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
|
{
Copyright (c) 1998-2002 by Florian Klaempfl
Generate ARM assembler for math 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 narmmat;
{$i fpcdefs.inc}
interface
uses
node,nmat,ncgmat;
type
tarmmoddivnode = class(tmoddivnode)
function first_moddivint: tnode;override;
procedure pass_generate_code;override;
end;
tarmnotnode = class(tcgnotnode)
procedure second_boolean;override;
end;
tarmunaryminusnode = class(tcgunaryminusnode)
procedure second_float;override;
end;
implementation
uses
globtype,systems,
cutils,verbose,globals,constexp,
aasmbase,aasmcpu,aasmtai,aasmdata,
defutil,
cgbase,cgobj,hlcgobj,cgutils,
pass_2,procinfo,
ncon,
cpubase,cpuinfo,
ncgutil,cgcpu,
nadd,pass_1,symdef;
{*****************************************************************************
TARMMODDIVNODE
*****************************************************************************}
function tarmmoddivnode.first_moddivint: tnode;
var
power : longint;
begin
if (right.nodetype=ordconstn) and
(nodetype=divn) and
(ispowerof2(tordconstnode(right).value,power) or
(tordconstnode(right).value=1) or
(tordconstnode(right).value=int64(-1))
) and
not(is_64bitint(resultdef)) then
result:=nil
else if (current_settings.cputype in [cpu_armv7m]) and
(nodetype=divn) and
not(is_64bitint(resultdef)) then
result:=nil
else if (current_settings.cputype in [cpu_armv7m]) and
(nodetype=modn) and
not(is_64bitint(resultdef)) then
begin
if (right.nodetype=ordconstn) and
ispowerof2(tordconstnode(right).value,power) and
(tordconstnode(right).value<=256) and
(tordconstnode(right).value>0) then
result:=caddnode.create(andn,left,cordconstnode.create(tordconstnode(right).value-1,sinttype,false))
else
begin
result:=caddnode.create(subn,left,caddnode.create(muln,right.getcopy, cmoddivnode.Create(divn,left.getcopy,right.getcopy)));
right:=nil;
end;
left:=nil;
end
else
result:=inherited first_moddivint;
end;
procedure tarmmoddivnode.pass_generate_code;
var
power : longint;
numerator,
helper1,
helper2,
resultreg : tregister;
size : Tcgsize;
so : tshifterop;
procedure genOrdConstNodeDiv;
begin
if tordconstnode(right).value=0 then
internalerror(2005061701)
else if tordconstnode(right).value=1 then
cg.a_load_reg_reg(current_asmdata.CurrAsmList, OS_INT, OS_INT, numerator, resultreg)
else if (tordconstnode(right).value = int64(-1)) then
begin
// note: only in the signed case possible..., may overflow
current_asmdata.CurrAsmList.concat(setoppostfix(taicpu.op_reg_reg(A_MVN,
resultreg,numerator),toppostfix(ord(cs_check_overflow in current_settings.localswitches)*ord(PF_S))));
end
else if ispowerof2(tordconstnode(right).value,power) then
begin
if (is_signed(right.resultdef)) then
begin
helper1:=cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
helper2:=cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
shifterop_reset(so);
so.shiftmode:=SM_ASR;
so.shiftimm:=31;
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_shifterop(A_MOV,helper1,numerator,so));
shifterop_reset(so);
so.shiftmode:=SM_LSR;
so.shiftimm:=32-power;
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_reg_shifterop(A_ADD,helper2,numerator,helper1,so));
shifterop_reset(so);
so.shiftmode:=SM_ASR;
so.shiftimm:=power;
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_shifterop(A_MOV,resultreg,helper2,so));
end
else
cg.a_op_const_reg_reg(current_asmdata.CurrAsmList,OP_SHR,OS_INT,power,numerator,resultreg)
end;
end;
{
procedure genOrdConstNodeMod;
var
modreg, maskreg, tempreg : tregister;
begin
if (tordconstnode(right).value = 0) then begin
internalerror(2005061702);
end
else if (abs(tordconstnode(right).value.svalue) = 1) then
begin
// x mod +/-1 is always zero
cg.a_load_const_reg(current_asmdata.CurrAsmList, OS_INT, 0, resultreg);
end
else if (ispowerof2(tordconstnode(right).value, power)) then
begin
if (is_signed(right.resultdef)) then begin
tempreg := cg.getintregister(current_asmdata.CurrAsmList, OS_INT);
maskreg := cg.getintregister(current_asmdata.CurrAsmList, OS_INT);
modreg := cg.getintregister(current_asmdata.CurrAsmList, OS_INT);
cg.a_load_const_reg(current_asmdata.CurrAsmList, OS_INT, abs(tordconstnode(right).value.svalue)-1, modreg);
cg.a_op_const_reg_reg(current_asmdata.CurrAsmList, OP_SAR, OS_INT, 31, numerator, maskreg);
cg.a_op_reg_reg_reg(current_asmdata.CurrAsmList, OP_AND, OS_INT, numerator, modreg, tempreg);
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_reg(A_ANDC, maskreg, maskreg, modreg));
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_const(A_SUBFIC, modreg, tempreg, 0));
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg_reg(A_SUBFE, modreg, modreg, modreg));
cg.a_op_reg_reg_reg(current_asmdata.CurrAsmList, OP_AND, OS_INT, modreg, maskreg, maskreg);
cg.a_op_reg_reg_reg(current_asmdata.CurrAsmList, OP_OR, OS_INT, maskreg, tempreg, resultreg);
end else begin
cg.a_op_const_reg_reg(current_asmdata.CurrAsmList, OP_AND, OS_INT, tordconstnode(right).value.svalue-1, numerator, resultreg);
end;
end else begin
genOrdConstNodeDiv();
cg.a_op_const_reg_reg(current_asmdata.CurrAsmList, OP_MUL, OS_INT, tordconstnode(right).value.svalue, resultreg, resultreg);
cg.a_op_reg_reg_reg(current_asmdata.CurrAsmList, OP_SUB, OS_INT, resultreg, numerator, resultreg);
end;
end;
}
begin
secondpass(left);
secondpass(right);
if (current_settings.cputype in [cpu_armv7m]) and
(nodetype=divn) and
not(is_64bitint(resultdef)) then
begin
size:=def_cgsize(left.resultdef);
hlcg.location_force_reg(current_asmdata.CurrAsmList,left.location,left.resultdef,left.resultdef,true);
location_copy(location,left.location);
location.loc := LOC_REGISTER;
location.register := cg.getintregister(current_asmdata.CurrAsmList,size);
resultreg:=location.register;
if (right.nodetype=ordconstn) and
((tordconstnode(right).value=1) or
(tordconstnode(right).value=int64(-1)) or
(tordconstnode(right).value=0) or
ispowerof2(tordconstnode(right).value,power)) then
begin
numerator:=left.location.register;
genOrdConstNodeDiv;
end
else
begin
hlcg.location_force_reg(current_asmdata.CurrAsmList,right.location,right.resultdef,left.resultdef,true);
if is_signed(left.resultdef) or
is_signed(right.resultdef) then
cg.a_op_reg_reg_reg(current_asmdata.CurrAsmList,OP_IDIV,OS_INT,right.location.register,left.location.register,location.register)
else
cg.a_op_reg_reg_reg(current_asmdata.CurrAsmList,OP_DIV,OS_INT,right.location.register,left.location.register,location.register);
end;
end
else
begin
location_copy(location,left.location);
{ put numerator in register }
size:=def_cgsize(left.resultdef);
hlcg.location_force_reg(current_asmdata.CurrAsmList,left.location,
left.resultdef,left.resultdef,true);
location_copy(location,left.location);
numerator:=location.register;
resultreg:=location.register;
if location.loc=LOC_CREGISTER then
begin
location.loc := LOC_REGISTER;
location.register := cg.getintregister(current_asmdata.CurrAsmList,size);
resultreg:=location.register;
end
else if (nodetype=modn) or (right.nodetype=ordconstn) then
begin
// for a modulus op, and for const nodes we need the result register
// to be an extra register
resultreg:=cg.getintregister(current_asmdata.CurrAsmList,size);
end;
if right.nodetype=ordconstn then
begin
if nodetype=divn then
genOrdConstNodeDiv
else
// genOrdConstNodeMod;
end;
location.register:=resultreg;
end;
{ unsigned division/module can only overflow in case of division by zero }
{ (but checking this overflow flag is more convoluted than performing a }
{ simple comparison with 0) }
if is_signed(right.resultdef) then
cg.g_overflowcheck(current_asmdata.CurrAsmList,location,resultdef);
end;
{*****************************************************************************
TARMNOTNODE
*****************************************************************************}
procedure tarmnotnode.second_boolean;
var
hl : tasmlabel;
begin
{ if the location is LOC_JUMP, we do the secondpass after the
labels are allocated
}
if left.expectloc=LOC_JUMP then
begin
hl:=current_procinfo.CurrTrueLabel;
current_procinfo.CurrTrueLabel:=current_procinfo.CurrFalseLabel;
current_procinfo.CurrFalseLabel:=hl;
secondpass(left);
maketojumpbool(current_asmdata.CurrAsmList,left,lr_load_regvars);
hl:=current_procinfo.CurrTrueLabel;
current_procinfo.CurrTrueLabel:=current_procinfo.CurrFalseLabel;
current_procinfo.CurrFalseLabel:=hl;
location.loc:=LOC_JUMP;
end
else
begin
secondpass(left);
case left.location.loc of
LOC_FLAGS :
begin
location_copy(location,left.location);
inverse_flags(location.resflags);
end;
LOC_REGISTER,LOC_CREGISTER,LOC_REFERENCE,LOC_CREFERENCE,
LOC_SUBSETREG,LOC_CSUBSETREG,LOC_SUBSETREF,LOC_CSUBSETREF :
begin
hlcg.location_force_reg(current_asmdata.CurrAsmList,left.location,left.resultdef,left.resultdef,true);
current_asmdata.CurrAsmList.concat(taicpu.op_reg_const(A_CMP,left.location.register,0));
location_reset(location,LOC_FLAGS,OS_NO);
location.resflags:=F_EQ;
end;
else
internalerror(2003042401);
end;
end;
end;
{*****************************************************************************
TARMUNARYMINUSNODE
*****************************************************************************}
procedure tarmunaryminusnode.second_float;
var
op: tasmop;
begin
secondpass(left);
case current_settings.fputype of
fpu_fpa,
fpu_fpa10,
fpu_fpa11:
begin
location_force_fpureg(current_asmdata.CurrAsmList,left.location,false);
location:=left.location;
current_asmdata.CurrAsmList.concat(setoppostfix(taicpu.op_reg_reg_const(A_RSF,
location.register,left.location.register,0),
cgsize2fpuoppostfix[def_cgsize(resultdef)]));
end;
fpu_vfpv2,
fpu_vfpv3,
fpu_vfpv3_d16:
begin
location_force_mmregscalar(current_asmdata.CurrAsmList,left.location,true);
location:=left.location;
if (left.location.loc=LOC_CMMREGISTER) then
location.register:=cg.getmmregister(current_asmdata.CurrAsmList,location.size);
if (location.size=OS_F32) then
op:=A_FNEGS
else
op:=A_FNEGD;
current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg(op,
location.register,left.location.register));
end;
else
internalerror(2009112602);
end;
end;
begin
cmoddivnode:=tarmmoddivnode;
cnotnode:=tarmnotnode;
cunaryminusnode:=tarmunaryminusnode;
end.
|