) then begin if is_widestring(rd) or is_widestring(ld) then begin if not(is_widestring(rd)) then inserttypeconv(right,cwidestringtype); if not(is_widestring(ld)) then inserttypeconv(left,cwidestringtype); end else if is_ansistring(rd) or is_ansistring(ld) then begin if not(is_ansistring(rd)) then inserttypeconv(right,cansistringtype); if not(is_ansistring(ld)) then inserttypeconv(left,cansistringtype); end else if is_longstring(rd) or is_longstring(ld) then begin if not(is_longstring(rd)) then inserttypeconv(right,clongstringtype); if not(is_longstring(ld)) then inserttypeconv(left,clongstringtype); end else begin if not(is_shortstring(ld)) then inserttypeconv(left,cshortstringtype); { don't convert char, that can be handled by the optimized node } if not(is_shortstring(rd) or is_char(rd)) then inserttypeconv(right,cshortstringtype); end; end { class or interface equation } else if is_class_or_interface(rd) or is_class_or_interface(ld) then begin if is_class_or_interface(rd) and is_class_or_interface(ld) then begin if tobjectdef(rd).is_related(tobjectdef(ld)) then inserttypeconv(right,left.resulttype) else inserttypeconv(left,right.resulttype); end else if is_class_or_interface(rd) then inserttypeconv(left,right.resulttype) else inserttypeconv(right,left.resulttype); if not(nodetype in [equaln,unequaln]) then CGMessage(type_e_mismatch); end else if (rd.deftype=classrefdef) and (ld.deftype=classrefdef) then begin if tobjectdef(tclassrefdef(rd).pointertype.def).is_related( tobjectdef(tclassrefdef(ld).pointertype.def)) then inserttypeconv(right,left.resulttype) else inserttypeconv(left,right.resulttype); if not(nodetype in [equaln,unequaln]) then CGMessage(type_e_mismatch); end { allows comperasion with nil pointer } else if is_class_or_interface(rd) or (rd.deftype=classrefdef) then begin inserttypeconv(left,right.resulttype); if not(nodetype in [equaln,unequaln]) then CGMessage(type_e_mismatch); end else if is_class_or_interface(ld) or (ld.deftype=classrefdef) then begin inserttypeconv(right,left.resulttype); if not(nodetype in [equaln,unequaln]) then CGMessage(type_e_mismatch); end { support procvar=nil,procvar<>nil } else if ((ld.deftype=procvardef) and (rt=niln)) or ((rd.deftype=procvardef) and (lt=niln)) then begin if not(nodetype in [equaln,unequaln]) then CGMessage(type_e_mismatch); end { support dynamicarray=nil,dynamicarray<>nil } else if (is_dynamic_array(ld) and (rt=niln)) or (is_dynamic_array(rd) and (lt=niln)) then begin if not(nodetype in [equaln,unequaln]) then CGMessage(type_e_mismatch); end {$ifdef SUPPORT_MMX} { mmx support, this must be before the zero based array check } else if (cs_mmx in aktlocalswitches) and is_mmx_able_array(ld) and is_mmx_able_array(rd) and equal_defs(ld,rd) then begin case nodetype of addn,subn,xorn,orn,andn: ; { mul is a little bit restricted } muln: if not(mmx_type(ld) in [mmxu16bit,mmxs16bit,mmxfixed16]) then CGMessage(type_e_mismatch); else CGMessage(type_e_mismatch); end; end {$endif SUPPORT_MMX} { this is a little bit dangerous, also the left type } { pointer to should be checked! This broke the mmx support } else if (rd.deftype=pointerdef) or is_zero_based_array(rd) then begin if is_zero_based_array(rd) then begin resulttype.setdef(tpointerdef.create(tarraydef(rd).elementtype)); inserttypeconv(right,resulttype); end; inserttypeconv(left,s32bittype); if nodetype=addn then begin if not(cs_extsyntax in aktmoduleswitches) or (not(is_pchar(ld)) and not(m_add_pointer in aktmodeswitches)) then CGMessage(type_e_mismatch); if (rd.deftype=pointerdef) and (tpointerdef(rd).pointertype.def.size>1) then left:=caddnode.create(muln,left, cordconstnode.create(tpointerdef(rd).pointertype.def.size,s32bittype,true)); end else CGMessage(type_e_mismatch); end else if (ld.deftype=pointerdef) or is_zero_based_array(ld) then begin if is_zero_based_array(ld) then begin resulttype.setdef(tpointerdef.create(tarraydef(ld).elementtype)); inserttypeconv(left,resulttype); end; inserttypeconv(right,s32bittype); if nodetype in [addn,subn] then begin if not(cs_extsyntax in aktmoduleswitches) or (not(is_pchar(ld)) and not(m_add_pointer in aktmodeswitches)) then CGMessage(type_e_mismatch); if (ld.deftype=pointerdef) and (tpointerdef(ld).pointertype.def.size>1) then right:=caddnode.create(muln,right, cordconstnode.create(tpointerdef(ld).pointertype.def.size,s32bittype,true)); end else CGMessage(type_e_mismatch); end else if (rd.deftype=procvardef) and (ld.deftype=procvardef) and equal_defs(rd,ld) then begin if not (nodetype in [equaln,unequaln]) then CGMessage(type_e_mismatch); end { enums } else if (ld.deftype=enumdef) and (rd.deftype=enumdef) then begin if not(equal_defs(ld,rd)) then inserttypeconv(right,left.resulttype); if not(nodetype in [equaln,unequaln,ltn,lten,gtn,gten]) then CGMessage(type_e_mismatch); end { generic conversion, this is for error recovery } else begin inserttypeconv(left,s32bittype); inserttypeconv(right,s32bittype); end; { set resulttype if not already done } if not assigned(resulttype.def) then begin case nodetype of ltn,lten,gtn,gten,equaln,unequaln : resulttype:=booltype; slashn : resulttype:=resultrealtype; addn: begin { for strings, return is always a 255 char string } if is_shortstring(left.resulttype.def) then resulttype:=cshortstringtype else resulttype:=left.resulttype; end; else resulttype:=left.resulttype; end; end; { when the result is currency we need some extra code for multiplication and division. this should not be done when the muln or slashn node is created internally } if not(nf_is_currency in flags) and is_currency(resulttype.def) then begin case nodetype of slashn : begin { slashn will only work with floats } hp:=caddnode.create(muln,getcopy,crealconstnode.create(10000.0,s64currencytype)); include(hp.flags,nf_is_currency); result:=hp; end; muln : begin if s64currencytype.def.deftype=floatdef then hp:=caddnode.create(slashn,getcopy,crealconstnode.create(10000.0,s64currencytype)) else hp:=cmoddivnode.create(divn,getcopy,cordconstnode.create(10000,s64currencytype,false)); include(hp.flags,nf_is_currency); result:=hp end; end; end; end; function taddnode.first_addstring: tnode; var p: tnode; begin { when we get here, we are sure that both the left and the right } { node are both strings of the same stringtype (JM) } case nodetype of addn: begin { create the call to the concat routine both strings as arguments } result := ccallnode.createintern('fpc_'+ tstringdef(resulttype.def).stringtypname+'_concat', ccallparanode.create(right,ccallparanode.create(left,nil))); { we reused the arguments } left := nil; right := nil; end; ltn,lten,gtn,gten,equaln,unequaln : begin { generate better code for s='' and s<>'' } if (nodetype in [equaln,unequaln]) and (((left.nodetype=stringconstn) and (str_length(left)=0)) or ((right.nodetype=stringconstn) and (str_length(right)=0))) then begin { switch so that the constant is always on the right } if left.nodetype = stringconstn then begin p := left; left := right; right := p; end; if is_shortstring(left.resulttype.def) then { compare the length with 0 } result := caddnode.create(nodetype, cinlinenode.create(in_length_x,false,left), cordconstnode.create(0,s32bittype,false)) else begin { compare the pointer with nil (for ansistrings etc), } { faster than getting the length (JM) } result:= caddnode.create(nodetype, ctypeconvnode.create_explicit(left,voidpointertype), cpointerconstnode.create(0,voidpointertype)); end; { left is reused } left := nil; { right isn't } right.free; right := nil; exit; end; { no string constant -> call compare routine } result := ccallnode.createintern('fpc_'+ tstringdef(left.resulttype.def).stringtypname+'_compare', ccallparanode.create(right,ccallparanode.create(left,nil))); { and compare its result with 0 according to the original operator } result := caddnode.create(nodetype,result, cordconstnode.create(0,s32bittype,false)); left := nil; right := nil; end; end; end; function taddnode.first_addset: tnode; var procname: string[31]; tempn: tnode; paras: tcallparanode; srsym: ttypesym; begin { get the sym that represents the fpc_normal_set type } if not searchsystype('FPC_NORMAL_SET',srsym) then internalerror(200108313); case nodetype of equaln,unequaln,lten,gten: begin case nodetype of equaln,unequaln: procname := 'fpc_set_comp_sets'; lten,gten: begin procname := 'fpc_set_contains_sets'; { (left >= right) = (right <= left) } if nodetype = gten then begin tempn := left; left := right; right := tempn; end; end; end; { convert the arguments (explicitely) to fpc_normal_set's } left := ctypeconvnode.create_explicit(left,srsym.restype); right := ctypeconvnode.create_explicit(right,srsym.restype); result := ccallnode.createintern(procname,ccallparanode.create(right, ccallparanode.create(left,nil))); { left and right are reused as parameters } left := nil; right := nil; { for an unequaln, we have to negate the result of comp_sets } if nodetype = unequaln then result := cnotnode.create(result); end; addn: begin { optimize first loading of a set } if (right.nodetype=setelementn) and not(assigned(tsetelementnode(right).right)) and is_emptyset(left) then begin { type cast the value to pass as argument to a byte, } { since that's what the helper expects } tsetelementnode(right).left := ctypeconvnode.create_explicit(tsetelementnode(right).left,u8bittype); { set the resulttype to the actual one (otherwise it's } { "fpc_normal_set") } result := ccallnode.createinternres('fpc_set_create_element', ccallparanode.create(tsetelementnode(right).left,nil), resulttype); { reused } tsetelementnode(right).left := nil; end else begin if right.nodetype=setelementn then begin { convert the arguments to bytes, since that's what } { the helper expects } tsetelementnode(right).left := ctypeconvnode.create_explicit(tsetelementnode(right).left, u8bittype); { convert the original set (explicitely) to an } { fpc_normal_set so we can pass it to the helper } left := ctypeconvnode.create_explicit(left,srsym.restype); { add a range or a single element? } if assigned(tsetelementnode(right).right) then begin tsetelementnode(right).right := ctypeconvnode.create_explicit(tsetelementnode(right).right, u8bittype); { create the call } result := ccallnode.createinternres('fpc_set_set_range', ccallparanode.create(tsetelementnode(right).right, ccallparanode.create(tsetelementnode(right).left, ccallparanode.create(left,nil))),resulttype); end else begin result := ccallnode.createinternres('fpc_set_set_byte', ccallparanode.create(tsetelementnode(right).left, ccallparanode.create(left,nil)),resulttype); end; { remove reused parts from original node } tsetelementnode(right).right := nil; tsetelementnode(right).left := nil; left := nil; end else begin { add two sets } { convert the sets to fpc_normal_set's } result := ccallnode.createinternres('fpc_set_add_sets', ccallparanode.create( ctypeconvnode.create_explicit(right,srsym.restype), ccallparanode.create( ctypeconvnode.create_explicit(left,srsym.restype),nil)),resulttype); { remove reused parts from original node } left := nil; right := nil; end; end end; subn,symdifn,muln: begin { convert the sets to fpc_normal_set's } paras := ccallparanode.create(ctypeconvnode.create_explicit(right,srsym.restype), ccallparanode.create(ctypeconvnode.create_explicit(left,srsym.restype),nil)); case nodetype of subn: result := ccallnode.createinternres('fpc_set_sub_sets', paras,resulttype); symdifn: result := ccallnode.createinternres('fpc_set_symdif_sets', paras,resulttype); muln: result := ccallnode.createinternres('fpc_set_mul_sets', paras,resulttype); end; { remove reused parts from original node } left := nil; right := nil; end; else internalerror(200108311); end; end; function taddnode.first_add64bitint: tnode; var procname: string[31]; temp: tnode; power: longint; begin result := nil; { create helper calls mul } if nodetype <> muln then exit; { make sure that if there is a constant, that it's on the right } if left.nodetype = ordconstn then begin temp := right; right := left; left := temp; end; { can we use a shift instead of a mul? } if (right.nodetype = ordconstn) and ispowerof2(tordconstnode(right).value,power) then begin tordconstnode(right).value := power; result := cshlshrnode.create(shln,left,right); { left and right are reused } left := nil; right := nil; { return firstpassed new node } exit; end; { when currency is used set the result of the parameters to s64bit, so they are not converted } if is_currency(resulttype.def) then begin left.resulttype:=cs64bittype; right.resulttype:=cs64bittype; end; { otherwise, create the parameters for the helper } right := ccallparanode.create( cordconstnode.create(ord(cs_check_overflow in aktlocalswitches),booltype,true), ccallparanode.create(right,ccallparanode.create(left,nil))); left := nil; { only qword needs the unsigned code, the signed code is also used for currency } if is_signed(resulttype.def) then procname := 'fpc_mul_int64' else procname := 'fpc_mul_qword'; result := ccallnode.createintern(procname,right); right := nil; end; {$ifdef cpufpemu} function taddnode.first_addfloat: tnode; var procname: string[31]; temp: tnode; power: longint; { do we need to reverse the result ? } notnode : boolean; begin result := nil; notnode := false; { In non-emulation mode, real opcodes are emitted for floating point values. } if not (cs_fp_emulation in aktmoduleswitches) then exit; case nodetype of addn : procname := 'fpc_single_add'; muln : procname := 'fpc_single_mul'; subn : procname := 'fpc_single_sub'; slashn : procname := 'fpc_single_div'; ltn : procname := 'fpc_single_lt'; lten: procname := 'fpc_single_le'; gtn: begin procname := 'fpc_single_le'; notnode := true; end; gten: begin procname := 'fpc_single_lt'; notnode := true; end; equaln: procname := 'fpc_single_eq'; unequaln : begin procname := 'fpc_single_eq'; notnode := true; end; else CGMessage(type_e_mismatch); end; { convert the arguments (explicitely) to fpc_normal_set's } result := ccallnode.createintern(procname,ccallparanode.create(right, ccallparanode.create(left,nil))); left:=nil; right:=nil; { do we need to reverse the result } if notnode then result := cnotnode.create(result); end; {$endif cpufpemu} function taddnode.pass_1 : tnode; var hp : tnode; lt,rt : tnodetype; rd,ld : tdef; begin result:=nil; { first do the two subtrees } firstpass(left); firstpass(right); if codegenerror then exit; { load easier access variables } rd:=right.resulttype.def; ld:=left.resulttype.def; rt:=right.nodetype; lt:=left.nodetype; { int/int gives real/real! } if nodetype=slashn then begin {$ifdef cpufpemu} result := first_addfloat; if assigned(result) then exit; {$endif cpufpemu} expectloc:=LOC_FPUREGISTER; { maybe we need an integer register to save } { a reference } if ((left.expectloc<>LOC_FPUREGISTER) or (right.expectloc<>LOC_FPUREGISTER)) and (left.registers32=right.registers32) then calcregisters(self,1,1,0) else calcregisters(self,0,1,0); { an add node always first loads both the left and the } { right in the fpu before doing the calculation. However, } { calcregisters(0,2,0) will overestimate the number of } { necessary registers (it will make it 3 in case one of } { the operands is already in the fpu) (JM) } if ((left.expectloc<>LOC_FPUREGISTER) or (right.expectloc<>LOC_FPUREGISTER)) and (registersfpu < 2) then inc(registersfpu); end { if both are orddefs then check sub types } else if (ld.deftype=orddef) and (rd.deftype=orddef) then begin { 2 booleans ? } if is_boolean(ld) and is_boolean(rd) then begin if not(cs_full_boolean_eval in aktlocalswitches) and (nodetype in [andn,orn]) then begin expectloc:=LOC_JUMP; calcregisters(self,0,0,0); end else begin expectloc:=LOC_FLAGS; if (left.expectloc in [LOC_JUMP,LOC_FLAGS]) and (left.expectloc in [LOC_JUMP,LOC_FLAGS]) then calcregisters(self,2,0,0) else calcregisters(self,1,0,0); end; end else { Both are chars? only convert to shortstrings for addn } if is_char(ld) then begin if nodetype=addn then internalerror(200103291); expectloc:=LOC_FLAGS; calcregisters(self,1,0,0); end { is there a 64 bit type ? } else if (torddef(ld).typ in [s64bit,u64bit,scurrency]) then begin result := first_add64bitint; if assigned(result) then exit; if nodetype in [addn,subn,muln,andn,orn,xorn] then expectloc:=LOC_REGISTER else expectloc:=LOC_JUMP; calcregisters(self,2,0,0) end { is there a cardinal? } else if (torddef(ld).typ=u32bit) then begin if nodetype in [addn,subn,muln,andn,orn,xorn] then expectloc:=LOC_REGISTER else expectloc:=LOC_FLAGS; calcregisters(self,1,0,0); { for unsigned mul we need an extra register } if nodetype=muln then inc(registers32); end { generic s32bit conversion } else begin if nodetype in [addn,subn,muln,andn,orn,xorn] then expectloc:=LOC_REGISTER else expectloc:=LOC_FLAGS; calcregisters(self,1,0,0); end; end { left side a setdef, must be before string processing, else array constructor can be seen as array of char (PFV) } else if (ld.deftype=setdef) then begin if tsetdef(ld).settype=smallset then begin if nodetype in [ltn,lten,gtn,gten,equaln,unequaln] then expectloc:=LOC_FLAGS else expectloc:=LOC_REGISTER; { are we adding set elements ? } if right.nodetype=setelementn then calcregisters(self,2,0,0) else calcregisters(self,1,0,0); end else begin result := first_addset; if assigned(result) then exit; expectloc:=LOC_CREFERENCE; calcregisters(self,0,0,0); { here we call SET... } include(current_procinfo.flags,pi_do_call); end; end { compare pchar by addresses like BP/Delphi } else if is_pchar(ld) then begin if nodetype in [addn,subn,muln,andn,orn,xorn] then expectloc:=LOC_REGISTER else expectloc:=LOC_FLAGS; calcregisters(self,1,0,0); end { is one of the operands a string } else if (ld.deftype=stringdef) then begin if is_widestring(ld) then begin { this is only for add, the comparisaion is handled later } expectloc:=LOC_REGISTER; end else if is_ansistring(ld) then begin { this is only for add, the comparisaion is handled later } expectloc:=LOC_REGISTER; end else if is_longstring(ld) then begin { this is only for add, the comparisaion is handled later } expectloc:=LOC_REFERENCE; end else begin {$ifdef addstringopt} { can create a call which isn't handled by callparatemp } if canbeaddsstringcharoptnode(self) then begin hp := genaddsstringcharoptnode(self); pass_1 := hp; exit; end else {$endif addstringopt} begin { Fix right to be shortstring } if is_char(right.resulttype.def) then begin inserttypeconv(right,cshortstringtype); firstpass(right); end; end; {$ifdef addstringopt} { can create a call which isn't handled by callparatemp } if canbeaddsstringcsstringoptnode(self) then begin hp := genaddsstringcsstringoptnode(self); pass_1 := hp; exit; end; {$endif addstringopt} end; { otherwise, let addstring convert everything } result := first_addstring; exit; end { is one a real float ? } else if (rd.deftype=floatdef) or (ld.deftype=floatdef) then begin {$ifdef cpufpemu} result := first_addfloat; if assigned(result) then exit; {$endif cpufpemu} if nodetype in [addn,subn,muln,andn,orn,xorn] then expectloc:=LOC_FPUREGISTER else expectloc:=LOC_FLAGS; calcregisters(self,0,1,0); { an add node always first loads both the left and the } { right in the fpu before doing the calculation. However, } { calcregisters(0,2,0) will overestimate the number of } { necessary registers (it will make it 3 in case one of } { the operands is already in the fpu) (JM) } if ((left.expectloc<>LOC_FPUREGISTER) or (right.expectloc<>LOC_FPUREGISTER)) and (registersfpu < 2) then inc(registersfpu); end { pointer comperation and subtraction } else if (ld.deftype=pointerdef) then begin if nodetype in [addn,subn,muln,andn,orn,xorn] then expectloc:=LOC_REGISTER else expectloc:=LOC_FLAGS; calcregisters(self,1,0,0); end else if is_class_or_interface(ld) then begin expectloc:=LOC_FLAGS; calcregisters(self,1,0,0); end else if (ld.deftype=classrefdef) then begin expectloc:=LOC_FLAGS; calcregisters(self,1,0,0); end { support procvar=nil,procvar<>nil } else if ((ld.deftype=procvardef) and (rt=niln)) or ((rd.deftype=procvardef) and (lt=niln)) then begin expectloc:=LOC_FLAGS; calcregisters(self,1,0,0); end {$ifdef SUPPORT_MMX} { mmx support, this must be before the zero based array check } else if (cs_mmx in aktlocalswitches) and is_mmx_able_array(ld) and is_mmx_able_array(rd) then begin expectloc:=LOC_MMXREGISTER; calcregisters(self,0,0,1); end {$endif SUPPORT_MMX} else if (rd.deftype=pointerdef) or (ld.deftype=pointerdef) then begin expectloc:=LOC_REGISTER; calcregisters(self,1,0,0); end else if (rd.deftype=procvardef) and (ld.deftype=procvardef) and equal_defs(rd,ld) then begin expectloc:=LOC_FLAGS; calcregisters(self,1,0,0); end else if (ld.deftype=enumdef) then begin expectloc:=LOC_FLAGS; calcregisters(self,1,0,0); end {$ifdef SUPPORT_MMX} else if (cs_mmx in aktlocalswitches) and is_mmx_able_array(ld) and is_mmx_able_array(rd) then begin expectloc:=LOC_MMXREGISTER; calcregisters(self,0,0,1); end {$endif SUPPORT_MMX} { the general solution is to convert to 32 bit int } else begin expectloc:=LOC_REGISTER; calcregisters(self,1,0,0); end; end; {$ifdef state_tracking} function Taddnode.track_state_pass(exec_known:boolean):boolean; var factval:Tnode; begin track_state_pass:=false; if left.track_state_pass(exec_known) then begin track_state_pass:=true; left.resulttype.def:=nil; do_resulttypepass(left); end; factval:=aktstate.find_fact(left); if factval<>nil then begin track_state_pass:=true; left.destroy; left:=factval.getcopy; end; if right.track_state_pass(exec_known) then begin track_state_pass:=true; right.resulttype.def:=nil; do_resulttypepass(right); end; factval:=aktstate.find_fact(right); if factval<>nil then begin track_state_pass:=true; right.destroy; right:=factval.getcopy; end; end; {$endif} begin caddnode:=taddnode; end. { $Log$ Revision 1.92 2003-06-03 21:04:43 peter * widen cardinal+signed operations Revision 1.91 2003/05/26 21:15:18 peter * disable string node optimizations for the moment Revision 1.90 2003/05/26 19:38:28 peter * generic fpc_shorstr_concat + fpc_shortstr_append_shortstr optimization Revision 1.89 2003/05/24 21:12:57 florian * if something doesn't work with callparatemp, the define callparatemp should be used because other processors with reigster calling conventions depend on this as well Revision 1.88 2003/05/23 22:57:38 jonas - disable addoptnodes for powerpc, because they can generate calls in pass_2, so -dcallparatemp can't detect them as nested calls Revision 1.87 2003/04/27 11:21:32 peter * aktprocdef renamed to current_procdef * procinfo renamed to current_procinfo * procinfo will now be stored in current_module so it can be cleaned up properly * gen_main_procsym changed to create_main_proc and release_main_proc to also generate a tprocinfo structure * fixed unit implicit initfinal Revision 1.86 2003/04/26 09:12:55 peter * add string returns in LOC_REFERENCE Revision 1.85 2003/04/24 22:29:57 florian * fixed a lot of PowerPC related stuff Revision 1.84 2003/04/23 20:16:04 peter + added currency support based on int64 + is_64bit for use in cg units instead of is_64bitint * removed cgmessage from n386add, replace with internalerrors Revision 1.83 2003/04/23 10:10:07 peter * expectloc fixes Revision 1.82 2003/04/22 23:50:22 peter * firstpass uses expectloc * checks if there are differences between the expectloc and location.loc from secondpass in EXTDEBUG Revision 1.81 2003/02/15 22:20:14 carl * bugfix for generic calls to FPU emulation code Revision 1.80 2003/02/12 22:10:07 carl * load_frame_pointer is now generic * change fpu emulation routine names Revision 1.79 2003/01/02 22:19:54 peter * support pchar-char operations converting to string first * support chararray-nil Revision 1.78 2002/12/11 22:41:03 peter * stop processing assignment node when the binaryoverload generates a codegenerror Revision 1.77 2002/12/06 16:56:57 peter * only compile cs_fp_emulation support when cpufpuemu is defined * define cpufpuemu for m68k only Revision 1.76 2002/11/30 21:32:24 carl + Add loading of softfpu in emulation mode + Correct routine call for softfpu * Extended type must also be defined even with softfpu Revision 1.75 2002/11/27 13:11:38 peter * more currency fixes, taddcurr runs now successfull Revision 1.74 2002/11/27 11:28:40 peter * when both flaottypes are the same then handle the addnode using that floattype instead of bestrealtype Revision 1.73 2002/11/25 18:43:32 carl - removed the invalid if <> checking (Delphi is strange on this) + implemented abstract warning on instance creation of class with abstract methods. * some error message cleanups Revision 1.72 2002/11/25 17:43:17 peter * splitted defbase in defutil,symutil,defcmp * merged isconvertable and is_equal into compare_defs(_ext) * made operator search faster by walking the list only once Revision 1.71 2002/11/23 22:50:06 carl * some small speed optimizations + added several new warnings/hints Revision 1.70 2002/11/16 14:20:22 peter * fix tbs0417 Revision 1.69 2002/11/15 01:58:50 peter * merged changes from 1.0.7 up to 04-11 - -V option for generating bug report tracing - more tracing for option parsing - errors for cdecl and high() - win32 import stabs - win32 records<=8 are returned in eax:edx (turned off by default) - heaptrc update - more info for temp management in .s file with EXTDEBUG Revision 1.68 2002/10/08 16:50:43 jonas * fixed web bug 2136 Revision 1.67 2002/10/05 00:47:03 peter * support dynamicarray<>nil Revision 1.66 2002/10/04 21:19:28 jonas * fixed web bug 2139: checking for division by zero fixed Revision 1.65 2002/09/07 15:25:02 peter * old logs removed and tabs fixed Revision 1.64 2002/09/07 12:16:05 carl * second part bug report 1996 fix, testrange in cordconstnode only called if option is set (also make parsing a tiny faster) Revision 1.63 2002/09/04 19:32:56 jonas * changed some ctypeconvnode/toggleflag(nf_explizit) combo's to ctypeconvnode.create_explicit() statements Revision 1.62 2002/08/17 09:23:34 florian * first part of current_procinfo rewrite Revision 1.61 2002/08/15 15:15:55 carl * jmpbuf size allocation for exceptions is now cpu specific (as it should) * more generic nodes for maths * several fixes for better m68k support Revision 1.60 2002/08/12 15:08:39 carl + stab register indexes for powerpc (moved from gdb to cpubase) + tprocessor enumeration moved to cpuinfo + linker in target_info is now a class * many many updates for m68k (will soon start to compile) - removed some ifdef or correct them for correct cpu Revision 1.59 2002/08/02 07:44:30 jonas * made assigned() handling generic * add nodes now can also evaluate constant expressions at compile time that contain nil nodes Revision 1.58 2002/07/26 11:17:52 jonas * the optimization of converting a multiplication with a power of two to a shl is moved from n386add/secondpass to nadd/resulttypepass Revision 1.57 2002/07/23 13:08:16 jonas * fixed constant set evaluation of new set handling for non-commutative operators Revision 1.56 2002/07/23 12:34:29 daniel * Readded old set code. To use it define 'oldset'. Activated by default for ppc. Revision 1.55 2002/07/22 11:48:04 daniel * Sets are now internally sets. Revision 1.54 2002/07/20 11:57:53 florian * types.pas renamed to defbase.pas because D6 contains a types unit so this would conflicts if D6 programms are compiled + Willamette/SSE2 instructions to assembler added Revision 1.53 2002/07/19 11:41:34 daniel * State tracker work * The whilen and repeatn are now completely unified into whilerepeatn. This allows the state tracker to change while nodes automatically into repeat nodes. * Resulttypepass improvements to the notn. 'not not a' is optimized away and 'not(a>b)' is optimized into 'a<=b'. * Resulttypepass improvements to the whilerepeatn. 'while not a' is optimized by removing the notn and later switchting the true and falselabels. The same is done with 'repeat until not a'. Revision 1.52 2002/07/14 18:00:43 daniel + Added the beginning of a state tracker. This will track the values of variables through procedures and optimize things away. Revision 1.51 2002/05/18 13:34:08 peter * readded missing revisions Revision 1.50 2002/05/16 19:46:37 carl + defines.inc -> fpcdefs.inc to avoid conflicts if compiling by hand + try to fix temp allocation (still in ifdef) + generic constructor calls + start of tassembler / tmodulebase class cleanup Revision 1.48 2002/05/13 19:54:36 peter * removed n386ld and n386util units * maybe_save/maybe_restore added instead of the old maybe_push Revision 1.47 2002/05/12 16:53:06 peter * moved entry and exitcode to ncgutil and cgobj * foreach gets extra argument for passing local data to the iterator function * -CR checks also class typecasts at runtime by changing them into as * fixed compiler to cycle with the -CR option * fixed stabs with elf writer, finally the global variables can be watched * removed a lot of routines from cga unit and replaced them by calls to cgobj * u32bit-s32bit updates for and,or,xor nodes. When one element is u32bit then the other is typecasted also to u32bit without giving a rangecheck warning/error. * fixed pascal calling method with reversing also the high tree in the parast, detected by tcalcst3 test Revision 1.46 2002/04/23 19:16:34 peter * add pinline unit that inserts compiler supported functions using one or more statements * moved finalize and setlength from ninl to pinline Revision 1.45 2002/04/04 19:05:56 peter * removed unused units * use tlocation.size in cg.a_*loc*() routines Revision 1.44 2002/04/02 17:11:28 peter * tlocation,treference update * LOC_CONSTANT added for better constant handling * secondadd splitted in multiple routines * location_force_reg added for loading a location to a register of a specified size * secondassignment parses now first the right and then the left node (this is compatible with Kylix). This saves a lot of push/pop especially with string operations * adapted some routines to use the new cg methods }