未验证 提交 abb9b25f 编写于 作者: A alexey-milovidov 提交者: GitHub

Merge pull request #16343 from ClickHouse/fix-perf-test-functions-logical

Fix performance test "functions logical" after move to clang
...@@ -290,10 +290,9 @@ private: ...@@ -290,10 +290,9 @@ private:
/// Apply target function by feeding it "batches" of N columns /// Apply target function by feeding it "batches" of N columns
/// Combining 10 columns per pass is the fastest for large columns sizes. /// Combining 8 columns per pass is the fastest method, because it's the maximum when clang vectorizes a loop.
/// For small columns sizes - more columns is faster.
template < template <
typename Op, template <typename, size_t> typename OperationApplierImpl, size_t N = 10> typename Op, template <typename, size_t> typename OperationApplierImpl, size_t N = 8>
struct OperationApplier struct OperationApplier
{ {
template <typename Columns, typename ResultData> template <typename Columns, typename ResultData>
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册