1//=-- llvm/CodeGen/DwarfAccelTable.cpp - Dwarf Accelerator Tables -*- C++ -*-=//
2//
3//                     The LLVM Compiler Infrastructure
4//
5// This file is distributed under the University of Illinois Open Source
6// License. See LICENSE.TXT for details.
7//
8//===----------------------------------------------------------------------===//
9//
10// This file contains support for writing dwarf accelerator tables.
11//
12//===----------------------------------------------------------------------===//
13
14#include "DwarfAccelTable.h"
15#include "DwarfDebug.h"
16#include "DIE.h"
17#include "llvm/ADT/Twine.h"
18#include "llvm/ADT/STLExtras.h"
19#include "llvm/CodeGen/AsmPrinter.h"
20#include "llvm/MC/MCExpr.h"
21#include "llvm/MC/MCStreamer.h"
22#include "llvm/MC/MCSymbol.h"
23#include "llvm/Support/Debug.h"
24
25using namespace llvm;
26
27const char *DwarfAccelTable::Atom::AtomTypeString(enum AtomType AT) {
28  switch (AT) {
29  case eAtomTypeNULL: return "eAtomTypeNULL";
30  case eAtomTypeDIEOffset: return "eAtomTypeDIEOffset";
31  case eAtomTypeCUOffset: return "eAtomTypeCUOffset";
32  case eAtomTypeTag: return "eAtomTypeTag";
33  case eAtomTypeNameFlags: return "eAtomTypeNameFlags";
34  case eAtomTypeTypeFlags: return "eAtomTypeTypeFlags";
35  }
36  llvm_unreachable("invalid AtomType!");
37}
38
39// The length of the header data is always going to be 4 + 4 + 4*NumAtoms.
40DwarfAccelTable::DwarfAccelTable(ArrayRef<DwarfAccelTable::Atom> atomList) :
41  Header(8 + (atomList.size() * 4)),
42  HeaderData(atomList),
43  Entries(Allocator) { }
44
45DwarfAccelTable::~DwarfAccelTable() { }
46
47void DwarfAccelTable::AddName(StringRef Name, DIE* die, char Flags) {
48  assert(Data.empty() && "Already finalized!");
49  // If the string is in the list already then add this die to the list
50  // otherwise add a new one.
51  DataArray &DIEs = Entries[Name];
52  DIEs.push_back(new (Allocator) HashDataContents(die, Flags));
53}
54
55void DwarfAccelTable::ComputeBucketCount(void) {
56  // First get the number of unique hashes.
57  std::vector<uint32_t> uniques(Data.size());
58  for (size_t i = 0, e = Data.size(); i < e; ++i)
59    uniques[i] = Data[i]->HashValue;
60  array_pod_sort(uniques.begin(), uniques.end());
61  std::vector<uint32_t>::iterator p =
62    std::unique(uniques.begin(), uniques.end());
63  uint32_t num = std::distance(uniques.begin(), p);
64
65  // Then compute the bucket size, minimum of 1 bucket.
66  if (num > 1024) Header.bucket_count = num/4;
67  if (num > 16) Header.bucket_count = num/2;
68  else Header.bucket_count = num > 0 ? num : 1;
69
70  Header.hashes_count = num;
71}
72
73// compareDIEs - comparison predicate that sorts DIEs by their offset.
74static bool compareDIEs(const DwarfAccelTable::HashDataContents *A,
75                        const DwarfAccelTable::HashDataContents *B) {
76  return A->Die->getOffset() < B->Die->getOffset();
77}
78
79void DwarfAccelTable::FinalizeTable(AsmPrinter *Asm, const char *Prefix) {
80  // Create the individual hash data outputs.
81  for (StringMap<DataArray>::iterator
82         EI = Entries.begin(), EE = Entries.end(); EI != EE; ++EI) {
83
84    // Unique the entries.
85    std::stable_sort(EI->second.begin(), EI->second.end(), compareDIEs);
86    EI->second.erase(std::unique(EI->second.begin(), EI->second.end()),
87                       EI->second.end());
88
89    HashData *Entry = new (Allocator) HashData(EI->getKey(), EI->second);
90    Data.push_back(Entry);
91  }
92
93  // Figure out how many buckets we need, then compute the bucket
94  // contents and the final ordering. We'll emit the hashes and offsets
95  // by doing a walk during the emission phase. We add temporary
96  // symbols to the data so that we can reference them during the offset
97  // later, we'll emit them when we emit the data.
98  ComputeBucketCount();
99
100  // Compute bucket contents and final ordering.
101  Buckets.resize(Header.bucket_count);
102  for (size_t i = 0, e = Data.size(); i < e; ++i) {
103    uint32_t bucket = Data[i]->HashValue % Header.bucket_count;
104    Buckets[bucket].push_back(Data[i]);
105    Data[i]->Sym = Asm->GetTempSymbol(Prefix, i);
106  }
107}
108
109// Emits the header for the table via the AsmPrinter.
110void DwarfAccelTable::EmitHeader(AsmPrinter *Asm) {
111  Asm->OutStreamer.AddComment("Header Magic");
112  Asm->EmitInt32(Header.magic);
113  Asm->OutStreamer.AddComment("Header Version");
114  Asm->EmitInt16(Header.version);
115  Asm->OutStreamer.AddComment("Header Hash Function");
116  Asm->EmitInt16(Header.hash_function);
117  Asm->OutStreamer.AddComment("Header Bucket Count");
118  Asm->EmitInt32(Header.bucket_count);
119  Asm->OutStreamer.AddComment("Header Hash Count");
120  Asm->EmitInt32(Header.hashes_count);
121  Asm->OutStreamer.AddComment("Header Data Length");
122  Asm->EmitInt32(Header.header_data_len);
123  Asm->OutStreamer.AddComment("HeaderData Die Offset Base");
124  Asm->EmitInt32(HeaderData.die_offset_base);
125  Asm->OutStreamer.AddComment("HeaderData Atom Count");
126  Asm->EmitInt32(HeaderData.Atoms.size());
127  for (size_t i = 0; i < HeaderData.Atoms.size(); i++) {
128    Atom A = HeaderData.Atoms[i];
129    Asm->OutStreamer.AddComment(Atom::AtomTypeString(A.type));
130    Asm->EmitInt16(A.type);
131    Asm->OutStreamer.AddComment(dwarf::FormEncodingString(A.form));
132    Asm->EmitInt16(A.form);
133  }
134}
135
136// Walk through and emit the buckets for the table. This will look
137// like a list of numbers of how many elements are in each bucket.
138void DwarfAccelTable::EmitBuckets(AsmPrinter *Asm) {
139  unsigned index = 0;
140  for (size_t i = 0, e = Buckets.size(); i < e; ++i) {
141    Asm->OutStreamer.AddComment("Bucket " + Twine(i));
142    if (Buckets[i].size() != 0)
143      Asm->EmitInt32(index);
144    else
145      Asm->EmitInt32(UINT32_MAX);
146    index += Buckets[i].size();
147  }
148}
149
150// Walk through the buckets and emit the individual hashes for each
151// bucket.
152void DwarfAccelTable::EmitHashes(AsmPrinter *Asm) {
153  for (size_t i = 0, e = Buckets.size(); i < e; ++i) {
154    for (HashList::const_iterator HI = Buckets[i].begin(),
155           HE = Buckets[i].end(); HI != HE; ++HI) {
156      Asm->OutStreamer.AddComment("Hash in Bucket " + Twine(i));
157      Asm->EmitInt32((*HI)->HashValue);
158    }
159  }
160}
161
162// Walk through the buckets and emit the individual offsets for each
163// element in each bucket. This is done via a symbol subtraction from the
164// beginning of the section. The non-section symbol will be output later
165// when we emit the actual data.
166void DwarfAccelTable::EmitOffsets(AsmPrinter *Asm, MCSymbol *SecBegin) {
167  for (size_t i = 0, e = Buckets.size(); i < e; ++i) {
168    for (HashList::const_iterator HI = Buckets[i].begin(),
169           HE = Buckets[i].end(); HI != HE; ++HI) {
170      Asm->OutStreamer.AddComment("Offset in Bucket " + Twine(i));
171      MCContext &Context = Asm->OutStreamer.getContext();
172      const MCExpr *Sub =
173        MCBinaryExpr::CreateSub(MCSymbolRefExpr::Create((*HI)->Sym, Context),
174                                MCSymbolRefExpr::Create(SecBegin, Context),
175                                Context);
176      Asm->OutStreamer.EmitValue(Sub, sizeof(uint32_t), 0);
177    }
178  }
179}
180
181// Walk through the buckets and emit the full data for each element in
182// the bucket. For the string case emit the dies and the various offsets.
183// Terminate each HashData bucket with 0.
184void DwarfAccelTable::EmitData(AsmPrinter *Asm, DwarfDebug *D) {
185  uint64_t PrevHash = UINT64_MAX;
186  for (size_t i = 0, e = Buckets.size(); i < e; ++i) {
187    for (HashList::const_iterator HI = Buckets[i].begin(),
188           HE = Buckets[i].end(); HI != HE; ++HI) {
189      // Remember to emit the label for our offset.
190      Asm->OutStreamer.EmitLabel((*HI)->Sym);
191      Asm->OutStreamer.AddComment((*HI)->Str);
192      Asm->EmitSectionOffset(D->getStringPoolEntry((*HI)->Str),
193                             D->getStringPool());
194      Asm->OutStreamer.AddComment("Num DIEs");
195      Asm->EmitInt32((*HI)->Data.size());
196      for (ArrayRef<HashDataContents*>::const_iterator
197             DI = (*HI)->Data.begin(), DE = (*HI)->Data.end();
198           DI != DE; ++DI) {
199        // Emit the DIE offset
200        Asm->EmitInt32((*DI)->Die->getOffset());
201        // If we have multiple Atoms emit that info too.
202        // FIXME: A bit of a hack, we either emit only one atom or all info.
203        if (HeaderData.Atoms.size() > 1) {
204          Asm->EmitInt16((*DI)->Die->getTag());
205          Asm->EmitInt8((*DI)->Flags);
206        }
207      }
208      // Emit a 0 to terminate the data unless we have a hash collision.
209      if (PrevHash != (*HI)->HashValue)
210        Asm->EmitInt32(0);
211      PrevHash = (*HI)->HashValue;
212    }
213  }
214}
215
216// Emit the entire data structure to the output file.
217void DwarfAccelTable::Emit(AsmPrinter *Asm, MCSymbol *SecBegin,
218                           DwarfDebug *D) {
219  // Emit the header.
220  EmitHeader(Asm);
221
222  // Emit the buckets.
223  EmitBuckets(Asm);
224
225  // Emit the hashes.
226  EmitHashes(Asm);
227
228  // Emit the offsets.
229  EmitOffsets(Asm, SecBegin);
230
231  // Emit the hash data.
232  EmitData(Asm, D);
233}
234
235#ifndef NDEBUG
236void DwarfAccelTable::print(raw_ostream &O) {
237
238  Header.print(O);
239  HeaderData.print(O);
240
241  O << "Entries: \n";
242  for (StringMap<DataArray>::const_iterator
243         EI = Entries.begin(), EE = Entries.end(); EI != EE; ++EI) {
244    O << "Name: " << EI->getKeyData() << "\n";
245    for (DataArray::const_iterator DI = EI->second.begin(),
246           DE = EI->second.end();
247         DI != DE; ++DI)
248      (*DI)->print(O);
249  }
250
251  O << "Buckets and Hashes: \n";
252  for (size_t i = 0, e = Buckets.size(); i < e; ++i)
253    for (HashList::const_iterator HI = Buckets[i].begin(),
254           HE = Buckets[i].end(); HI != HE; ++HI)
255      (*HI)->print(O);
256
257  O << "Data: \n";
258    for (std::vector<HashData*>::const_iterator
259           DI = Data.begin(), DE = Data.end(); DI != DE; ++DI)
260      (*DI)->print(O);
261
262
263}
264#endif
265