# Optimization for number of goroutines using feedback control Yusuke MIYAKE / Pepabo R&D Institute, GMO Pepabo, Inc. 2019.07.25 GopherCon 2019 ![Image](/uploads/documents/4/2/d/f/42dfd7f07b1ef4

target - Uses type-erasure under the hood to gain all the magical powers • Utilizes small-size optimization in case target size is within certain limits ## Our Goals • Is to schedule function on the task

Kyuubi and nothing more. The Kyuubi server-side or the corresponding engines could do most of the optimization. On the other hand, we don’t wholly restrict end-users to special handling of specific cases to in Kyuubi ##### 2.2.1. The Basics of AQE Spark Adaptive Query Execution (AQE) is a query re-optimization that occurs during query execution. In terms of technical architecture, the AQE is a framework storage, actually without performing the shuffle across the network. The local shuffle reader optimization consists of avoiding shuffle when the SortMerge Join transforms to BroadcastHash Join after applying

rows By default peewee will cache the rows returned when iterating of a SelectQuery. This is an optimization to allow multiple iterations as well as indexing and slicing without causing additional queries peewee model instances from raw cursor data. See the naive() documentation for more details on this optimization. for stat in stats.naive().iterator(): 序列izer.serialize_object(stat) You can also see performance

c6ad37b/p9_1.jpg) Andreas Fertig Fast and small C++ 12 ## I mplementing the Small String Optimization struct string { size_t mSize(); size_t mCapacity(); char* mData(); char mSSOBuffer[16]; 7f4d50006c6ad37b/p9_2.jpg) Andreas Fertig ## I mplementing the Small String Optimization - libstdc++ struct string { char* mPtr; size_t mSize(); union { size_t mCapacity; 48c47c017b7f4d50006c6ad37b/p10_1.jpg) Fast and small C++ ## I mplementing the Small String Optimization - MS STL struct string { union { char* mPtr; char mBuf[8]; }; size_t

## BENCHMARK • 03a, b – branch is not predicted, optimization works • 03c, d – branch is well-predicted, no optimization ## ADVANCED OPTIMIZATION – ALWAYS MEASURE • Sometimes the compiler will do a - Often using “conditional move” instructions (they are not branches) • Compiler’s branchless optimization is usually better than yours • In particular, this is almost always branchless in reality: return Optimizations depend on the compiler! ## BENCHMARK • 04c, d – optimization does not work with GCC • with perf – no branch ## ADVANCED OPTIMIZATION – ALWAYS MEASURE • Sometimes the compiler will not do a branchless

feature! ## Conclusion: Lessons Learned ## std::�mod is up and running ■ Compilers make a decent optimization job ■ Multiply your native simd size with a factor depending on your algorithm for best results

I MIGHT NOT BE RUNNING THE CODE ON JUST ONE ARCHITECTURE - IN ORDER TO GET THE BEST POSSIBLE OPTIMIZATION, I NEED THE BEST POSSIBLE INFORMATION ABOUT ALL OF THE CODE ## SOMETIMES JIT IS MORE PRODUCTIVE