fpc/compiler/powerpc/nppcadd.pas
2014-08-20 14:31:36 +00:00

785 lines
31 KiB
ObjectPascal

{
Copyright (c) 2000-2002 by Florian Klaempfl and Jonas Maebe
Code generation for add nodes on the PowerPC
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 nppcadd;
{$i fpcdefs.inc}
interface
uses
node,nadd,ncgadd,ngppcadd,cpubase;
type
tppcaddnode = class(tgenppcaddnode)
procedure pass_generate_code;override;
protected
function use_generic_mul32to64: boolean; override;
private
procedure emit_compare(unsigned : boolean); override;
{$ifdef SUPPORT_MMX}
procedure second_addmmx;override;
{$endif SUPPORT_MMX}
procedure second_add64bit;override;
end;
implementation
uses
globtype,systems,
cutils,verbose,globals,
symconst,symdef,paramgr,
aasmbase,aasmtai,aasmdata,aasmcpu,defutil,htypechk,
cgbase,cpuinfo,pass_1,pass_2,regvars,
cpupara,cgcpu,cgutils,procinfo,
ncon,nset,
ncgutil,tgobj,rgobj,rgcpu,cgobj,hlcgobj,cg64f32;
{*****************************************************************************
Pass 1
*****************************************************************************}
function tppcaddnode.use_generic_mul32to64: boolean;
begin
result := false;
end;
{*****************************************************************************
Helpers
*****************************************************************************}
procedure tppcaddnode.emit_compare(unsigned: boolean);
var
op : tasmop;
tmpreg : tregister;
useconst : boolean;
begin
tmpreg:=NR_NO;
// get the constant on the right if there is one
if (left.location.loc = LOC_CONSTANT) then
swapleftright;
// can we use an immediate, or do we have to load the
// constant in a register first?
if (right.location.loc = LOC_CONSTANT) then
begin
{$ifdef dummy}
if (right.location.size in [OS_64,OS_S64]) and (hi(right.location.value64)<>0) and ((hi(right.location.value64)<>$ffffffff) or unsigned) then
internalerror(2002080301);
{$endif extdebug}
if (nodetype in [equaln,unequaln]) then
if (unsigned and
(aword(right.location.value) > high(word))) or
(not unsigned and
(aint(right.location.value) < low(smallint)) or
(aint(right.location.value) > high(smallint))) then
{ we can then maybe use a constant in the 'othersigned' case
(the sign doesn't matter for // equal/unequal)}
unsigned := not unsigned;
if (unsigned and
(aword(right.location.value) <= high(word))) or
(not(unsigned) and
(aint(right.location.value) >= low(smallint)) and
(aint(right.location.value) <= high(smallint))) then
useconst := true
else
begin
useconst := false;
tmpreg := cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
cg.a_load_const_reg(current_asmdata.CurrAsmList,OS_INT,
right.location.value,tmpreg);
end
end
else
useconst := false;
location.loc := LOC_FLAGS;
location.resflags := getresflags;
if not unsigned then
if useconst then
op := A_CMPWI
else
op := A_CMPW
else
if useconst then
op := A_CMPLWI
else
op := A_CMPLW;
if (right.location.loc = LOC_CONSTANT) then
begin
if useconst then
current_asmdata.CurrAsmList.concat(taicpu.op_reg_const(op,left.location.register,longint(right.location.value)))
else
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg(op,left.location.register,tmpreg));
end
else
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg(op,
left.location.register,right.location.register));
end;
{*****************************************************************************
Add64bit
*****************************************************************************}
procedure tppcaddnode.second_add64bit;
var
op : TOpCG;
op1,op2 : TAsmOp;
cmpop,
unsigned : boolean;
procedure emit_cmp64_hi;
var
oldleft, oldright: tlocation;
begin
// put the high part of the location in the low part
location_copy(oldleft,left.location);
location_copy(oldright,right.location);
if left.location.loc = LOC_CONSTANT then
left.location.value64 := left.location.value64 shr 32
else
left.location.register64.reglo := left.location.register64.reghi;
if right.location.loc = LOC_CONSTANT then
right.location.value64 := right.location.value64 shr 32
else
right.location.register64.reglo := right.location.register64.reghi;
// and call the normal emit_compare
emit_compare(unsigned);
location_copy(left.location,oldleft);
location_copy(right.location,oldright);
end;
procedure emit_cmp64_lo;
begin
emit_compare(true);
end;
procedure firstjmp64bitcmp;
var
oldnodetype: tnodetype;
begin
{$ifdef OLDREGVARS}
load_all_regvars(current_asmdata.CurrAsmList);
{$endif OLDREGVARS}
{ the jump the sequence is a little bit hairy }
case nodetype of
ltn,gtn:
begin
cg.a_jmp_flags(current_asmdata.CurrAsmList,getresflags,current_procinfo.CurrTrueLabel);
{ cheat a little bit for the negative test }
toggleflag(nf_swapped);
cg.a_jmp_flags(current_asmdata.CurrAsmList,getresflags,current_procinfo.CurrFalseLabel);
toggleflag(nf_swapped);
end;
lten,gten:
begin
oldnodetype:=nodetype;
if nodetype=lten then
nodetype:=ltn
else
nodetype:=gtn;
cg.a_jmp_flags(current_asmdata.CurrAsmList,getresflags,current_procinfo.CurrTrueLabel);
{ cheat for the negative test }
if nodetype=ltn then
nodetype:=gtn
else
nodetype:=ltn;
cg.a_jmp_flags(current_asmdata.CurrAsmList,getresflags,current_procinfo.CurrFalseLabel);
nodetype:=oldnodetype;
end;
equaln:
begin
nodetype := unequaln;
cg.a_jmp_flags(current_asmdata.CurrAsmList,getresflags,current_procinfo.CurrFalseLabel);
nodetype := equaln;
end;
unequaln:
begin
cg.a_jmp_flags(current_asmdata.CurrAsmList,getresflags,current_procinfo.CurrTrueLabel);
end;
end;
end;
procedure secondjmp64bitcmp;
begin
{ the jump the sequence is a little bit hairy }
case nodetype of
ltn,gtn,lten,gten:
begin
{ the comparison of the low dword always has }
{ to be always unsigned! }
cg.a_jmp_flags(current_asmdata.CurrAsmList,getresflags,current_procinfo.CurrTrueLabel);
cg.a_jmp_always(current_asmdata.CurrAsmList,current_procinfo.CurrFalseLabel);
end;
equaln:
begin
nodetype := unequaln;
cg.a_jmp_flags(current_asmdata.CurrAsmList,getresflags,current_procinfo.CurrFalseLabel);
cg.a_jmp_always(current_asmdata.CurrAsmList,current_procinfo.CurrTrueLabel);
nodetype := equaln;
end;
unequaln:
begin
cg.a_jmp_flags(current_asmdata.CurrAsmList,getresflags,current_procinfo.CurrTrueLabel);
cg.a_jmp_always(current_asmdata.CurrAsmList,current_procinfo.CurrFalseLabel);
end;
end;
end;
var
tempreg64: tregister64;
begin
firstcomplex(self);
pass_left_and_right;
cmpop:=false;
unsigned:=((left.resultdef.typ=orddef) and
(torddef(left.resultdef).ordtype=u64bit)) or
((right.resultdef.typ=orddef) and
(torddef(right.resultdef).ordtype=u64bit));
case nodetype of
addn :
begin
op:=OP_ADD;
end;
subn :
begin
op:=OP_SUB;
if (nf_swapped in flags) then
swapleftright;
end;
ltn,lten,
gtn,gten,
equaln,unequaln:
begin
op:=OP_NONE;
cmpop:=true;
end;
xorn:
op:=OP_XOR;
orn:
op:=OP_OR;
andn:
op:=OP_AND;
muln:
begin
{ should be handled in pass_1 (JM) }
if not(torddef(left.resultdef).ordtype in [U32bit,s32bit]) or
(torddef(left.resultdef).typ <> torddef(right.resultdef).typ) then
internalerror(200109051);
{ handled separately }
op := OP_NONE;
end;
else
internalerror(2002072705);
end;
if not cmpop then
location_reset(location,LOC_REGISTER,def_cgsize(resultdef));
load_left_right(cmpop,((cs_check_overflow in current_settings.localswitches) and
(nodetype in [addn,subn])) or (nodetype = muln));
if (nodetype <> muln) and
(not(cs_check_overflow in current_settings.localswitches) or
not(nodetype in [addn,subn])) then
begin
case nodetype of
ltn,lten,
gtn,gten:
begin
emit_cmp64_hi;
firstjmp64bitcmp;
emit_cmp64_lo;
secondjmp64bitcmp;
end;
equaln,unequaln:
begin
// instead of doing a complicated compare, do
// (left.hi xor right.hi) or (left.lo xor right.lo)
// (somewhate optimized so that no superfluous 'mr's are
// generated)
if (left.location.loc = LOC_CONSTANT) then
swapleftright;
if (right.location.loc = LOC_CONSTANT) then
begin
if left.location.loc = LOC_REGISTER then
begin
tempreg64.reglo := left.location.register64.reglo;
tempreg64.reghi := left.location.register64.reghi;
end
else
begin
if (aint(right.location.value64) <> 0) then
tempreg64.reglo := cg.getintregister(current_asmdata.CurrAsmList,OS_32)
else
tempreg64.reglo := left.location.register64.reglo;
if ((right.location.value64 shr 32) <> 0) then
tempreg64.reghi := cg.getintregister(current_asmdata.CurrAsmList,OS_32)
else
tempreg64.reghi := left.location.register64.reghi;
end;
if (aint(right.location.value64) <> 0) then
{ negative values can be handled using SUB, }
{ positive values < 65535 using XOR. }
if (longint(right.location.value64) >= -32767) and
(longint(right.location.value64) < 0) then
cg.a_op_const_reg_reg(current_asmdata.CurrAsmList,OP_SUB,OS_INT,
aint(right.location.value64),
left.location.register64.reglo,tempreg64.reglo)
else
cg.a_op_const_reg_reg(current_asmdata.CurrAsmList,OP_XOR,OS_INT,
aint(right.location.value64),
left.location.register64.reglo,tempreg64.reglo);
if ((right.location.value64 shr 32) <> 0) then
if (longint(right.location.value64 shr 32) >= -32767) and
(longint(right.location.value64 shr 32) < 0) then
cg.a_op_const_reg_reg(current_asmdata.CurrAsmList,OP_SUB,OS_INT,
aint(right.location.value64 shr 32),
left.location.register64.reghi,tempreg64.reghi)
else
cg.a_op_const_reg_reg(current_asmdata.CurrAsmList,OP_XOR,OS_INT,
aint(right.location.value64 shr 32),
left.location.register64.reghi,tempreg64.reghi);
end
else
begin
tempreg64.reglo := cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
tempreg64.reghi := cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
cg64.a_op64_reg_reg_reg(current_asmdata.CurrAsmList,OP_XOR,location.size,
left.location.register64,right.location.register64,
tempreg64);
end;
cg.a_reg_alloc(current_asmdata.CurrAsmList,NR_R0);
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_reg(A_OR_,NR_R0,
tempreg64.reglo,tempreg64.reghi));
cg.a_reg_dealloc(current_asmdata.CurrAsmList,NR_R0);
location_reset(location,LOC_FLAGS,OS_NO);
location.resflags := getresflags;
end;
xorn,orn,andn,addn:
begin
location.register64.reglo := cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
location.register64.reghi := cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
if (left.location.loc = LOC_CONSTANT) then
swapleftright;
if (right.location.loc = LOC_CONSTANT) then
cg64.a_op64_const_reg_reg(current_asmdata.CurrAsmList,op,location.size,right.location.value64,
left.location.register64,location.register64)
else
cg64.a_op64_reg_reg_reg(current_asmdata.CurrAsmList,op,location.size,right.location.register64,
left.location.register64,location.register64);
end;
subn:
begin
location.register64.reglo := cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
location.register64.reghi := cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
if left.location.loc <> LOC_CONSTANT then
begin
if right.location.loc <> LOC_CONSTANT then
// reg64 - reg64
cg64.a_op64_reg_reg_reg(current_asmdata.CurrAsmList,OP_SUB,location.size,
right.location.register64,left.location.register64,
location.register64)
else
// reg64 - const64
cg64.a_op64_const_reg_reg(current_asmdata.CurrAsmList,OP_SUB,location.size,
right.location.value64,left.location.register64,
location.register64)
end
else if ((left.location.value64 shr 32) = 0) then
begin
if (int64(left.location.value64) >= low(smallint)) and
(int64(left.location.value64) <= high(smallint)) then
begin
// consts16 - reg64
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_const(A_SUBFIC,
location.register64.reglo,right.location.register64.reglo,
left.location.value));
end
else
begin
// const32 - reg64
hlcg.location_force_reg(current_asmdata.CurrAsmList,left.location,
left.resultdef,u32inttype,true);
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_reg(A_SUBC,
location.register64.reglo,left.location.register64.reglo,
right.location.register64.reglo));
end;
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg(A_SUBFZE,
location.register64.reghi,right.location.register64.reghi));
end
else if (aint(left.location.value64) = 0) then
begin
// (const32 shl 32) - reg64
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_const(A_SUBFIC,
location.register64.reglo,right.location.register64.reglo,0));
left.location.value64 := left.location.value64 shr 32;
hlcg.location_force_reg(current_asmdata.CurrAsmList,left.location,left.resultdef,u32inttype,true);
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_reg(A_SUBFE,
location.register64.reghi,right.location.register64.reghi,
left.location.register));
end
else
begin
// const64 - reg64
hlcg.location_force_reg(current_asmdata.CurrAsmList,left.location,
left.resultdef,left.resultdef,false);
cg64.a_op64_reg_reg_reg(current_asmdata.CurrAsmList,OP_SUB,location.size,
right.location.register64,left.location.register64,
location.register64);
end;
end;
else
internalerror(2002072803);
end;
end
else
begin
if is_signed(resultdef) then
begin
case nodetype of
addn:
begin
op1 := A_ADDC;
op2 := A_ADDEO;
end;
subn:
begin
op1 := A_SUBC;
op2 := A_SUBFEO;
end;
muln:
begin
op1 := A_MULLW;
op2 := A_MULHW
end;
else
internalerror(2002072806);
end
end
else
begin
case nodetype of
addn:
begin
op1 := A_ADDC;
op2 := A_ADDE;
end;
subn:
begin
op1 := A_SUBC;
op2 := A_SUBFE;
end;
muln:
begin
op1 := A_MULLW;
op2 := A_MULHWU
end;
else
internalerror(2014082040);
end;
end;
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_reg(op1,location.register64.reglo,
left.location.register64.reglo,right.location.register64.reglo));
if (nodetype <> muln) then
begin
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_reg(op2,location.register64.reghi,
right.location.register64.reghi,left.location.register64.reghi));
if not(is_signed(resultdef)) then
if nodetype = addn then
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg(A_CMPLW,location.register64.reghi,left.location.register64.reghi))
else
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg(A_CMPLW,left.location.register64.reghi,location.register64.reghi));
cg.g_overflowcheck(current_asmdata.CurrAsmList,location,resultdef);
end
else
begin
{ 32 * 32 -> 64 cannot overflow }
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_reg(op2,location.register64.reghi,
left.location.register64.reglo,right.location.register64.reglo));
end
end;
{ set result location }
{ (emit_compare sets it to LOC_FLAGS for compares, so set the }
{ real location only now) (JM) }
if cmpop and
not(nodetype in [equaln,unequaln]) then
location_reset(location,LOC_JUMP,OS_NO);
end;
{*****************************************************************************
pass_2
*****************************************************************************}
procedure tppcaddnode.pass_generate_code;
{ is also being used for xor, and "mul", "sub, or and comparative }
{ operators }
var
cgop : topcg;
op : tasmop;
tmpreg : tregister;
hl : tasmlabel;
cmpop : boolean;
{ true, if unsigned types are compared }
unsigned : boolean;
checkoverflow : boolean;
begin
{ to make it more readable, string and set (not smallset!) have their
own procedures }
case left.resultdef.typ of
orddef :
begin
{ handling boolean expressions }
if is_boolean(left.resultdef) and
is_boolean(right.resultdef) then
begin
second_addboolean;
exit;
end
{ 64bit operations }
else if is_64bit(resultdef) or
is_64bit(left.resultdef) then
begin
second_add64bit;
exit;
end;
end;
stringdef :
begin
internalerror(2002072402);
exit;
end;
setdef :
begin
{ normalsets are already handled in pass1 }
if not is_smallset(left.resultdef) then
internalerror(200109042);
second_addsmallset;
exit;
end;
arraydef :
begin
{$ifdef SUPPORT_MMX}
if is_mmx_able_array(left.resultdef) then
begin
second_addmmx;
exit;
end;
{$endif SUPPORT_MMX}
end;
floatdef :
begin
second_addfloat;
exit;
end;
end;
{ defaults }
cmpop:=nodetype in [ltn,lten,gtn,gten,equaln,unequaln];
unsigned:=not(is_signed(left.resultdef)) or
not(is_signed(right.resultdef));
pass_left_and_right;
{ Convert flags to register first }
{ can any of these things be in the flags actually?? (JM) }
if (left.location.loc = LOC_FLAGS) or
(right.location.loc = LOC_FLAGS) then
internalerror(2002072602);
{ set result location }
if not cmpop then
location_reset(location,LOC_REGISTER,def_cgsize(resultdef))
else
location_reset(location,LOC_FLAGS,OS_NO);
checkoverflow:=
(nodetype in [addn,subn,muln]) and
(cs_check_overflow in current_settings.localswitches) and
(left.resultdef.typ<>pointerdef) and
(right.resultdef.typ<>pointerdef);
load_left_right(cmpop, checkoverflow);
if not(cmpop) then
location.register := cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
if not(checkoverflow) then
begin
case nodetype of
addn, muln, xorn, orn, andn:
begin
case nodetype of
addn:
cgop := OP_ADD;
muln:
if unsigned then
cgop := OP_MUL
else
cgop := OP_IMUL;
xorn:
cgop := OP_XOR;
orn:
cgop := OP_OR;
andn:
cgop := OP_AND;
else
internalerror(2014082041);
end;
if (left.location.loc = LOC_CONSTANT) then
swapleftright;
if (right.location.loc <> LOC_CONSTANT) then
cg.a_op_reg_reg_reg(current_asmdata.CurrAsmList,cgop,OS_INT,
left.location.register,right.location.register,
location.register)
else
cg.a_op_const_reg_reg(current_asmdata.CurrAsmList,cgop,OS_INT,
right.location.value,left.location.register,
location.register);
end;
subn:
begin
if (nf_swapped in flags) then
swapleftright;
if left.location.loc <> LOC_CONSTANT then
if right.location.loc <> LOC_CONSTANT then
cg.a_op_reg_reg_reg(current_asmdata.CurrAsmList,OP_SUB,OS_INT,
right.location.register,left.location.register,
location.register)
else
cg.a_op_const_reg_reg(current_asmdata.CurrAsmList,OP_SUB,OS_INT,
right.location.value,left.location.register,
location.register)
else
if (longint(left.location.value) >= low(smallint)) and
(longint(left.location.value) <= high(smallint)) then
begin
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_const(A_SUBFIC,
location.register,right.location.register,
longint(left.location.value)));
end
else
begin
tmpreg := cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
cg.a_load_const_reg(current_asmdata.CurrAsmList,OS_INT,
left.location.value,tmpreg);
cg.a_op_reg_reg_reg(current_asmdata.CurrAsmList,OP_SUB,OS_INT,
right.location.register,tmpreg,location.register);
end;
end;
ltn,lten,gtn,gten,equaln,unequaln :
begin
emit_compare(unsigned);
end;
end;
end
else
// overflow checking is on and we have an addn, subn or muln
begin
if is_signed(resultdef) then
begin
case nodetype of
addn:
op := A_ADDO;
subn:
begin
op := A_SUBO;
if (nf_swapped in flags) then
swapleftright;
end;
muln:
op := A_MULLWO;
else
internalerror(2002072601);
end;
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_reg(op,location.register,
left.location.register,right.location.register));
cg.g_overflowcheck(current_asmdata.CurrAsmList,location,resultdef);
end
else
begin
case nodetype of
addn:
begin
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_reg(A_ADD,location.register,
left.location.register,right.location.register));
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg(A_CMPLW,location.register,left.location.register));
cg.g_overflowcheck(current_asmdata.CurrAsmList,location,resultdef);
end;
subn:
begin
if nf_swapped in flags then
swapleftright;
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_reg(A_SUB,location.register,
left.location.register,right.location.register));
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg(A_CMPLW,left.location.register,location.register));
cg.g_overflowcheck(current_asmdata.CurrAsmList,location,resultdef);
end;
muln:
begin
{ calculate the upper 32 bits of the product, = 0 if no overflow }
cg.a_reg_alloc(current_asmdata.CurrAsmList,NR_R0);
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_reg(A_MULHWU_,NR_R0,
left.location.register,right.location.register));
cg.a_reg_dealloc(current_asmdata.CurrAsmList,NR_R0);
{ calculate the real result }
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_reg(A_MULLW,location.register,
left.location.register,right.location.register));
{ g_overflowcheck generates a OC_AE instead of OC_EQ :/ }
current_asmdata.getjumplabel(hl);
tcgppc(cg).a_jmp_cond(current_asmdata.CurrAsmList,OC_EQ,hl);
cg.a_call_name(current_asmdata.CurrAsmList,'FPC_OVERFLOW',false);
cg.a_label(current_asmdata.CurrAsmList,hl);
end;
end;
end;
end;
end;
begin
caddnode:=tppcaddnode;
end.