/
abi-examples.html
907 lines (778 loc) · 23 KB
/
abi-examples.html
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
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
<HTML>
<HEAD>
<title>C++ ABI for IA-64: Code and Implementation Examples</title>
<link rel=stylesheet href=small-table.css type="text/css">
<link rel=stylesheet href=code.css type="text/css">
<hr>
<font size=6><i><b>
<p>
C++ ABI for IA-64: Code and Implementation Examples
</b></i></font>
<font size=-1>
<p>
<i>Revised 5 September 2000</i>
</center>
</HEAD>
<BODY>
<p> <hr> <p>
<a name=intro>
<h3> Introduction </h3>
In this document,
we provide a number of code examples as illustration or tests
for the ABI specifications.
<p> <hr> <p>
<h3> Contents </h3>
<ul>
<li> <a href=#bases> Base Class Layout </a>
<li> <a href=#vcall> Virtual Function Calls </a>
<li> <a href=#vtable-ctor> Virtual Tables During Object Construction </a>
<li> <a href=#mangling> External Name Mangling </a>
<li> <a href=#vague> Vague Linkage </a>
</ul>
<p> <hr> <p>
<a name=bases>
<h3> Base Class Layout </h3>
<p>
The following test programs check for correct sharing of the virtual
pointer with a virtual base class.
<p><hr width=50%><p>
<code><pre>
/*
Test case for sharing virtual bases.
In Derived_too,
the primary base class is NewShareme,
The bases Base and Shareme share vptr's
with Derived and are allocated at the
same offset as Derived.
Should get:
60% a.out
(long)(NewShareme *)dd - (long)dd = 0
(long)(Derived *)dd - (long)dd = 8
(long)(Base *)dd - (long)dd = 8
(long)(Shareme *)dd - (long)dd = 8
*/
struct Shareme {
virtual void foo();
};
struct Base : virtual Shareme {
virtual void bar();
};
struct Derived : virtual Base {
virtual void baz();
};
struct NewShareme {
virtual void foo();
};
struct Derived_too : virtual NewShareme, virtual Derived {
virtual void bar();
};
void Shareme::foo() { }
void Base::bar() { }
void Derived::baz() { }
void NewShareme::foo() { }
void Derived_too::bar() { }
extern "C" int printf(const char *,...);
#define EVAL(EXPR) printf( #EXPR " = %d\n", (EXPR) );
main()
{
Derived_too *dd = new Derived_too;
EVAL((long)(NewShareme *)dd - (long)dd);
EVAL((long)(Derived *)dd - (long)dd);
EVAL((long)(Base *)dd - (long)dd);
EVAL((long)(Shareme *)dd - (long)dd);
}
</code></pre>
<p><hr width=50%><p>
<code><pre>
/*
Test case for sharing virtual bases.
In Most_Derived,
the primary base class is Nonvirt1,
Nonvirt2 and Nonvirt3 share vptrs with
virtual base Shared_Virt. Shared_Virt
should be at the same offset as Nonvirt2.
Should get:
67% a.out
(long)(Nonvirt1 *)dd - (long)dd = 0
(long)(Nonvirt2 *)dd - (long)dd = 8
(long)(Nonvirt3 *)dd - (long)dd = 16
(long)(Shared_Virt *)dd - (long)dd = 8
*/
struct Shared_Virt {
virtual void foo();
};
struct Nonvirt2 : virtual Shared_Virt {
virtual void bar();
};
struct Nonvirt3 : virtual Shared_Virt {
virtual void baz();
};
struct Nonvirt1 {
virtual void foo();
};
struct Most_Derived : Nonvirt1, Nonvirt2, Nonvirt3 {
virtual void bar();
};
void Shared_Virt::foo() { }
void Nonvirt2::bar() { }
void Nonvirt3::baz() { }
void Nonvirt1::foo() { }
void Most_Derived::bar() { }
extern "C" int printf(const char *,...);
#define EVAL(EXPR) printf( #EXPR " = %d\n", (EXPR) );
main()
{
Most_Derived *dd = new Most_Derived;
EVAL((long)(Nonvirt1 *)dd - (long)dd);
EVAL((long)(Nonvirt2 *)dd - (long)dd);
EVAL((long)(Nonvirt3 *)dd - (long)dd);
EVAL((long)(Shared_Virt *)dd - (long)dd);
}
</code></pre>
<p><hr width=50%><p>
<code><pre>
/*
Test case for sharing virtual bases.
In Most_Derived, share the vptr with
Interface1 but not Interface3, since
Interface3 is indirectly inherited.
Should get:
(long)(Interface1 *)dd - (long)dd = 0
(long)(Interface2 *)dd - (long)dd = 8
(long)(Interface3 *)dd - (long)dd = 8
(long)(Concrete1 *)dd - (long)dd = 8
*/
struct Interface1 {
virtual void foo();
};
struct Interface2 : virtual Interface1 {
virtual void bar();
};
struct Interface3 : virtual Interface2 {
virtual void baz();
};
struct Concrete1 : virtual Interface3 {
virtual void foo();
int i; // important.
};
struct Most_Derived : virtual Interface1,
virtual Interface2,
virtual Concrete1 {
virtual void bar();
};
void Interface1::foo() { }
void Interface2::bar() { }
void Interface3::baz() { }
void Concrete1::foo() { }
void Most_Derived::bar() { }
extern "C" int printf(const char *,...);
#define EVAL(EXPR) printf( #EXPR " = %d\n", (EXPR) );
main()
{
Most_Derived *dd = new Most_Derived;
EVAL((long)(Interface1 *)dd - (long)dd);
EVAL((long)(Interface2 *)dd - (long)dd);
EVAL((long)(Interface3 *)dd - (long)dd);
EVAL((long)(Concrete1 *)dd - (long)dd);
}
</code></pre>
<p> <hr> <p>
<a name=vcall>
<h3> Virtual Function Calls </h3>
<p>
The following class definitions are intended to illustrate various cases
relevant to the entries found in vtables to support virtual functions.
We define a number of classes, many of which define a virtual function f.
We are interested only in the vtable contents supporting calls of f;
other contents of the classes are generally elided,
except where necessary to emphasize some situation,
and we do not attempt to describe other contents of the vtables,
including precise layout.
<p>
The left column of this table contains the class definitions.
The right column(s) describe the corresponding vtable contents as they
are required to support calls to f.
We use the notation <code>&C::f</code> to mean a vtable entry for the
instance of f defined in class C,
in the form defined by the ABI (currently a function pointer/GP pair).
We use the notation <code>&thunk(A*,C::f)</code> to mean a vtable entry
for a thunk which converts <code>this</code> from A* to C* and then
calls C::f.
We use the notation <code>vcall(A*,C*,f)</code> to mean a vcall offset
that is added to an A* to produce a C* for use by an f thunk.
<p>
<table border=1>
<tr>
<th> Class definition </th>
<th> Vtable 1 contents </th>
<th> Vtable 2 contents </th>
</tr>
<tr class=small>
<td><code><pre>
struct X {
virtual void u();
}
</pre></code></td>
<td colspan=2 class=tiny>
Class X is an auxiliary class to be used only to prevent interesting
classes from being primary bases later.
</td>
</tr>
<tr class=small>
<td rowspan=2><code><pre>
struct A {
virtual void f();
}
</pre></code></td>
<td> <i>Vtable A</i> </td>
</tr>
<tr class=small>
<td> &A::f </td>
</tr>
<tr class=small>
<td rowspan=2><code><pre>
struct B: public A { }
</pre></code></td>
<td> <i>Vtable A_in_B</i> </td>
</tr>
<tr class=small>
<td> &A::f </td>
</tr>
<tr class=small>
<td rowspan=2><code><pre>
struct C: public A {
virtual void f();
}
</pre></code></td>
<td> <i>Vtable A_in_C (primary)</i> </td>
</tr>
<tr class=small>
<td> &C::f </td>
</tr>
<tr class=small>
<td rowspan=2><code><pre>
struct D:
public X, public A
{ }
</pre></code></td>
<td> <i>Vtable A_in_D</i> </td>
</tr>
<tr class=small>
<td> &A::f </td>
</tr>
<tr class=small>
<td rowspan=2><code><pre>
struct E:
public X, public A
{
virtual void f();
}
</pre></code></td>
<td> <i>Vtable A_in_E</i> </td>
<td> <i>Vtable E</i> </td>
</tr>
<tr class=small>
<td> &thunk(A*,E::f) </td>
<td> &E::f </td>
</tr>
<tr class=small>
<td rowspan=2><code><pre>
struct G:
public virtual A
{ }
</pre></code></td>
<td> <i>Vtable A_in_G</i> </td>
</tr>
<tr class=small>
<td> &A::f </td>
</tr>
<tr class=small>
<td rowspan=2><code><pre>
struct H:
public X,
public virtual A
{
virtual void f();
}
</pre></code></td>
<td> <i>Vtable A_in_H</i> </td>
<td> <i>Vtable H</i> </td>
</tr>
<tr class=small>
<td> &thunk(A*,H::f)<br>vcall(H::A*,H*,f) </td>
<td> &H::f </td>
</tr>
<tr class=small>
<td rowspan=2><code><pre>
struct I:
public H
{ int i; }
</pre></code></td>
<td> <i>Vtable A_in_I</i> </td>
<td> <i>Vtable H_in_I</i> </td>
</tr>
<tr class=small>
<td> &thunk(A*,H::f)<br>vcall(I::A*,I::H*,f) </td>
<td> &H::f </td>
</tr>
</table>
<p>
<a name=vcall-impl>
<h4> Implementation </h4>
<p>
There are several possible implementations of the thunks given the
above information.
Note in the following that we assume that prior to calling any vtable entry,
the <code>this</code> pointer has been adjusted to point to the
subobject corresponding to the vtable from which the vptr is fetched.
<ol type=A>
<p>
<li>
Since the offsets are always known at compile time,
even for virtual bases,
each thunk could be distinct,
adding the known offset to <code>this</code> and branching to the
target function.
<p>
This would result in a thunk for each overrider at a distinct offset.
As a result, a branch mispredict and possibly an instruction cache miss
would occur each time the actual type changed for a reference at any
given point in the code.
<p>
<li>
In the case of virtual inheritance,
the offset, although known when the overrider is declared,
may differ depending on derivations from the overrider's class.
H and I above are the simplest example.
H is a primary base for I,
but the int member of I means that A is at a different offset from H in
I than it was from a standalone H.
Because of this, the ABI specifies that the secondary vtable for a
virtual base A contain a vcall offset to H,
so that a shared thunk can load the vcall offset,
adding it to <code>this</code>,
and branch to the target function H::f.
<p>
This would result in fewer thunks,
since for a inheritance hierarchy where A is a virtual base of H,
and H::f overrides A::f,
all instances of H in a larger hierarchy can use the same thunk.
As a result, these thunks will cause fewer branch mispredictions and
instruction cache misses.
The tradeoff is that they must do a load before the offset add.
Since the offset is smaller than the code for a thunk,
the load should miss in cache less frequently,
so better cache miss behavior should produce better results in spite of
the 2 or more cycles required for the vcall offset load.
<p>
<li>
In the case of non-virtual inheritance,
when an overrider is declared,
the entire set of associated overriding and overridden functions f is
known along with their associated vcall offsets,
and they are all constants.
Therefore, it is possible to cascade the offset adds as follows.
Suppose we have A::f, overridden by B::f at offset -16,
and finally overridden by C::f at offset -32 from B, -48 from A.
Assuming that <code>this</code> is in register <code>out0</code>,
the IA-64 code can look like:
<code><pre>
thunk-A-to-C::f :
add out0 = -16,out0 ;;
thunk-A-to-C::f :
add out0 = -32,out0 ;;
C::f :
# normal entry code...
</pre></code>
<p>
This code will leave us with about the same number of branch
mispredictions for the thunk calls,
but should eliminate a large proportion of the instruction cache misses,
and costs one cycle per level in the adjustment sequence.
Note that there is no particular required order of the thunk entries,
so the compiler could optimize by putting the most frequent one closest
to the target function and so on.
Also, if the sequence gets too long,
any entry can add its entire offset at once and branch directly to the
</ol>
<p>
In order to make these implementations practical,
the ABI must specify:
<ul>
<li> where the vcall offsets are allocated in the secondary vtables,
<li> where the thunks are required to be emitted, and
<li> the names (mangled) of the thunks for reference elsewhere.
</ul>
The emission of the thunks is the most relevant issue here --
the others are independent of this discussion.
Cascading thunks as in (C) only works if they appear with the
definition of the overriding function,
so that seems to be the right answer.
<p>
Assuming that all of the above optimizations are implemented,
we note a number of benefits of the design specified.
<ul>
<p>
<li>
In the case of single, non-virtual inheritance,
calling a virtual function requires no adjustment to the `this' pointer,
and no vcall offsets.
This is in keeping with the guiding principle that
"you shouldn't pay for features you don't use."
<p>
<li>
If the static type of an object is the same as its dynamic type,
then no adjustment to the `this' pointer is required.
<p>
<li>
All thunks can be emitted in the same translation unit as the
overriding function.
<p>
<li>
Thunks from classes that are not morally virtual bases do not
require a branch to the non-adjusting entry point.
(We say that a subobject X is a "morally virtual" base of Y if X is an
indirect or direct base class of Y,
and if X is either a virtual base of Y,
or the direct or indirect base of a virtual base of Y.)
<p>
Suppose we have classes M1 and M2, base classes of a third class V.
Suppose V is a virtual base of C,
and A2 and A1 are non-virtual bases of C.
Suppose all of these classes have definitions of virtual function f.
We suggest the following implementation:
<code><pre>
m2: /* Thunk for morally virtual base M2. */
this += offsetof (V, M1) - offsetof (V, M2)
m1: /* Thunk for morally virtual base M1. */
this -= offsetof (V, M1)
v: /* Thunk for virtual base V. */
this += vcall offset stored in V vtable
goto f;
a2: /* Thunk for non-virtual base A2. */
this += offsetof (B, A1) - offsetof (B, A2)
/* Fall through. */
a1: /* Thunk for non-virtual base A1. */
this += offsetof (B, C) - offsetof (B, A1)
/* Fall through. */
f: /* Non-adjusting entry point. */
</code></pre>
<p>
(Here `offsetof' is a compile-time computable function that gives
the offset of its second parameter in its first parameter.)
<p>
(Alternatively the `v' entry point above could be of the form:
<code><pre>
v: /* Thunk for virtual base V. */
this += vcall offset stored in V vtable
- offsetof (B, C) + offsetof (B, A2)
/* Fall through. */
</code></pre>
Which alternative is better depends on how many adds follow at this
point.
In general, if many adds remain before the non-adjusting entry point,
it may be better to suffer the consequences of the indirect branch.)
<p>
In this way, a virtual call through a base A1, A2, etc.,
that is not a virtual base of C
(or a direct or indirect base of a virtual base of C),
does not require an additional branch,
and is therefore more likely to avoid icache misses.
Even the thunk for V may avoid severe icache penalties
since it is located near the non-adjusting entry point for f.
Furthermore, if there are no non-virtual bases,
then the sequence can become just:
<code><pre>
v: /* Thunk for virtual base V. */
this += vcall offset stored in V vtable
f: /* Non-adjusting entry point. */
</code></pre>
</ul>
<p> <hr> <p>
<a name=vtable-ctor>
<h3> Vtables During Object Construction</h3>
<p>
Following is a test program from Compaq, that breaks on many compilers.
<code><pre>
/*
This test program should output:
V1 called
V2 called
C called
C::foo called 7
PASSED this == cp
D called
~C called
C::foo called 7
PASSED this == cp
~V2 called
~V1 called
Int caught
*/
extern "C" int printf(const char *,...);
struct V1 {
int v;
virtual int foo();
V1();
~V1();
};
struct V2 : virtual V1 {
int v2;
virtual int foo();
V2();
~V2();
};
struct C : virtual V1, virtual V2 {
int c;
virtual int foo();
C();
~C();
};
struct B {
int b; };
struct D : B, C {
int d;
virtual int bar();
D();
~D();
};
extern "C" int printf(const char *,...);
main()
{
try {
D *d = new D;
delete d;
} catch (int) {
printf("Int caught\n");
}
}
int V1::foo() {
printf("V1::foo called\n"); return 1; }
V1::V1() : v(5) {
printf("V1 called\n"); }
V1::~V1() {
printf("~V1 called\n"); }
int V2::foo() {
printf("V2::foo called\n"); return 1; }
V2::V2() : v2(6) {
printf("V2 called\n"); }
V2::~V2() {
printf("~V2 called\n"); }
int C::foo() {
printf("C::foo called %d\n", c); return 1; }
C::C() : c(7) {
printf("C called\n");
V1 *vv = this; vv->foo();
C *cp = dynamic_cast<C *>(vv);
if (this == cp) {
printf("PASSED this == cp\n");
} else {
printf("FAILED this != cp\n");
}
}
C::~C() {
printf("~C called\n");
V1 *vv = this; vv->foo();
C *cp = dynamic_cast<C *>(vv);
if (this == cp) {
printf("PASSED this == cp\n");
} else {
printf("FAILED this != cp\n");
}
}
int D::bar() {
printf("D::bar called\n"); return 1; }
D::D() : d(8) {
printf("D called\n"); throw 5; }
D::~D() {
printf("~D called\n"); }
</pre></code>
</font>
<p> <hr> <p>
<a name=mangling>
<h3> External Name Mangling</h3>
<p>
In the table below,
<code>Ret?</code> in the source name indicates a function for which
the return type is not part of the mangling.
<code>Type?</code> indicates a data object
(for which the type is never part of the mangling).
Spaces appear in some of the mangled names to assist a human parser --
they are not part of the actual mangled name, and should be ignored.
<p>
<table border=1 cellpadding=4>
<tr><th> Mangled name <br><i>(ignore spaces)</i> </th>
<th> Source name </th> </tr>
<tr class=small><td> f </td>
<td> C function or variable "f" or a global namespace variable "f"</td></tr>
<tr class=small><td> _Z1fv </td>
<td> Ret? f(); <i>or</i> Ret? f(void); </td></tr>
<tr class=small><td> _Z1fi </td>
<td> Ret? f(int); </td></tr>
<tr class=small><td> _Z3foo3bar </td>
<td> Ret? foo(bar); </td></tr>
<tr class=small><td> _Zrm1XS_ </td>
<td> Ret? operator%(X, X); </td></tr>
<tr class=small><td> _ZplR1XS0_ </td>
<td> Ret? operator+(X&, X&); </td></tr>
<tr class=small><td> _ZlsRK1XS1_ </td>
<td> Ret? operator<< (X const&, X const&);
<br> (Note: X is S_, X const is S0_)
</td></tr>
<tr class=small><td> _ZN3FooIA4_iE3barE </td>
<td> Type? Foo<int[4]>::bar; </td></tr>
<tr class=small><td> _Z1fIiEvi </td>
<td> void f<int>(/*nondependent*/int);
<br> (Note: the return type is always explicitly encoded for template
functions taking parameters.)</td></tr>
<tr class=small><td> _Z5firstI3DuoEvS0_ </td>
<td> void first<Duo>(/*nondependent*/Duo);
<br>
(Note: first template is S_, Duo is S0_, first<Duo>(Duo) is S1_.
<br>Since the function parameter is not dependent, don't use T_.)
</td></tr>
<tr class=small><td> _Z5firstI3DuoEvT_ </td>
<td> void first<Duo>(/*T1=*/Duo); </td></tr>
<tr class=small><td> _Z3fooIiPFidEiEvv </td>
<td> void foo<int,int(*)(double),int>();
<br>(Note: return type encoded for template function.)</td></tr>
<tr class=small><td> _ZN1N1fE </td>
<td> Type? N::f </td></tr>
<tr class=small><td> _ZN6System5Sound4beepEv </td>
<td> Ret? System::Sound::beep(); </td></tr>
<tr class=small><td> _ZN5Arena5levelE </td>
<td> Type? Arena::level; </td></tr>
<tr class=small><td> _ZN5StackIiiE5levelE </td>
<td> Type? Stack<int, int>::level; </td></tr>
<tr class=small><td> _Z1fI1XE vPV N1AIT_E1TE </td>
<td> void f<X>(A</*T1=*/X>::T volatile*); </td></tr>
<tr class=small><td>
_ZngILi42EE v N1A I XplT_Li2EE E 1TE </td>
<td> void operator-</*int J=*/42>(A<J+2>::T); </td></tr>
<tr class=small><td> _Z4makeI7FactoryiE T_IT0_E v </td>
<td> /*T1=*/Factory</*T2=*/int> make<Factory, int>();
<br> (Note: T_ = factory (a template), T0_ = int) </td></tr>
<tr class=small><td> _Z3foo 5Hello5WorldS0_S_ </td>
<td> Type? foo(Hello,World,World,Hello)
<br> (Note: Hello is S_, World is S0_, foo(...) is S1_)
</td></tr>
<tr class=small><td> _Z3fooPM2ABi </td>
<td> foo(int AB::**)
<br> // M is a pointer, P adds another level </td></tr>
<tr class=small><td> _ZlsRSoRKSs </td>
<td> operator<< (std::ostream&,std::string const&) </td></tr>
<tr class=small><td> _ZTI7a_class </td>
<td> typeid(class a_class) </td></tr>
</table>
<font color=red>
<p>
<a name=mangling-ex>
<h4> Local Scope Mangling</h4>
<p>
Following are several examples illustrating local scope mangling,
all based on the following code snippet:
<pre><code>
class A {
void foo (int) {
class B { };
}
};
void foo () {
class C {
class D { };
void bar () {
struct E {
void baz() { }
};
}
};
}
</code></pre>
<p>
First consider the mangling of A::foo::B,
which is the case of a simple local name (B) in a nested function.
<pre><code>
<encoding> ::- <name>
::- <local-name>
::- Z <function encoding> E <entity name>
</code></pre>
where:
<pre><code>
<function encoding> ::- <name> <bare-function-type>
::- <nested-name> <bare-function-type>
...
::- N <unqualified-name> <unqualified-name> E <bare-function-type>
::- N 1A 3foo E i
<entity name> ::- <name>
::- 1B
</code></pre>
<p>
So the final result is "Z N 1A 3foo E i E 1B".
The next example is foo::C::D, i.e. a nested local name in a simple function:
<pre><code>
<encoding> ::- <name>
::- <local-name>
::- Z <function encoding> E <entity name>
</code></pre>
where:
<pre><code>
<function encoding> ::- <name> <bare-function-type>
::- 3foo v
<entity name> ::- <nested-name>
...
::- N <unqualified-name> <unqualified-name> E
::- N 1C 1D E
</code></pre>
<p>
So the final result is "Z 3foo v E N 1C 1D E".
Finally, as an example of nested local scopes, consider foo::C::bar::E::baz:
<pre><code>
<encoding> ::- <name>
::- <local-name>
::- Z <function encoding> E <entity name>
</code></pre>
where:
<pre><code>
<function encoding> ::- <name> <bare-function-type> (bar)
::- <local-name> v
::- Z <function encoding> E <entity name> v
::- Z <name> <bare-function-type> E <nested-name> v
...
::- Z 3foo v E N 1C 3bar E v
<entity name> ::- <function name>
::- <function name> <bare-function-type>
::- <nested-name> <bare-function-type>
...
::- N <unqualified-name> <unqualified-name> E <bare-function-type>
::- N 1E 3baz E v
</code></pre>
yielding the final result: <pre><code>
Z Z 3foo v E N 1C 3bar E v E N 1E 3baz E v
</code></pre>
<p> <hr> <p>
<a name=vague>
<h3> Vague Linkage</h3>
<p>
COMDAT groups are a new gABI feature specified during the IA-64 ABI
definition, and may not be implemented by all vendors immediately.
Pending their availability,
other implementations may be required for the features specified to use
COMDAT in the ABI.
Using weak symbols for the objects requiring weak linkage instead of
putting them in COMDAT groups will normally work,
even in interaction with correctly-defined COMDAT group representations,
although it will not have the space efficiency of COMDAT.
However,
weak symbol semantics are not consistent among Unix implementations,
and implementors should be careful that whenever multiple symbols are
specified to reside in a single COMDAT, and might interact
(e.g. thunks falling through to primary function definitions),
all are generated together as weak symbols as well.
</font>
<p>
<hr>
<p>
Please send corrections to <a href="mailto:mark@codesourcery.com">Mark
Mitchell</a>.
</BODY>
</HTML>