fpc/compiler/wasm32/nwasmset.pas

482 lines
18 KiB
ObjectPascal

{
Copyright (c) 1998-2002, 2021 by Florian Klaempfl and Nikolay Nikolov
Generate WebAssembly code for in/case 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 nwasmset;
{$i fpcdefs.inc}
interface
uses
node,nset,ncgset,
aasmbase;
type
{ twasminnode }
twasminnode = class(tcginnode)
protected
function checkgenjumps(out setparts: Tsetparts; out numparts: byte; out use_small: boolean): boolean;override;
end;
{ twasmcasenode }
twasmcasenode = class(tcgcasenode)
private
function GetBranchLabel(Block: TNode; out _Label: TAsmLabel): Boolean;
protected
procedure genlinearlist(hp : pcaselabel);override;
procedure genlinearcmplist(hp : pcaselabel);override;
public
procedure pass_generate_code;override;
end;
implementation
uses
globtype,globals,
cpubase,
cgbase,cgutils,
aasmdata,aasmcpu,
hlcgobj,hlcgcpu,
nbas,
symtype,
pass_2,defutil,verbose,constexp;
{*****************************************************************************
TWASMINNODE
*****************************************************************************}
function twasminnode.checkgenjumps(out setparts: Tsetparts; out numparts: byte; out use_small: boolean): boolean;
begin
{ call inherited to initialize use_small }
inherited;
result:=false;
end;
{*****************************************************************************
TWASMCASENODE
*****************************************************************************}
function twasmcasenode.GetBranchLabel(Block: TNode; out _Label: TAsmLabel): Boolean;
begin
Result := True;
if not Assigned(Block) then
begin
{ Block doesn't exist / is empty }
_Label := endlabel;
Exit;
end;
{ These optimisations aren't particularly debugger friendly }
if not (cs_opt_level2 in current_settings.optimizerswitches) then
begin
Result := False;
current_asmdata.getjumplabel(_Label);
current_asmdata.CurrAsmList.concat(taicpu.op_none(a_block));
Exit;
end;
while Assigned(Block) do
begin
case Block.nodetype of
nothingn:
begin
_Label := endlabel;
Exit;
end;
blockn:
begin
Block := TBlockNode(Block).Left;
Continue;
end;
statementn:
begin
{ If the right node is assigned, then it's a compound block
that can't be simplified, so fall through, set Result to
False and make a new label }
if Assigned(TStatementNode(Block).right) then
Break;
Block := TStatementNode(Block).Left;
Continue;
end;
else
;
end;
Break;
end;
{ Create unique label }
Result := False;
current_asmdata.CurrAsmList.concat(taicpu.op_none(a_block));
current_asmdata.getjumplabel(_Label);
end;
procedure twasmcasenode.genlinearlist(hp: pcaselabel);
var
first : boolean;
last : TConstExprInt;
scratch_reg: tregister;
newsize: tcgsize;
newdef: tdef;
procedure gensub(value:tcgint);
begin
{ here, since the sub and cmp are separate we need
to move the result before subtract to help
the register allocator
}
hlcg.a_load_reg_reg(current_asmdata.CurrAsmList, opsize, opsize, hregister, scratch_reg);
hlcg.a_op_const_reg(current_asmdata.CurrAsmList, OP_SUB, opsize, value, hregister);
end;
procedure genitem(t : pcaselabel);
begin
if assigned(t^.less) then
genitem(t^.less);
{ do we need to test the first value? }
if first and (t^._low>get_min_value(left.resultdef)) then
thlcgwasm(hlcg).a_cmp_const_reg_label(current_asmdata.CurrAsmList,opsize,jmp_lt,tcgint(t^._low.svalue),hregister,elselabel);
if t^._low=t^._high then
begin
if t^._low-last=0 then
thlcgwasm(hlcg).a_cmp_const_reg_label(current_asmdata.CurrAsmList,opsize,OC_EQ,0,hregister,blocklabel(t^.blockid))
else
begin
gensub(tcgint(t^._low.svalue-last.svalue));
thlcgwasm(hlcg).a_cmp_const_reg_label(current_asmdata.CurrAsmList,opsize,
OC_EQ,tcgint(t^._low.svalue-last.svalue),scratch_reg,blocklabel(t^.blockid));
end;
last:=t^._low;
end
else
begin
{ it begins with the smallest label, if the value }
{ is even smaller then jump immediately to the }
{ ELSE-label }
if first then
begin
{ have we to ajust the first value ? }
if (t^._low>get_min_value(left.resultdef)) or (get_min_value(left.resultdef)<>0) then
gensub(tcgint(t^._low.svalue));
end
else
begin
{ if there is no unused label between the last and the }
{ present label then the lower limit can be checked }
{ immediately. else check the range in between: }
gensub(tcgint(t^._low.svalue-last.svalue));
thlcgwasm(hlcg).a_cmp_const_reg_label(current_asmdata.CurrAsmList, opsize,jmp_lt,tcgint(t^._low.svalue-last.svalue),scratch_reg,elselabel);
end;
gensub(tcgint(t^._high.svalue-t^._low.svalue));
thlcgwasm(hlcg).a_cmp_const_reg_label(current_asmdata.CurrAsmList,opsize,jmp_le,tcgint(t^._high.svalue-t^._low.svalue),scratch_reg,blocklabel(t^.blockid));
last:=t^._high;
end;
first:=false;
if assigned(t^.greater) then
genitem(t^.greater);
end;
begin
{ do we need to generate cmps? }
if (with_sign and (min_label<0)) then
genlinearcmplist(hp)
else
begin
{ sign/zero extend the value to a full register before starting to
subtract values, so that on platforms that don't have
subregisters of the same size as the value we don't generate
sign/zero-extensions after every subtraction
make newsize always signed, since we only do this if the size in
bytes of the register is larger than the original opsize, so
the value can always be represented by a larger signed type }
newsize:=tcgsize2signed[reg_cgsize(hregister)];
if tcgsize2size[newsize]>opsize.size then
begin
newdef:=cgsize_orddef(newsize);
scratch_reg:=hlcg.getintregister(current_asmdata.CurrAsmList,newdef);
hlcg.a_load_reg_reg(current_asmdata.CurrAsmList,opsize,newdef,hregister,scratch_reg);
hregister:=scratch_reg;
opsize:=newdef;
end;
if (labelcnt>1) or not(cs_opt_level1 in current_settings.optimizerswitches) then
begin
last:=0;
first:=true;
scratch_reg:=hlcg.getintregister(current_asmdata.CurrAsmList,opsize);
genitem(hp);
end
else
begin
{ If only one label exists, we can greatly simplify the checks to a simple comparison }
if hp^._low=hp^._high then
thlcgwasm(hlcg).a_cmp_const_reg_label(current_asmdata.CurrAsmList, opsize, OC_EQ, tcgint(hp^._low.svalue), hregister,blocklabel(hp^.blockid))
else
begin
scratch_reg:=hlcg.getintregister(current_asmdata.CurrAsmList,opsize);
gensub(tcgint(hp^._low.svalue));
thlcgwasm(hlcg).a_cmp_const_reg_label(current_asmdata.CurrAsmList, opsize, OC_BE, tcgint(hp^._high.svalue-hp^._low.svalue), hregister,blocklabel(hp^.blockid))
end;
end;
current_asmdata.CurrAsmList.concat(taicpu.op_sym(a_br,elselabel));
end;
end;
procedure twasmcasenode.genlinearcmplist(hp : pcaselabel);
var
last : TConstExprInt;
lastwasrange: boolean;
procedure genitem(t : pcaselabel);
begin
if assigned(t^.less) then
genitem(t^.less);
if t^._low=t^._high then
begin
thlcgwasm(hlcg).a_cmp_const_reg_label(current_asmdata.CurrAsmList, opsize, OC_EQ, tcgint(t^._low.svalue),hregister, blocklabel(t^.blockid));
{ Reset last here, because we've only checked for one value and need to compare
for the next range both the lower and upper bound }
lastwasrange := false;
end
else
begin
{ it begins with the smallest label, if the value }
{ is even smaller then jump immediately to the }
{ ELSE-label }
if not lastwasrange or (t^._low-last>1) then
thlcgwasm(hlcg).a_cmp_const_reg_label(current_asmdata.CurrAsmList, opsize, jmp_lt, tcgint(t^._low.svalue), hregister, elselabel);
thlcgwasm(hlcg).a_cmp_const_reg_label(current_asmdata.CurrAsmList, opsize, jmp_le, tcgint(t^._high.svalue), hregister, blocklabel(t^.blockid));
last:=t^._high;
lastwasrange := true;
end;
if assigned(t^.greater) then
genitem(t^.greater);
end;
begin
last:=0;
lastwasrange:=false;
genitem(hp);
current_asmdata.CurrAsmList.concat(taicpu.op_sym(a_br,elselabel));
end;
procedure twasmcasenode.pass_generate_code;
var
oldflowcontrol: tflowcontrol;
ShortcutElse: Boolean;
i: Integer;
begin
location_reset(location,LOC_VOID,OS_NO);
oldflowcontrol := flowcontrol;
include(flowcontrol,fc_inflowcontrol);
current_asmdata.getjumplabel(endlabel);
current_asmdata.CurrAsmList.concat(taicpu.op_none(a_block));
{ Do some optimisation to deal with empty else blocks }
ShortcutElse := GetBranchLabel(elseblock, elselabel);
for i:=blocks.count-1 downto 0 do
with pcaseblock(blocks[i])^ do
shortcut := GetBranchLabel(statement, blocklabel);
with_sign:=is_signed(left.resultdef);
if with_sign then
begin
jmp_gt:=OC_GT;
jmp_lt:=OC_LT;
jmp_le:=OC_LTE;
end
else
begin
jmp_gt:=OC_A;
jmp_lt:=OC_B;
jmp_le:=OC_BE;
end;
secondpass(left);
if (left.expectloc=LOC_JUMP)<>
(left.location.loc=LOC_JUMP) then
internalerror(2006050501);
{ determines the size of the operand }
opsize:=left.resultdef;
{ copy the case expression to a register }
hlcg.location_force_reg(current_asmdata.CurrAsmList,left.location,left.resultdef,opsize,false);
{$if not defined(cpu64bitalu)}
if def_cgsize(opsize) in [OS_S64,OS_64] then
begin
hregister:=left.location.register64.reglo;
hregister2:=left.location.register64.reghi;
end
else
{$endif not cpu64bitalu and not cpuhighleveltarget}
hregister:=left.location.register;
{ we need the min_label always to choose between }
{ cmps and subs/decs }
min_label:=case_get_min(labels);
{ Generate the jumps }
{$if not defined(cpu64bitalu)}
if def_cgsize(opsize) in [OS_64,OS_S64] then
genlinearcmplist(labels)
else
{$endif not cpu64bitalu and not cpuhighleveltarget}
begin
//if cs_opt_level1 in current_settings.optimizerswitches then
// begin
// { procedures are empirically passed on }
// { consumption can also be calculated }
// { but does it pay on the different }
// { processors? }
// { moreover can the size only be appro- }
// { ximated as it is not known if rel8, }
// { rel16 or rel32 jumps are used }
//
// max_label := case_get_max(labels);
//
// { can we omit the range check of the jump table ? }
// getrange(left.resultdef,lv,hv);
// jumptable_no_range:=(lv=min_label) and (hv=max_label);
//
// distv:=max_label-min_label;
// if distv>=0 then
// dist:=distv.uvalue
// else
// dist:=asizeuint(-distv.svalue);
//
// { optimize for size ? }
// if cs_opt_size in current_settings.optimizerswitches then
// begin
// if has_jumptable and
// (min_label>=int64(low(aint))) and
// (max_label<=high(aint)) and
// not((labelcnt<=2) or
// (distv.svalue<0) or
// (dist>3*labelcnt)) then
// begin
// { if the labels less or more a continuum then }
// genjumptable(labels,min_label.svalue,max_label.svalue);
// end
// else
// begin
// { a linear list is always smaller than a jump tree }
// genlinearlist(labels);
// end;
// end
// else
// begin
// max_dist:=4*labelcoverage;
//
// { Don't allow jump tables to get too large }
// if max_dist>4*labelcnt then
// max_dist:=min(max_dist,2048);
//
// if jumptable_no_range then
// max_linear_list:=4
// else
// max_linear_list:=2;
//
// { allow processor specific values }
// optimizevalues(max_linear_list,max_dist);
//
// if (labelcnt<=max_linear_list) then
// genlinearlist(labels)
// else
// begin
// if (has_jumptable) and
// (dist<max_dist) and
// (min_label>=int64(low(aint))) and
// (max_label<=high(aint)) then
// genjumptable(labels,min_label.svalue,max_label.svalue)
// { value has been determined on an i7-4770 using a random case with random values
// if more values are known, this can be handled depending on the target CPU
//
// Testing on a Core 2 Duo E6850 as well as on a Raspi3 showed also, that 64 is
// a good value }
// else if labelcnt>=64 then
// genjmptree(labels)
// else
// genlinearlist(labels);
// end;
// end;
// end
//else
{ it's always not bad }
genlinearlist(labels);
end;
{ generate the instruction blocks }
for i:=0 to blocks.count-1 do with pcaseblock(blocks[i])^ do
begin
{ If the labels are not equal, then the block label has been shortcut to point elsewhere,
so there's no need to implement it }
if not shortcut then
begin
current_asmdata.CurrAsmList.concat(taicpu.op_none(a_end_block));
hlcg.a_label(current_asmdata.CurrAsmList,blocklabel);
secondpass(statement);
{ don't come back to case line }
current_filepos:=current_asmdata.CurrAsmList.getlasttaifilepos^;
current_asmdata.CurrAsmList.concat(taicpu.op_sym(a_br,endlabel));
end;
end;
{ ...and the else block }
if not ShortcutElse then
begin
current_asmdata.CurrAsmList.concat(taicpu.op_none(a_end_block));
hlcg.a_label(current_asmdata.CurrAsmList,elselabel);
end;
if Assigned(elseblock) then
begin
secondpass(elseblock);
end;
current_asmdata.CurrAsmList.concat(taicpu.op_none(a_end_block));
hlcg.a_label(current_asmdata.CurrAsmList,endlabel);
flowcontrol := oldflowcontrol + (flowcontrol - [fc_inflowcontrol]);
end;
begin
cinnode:=twasminnode;
ccasenode:=twasmcasenode;
end.