[翻译完成] 树莓派性能测试

转自：http://my.oschina.net/funnky/blog/141251?p=1#OSC_h3_3

CPU

Linpack基准测试

已经完成对Arm的linpack基准测试, 选用gcc带-O3 (优化等级3)编译代码[1]. 用200大小的数组运行.

包括软件浮点

源码

编译/运行

cc -O3 -o linpack linpack.c -lm  linpack.c: In function ‘main’:  linpack.c:69: warning: return type of ‘main’ is not ‘int’./linpack  Enter array size (q to quit) [200]: 200

结果

Crippled

Memory required:  315K.LINPACK benchmark, Double precision.Machine precision:  15 digits.Array size 200 X 200.Average rolled and unrolled performance:    Reps Time(s) DGEFA   DGESL  OVERHEAD    KFLOPS       2   0.53  92.45%   1.89%   5.66%   5493.333       4   1.07  92.52%   2.80%   4.67%   5385.621       8   2.12  92.45%   2.36%   5.19%   5466.003      16   4.24  92.45%   2.83%   4.72%   5438.944      32   8.49  92.11%   2.71%   5.18%   5459.213      64  16.98  92.05%   2.89%   5.06%   5452.440

硬件浮点 (-mfloat-abi=softfp)

Memory required:  315K.LINPACK benchmark, Double precision.Machine precision:  15 digits.Array size 200 X 200.Average rolled and unrolled performance:    Reps Time(s) DGEFA   DGESL  OVERHEAD    KFLOPS       8   0.51  90.20%   3.92%   5.88%  22888.889      16   1.02  89.22%   4.90%   5.88%  22888.889      32   2.05  90.24%   3.41%   6.34%  22888.889      64   4.08  91.42%   2.94%   5.64%  22829.437     128   8.16  91.54%   2.94%   5.51%  22799.827     256  16.31  91.35%   2.76%   5.89%  22903.800

Raspbian下的全硬件浮点 (-mfloat-abi=hard -mfpu=vfp), 频率arm_freq=700

Memory required:  315K.LINPACK benchmark, Double precision.Machine precision:  15 digits.Array size 200 X 200.Average rolled and unrolled performance:    Reps Time(s) DGEFA   DGESL  OVERHEAD    KFLOPS      16   0.58  89.66%   3.45%   6.90%  40691.358      32   1.17  87.18%   4.27%   8.55%  41071.651      64   2.32  88.36%   3.02%   8.62%  41459.119     128   4.67  88.22%   3.43%   8.35%  41071.651     256   9.33  88.85%   3.32%   7.82%  40880.620     512  18.63  89.00%   2.95%   8.05%  41047.675

Raspbian下的全硬件浮点 (-mfloat-abi=hard -mfpu=vfp), 频率arm_freq=1000 core_freq=500

Memory required:  315K.LINPACK benchmark, Double precision.Machine precision:  15 digits.Array size 200 X 200.Average rolled and unrolled performance:    Reps Time(s) DGEFA   DGESL  OVERHEAD    KFLOPS      32   0.79  89.87%   0.00%  10.13%  61896.714      64   1.58  89.24%   1.27%   9.49%  61463.869     128   3.16  90.19%   1.90%   7.91%  60407.789     256   6.32  88.13%   3.80%   8.07%  60511.761     512  12.65  87.83%   3.56%   8.62%  60825.836

Gentoo下的全硬件浮点, 带编译器优化(gcc-4.6.3 -Ofast -fno-fast-math), 默认时序

Memory required:  315K.LINPACK benchmark, Double precision.Machine precision:  15 digits.Array size 200 X 200.Average rolled and unrolled performance:    Reps Time(s) DGEFA   DGESL  OVERHEAD    KFLOPS----------------------------------------------------      16   0.56  89.29%   1.79%   8.93%  43084.967      32   1.13  91.15%   4.42%   4.42%  40691.358      64   2.25  89.78%   3.56%   6.67%  41853.968     128   4.51  87.80%   4.21%   7.98%  42358.233     256   9.01  88.68%   3.88%   7.44%  42155.076     512  18.01  89.23%   2.78%   8.00%  42434.923

Whetstone/Dhrystone综合基准测试

用gcc带参数-float-abi=softfp -O3编译全部代码

源码

测试代码在http://www.rowley.co.uk/arm/whet_dhry.zip.

备用地址http://freespace.virgin.net/roy.longbottom/benchnt.zip

编译/运行

?

结果

Dhrystone

Microseconds for one run through Dhrystone: 1.2Dhrystones per Second: 809061.5

Whetstone Crippled

Loops: 1000, Iterations: 10, Duration: 24 sec.C Converted Double Precision Whetstones: 41.7 MIPS

用'gcc -mfpu -float-abi=softfp'重新编译Whetstone, 结果更好:

Loops: 1000, Iterations: 100, Duration: 106 sec.C Converted Double Precision Whetstones: 94.3 MIPS

上面的测试没有带-mfpu=vfp编译, 所以大部分运算时间都花费在SQRT方法上. 用了vfp后提升很大:

Loops: 1000, Iterations: 100, Duration: 15 sec.C Converted Double Precision Whetstones: 666.7 MIPS

OpenSSL安全协议测试

源码

[3]

编译/运行

openssl version;openssl speed;

结果

关闭汇编优化:

OpenSSL 0.9.8o 01 Jun 2010built on: Thu Aug 26 18:56:26 UTC 2010options:bn(64,32) md2(int) rc4(ptr,int) des(idx,risc1,4,long) aes(partial) blowfish(idx)compiler: gcc -fPIC -DOPENSSL_PIC -DZLIB -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DL_ENDIAN -DTERMIO -O2 -Wa,--noexecstack -g -Wallavailable timing options: TIMES TIMEB HZ=100 [sysconf value]timing function used: timesThe 'numbers' are in 1000s of bytes per second processed.type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytesmd2                148.81k      372.18k      624.81k      769.95k      832.90kmdc2                 0.00         0.00         0.00         0.00         0.00md4                615.30k     2468.76k     7612.19k    16707.01k    28104.86kmd5                380.13k     1501.12k     4800.77k    11312.81k    21682.77khmac(md5)         1022.28k     3480.23k     9587.80k    17492.25k    25441.78ksha1               303.72k     1092.39k     3106.50k     6302.57k     9852.39krmd160             244.29k      849.04k     2414.53k     4747.26k     7513.00krc4              14658.70k    16836.49k    17462.03k    17628.21k    17522.08kdes cbc           2913.17k     3221.30k     3289.77k     3360.09k     3367.21kdes ede3          1149.87k     1188.59k     1198.46k     1206.00k     1208.25kidea cbc             0.00         0.00         0.00         0.00         0.00seed cbc             0.00         0.00         0.00         0.00         0.00rc2 cbc           2812.71k     3012.02k     3054.19k     3077.82k     3076.12krc5-32/12 cbc        0.00         0.00         0.00         0.00         0.00blowfish cbc      6091.32k     7007.89k     7250.62k     7288.21k     7163.88kcast cbc          5068.25k     6020.03k     6345.71k     6367.64k     6260.44kaes-128 cbc       3205.76k     3497.72k     3616.00k     3652.49k     3665.85kaes-192 cbc       2730.65k     2981.88k     3073.20k     3102.38k     3111.86kaes-256 cbc       2383.90k     2596.12k     2659.91k     2702.13k     2732.50kcamellia-128 cbc     0.00         0.00         0.00         0.00         0.00camellia-192 cbc     0.00         0.00         0.00         0.00         0.00camellia-256 cbc     0.00         0.00         0.00         0.00         0.00sha256             679.98k     1629.47k     2905.43k     3708.32k     4175.45ksha512              41.02k      163.83k      232.63k      318.20k      353.81kaes-128 ige       3089.03k     3579.08k     3698.68k     3689.14k     3578.18kaes-192 ige       2641.68k     3019.45k     3111.38k     3144.95k     3035.70kaes-256 ige       2334.50k     2632.35k     2705.04k     2735.69k     2687.74k                  sign    verify    sign/s verify/srsa  512 bits 0.013747s 0.001193s     72.7    838.4rsa 1024 bits 0.063481s 0.002742s     15.8    364.7rsa 2048 bits 0.321250s 0.007378s      3.1    135.5rsa 4096 bits 1.805000s 0.022528s      0.6     44.4                  sign    verify    sign/s verify/sdsa  512 bits 0.011690s 0.013597s     85.5     73.5dsa 1024 bits 0.027233s 0.031683s     36.7     31.6dsa 2048 bits 0.073897s 0.087304s     13.5     11.5

开启汇编优化:

OpenSSL 1.0.1c 10 May 2012built on: Sun Jul 29 00:43:16 CEST 2012options:bn(64,32) rc4(ptr,char) des(idx,cisc,16,long) aes(partial) idea(int) blowfish(ptr)compiler: armv6j-hardfloat-linux-gnueabi-gcc -fPIC -DOPENSSL_PIC -DZLIB -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN \-DHAVE_DLFCN_H -DL_ENDIAN -DTERMIO -Wall -DOPENSSL_BN_ASM_MONT -DOPENSSL_BN_ASM_GF2m -DSHA1_ASM -DSHA256_ASM \-DSHA512_ASM -DAES_ASM -DGHASH_ASM -O2 -march=armv6j -mfpu=vfp -mfloat-abi=hard -fno-strict-aliasing -Wa,--noexecstackThe 'numbers' are in 1000s of bytes per second processed.type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytesmd2                  0.00         0.00         0.00         0.00         0.00mdc2               724.41k      933.06k     1024.68k     1063.59k     1075.88kmd4               2547.83k     9441.28k    27284.48k    51824.98k    69894.14kmd5               1954.05k     7217.96k    20805.95k    39365.29k    53226.15khmac(md5)         3075.61k    10241.88k    26669.65k    44729.00k    55386.11ksha1              2115.34k     6823.83k    16264.45k    25053.18k    30121.35krmd160            1487.88k     4783.96k    10707.71k    15800.32k    19303.08krc4              34205.43k    39535.98k    41215.83k    41561.43k    41570.04kdes cbc           6251.12k     6605.08k     6686.81k     6713.01k     6707.54kdes ede3          2326.45k     2368.36k     2385.83k     2397.53k     2391.84kidea cbc          8758.77k     9421.31k     9607.34k     9653.93k     9687.93kseed cbc          8274.52k     9036.46k     9264.64k     9321.47k     9284.27krc2 cbc           6047.90k     6354.82k     6458.82k     6465.19k     6485.33krc5-32/12 cbc    16204.47k    18649.32k    19367.94k    19560.11k    19649.84kblowfish cbc     11934.03k    13189.85k    13546.92k    13633.19k    13486.76kcast cbc         10797.59k    11828.46k    12156.58k    12187.65k    12050.43kaes-128 cbc      12978.72k    14708.69k    15387.40k    15472.93k    15529.06kaes-192 cbc      11441.49k    12834.60k    13315.69k    13453.78k    13430.80kaes-256 cbc      10267.01k    11409.83k    11744.41k    11812.86k    11859.64kcamellia-128 cbc     9312.98k    10278.89k    10572.46k    10646.19k    10657.82kcamellia-192 cbc     7541.38k     8140.71k     8325.63k     8370.18k     8361.30kcamellia-256 cbc     7513.97k     8138.65k     8297.98k     8351.40k     8347.65ksha256            3598.03k     8377.26k    14605.57k    17979.39k    19300.35ksha512            1080.74k     4322.82k     6151.85k     8416.32k     9418.07kwhirlpool          361.82k      729.24k     1186.42k     1425.38k     1512.79kaes-128 ige      11702.57k    13853.45k    14429.53k    14671.38k    14057.47kaes-192 ige      10468.67k    12165.24k    12628.24k    12743.72k    12331.69kaes-256 ige       9505.78k    10831.25k    11205.36k    11333.43k    10982.74kghash            15681.70k    17279.32k    17770.84k    17894.06k    17940.48k                  sign    verify    sign/s verify/srsa  512 bits 0.002185s 0.000217s    457.6   4611.1rsa 1024 bits 0.011325s 0.000640s     88.3   1563.5rsa 2048 bits 0.074296s 0.002289s     13.5    436.8rsa 4096 bits 0.544211s 0.008741s      1.8    114.4                  sign    verify    sign/s verify/sdsa  512 bits 0.002157s 0.002262s    463.5    442.0dsa 1024 bits 0.006234s 0.007123s    160.4    140.4dsa 2048 bits 0.022247s 0.025884s     44.9     38.6

GPU

树莓派从USB到HDMI处理h264 1080p电影至少能达到4MB/s.

管理员"JamesH"说"基本上1080p30高清的都>40Mb/s." h264也有5MB/s 

包括WVGA(480p30) 或 720p20的VP8/WEBM

ioquake3(雷神之锤3修改版)

源码

https://github.com/raspberrypi/quake3

编译/运行

- Download source, compile as delivered- Start game- Runs at display's native res, in my case 1280x1024- Bitdepth stuck at 16bpp, not sure how to change, values in q3config.cfg seem to be ignored- In-game console commands:\timedemo 1\demo four

结果

armel "driver info" : http://i.imgur.com/wtYhB.jpg armel timedemo score: http://i.imgur.com/i2TkN.jpg 20.2fps

armhf "driver info" : http://i.imgur.com/8nqa1.jpg armhf timedemo score: http://i.imgur.com/dUu0g.jpg 28.5fps

IO

USB总线

• 所有的IO都用同一个总线, 所以所有IO的总和不可能超过设计的60MB/s速度;
• 高速USB盘的测试结果显示速度可达30 MB/s:

root@raspberrypi:~# dd if=/dev/sda of=/dev/null bs=32M count=10 iflag=direct10+0 records in10+0 records out335544320 bytes (336 MB) copied, 10.6428 s, 31.5 MB/s

SD卡

这部分已经移到RPi_SD_cards#Performance

NIC(网卡)

编译/运行

局域网中:

iperf -s

树莓派中:

iperf -t 60 -c <SERVER_IP_ADDRESS> -d

结果

宽带 (Mbit/s)CPU使用率 (峰值)发行版内核备注52.1 + 46.45.1%us, 66.2%sy, 28.7%siDebian Squeeze "debian6-19-04-2012"Linux raspberrypi 3.1.9+ #95 PREEMPT 91.8 + 36.81.6%us, 60.8%sy, 37.5%siDebian Wheezy "Raspbian"Linux raspbian 3.1.9+ #101 PREEMPT 65.1 + 48.81.3%us, 61.9%sy, 36.8%siArch Linux 2012-04-29Linux alarmpi 3.1.9-12+ #5 Sat Apr 28 04:49:38 UTC 2012 armv6l ARMv6-compatible processor rev 7 (v6l) BCM2708 GNU/LinuxRemote host connected at gigabit69.5 + 29.10.6%us, 55.5%sy, 40.0%siDebian Wheezy "Raspbian"Linux rpi 3.1.9+ #168 PREEMPTRemote connected at gigabit, values for si between 30 and 55 %90.8 + 91.40.3%us, 62.2%sy, 37.5%siGentoo Linux ARMLinux genpi 3.2.23-bootc #1Remote host connected at gigabit, vm.min_free_kbytes = 4096

路由

方法

• 一台带两个lan口的OpenBSD 5.2电脑在不同域用两个lan口直接连接树莓派.
• 在OpenBSD上的一个接口跑iperf -s, 另一个跑iperf -t 300 -c, 树莓派负责在中间路由.
• iperf v2.0.5.
• 测试150秒后树莓派系统负载.

结果

宽带 (Mbit/s)CPU使用率 (峰值)发行版内核备注50.2 Mbps0.0%us, 0.1%sy, 99.8%siDebian Wheezy "Raspbian"Linux raspberrypi 3.6.11+ #366 PREEMPTStock clock85.5 Mbps0.0%us, 2.8%sy, 69.7%siDebian Wheezy "Raspbian"Linux raspberrypi 3.6.11+ #366 PREEMPTOverclocked via raspi_config "turbo"87.1 Mbps4.3%us, 4.3%sy, 52.1%siDebian Wheezy "Raspbian"Linux raspberrypi 3.6.11+ #366 PREEMPTOverclocked arm_freq 1100 core_freq 500 sdram_freq 600 over_voltage 662.1 Mbps0.0%us, 0.1%sy, 99.2%siDebian Wheezy "Raspbian"Linux raspberrypi 3.6.11+ #366 PREEMPTOverclocked arm_freq 1100 core_freq 500 sdram_freq 600 over_voltage 6 - SNAT enabled

电源

表格中列出了用5V电源线时同时执行各种任务所消耗电量.

进程电量使用 (mA)备注启动 (无外设)120-400Taken from Agilent lab power supply readings. No composite/keyboard/mouse/network connected. HDMI was enabled but the cable was disconnected.待机(HDMI开启,网络开启)370Taken from Agilent lab power supply readings. No composite/keyboard/mouse connected.待机(HDMI开启,网络关闭)320Taken from Agilent lab power supply readings. No composite/keyboard/mouse/network connected.播放1080p视频750About 3h on 4 AA batteries编辑文本-Same as idling编译C代码(QuakeIII)364Measured with a Fluke 87V multimeter. No composite, keyboard, mouse, or network connected. HDMI was connected. Current peaked at 418mA.运行Python程序 ? 玩Quake III461Measured with a Fluke 87V multimeter. No composite or network connected. HDMI was connected. Current peaks at 551mA.