高通 CPU 浮点这么快吗。。。

[lwjef]

wget https://github.com/xiexiexx/PPLA/raw/main/billionsort/billionsort.cpp

clang++ -O3 billionsort.cpp

./a.out

• redmi note12t 7+gen2 16+1t miui14.0.25 termux witout root

有没有 8gen3 和 9300 的老哥跑一下让我长长见识

• M1 Max https://weibo.com/5819320755/LlO4Cq4Jc

• iphone11 a13 4+128g ios17.1.1 a-shell 跑的 1 亿

[Documents]$ ./1x_millionsort_with_output 100
Data size: 100000000
0.0027 minutes
0.02695 minutes
0.478667 minutes

• 7700k 32g xubuntu2204

[pubby]

$ uname -a
Darwin Mac-mini.local 22.6.0 Darwin Kernel Version 22.6.0: Wed Oct 4 21:25:26 PDT 2023; root:xnu-8796.141.3.701.17~4/RELEASE_X86_64 x86_64
$ sysctl machdep.cpu.brand_string
machdep.cpu.brand_string: Intel(R) Core(TM) i9-9900T CPU @ 2.10GHz
$ ./a.out
0.0942143 minutes
0.277907 minutes
2.31389 minutes

[lwjef]

@pubby #1 这个 i9 也太慢了。。。

[pubby]

@lwjef 低压 u, 内存频率也不高 :D

[kikitte]

AMD 5950x ArchLinux

[kikitte@archlinux test]$ clang++ -O3 billionsort.cpp
[kikitte@archlinux test]$ ./a.out
0.0128345 minutes
0.954518 minutes
1.46679 minutes

[hefish]

这个 sort 也跟 cpu 线程数有关吧

[maggch97]

首先这个 sort 是单线程的

[lslqtz]

M1 Pro 感觉差不多.
这是单核浮点吧.

[lwjef]

@maggch97 #6 不是 io 项目单核下多线程也没用吧

[lwjef]

@lslqtz #7 嗯嗯，就是高通的 fpu 有点离谱

[beetlerx]

OS: Debian GNU/Linux trixie/sid x86_64
CPU: AMD Ryzen 7 7840H w/ Radeon 780M Graphics (16) @ 5.137GHz

$ clang++-16 -std=c++2a -O3 ./billionsort.cpp
$ ./a.out
0.00709458 minutes
0.757148 minutes
1.45406 minutes
看起来 7840H 单核浮点可以和 5950x 掰掰

[Philippa]

.., larry@DESKTOP-OVBIJM9
....,,:;+ccllll ---------------------
...,,+:; cllllllllllllllllll OS: Windows 10 רҵ□□ x86_64
,cclllllllllll lllllllllllllllllll Host: ASUS
llllllllllllll lllllllllllllllllll Kernel: 10.0.19045
llllllllllllll lllllllllllllllllll Uptime: 14 hours, 59 mins
llllllllllllll lllllllllllllllllll Packages: 3 (scoop)
llllllllllllll lllllllllllllllllll Shell: bash 5.2.15
llllllllllllll lllllllllllllllllll Resolution: 3840x2160
DE: Aero
llllllllllllll lllllllllllllllllll WM: Explorer
llllllllllllll lllllllllllllllllll WM Theme: Custom
llllllllllllll lllllllllllllllllll Terminal: Windows Terminal
llllllllllllll lllllllllllllllllll CPU: AMD Ryzen 9 7950X3D (32) @ 4.200GHz
llllllllllllll lllllllllllllllllll GPU: Caption
`'ccllllllllll lllllllllllllllllll GPU: NVIDIA GeForce RTX 4090
`' \*:: :ccllllllllllllllll GPU
````''*::cll Memory: 16818MiB / 130802MiB
``

clang++ -std=c++2a -O3 main.cpp

0.0153925 minutes
0.683383 minutes
1.30065 minutes

不是很懂，为什么这么慢

[lwjef]

@Philippa #11 还有 15 个核在围观 QAQ

[Philippa]

@lwjef 哈哈哈，那要怎么设

[maggch97]

@lwjef ? 我不知道你是不是了解 CPU 性能，了解代码。

你这个代码只能测出 CPU 跑 std::sort 这样的单线程负载的性能，速度受到 Memory, Cache 的影响

[lwjef]

@maggch97 #14 那么是高通不够快吗 （狗头

[maggch97]

这能叫做"浮点性能"吗，下面的回帖很多人也被误导了

[lwjef]

@Philippa #13 就是跑的一个核，已经是贴子里 x86 最快的了

[lwjef]

@maggch97 #16 那为啥高通这么快啊，请教下。。。

[Philippa]

@lwjef 感谢科普

[bigtan]

我这个 14700KF 怎么这么慢

0.0179123 minutes
0.263897 minutes
3.26288 minutes