1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
|
{
Copyright (c) 2000-2002 by Florian Klaempfl
Code generation for add nodes on the i386
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 n386add;
{$i fpcdefs.inc}
interface
uses
node,nadd,cpubase,nx86add;
type
ti386addnode = class(tx86addnode)
function use_generic_mul32to64: boolean; override;
procedure second_addordinal; override;
procedure second_add64bit;override;
procedure second_cmp64bit;override;
procedure second_mul(unsigned: boolean);
end;
implementation
uses
globtype,systems,
cutils,verbose,globals,
symconst,symdef,paramgr,defutil,
aasmbase,aasmtai,aasmdata,aasmcpu,
cgbase,procinfo,
ncon,nset,cgutils,tgobj,
cga,ncgutil,cgobj,cg64f32,cgx86;
{*****************************************************************************
use_generic_mul32to64
*****************************************************************************}
function ti386addnode.use_generic_mul32to64: boolean;
begin
result := False;
end;
{ handles all unsigned multiplications, and 32->64 bit signed ones.
32bit-only signed mul is handled by generic codegen }
procedure ti386addnode.second_addordinal;
var
unsigned: boolean;
begin
unsigned:=not(is_signed(left.resultdef)) or
not(is_signed(right.resultdef));
if (nodetype=muln) and (unsigned or is_64bit(resultdef)) then
second_mul(unsigned)
else
inherited second_addordinal;
end;
{*****************************************************************************
Add64bit
*****************************************************************************}
procedure ti386addnode.second_add64bit;
var
op : TOpCG;
op1,op2 : TAsmOp;
opsize : TOpSize;
hregister,
hregister2 : tregister;
hl4 : tasmlabel;
mboverflow,
unsigned:boolean;
r:Tregister;
begin
firstcomplex(self);
pass_left_right;
op1:=A_NONE;
op2:=A_NONE;
mboverflow:=false;
opsize:=S_L;
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;
mboverflow:=true;
end;
subn :
begin
op:=OP_SUB;
op1:=A_SUB;
op2:=A_SBB;
mboverflow:=true;
end;
xorn:
op:=OP_XOR;
orn:
op:=OP_OR;
andn:
op:=OP_AND;
else
begin
{ everything should be handled in pass_1 (JM) }
internalerror(200109051);
end;
end;
{ left and right no register? }
{ then one must be demanded }
if (left.location.loc<>LOC_REGISTER) then
begin
if (right.location.loc<>LOC_REGISTER) then
begin
hregister:=cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
hregister2:=cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
cg64.a_load64_loc_reg(current_asmdata.CurrAsmList,left.location,joinreg64(hregister,hregister2));
location_reset(left.location,LOC_REGISTER,left.location.size);
left.location.register64.reglo:=hregister;
left.location.register64.reghi:=hregister2;
end
else
begin
location_swap(left.location,right.location);
toggleflag(nf_swapped);
end;
end;
{ at this point, left.location.loc should be LOC_REGISTER }
if right.location.loc=LOC_REGISTER then
begin
{ when swapped another result register }
if (nodetype=subn) and (nf_swapped in flags) then
begin
cg64.a_op64_reg_reg(current_asmdata.CurrAsmList,op,location.size,
left.location.register64,
right.location.register64);
location_swap(left.location,right.location);
toggleflag(nf_swapped);
end
else
begin
cg64.a_op64_reg_reg(current_asmdata.CurrAsmList,op,location.size,
right.location.register64,
left.location.register64);
end;
end
else
begin
{ right.location<>LOC_REGISTER }
if (nodetype=subn) and (nf_swapped in flags) then
begin
r:=cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
cg64.a_load64low_loc_reg(current_asmdata.CurrAsmList,right.location,r);
emit_reg_reg(op1,opsize,left.location.register64.reglo,r);
emit_reg_reg(A_MOV,opsize,r,left.location.register64.reglo);
cg64.a_load64high_loc_reg(current_asmdata.CurrAsmList,right.location,r);
{ the carry flag is still ok }
emit_reg_reg(op2,opsize,left.location.register64.reghi,r);
emit_reg_reg(A_MOV,opsize,r,left.location.register64.reghi);
end
else
begin
cg64.a_op64_loc_reg(current_asmdata.CurrAsmList,op,location.size,right.location,
left.location.register64);
end;
location_freetemp(current_asmdata.CurrAsmList,right.location);
end;
{ only in case of overflow operations }
{ produce overflow code }
{ we must put it here directly, because sign of operation }
{ is in unsigned VAR!! }
if mboverflow then
begin
if cs_check_overflow in current_settings.localswitches then
begin
current_asmdata.getjumplabel(hl4);
if unsigned then
cg.a_jmp_flags(current_asmdata.CurrAsmList,F_AE,hl4)
else
cg.a_jmp_flags(current_asmdata.CurrAsmList,F_NO,hl4);
cg.a_call_name(current_asmdata.CurrAsmList,'FPC_OVERFLOW',false);
cg.a_label(current_asmdata.CurrAsmList,hl4);
end;
end;
location_copy(location,left.location);
end;
procedure ti386addnode.second_cmp64bit;
var
hregister,
hregister2 : tregister;
href : treference;
unsigned : boolean;
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(unsigned),current_procinfo.CurrTrueLabel);
{ cheat a little bit for the negative test }
toggleflag(nf_swapped);
cg.a_jmp_flags(current_asmdata.CurrAsmList,getresflags(unsigned),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(unsigned),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(unsigned),current_procinfo.CurrFalseLabel);
nodetype:=oldnodetype;
end;
equaln:
cg.a_jmp_flags(current_asmdata.CurrAsmList,F_NE,current_procinfo.CurrFalseLabel);
unequaln:
cg.a_jmp_flags(current_asmdata.CurrAsmList,F_NE,current_procinfo.CurrTrueLabel);
end;
end;
procedure secondjmp64bitcmp;
begin
{ the jump the sequence is a little bit hairy }
case nodetype of
ltn,gtn,lten,gten:
begin
{ the comparisaion of the low dword have to be }
{ always unsigned! }
cg.a_jmp_flags(current_asmdata.CurrAsmList,getresflags(true),current_procinfo.CurrTrueLabel);
cg.a_jmp_always(current_asmdata.CurrAsmList,current_procinfo.CurrFalseLabel);
end;
equaln:
begin
cg.a_jmp_flags(current_asmdata.CurrAsmList,F_NE,current_procinfo.CurrFalseLabel);
cg.a_jmp_always(current_asmdata.CurrAsmList,current_procinfo.CurrTrueLabel);
end;
unequaln:
begin
cg.a_jmp_flags(current_asmdata.CurrAsmList,F_NE,current_procinfo.CurrTrueLabel);
cg.a_jmp_always(current_asmdata.CurrAsmList,current_procinfo.CurrFalseLabel);
end;
end;
end;
begin
firstcomplex(self);
pass_left_right;
unsigned:=((left.resultdef.typ=orddef) and
(torddef(left.resultdef).ordtype=u64bit)) or
((right.resultdef.typ=orddef) and
(torddef(right.resultdef).ordtype=u64bit));
{ left and right no register? }
{ then one must be demanded }
if (left.location.loc<>LOC_REGISTER) then
begin
if (right.location.loc<>LOC_REGISTER) then
begin
{ we can reuse a CREGISTER for comparison }
if (left.location.loc<>LOC_CREGISTER) then
begin
hregister:=cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
hregister2:=cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
cg64.a_load64_loc_reg(current_asmdata.CurrAsmList,left.location,joinreg64(hregister,hregister2));
location_freetemp(current_asmdata.CurrAsmList,left.location);
location_reset(left.location,LOC_REGISTER,left.location.size);
left.location.register64.reglo:=hregister;
left.location.register64.reghi:=hregister2;
end;
end
else
begin
location_swap(left.location,right.location);
toggleflag(nf_swapped);
end;
end;
{ at this point, left.location.loc should be LOC_REGISTER }
if right.location.loc=LOC_REGISTER then
begin
emit_reg_reg(A_CMP,S_L,right.location.register64.reghi,left.location.register64.reghi);
firstjmp64bitcmp;
emit_reg_reg(A_CMP,S_L,right.location.register64.reglo,left.location.register64.reglo);
secondjmp64bitcmp;
end
else
begin
case right.location.loc of
LOC_CREGISTER :
begin
emit_reg_reg(A_CMP,S_L,right.location.register64.reghi,left.location.register64.reghi);
firstjmp64bitcmp;
emit_reg_reg(A_CMP,S_L,right.location.register64.reglo,left.location.register64.reglo);
secondjmp64bitcmp;
end;
LOC_CREFERENCE,
LOC_REFERENCE :
begin
tcgx86(cg).make_simple_ref(current_asmdata.CurrAsmList,right.location.reference);
href:=right.location.reference;
inc(href.offset,4);
emit_ref_reg(A_CMP,S_L,href,left.location.register64.reghi);
firstjmp64bitcmp;
emit_ref_reg(A_CMP,S_L,right.location.reference,left.location.register64.reglo);
secondjmp64bitcmp;
cg.a_jmp_always(current_asmdata.CurrAsmList,current_procinfo.CurrFalseLabel);
location_freetemp(current_asmdata.CurrAsmList,right.location);
end;
LOC_CONSTANT :
begin
current_asmdata.CurrAsmList.concat(taicpu.op_const_reg(A_CMP,S_L,aint(hi(right.location.value64)),left.location.register64.reghi));
firstjmp64bitcmp;
current_asmdata.CurrAsmList.concat(taicpu.op_const_reg(A_CMP,S_L,aint(lo(right.location.value64)),left.location.register64.reglo));
secondjmp64bitcmp;
end;
else
internalerror(200203282);
end;
end;
{ we have LOC_JUMP as result }
location_reset(location,LOC_JUMP,OS_NO)
end;
{*****************************************************************************
x86 MUL
*****************************************************************************}
procedure ti386addnode.second_mul(unsigned: boolean);
var reg:Tregister;
ref:Treference;
use_ref:boolean;
hl4 : tasmlabel;
const
asmops: array[boolean] of tasmop = (A_IMUL, A_MUL);
begin
pass_left_right;
{The location.register will be filled in later (JM)}
location_reset(location,LOC_REGISTER,def_cgsize(resultdef));
{ Mul supports registers and references, so if not register/reference,
load the location into a register.
The variant of IMUL which is capable of doing 32->64 bits has the same restrictions. }
use_ref:=false;
if left.location.loc in [LOC_REGISTER,LOC_CREGISTER] then
reg:=left.location.register
else if left.location.loc in [LOC_REFERENCE,LOC_CREFERENCE] then
begin
tcgx86(cg).make_simple_ref(current_asmdata.CurrAsmList,left.location.reference);
ref:=left.location.reference;
use_ref:=true;
end
else
begin
{LOC_CONSTANT for example.}
reg:=cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
cg.a_load_loc_reg(current_asmdata.CurrAsmList,OS_INT,left.location,reg);
end;
{Allocate EAX.}
cg.getcpuregister(current_asmdata.CurrAsmList,NR_EAX);
{Load the right value.}
cg.a_load_loc_reg(current_asmdata.CurrAsmList,OS_INT,right.location,NR_EAX);
{Also allocate EDX, since it is also modified by a mul (JM).}
cg.getcpuregister(current_asmdata.CurrAsmList,NR_EDX);
if use_ref then
emit_ref(asmops[unsigned],S_L,ref)
else
emit_reg(asmops[unsigned],S_L,reg);
if (cs_check_overflow in current_settings.localswitches) and
{ 32->64 bit cannot overflow }
(not is_64bit(resultdef)) then
begin
current_asmdata.getjumplabel(hl4);
cg.a_jmp_flags(current_asmdata.CurrAsmList,F_AE,hl4);
cg.a_call_name(current_asmdata.CurrAsmList,'FPC_OVERFLOW',false);
cg.a_label(current_asmdata.CurrAsmList,hl4);
end;
{Free EAX,EDX}
cg.ungetcpuregister(current_asmdata.CurrAsmList,NR_EDX);
if is_64bit(resultdef) then
begin
{Allocate a couple of registers and store EDX:EAX into it}
location.register64.reghi := cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
cg.a_load_reg_reg(current_asmdata.CurrAsmList, OS_INT, OS_INT, NR_EDX, location.register64.reghi);
cg.ungetcpuregister(current_asmdata.CurrAsmList,NR_EAX);
location.register64.reglo := cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
cg.a_load_reg_reg(current_asmdata.CurrAsmList, OS_INT, OS_INT, NR_EAX, location.register64.reglo);
end
else
begin
{Allocate a new register and store the result in EAX in it.}
location.register:=cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
cg.ungetcpuregister(current_asmdata.CurrAsmList,NR_EAX);
cg.a_load_reg_reg(current_asmdata.CurrAsmList,OS_INT,OS_INT,NR_EAX,location.register);
end;
location_freetemp(current_asmdata.CurrAsmList,left.location);
location_freetemp(current_asmdata.CurrAsmList,right.location);
end;
begin
caddnode:=ti386addnode;
end.
|
