RuntimeLibcalls: Fix building hash table with duplicate entries #153801
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This was referenced Aug 15, 2025
Contributor
Author
Member
|
@llvm/pr-subscribers-llvm-ir @llvm/pr-subscribers-llvm-binary-utilities Author: Matt Arsenault (arsenm) ChangesWe were sizing the table appropriately for the number of LibcallImpls, BM_LookupRuntimeLibcallByNameRandomCalls improves by ~25% and As a secondary change, align the table size up to the next Full diff: https://github.com/llvm/llvm-project/pull/153801.diff 1 Files Affected:
diff --git a/llvm/utils/TableGen/Basic/RuntimeLibcallsEmitter.cpp b/llvm/utils/TableGen/Basic/RuntimeLibcallsEmitter.cpp
index 1c5f38d0c24b8..05f2512e24a50 100644
--- a/llvm/utils/TableGen/Basic/RuntimeLibcallsEmitter.cpp
+++ b/llvm/utils/TableGen/Basic/RuntimeLibcallsEmitter.cpp
@@ -289,7 +289,6 @@ class RuntimeLibcallEmitter {
/// Helper struct for the name hash table.
struct LookupEntry {
- StringRef FuncName;
uint64_t Hash = 0;
unsigned TableValue = 0;
};
@@ -339,14 +338,17 @@ static void emitHashFunction(raw_ostream &OS) {
/// Return the table size, maximum number of collisions for the set of hashes
static std::pair<int, int>
computePerfectHashParameters(ArrayRef<uint64_t> Hashes) {
- const int SizeOverhead = 10;
- const int NumHashes = Hashes.size();
+ // Chosen based on experimentation with llvm/benchmarks/RuntimeLibcalls.cpp
+ const int SizeOverhead = 4;
// Index derived from hash -> number of collisions.
DenseMap<uint64_t, int> Table;
+ unsigned NumHashes = Hashes.size();
+
for (int MaxCollisions = 1;; ++MaxCollisions) {
- for (int N = NumHashes; N < SizeOverhead * NumHashes; ++N) {
+ for (unsigned N = NextPowerOf2(NumHashes - 1); N < SizeOverhead * NumHashes;
+ N <<= 1) {
Table.clear();
bool NeedResize = false;
@@ -367,22 +369,12 @@ computePerfectHashParameters(ArrayRef<uint64_t> Hashes) {
static std::vector<LookupEntry>
constructPerfectHashTable(ArrayRef<RuntimeLibcallImpl> Keywords,
- ArrayRef<uint64_t> Hashes, int Size, int Collisions,
- StringToOffsetTable &OffsetTable) {
- DenseSet<StringRef> Seen;
+ ArrayRef<uint64_t> Hashes,
+ ArrayRef<unsigned> TableValues, int Size,
+ int Collisions, StringToOffsetTable &OffsetTable) {
std::vector<LookupEntry> Lookup(Size * Collisions);
- for (const RuntimeLibcallImpl &LibCallImpl : Keywords) {
- StringRef ImplName = LibCallImpl.getLibcallFuncName();
-
- // We do not want to add repeated entries for cases with the same name, only
- // an entry for the first, with the name collision enum values immediately
- // following.
- if (!Seen.insert(ImplName).second)
- continue;
-
- uint64_t HashValue = Hashes[LibCallImpl.getEnumVal() - 1];
-
+ for (auto [HashValue, TableValue] : zip(Hashes, TableValues)) {
uint64_t Idx = (HashValue % static_cast<uint64_t>(Size)) *
static_cast<uint64_t>(Collisions);
@@ -390,8 +382,7 @@ constructPerfectHashTable(ArrayRef<RuntimeLibcallImpl> Keywords,
for (int J = 0; J < Collisions; ++J) {
LookupEntry &Entry = Lookup[Idx + J];
if (Entry.TableValue == 0) {
- Entry.FuncName = ImplName;
- Entry.TableValue = LibCallImpl.getEnumVal();
+ Entry.TableValue = TableValue;
Entry.Hash = HashValue;
Found = true;
break;
@@ -399,7 +390,7 @@ constructPerfectHashTable(ArrayRef<RuntimeLibcallImpl> Keywords,
}
if (!Found)
- reportFatalInternalError("failure to hash " + ImplName);
+ reportFatalInternalError("failure to hash");
}
return Lookup;
@@ -409,15 +400,25 @@ constructPerfectHashTable(ArrayRef<RuntimeLibcallImpl> Keywords,
void RuntimeLibcallEmitter::emitNameMatchHashTable(
raw_ostream &OS, StringToOffsetTable &OffsetTable) const {
std::vector<uint64_t> Hashes(RuntimeLibcallImplDefList.size());
+ std::vector<unsigned> TableValues(RuntimeLibcallImplDefList.size());
+ DenseSet<StringRef> SeenFuncNames;
size_t MaxFuncNameSize = 0;
size_t Index = 0;
+
for (const RuntimeLibcallImpl &LibCallImpl : RuntimeLibcallImplDefList) {
StringRef ImplName = LibCallImpl.getLibcallFuncName();
- MaxFuncNameSize = std::max(MaxFuncNameSize, ImplName.size());
- Hashes[Index++] = hash(ImplName);
+ if (SeenFuncNames.insert(ImplName).second) {
+ MaxFuncNameSize = std::max(MaxFuncNameSize, ImplName.size());
+ TableValues[Index] = LibCallImpl.getEnumVal();
+ Hashes[Index++] = hash(ImplName);
+ }
}
+ // Trim excess elements from non-unique entries.
+ Hashes.resize(SeenFuncNames.size());
+ TableValues.resize(SeenFuncNames.size());
+
LLVM_DEBUG({
for (const RuntimeLibcallImpl &LibCallImpl : RuntimeLibcallImplDefList) {
StringRef ImplName = LibCallImpl.getLibcallFuncName();
@@ -447,8 +448,9 @@ void RuntimeLibcallEmitter::emitNameMatchHashTable(
"#endif\n";
auto [Size, Collisions] = computePerfectHashParameters(Hashes);
- std::vector<LookupEntry> Lookup = constructPerfectHashTable(
- RuntimeLibcallImplDefList, Hashes, Size, Collisions, OffsetTable);
+ std::vector<LookupEntry> Lookup =
+ constructPerfectHashTable(RuntimeLibcallImplDefList, Hashes, TableValues,
+ Size, Collisions, OffsetTable);
LLVM_DEBUG(dbgs() << "Runtime libcall perfect hashing parameters: Size = "
<< Size << ", maximum collisions = " << Collisions << '\n');
@@ -464,10 +466,13 @@ void RuntimeLibcallEmitter::emitNameMatchHashTable(
OS << " static constexpr uint16_t HashTableNameToEnum[" << Lookup.size()
<< "] = {\n";
- for (auto [FuncName, Hash, TableVal] : Lookup) {
+ for (auto [Hash, TableVal] : Lookup) {
OS << " " << TableVal << ',';
- if (TableVal != 0)
+ if (TableVal != 0) {
+ StringRef FuncName =
+ RuntimeLibcallImplDefList[TableVal].getLibcallFuncName();
OS << " // " << format_hex(Hash, 16) << ", " << FuncName;
+ }
OS << '\n';
}
|
Member
|
@llvm/pr-subscribers-tablegen Author: Matt Arsenault (arsenm) ChangesWe were sizing the table appropriately for the number of LibcallImpls, BM_LookupRuntimeLibcallByNameRandomCalls improves by ~25% and As a secondary change, align the table size up to the next Full diff: https://github.com/llvm/llvm-project/pull/153801.diff 1 Files Affected:
diff --git a/llvm/utils/TableGen/Basic/RuntimeLibcallsEmitter.cpp b/llvm/utils/TableGen/Basic/RuntimeLibcallsEmitter.cpp
index 1c5f38d0c24b8..05f2512e24a50 100644
--- a/llvm/utils/TableGen/Basic/RuntimeLibcallsEmitter.cpp
+++ b/llvm/utils/TableGen/Basic/RuntimeLibcallsEmitter.cpp
@@ -289,7 +289,6 @@ class RuntimeLibcallEmitter {
/// Helper struct for the name hash table.
struct LookupEntry {
- StringRef FuncName;
uint64_t Hash = 0;
unsigned TableValue = 0;
};
@@ -339,14 +338,17 @@ static void emitHashFunction(raw_ostream &OS) {
/// Return the table size, maximum number of collisions for the set of hashes
static std::pair<int, int>
computePerfectHashParameters(ArrayRef<uint64_t> Hashes) {
- const int SizeOverhead = 10;
- const int NumHashes = Hashes.size();
+ // Chosen based on experimentation with llvm/benchmarks/RuntimeLibcalls.cpp
+ const int SizeOverhead = 4;
// Index derived from hash -> number of collisions.
DenseMap<uint64_t, int> Table;
+ unsigned NumHashes = Hashes.size();
+
for (int MaxCollisions = 1;; ++MaxCollisions) {
- for (int N = NumHashes; N < SizeOverhead * NumHashes; ++N) {
+ for (unsigned N = NextPowerOf2(NumHashes - 1); N < SizeOverhead * NumHashes;
+ N <<= 1) {
Table.clear();
bool NeedResize = false;
@@ -367,22 +369,12 @@ computePerfectHashParameters(ArrayRef<uint64_t> Hashes) {
static std::vector<LookupEntry>
constructPerfectHashTable(ArrayRef<RuntimeLibcallImpl> Keywords,
- ArrayRef<uint64_t> Hashes, int Size, int Collisions,
- StringToOffsetTable &OffsetTable) {
- DenseSet<StringRef> Seen;
+ ArrayRef<uint64_t> Hashes,
+ ArrayRef<unsigned> TableValues, int Size,
+ int Collisions, StringToOffsetTable &OffsetTable) {
std::vector<LookupEntry> Lookup(Size * Collisions);
- for (const RuntimeLibcallImpl &LibCallImpl : Keywords) {
- StringRef ImplName = LibCallImpl.getLibcallFuncName();
-
- // We do not want to add repeated entries for cases with the same name, only
- // an entry for the first, with the name collision enum values immediately
- // following.
- if (!Seen.insert(ImplName).second)
- continue;
-
- uint64_t HashValue = Hashes[LibCallImpl.getEnumVal() - 1];
-
+ for (auto [HashValue, TableValue] : zip(Hashes, TableValues)) {
uint64_t Idx = (HashValue % static_cast<uint64_t>(Size)) *
static_cast<uint64_t>(Collisions);
@@ -390,8 +382,7 @@ constructPerfectHashTable(ArrayRef<RuntimeLibcallImpl> Keywords,
for (int J = 0; J < Collisions; ++J) {
LookupEntry &Entry = Lookup[Idx + J];
if (Entry.TableValue == 0) {
- Entry.FuncName = ImplName;
- Entry.TableValue = LibCallImpl.getEnumVal();
+ Entry.TableValue = TableValue;
Entry.Hash = HashValue;
Found = true;
break;
@@ -399,7 +390,7 @@ constructPerfectHashTable(ArrayRef<RuntimeLibcallImpl> Keywords,
}
if (!Found)
- reportFatalInternalError("failure to hash " + ImplName);
+ reportFatalInternalError("failure to hash");
}
return Lookup;
@@ -409,15 +400,25 @@ constructPerfectHashTable(ArrayRef<RuntimeLibcallImpl> Keywords,
void RuntimeLibcallEmitter::emitNameMatchHashTable(
raw_ostream &OS, StringToOffsetTable &OffsetTable) const {
std::vector<uint64_t> Hashes(RuntimeLibcallImplDefList.size());
+ std::vector<unsigned> TableValues(RuntimeLibcallImplDefList.size());
+ DenseSet<StringRef> SeenFuncNames;
size_t MaxFuncNameSize = 0;
size_t Index = 0;
+
for (const RuntimeLibcallImpl &LibCallImpl : RuntimeLibcallImplDefList) {
StringRef ImplName = LibCallImpl.getLibcallFuncName();
- MaxFuncNameSize = std::max(MaxFuncNameSize, ImplName.size());
- Hashes[Index++] = hash(ImplName);
+ if (SeenFuncNames.insert(ImplName).second) {
+ MaxFuncNameSize = std::max(MaxFuncNameSize, ImplName.size());
+ TableValues[Index] = LibCallImpl.getEnumVal();
+ Hashes[Index++] = hash(ImplName);
+ }
}
+ // Trim excess elements from non-unique entries.
+ Hashes.resize(SeenFuncNames.size());
+ TableValues.resize(SeenFuncNames.size());
+
LLVM_DEBUG({
for (const RuntimeLibcallImpl &LibCallImpl : RuntimeLibcallImplDefList) {
StringRef ImplName = LibCallImpl.getLibcallFuncName();
@@ -447,8 +448,9 @@ void RuntimeLibcallEmitter::emitNameMatchHashTable(
"#endif\n";
auto [Size, Collisions] = computePerfectHashParameters(Hashes);
- std::vector<LookupEntry> Lookup = constructPerfectHashTable(
- RuntimeLibcallImplDefList, Hashes, Size, Collisions, OffsetTable);
+ std::vector<LookupEntry> Lookup =
+ constructPerfectHashTable(RuntimeLibcallImplDefList, Hashes, TableValues,
+ Size, Collisions, OffsetTable);
LLVM_DEBUG(dbgs() << "Runtime libcall perfect hashing parameters: Size = "
<< Size << ", maximum collisions = " << Collisions << '\n');
@@ -464,10 +466,13 @@ void RuntimeLibcallEmitter::emitNameMatchHashTable(
OS << " static constexpr uint16_t HashTableNameToEnum[" << Lookup.size()
<< "] = {\n";
- for (auto [FuncName, Hash, TableVal] : Lookup) {
+ for (auto [Hash, TableVal] : Lookup) {
OS << " " << TableVal << ',';
- if (TableVal != 0)
+ if (TableVal != 0) {
+ StringRef FuncName =
+ RuntimeLibcallImplDefList[TableVal].getLibcallFuncName();
OS << " // " << format_hex(Hash, 16) << ", " << FuncName;
+ }
OS << '\n';
}
|
arsenm
force-pushed
the
users/arsenm/runtime-libcalls/store-libcall-string-lengths-table
branch
from
b6af40f to
f0c374c
Compare
arsenm
force-pushed
the
users/arsenm/runtime-libcalls/shrink-libcall-impl-name-hash-table-unique-entries
branch
from
f445a16 to
6e2b170
Compare
Base automatically changed from
users/arsenm/runtime-libcalls/store-libcall-string-lengths-table
to
main
August 15, 2025 14:29
We were sizing the table appropriately for the number of LibcallImpls, but many of those have identical names which were pushing up the collision count unnecessarily. This ends up decreasing the table size slightly, and makes it a bit faster. BM_LookupRuntimeLibcallByNameRandomCalls improves by ~25% and BM_LookupRuntimeLibcallByNameSampleData by ~5%. As a secondary change, align the table size up to the next power of 2. This makes the table larger than before, but improves the sample data benchmark by an additional 5%.
arsenm
force-pushed
the
users/arsenm/runtime-libcalls/shrink-libcall-impl-name-hash-table-unique-entries
branch
from
6e2b170 to
aaec1db
Compare
Contributor
Author
|
ping |
efriedma-quic
approved these changes
Aug 25, 2025
Collaborator
efriedma-quic
left a comment
efriedma-quic
left a comment
There was a problem hiding this comment.
Using a power-of-two hashtable size means you're really relying on a strong hash function; allowing non-power-of-two functions gives a lot more chances to cover up any imperfection. We might want to consider adding a diagnostic if MaxCollisions gets excessively high.
Otherwise LGTM
arsenm
deleted the
users/arsenm/runtime-libcalls/shrink-libcall-impl-name-hash-table-unique-entries
branch
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/116/builds/17418 Here is the relevant piece of the build log for the reference |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
We were sizing the table appropriately for the number of LibcallImpls,
but many of those have identical names which were pushing up the
collision count unnecessarily. This ends up decreasing the table size
slightly, and makes it a bit faster.
BM_LookupRuntimeLibcallByNameRandomCalls improves by ~25% and
BM_LookupRuntimeLibcallByNameSampleData by ~5%.
As a secondary change, align the table size up to the next
power of 2. This makes the table larger than before, but improves
the sample data benchmark by an additional 5%.