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
|
{
Copyright (c) 1998-2002 by Florian Klaempfl
Generate Xtensa 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 ncpumat;
{$i fpcdefs.inc}
interface
uses
cgbase,node,nmat,ncgmat;
type
tcpumoddivnode = class(tcgmoddivnode)
function first_moddivint: tnode; override;
procedure emit_div_reg_reg(signed: boolean; denum, num: tregister); override;
procedure emit_mod_reg_reg(signed: boolean; denum, num: tregister); override;
end;
tcpunotnode = class(tcgnotnode)
procedure second_boolean;override;
function pass_1: tnode;override;
end;
tcpuunaryminusnode = class(tcgunaryminusnode)
function pass_1: tnode; override;
procedure second_float;override;
end;
tcpushlshrnode = class(tcgshlshrnode)
procedure second_64bit;override;
function pass_1: tnode;override;
end;
implementation
uses
globtype,compinnr,
cutils,verbose,globals,constexp,
aasmbase,aasmcpu,aasmtai,aasmdata,
defutil,
symtype,symconst,symtable,
cgobj,hlcgobj,cgutils,
pass_2,procinfo,
ncon,ncnv,ncal,ninl,
cpubase,cpuinfo,
ncgutil,
nadd,pass_1,symdef;
{*****************************************************************************
TCPUMODDIVNODE
*****************************************************************************}
procedure tcpumoddivnode.emit_div_reg_reg(signed: boolean; denum, num: tregister);
var
op: TAsmOp;
begin
if signed then
op:=A_QUOS
else
op:=A_QUOU;
current_asmdata.CurrAsmList.Concat(taicpu.op_reg_reg_reg(op,num,num,denum));
end;
procedure tcpumoddivnode.emit_mod_reg_reg(signed: boolean; denum, num: tregister);
var
op: TAsmOp;
begin
if signed then
op:=A_REMS
else
op:=A_REMU;
current_asmdata.CurrAsmList.Concat(taicpu.op_reg_reg_reg(op,num,num,denum));
end;
function tcpumoddivnode.first_moddivint: tnode;
begin
if (not is_64bitint(resultdef)) and
(CPUXTENSA_HAS_DIV in cpu_capabilities[current_settings.cputype]) then
Result:=nil
else
result:=inherited;
end;
{*****************************************************************************
TCPUNOTNODE
*****************************************************************************}
function tcpunotnode.pass_1 : tnode;
begin
result:=nil;
firstpass(left);
expectloc:=LOC_REGISTER;
end;
procedure tcpunotnode.second_boolean;
var
tmpreg, hreg1, hreg2, hreg3: TRegister;
instr: taicpu;
begin
secondpass(left);
if is_64bit(resultdef) then
begin
if not(left.location.loc in [LOC_CREGISTER,LOC_REGISTER]) then
hlcg.location_force_reg(current_asmdata.CurrAsmList,left.location,resultdef,resultdef,false);
hreg1:=cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
cg.a_op_reg_reg_reg(current_asmdata.CurrAsmList,OP_OR,OS_INT,left.location.register64.reglo,left.location.register64.reghi,hreg1);
hreg2:=cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
cg.a_load_const_reg(current_asmdata.CurrAsmList,OS_INT,0,hreg2);
hreg3:=cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
if is_cbool(resultdef) then
cg.a_load_const_reg(current_asmdata.CurrAsmList,OS_INT,-1,hreg3)
else
cg.a_load_const_reg(current_asmdata.CurrAsmList,OS_INT,1,hreg3);
location_reset(location, LOC_REGISTER, def_cgsize(resultdef));
location.register64.reglo:=cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
location.register64.reghi:=cg.getintregister(current_asmdata.CurrAsmList,OS_INT);
if is_cbool(resultdef) then
begin
instr:=taicpu.op_reg_reg_reg(A_MOV,location.register64.reglo,hreg3,hreg1);
instr.condition:=C_EQZ;
current_asmdata.CurrAsmList.concat(instr);
instr:=taicpu.op_reg_reg_reg(A_MOV,location.register64.reghi,hreg3,hreg1);
instr.condition:=C_EQZ;
current_asmdata.CurrAsmList.concat(instr);
instr:=taicpu.op_reg_reg_reg(A_MOV,location.register64.reglo,hreg2,hreg1);
instr.condition:=C_NEZ;
current_asmdata.CurrAsmList.concat(instr);
instr:=taicpu.op_reg_reg_reg(A_MOV,location.register64.reghi,hreg2,hreg1);
instr.condition:=C_NEZ;
current_asmdata.CurrAsmList.concat(instr);
end
else
begin
cg.a_load_const_reg(current_asmdata.CurrAsmList,OS_INT,0,location.register64.reghi);
instr:=taicpu.op_reg_reg_reg(A_MOV,location.register64.reglo,hreg3,hreg1);
instr.condition:=C_EQZ;
current_asmdata.CurrAsmList.concat(instr);
instr:=taicpu.op_reg_reg_reg(A_MOV,location.register64.reglo,hreg2,hreg1);
instr.condition:=C_NEZ;
current_asmdata.CurrAsmList.concat(instr);
end
end
else
begin
location:=left.location;
hlcg.location_force_reg(current_asmdata.CurrAsmList,location,resultdef,resultdef,false);
if is_cbool(resultdef) then
begin
{ normalize }
hreg3:=cg.getintregister(current_asmdata.CurrAsmList,def_cgsize(resultdef));
cg.a_load_const_reg(current_asmdata.CurrAsmList,def_cgsize(resultdef),-1,hreg3);
instr:=taicpu.op_reg_reg_reg(A_MOV,location.register,hreg3,location.register);
instr.condition:=C_NEZ;
current_asmdata.CurrAsmList.concat(instr);
cg.a_op_reg_reg(current_asmdata.CurrAsmList,OP_NOT,def_cgsize(resultdef), location.register, location.register)
end
else
cg.a_op_const_reg_reg(current_asmdata.CurrAsmList,OP_XOR,def_cgsize(resultdef),1, location.register, location.register)
end;
end;
{*****************************************************************************
TXTENSAUNARYMINUSNODE
*****************************************************************************}
function tcpuunaryminusnode.pass_1: tnode;
var
procname: string[31];
fdef : tdef;
begin
Result:=nil;
if (current_settings.fputype=fpu_soft) and
(left.resultdef.typ=floatdef) then
begin
result:=nil;
firstpass(left);
expectloc:=LOC_REGISTER;
exit;
end;
result:=nil;
firstpass(left);
if codegenerror then
exit;
expectloc:=LOC_REGISTER;
end;
procedure tcpuunaryminusnode.second_float;
var
ai : taicpu;
begin
secondpass(left);
if (current_settings.fputype=fpu_soft) or (tfloatdef(left.resultdef).floattype<>s32real) or
not(FPUXTENSA_SINGLE in fpu_capabilities[current_settings.fputype]) then
begin
if not(left.location.loc in [LOC_CREGISTER,LOC_REGISTER]) then
hlcg.location_force_reg(current_asmdata.CurrAsmList,left.location,left.resultdef,left.resultdef,false);
location_reset(location,LOC_REGISTER,def_cgsize(resultdef));
if location.size in [OS_64,OS_S64,OS_F64] then
begin
location.register64.reglo:=cg.getintregister(current_asmdata.CurrAsmList,OS_32);
location.register64.reghi:=cg.getintregister(current_asmdata.CurrAsmList,OS_32);
end
else
location.register:=cg.getintregister(current_asmdata.CurrAsmList,location.size);
case location.size of
OS_32:
cg.a_op_const_reg_reg(current_asmdata.CurrAsmList,OP_XOR,OS_32,tcgint($80000000),left.location.register,location.register);
OS_64:
begin
cg.a_op_const_reg_reg(current_asmdata.CurrAsmList,OP_XOR,OS_32,tcgint($80000000),left.location.registerhi,location.registerhi);
cg.a_load_reg_reg(current_asmdata.CurrAsmList,OS_32,OS_32,left.location.register64.reglo,location.register64.reglo);
end;
else
internalerror(2014033101);
end;
end
else
begin
if not(left.location.loc in [LOC_CFPUREGISTER,LOC_FPUREGISTER]) then
hlcg.location_force_fpureg(current_asmdata.CurrAsmList,left.location,left.resultdef,false);
location_reset(location,LOC_FPUREGISTER,def_cgsize(resultdef));
location.register:=cg.getfpuregister(current_asmdata.CurrAsmList,location.size);
ai:=taicpu.op_reg_reg(A_NEG,location.register,left.location.register);
ai.oppostfix := PF_S;
current_asmdata.CurrAsmList.Concat(ai);
end;
end;
function tcpushlshrnode.pass_1 : tnode;
begin
{ the xtensa code generator can handle 64 bit shifts by constants directly }
if is_constintnode(right) and is_64bit(resultdef) and
(((nodetype=shln) and (tordconstnode(right).value>=0) and (tordconstnode(right).value<=16)) or
((nodetype=shrn) and (tordconstnode(right).value>0) and (tordconstnode(right).value<16)) or
(tordconstnode(right).value=32)) then
begin
result:=nil;
firstpass(left);
firstpass(right);
if codegenerror then
exit;
expectloc:=LOC_REGISTER;
end
else
Result:=inherited pass_1;
end;
procedure tcpushlshrnode.second_64bit;
var
op: topcg;
opsize: TCgSize;
opdef: tdef;
shiftval: longint;
hcountreg: TRegister;
begin
{ determine operator }
case nodetype of
shln: op:=OP_SHL;
shrn: op:=OP_SHR;
else
internalerror(2020082208);
end;
opsize:=left.location.size;
opdef:=left.resultdef;
if not(left.location.loc in [LOC_CREGISTER,LOC_REGISTER]) or
{ location_force_reg can be also used to change the size of a register }
(left.location.size<>opsize) then
hlcg.location_force_reg(current_asmdata.CurrAsmList,left.location,left.resultdef,opdef,true);
location_reset(location,LOC_REGISTER,opsize);
location.register:=cg.getintregister(current_asmdata.CurrAsmList,OS_32);
location.registerhi:=cg.getintregister(current_asmdata.CurrAsmList,OS_32);
{ shifting by a constant directly coded: }
if right.nodetype=ordconstn then
begin
{ shl/shr must "wrap around", so use ... and 31 }
{ In TP, "byte/word shl 16 = 0", so no "and 15" in case of
a 16 bit ALU }
if tcgsize2size[opsize]<=4 then
shiftval:=tordconstnode(right).value.uvalue and 31
else
shiftval:=tordconstnode(right).value.uvalue and 63;
cg64.a_op64_const_reg_reg(current_asmdata.CurrAsmList,op,location.size,
shiftval,left.location.register64,location.register64)
end
else
begin
internalerror(2020082209);
{ load right operators in a register - this
is done since most target cpu which will use this
node do not support a shift count in a mem. location (cec)
}
hlcg.location_force_reg(current_asmdata.CurrAsmList,right.location,right.resultdef,sinttype,true);
hlcg.a_op_reg_reg_reg(current_asmdata.CurrAsmList,op,opdef,right.location.register,left.location.register,location.register);
end;
{ shl/shr nodes return the same type as left, which can be different
from opdef }
if opdef<>resultdef then
begin
hcountreg:=hlcg.getintregister(current_asmdata.CurrAsmList,resultdef);
hlcg.a_load_reg_reg(current_asmdata.CurrAsmList,opdef,resultdef,location.register,hcountreg);
location.register:=hcountreg;
end;
end;
begin
cmoddivnode:=tcpumoddivnode;
cnotnode:=tcpunotnode;
cunaryminusnode:=tcpuunaryminusnode;
cshlshrnode:=tcpushlshrnode;
end.
|