attractivechaos / benchmarks Goto Github PK

Summary

std::vector is very fast. It is almost making kvec.h obsolete. Although cl.exe optimization does influence the results greatly. As it should be expected. x64 release build results are:

C array_, preallocated: 0.397 sec
C array_, dynamic: 1.054 sec
C vector, dynamic(kv_a): 0.510 sec
C vector, dynamic(kv_push): 1.030 sec
C++ vector, preallocated: 0.453 sec
C++ vector, dynamic: 1.654 sec

cl.exe is not known as excellent C compiler so I tried with CLANG 9.0 coming pre packaged with Visual Studio 2019. These are the results :

C array_, preallocated: 0.000 sec
C array_, dynamic: 1.135 sec
C vector, dynamic(kv_a): 0.480 sec
C vector, dynamic(kv_push): 1.107 sec
C++ vector, preallocated: 0.737 sec
C++ vector, dynamic: 1.835 sec

C side is a bit more noticeably faster with CLANG.Machine is: Windows 10 pro, (old) i5, 8GB RAM, 60 GB SSD. I assume LINUX, Mac results might vary.

Minor changes

kvec_test.cc has one modernization needed. Instead of:

std::vector<int> array_ ) ;
array_.reserve(N);

it should be

std::vector<int> array_( size_t(N), int(0) ) ;
// array_.reserve(N);

I also renamed array to array_.
I am using

Microsoft (R) C/C++ Optimizing Compiler Version 19.25.28612 for x86
Copyright (C) Microsoft Corporation.  All rights reserved.

Of course instead of std::vector "pre-allocated", one would use std::array, which would make things even faster. But in this program we would have instant stack overflow.