This is an automated email from the ASF dual-hosted git repository.
yiguolei pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/doris.git
The following commit(s) were added to refs/heads/master by this push:
new 790b771a49 [improvement](execute) Eliminate virtual function calls
when serializing and deserializing aggregate functions (#21427)
790b771a49 is described below
commit 790b771a4934f7d17330a865fc44ec82ee14bdce
Author: ZenoYang <[email protected]>
AuthorDate: Tue Jul 4 09:26:37 2023 +0800
[improvement](execute) Eliminate virtual function calls when serializing
and deserializing aggregate functions (#21427)
Eliminate virtual function calls when serializing and deserializing
aggregate functions.
For example, in AggregateFunctionUniq::deserialize_and_merge method,
calling read_pod_binary(ref, buf) in the for loop generates a large number of
virtual function calls.
void deserialize_and_merge(AggregateDataPtr __restrict place,
BufferReadable& buf,
Arena* arena) const override {
auto& set = this->data(place).set;
UInt64 size;
read_var_uint(size, buf);
set.rehash(size + set.size());
for (size_t i = 0; i < size; ++i) {
KeyType ref;
read_pod_binary(ref, buf);
set.insert(ref);
}
}
template <typename Type>
void read_pod_binary(Type& x, BufferReadable& buf) {
buf.read(reinterpret_cast<char*>(&x), sizeof(x));
}
BufferReadable has only one subclass, VectorBufferReader, so it is better
to implement the BufferReadable class directly.
The following sql was tested on SSB-flat dataset:
SELECT COUNT (DISTINCT lo_partkey), COUNT (DISTINCT lo_suppkey) FROM
lineorder_flat;
before: MergeTime: 415.398ms
after opt: MergeTime: 174.660ms
---
be/src/vec/common/string_buffer.hpp | 52 ++++++++++++++++---------------------
1 file changed, 23 insertions(+), 29 deletions(-)
diff --git a/be/src/vec/common/string_buffer.hpp
b/be/src/vec/common/string_buffer.hpp
index 63da8544c2..9a760e3a6b 100644
--- a/be/src/vec/common/string_buffer.hpp
+++ b/be/src/vec/common/string_buffer.hpp
@@ -24,36 +24,30 @@
#include "vec/common/string_ref.h"
namespace doris::vectorized {
-class BufferWritable {
-public:
- virtual void write(const char* data, int len) = 0;
- virtual void commit() = 0;
- virtual ~BufferWritable() = default;
- template <typename T>
- void write_number(T data) {
- fmt::memory_buffer buffer;
- fmt::format_to(buffer, "{}", data);
- write(buffer.data(), buffer.size());
- }
-};
-
-class VectorBufferWriter final : public BufferWritable {
+class BufferWritable final {
public:
- explicit VectorBufferWriter(ColumnString& vector)
+ explicit BufferWritable(ColumnString& vector)
: _data(vector.get_chars()), _offsets(vector.get_offsets()) {}
- void write(const char* data, int len) override {
+ inline void write(const char* data, int len) {
_data.insert(data, data + len);
_now_offset += len;
}
- void commit() override {
+ inline void commit() {
_offsets.push_back(_offsets.back() + _now_offset);
_now_offset = 0;
}
- ~VectorBufferWriter() override { DCHECK(_now_offset == 0); }
+ ~BufferWritable() { DCHECK(_now_offset == 0); }
+
+ template <typename T>
+ void write_number(T data) {
+ fmt::memory_buffer buffer;
+ fmt::format_to(buffer, "{}", data);
+ write(buffer.data(), buffer.size());
+ }
private:
ColumnString::Chars& _data;
@@ -61,25 +55,22 @@ private:
size_t _now_offset = 0;
};
-class BufferReadable {
-public:
- virtual ~BufferReadable() = default;
- virtual void read(char* data, int len) = 0;
- virtual StringRef read(int len) = 0;
-};
+using VectorBufferWriter = BufferWritable;
+using BufferWriter = BufferWritable;
-class VectorBufferReader final : public BufferReadable {
+class BufferReadable {
public:
- explicit VectorBufferReader(StringRef& ref) : _data(ref.data) {}
- explicit VectorBufferReader(StringRef&& ref) : _data(ref.data) {}
+ explicit BufferReadable(StringRef& ref) : _data(ref.data) {}
+ explicit BufferReadable(StringRef&& ref) : _data(ref.data) {}
+ ~BufferReadable() = default;
- StringRef read(int len) override {
+ inline StringRef read(int len) {
StringRef ref(_data, len);
_data += len;
return ref;
}
- void read(char* data, int len) override {
+ inline void read(char* data, int len) {
memcpy(data, _data, len);
_data += len;
}
@@ -88,4 +79,7 @@ private:
const char* _data;
};
+using VectorBufferReader = BufferReadable;
+using BufferReader = BufferReadable;
+
} // namespace doris::vectorized
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]