ROC-RK3568-PC全面性能测试

Firefly搬运工

前言

板子的性能不仅仅和CPU相关,也和存储等相关,是一个综合体现,所以我们从CPU,存储,等几个关键的部分去进行性能测试。

CPU性能: Coremark跑分

git clone [https://github.com/eembc/coremark.git]
(https://github.com/eembc/coremark.git)

cd coremark/


单线程

make ITERATIONS=100000

打印信息如下

1. root@firefly:~/coremark# make ITERATIONS=100000
   
2. 
   
3. make XCFLAGS=" -DPERFORMANCE_RUN=1" load run1.log
   
4. 
   
5. make[1]: Entering directory '/root/coremark'
   
6. 
   
7. make port_prebuild
   
8. 
   
9. make[2]: Entering directory '/root/coremark'
   
10. 
    
11. make[2]: Nothing to be done for 'port_prebuild'.
    
12. 
    
13. make[2]: Leaving directory '/root/coremark'
    
14. 
    
15. make link
    
16. 
    
17. make[2]: Entering directory '/root/coremark'
    
18. 
    
19. cc -O2 -Ilinux -Iposix -I. -DFLAGS_STR=\""-O2 -DPERFORMANCE_RUN=1  -lrt"\" -DITERATIONS=100000 -DPERFORMANCE_RUN=1 core_list_join.c core_main.c core_matrix.c core_state.c core_util.c posix/core_portme.c -o ./coremark.exe -lrt
    
20. 
    
21. Link performed along with compile
    
22. 
    
23. make[2]: Leaving directory '/root/coremark'
    
24. 
    
25. make port_postbuild
    
26. 
    
27. make[2]: Entering directory '/root/coremark'
    
28. 
    
29. make[2]: Nothing to be done for 'port_postbuild'.
    
30. 
    
31. make[2]: Leaving directory '/root/coremark'
    
32. 
    
33. make port_preload
    
34. 
    
35. make[2]: Entering directory '/root/coremark'
    
36. 
    
37. make[2]: Nothing to be done for 'port_preload'.
    
38. 
    
39. make[2]: Leaving directory '/root/coremark'
    
40. 
    
41. echo Loading done ./coremark.exe
    
42. 
    
43. Loading done ./coremark.exe
    
44. 
    
45. make port_postload
    
46. 
    
47. make[2]: Entering directory '/root/coremark'
    
48. 
    
49. make[2]: Nothing to be done for 'port_postload'.
    
50. 
    
51. make[2]: Leaving directory '/root/coremark'
    
52. 
    
53. make port_prerun
    
54. 
    
55. make[2]: Entering directory '/root/coremark'
    
56. 
    
57. make[2]: Nothing to be done for 'port_prerun'.
    
58. 
    
59. make[2]: Leaving directory '/root/coremark'
    
60. 
    
61. ./coremark.exe  0x0 0x0 0x66 100000 7 1 2000 > ./run1.log
    
62. 
    
63. make port_postrun
    
64. 
    
65. make[2]: Entering directory '/root/coremark'
    
66. 
    
67. make[2]: Nothing to be done for 'port_postrun'.
    
68. 
    
69. make[2]: Leaving directory '/root/coremark'
    
70. 
    
71. make[1]: Leaving directory '/root/coremark'
    
72. 
    
73. make XCFLAGS=" -DVALIDATION_RUN=1" load run2.log
    
74. 
    
75. make[1]: Entering directory '/root/coremark'
    
76. 
    
77. make port_preload
    
78. 
    
79. make[2]: Entering directory '/root/coremark'
    
80. 
    
81. make[2]: Nothing to be done for 'port_preload'.
    
82. 
    
83. make[2]: Leaving directory '/root/coremark'
    
84. 
    
85. echo Loading done ./coremark.exe
    
86. 
    
87. Loading done ./coremark.exe
    
88. 
    
89. make port_postload
    
90. 
    
91. make[2]: Entering directory '/root/coremark'
    
92. 
    
93. make[2]: Nothing to be done for 'port_postload'.
    
94. 
    
95. make[2]: Leaving directory '/root/coremark'
    
96. 
    
97. make port_prerun
    
98. 
    
99. make[2]: Entering directory '/root/coremark'
    
100. 
     
101. make[2]: Nothing to be done for 'port_prerun'.
     
102. 
     
103. make[2]: Leaving directory '/root/coremark'
     
104. 
     
105. ./coremark.exe  0x3415 0x3415 0x66 100000 7 1 2000  > ./run2.log
     
106. 
     
107. make port_postrun
     
108. 
     
109. make[2]: Entering directory '/root/coremark'
     
110. 
     
111. make[2]: Nothing to be done for 'port_postrun'.
     
112. 
     
113. make[2]: Leaving directory '/root/coremark'
     
114. 
     
115. make[1]: Leaving directory '/root/coremark'
     
116. 
     
117. Check run1.log and run2.log for results.
     
118. 
     
119. See README.md for run and reporting rules.

复制代码

run1.log

1. root@firefly:~/coremark# vi run1.log
   
2. 
   
3. 2K performance run parameters for coremark.
   
4. 
   
5. CoreMark Size    : 666
   
6. 
   
7. Total ticks      : 14036
   
8. 
   
9. Total time (secs): 14.036000
   
10. 
    
11. Iterations/Sec   : 7124.536905
    
12. 
    
13. Iterations       : 100000
    
14. 
    
15. Compiler version : GCC9.4.0
    
16. 
    
17. Compiler flags   : -O2 -DPERFORMANCE_RUN=1  -lrt
    
18. 
    
19. Memory location  : Please put data memory location here
    
20. 
    
21. (e.g. code in flash, data on heap etc)
    
22. 
    
23. seedcrc          : 0xe9f5
    
24. 
    
25. [0]crclist       : 0xe714
    
26. 
    
27. [0]crcmatrix     : 0x1fd7
    
28. 
    
29. [0]crcstate      : 0x8e3a
    
30. 
    
31. [0]crcfinal      : 0xd340
    
32. 
    
33. Correct operation validated. See README.md for run and reporting rules.
    
34. 
    
35. CoreMark 1.0 : 7124.536905 / GCC9.4.0 -O2 -DPERFORMANCE_RUN=1  -lrt / Heap

复制代码

run2.log

1. root@firefly:~/coremark# vi run2.log
   
2. 
   
3. 2K validation run parameters for coremark.
   
4. 
   
5. CoreMark Size    : 666
   
6. 
   
7. Total ticks      : 14138
   
8. 
   
9. Total time (secs): 14.138000
   
10. 
    
11. Iterations/Sec   : 7073.136229
    
12. 
    
13. Iterations       : 100000
    
14. 
    
15. Compiler version : GCC9.4.0
    
16. 
    
17. Compiler flags   : -O2 -DPERFORMANCE_RUN=1  -lrt
    
18. 
    
19. Memory location  : Please put data memory location here
    
20. 
    
21. (e.g. code in flash, data on heap etc)
    
22. 
    
23. seedcrc          : 0x18f2
    
24. 
    
25. [0]crclist       : 0xe3c1
    
26. 
    
27. [0]crcmatrix     : 0x0747
    
28. 
    
29. [0]crcstate      : 0x8d84
    
30. 
    
31. [0]crcfinal      : 0x5c66
    
32. 
    
33. Correct operation validated. See README.md for run and reporting rules.

复制代码

多线程


make XCFLAGS="-DMULTITHREAD=4 -DUSE_FORK"

打印如下

1. root@firefly:~/coremark# make XCFLAGS="-DMULTITHREAD=4 -DUSE_FORK"
   
2. 
   
3. make XCFLAGS="-DMULTITHREAD=4 -DUSE_FORK -DPERFORMANCE_RUN=1" load run1.log
   
4. 
   
5. make[1]: Entering directory '/root/coremark'
   
6. 
   
7. make port_preload
   
8. 
   
9. make[2]: Entering directory '/root/coremark'
   
10. 
    
11. make[2]: Nothing to be done for 'port_preload'.
    
12. 
    
13. make[2]: Leaving directory '/root/coremark'
    
14. 
    
15. echo Loading done ./coremark.exe
    
16. 
    
17. Loading done ./coremark.exe
    
18. 
    
19. make port_postload
    
20. 
    
21. make[2]: Entering directory '/root/coremark'
    
22. 
    
23. make[2]: Nothing to be done for 'port_postload'.
    
24. 
    
25. make[2]: Leaving directory '/root/coremark'
    
26. 
    
27. make port_prerun
    
28. 
    
29. make[2]: Entering directory '/root/coremark'
    
30. 
    
31. make[2]: Nothing to be done for 'port_prerun'.
    
32. 
    
33. make[2]: Leaving directory '/root/coremark'
    
34. 
    
35. ./coremark.exe  0x0 0x0 0x66 0 7 1 2000 > ./run1.log
    
36. 
    
37. make port_postrun
    
38. 
    
39. make[2]: Entering directory '/root/coremark'
    
40. 
    
41. make[2]: Nothing to be done for 'port_postrun'.
    
42. 
    
43. make[2]: Leaving directory '/root/coremark'
    
44. 
    
45. make[1]: Leaving directory '/root/coremark'
    
46. 
    
47. make XCFLAGS="-DMULTITHREAD=4 -DUSE_FORK -DVALIDATION_RUN=1" load run2.log
    
48. 
    
49. make[1]: Entering directory '/root/coremark'
    
50. 
    
51. make port_preload
    
52. 
    
53. make[2]: Entering directory '/root/coremark'
    
54. 
    
55. make[2]: Nothing to be done for 'port_preload'.
    
56. 
    
57. make[2]: Leaving directory '/root/coremark'
    
58. 
    
59. echo Loading done ./coremark.exe
    
60. 
    
61. Loading done ./coremark.exe
    
62. 
    
63. make port_postload
    
64. 
    
65. make[2]: Entering directory '/root/coremark'
    
66. 
    
67. make[2]: Nothing to be done for 'port_postload'.
    
68. 
    
69. make[2]: Leaving directory '/root/coremark'
    
70. 
    
71. make port_prerun
    
72. 
    
73. make[2]: Entering directory '/root/coremark'
    
74. 
    
75. make[2]: Nothing to be done for 'port_prerun'.
    
76. 
    
77. make[2]: Leaving directory '/root/coremark'
    
78. 
    
79. ./coremark.exe  0x3415 0x3415 0x66 0 7 1 2000  > ./run2.log
    
80. 
    
81. make port_postrun
    
82. 
    
83. make[2]: Entering directory '/root/coremark'
    
84. 
    
85. make[2]: Nothing to be done for 'port_postrun'.
    
86. 
    
87. make[2]: Leaving directory '/root/coremark'
    
88. 
    
89. make[1]: Leaving directory '/root/coremark'
    
90. 
    
91. Check run1.log and run2.log for results.
    
92. 
    
93. See README.md for run and reporting rules.

复制代码

run1.log

1. root@firefly:~/coremark# vi run1.log
   
2. 
   
3. 2K performance run parameters for coremark.
   
4. 
   
5. CoreMark Size    : 666
   
6. 
   
7. Total ticks      : 15471
   
8. 
   
9. Total time (secs): 15.471000
   
10. 
    
11. Iterations/Sec   : 28440.307672
    
12. 
    
13. Iterations       : 440000
    
14. 
    
15. Compiler version : GCC9.4.0
    
16. 
    
17. Compiler flags   : -O2 -DMULTITHREAD=4 -DUSE_FORK -DPERFORMANCE_RUN=1  -lrt
    
18. 
    
19. Parallel Fork : 4
    
20. 
    
21. Memory location  : Please put data memory location here
    
22. 
    
23. (e.g. code in flash, data on heap etc)
    
24. 
    
25. seedcrc          : 0xe9f5
    
26. 
    
27. [0]crclist       : 0xe714
    
28. 
    
29. [1]crclist       : 0xe714
    
30. 
    
31. [2]crclist       : 0xe714
    
32. 
    
33. [3]crclist       : 0xe714
    
34. 
    
35. [0]crcmatrix     : 0x1fd7
    
36. 
    
37. [1]crcmatrix     : 0x1fd7
    
38. 
    
39. [2]crcmatrix     : 0x1fd7
    
40. 
    
41. [3]crcmatrix     : 0x1fd7
    
42. 
    
43. [0]crcstate      : 0x8e3a
    
44. 
    
45. [1]crcstate      : 0x8e3a
    
46. 
    
47. [2]crcstate      : 0x8e3a
    
48. 
    
49. [3]crcstate      : 0x8e3a
    
50. 
    
51. [0]crcfinal      : 0x33ff
    
52. 
    
53. [1]crcfinal      : 0x33ff
    
54. 
    
55. [2]crcfinal      : 0x33ff
    
56. 
    
57. [3]crcfinal      : 0x33ff
    
58. 
    
59. Correct operation validated. See README.md for run and reporting rules.
    
60. 
    
61. CoreMark 1.0 : 28440.307672 / GCC9.4.0 -O2 -DMULTITHREAD=4 -DUSE_FORK -DPERFORMANCE_RUN=1  -lrt / Heap / 4:Fork
    
62. 
    
63. ~

复制代码

run2.log

1. root@firefly:~/coremark# vi run2.log
   
2. 
   
3. 2K validation run parameters for coremark.
   
4. 
   
5. CoreMark Size    : 666
   
6. 
   
7. Total ticks      : 15582
   
8. 
   
9. Total time (secs): 15.582000
   
10. 
    
11. Iterations/Sec   : 28237.710178
    
12. 
    
13. Iterations       : 440000
    
14. 
    
15. Compiler version : GCC9.4.0
    
16. 
    
17. Compiler flags   : -O2 -DMULTITHREAD=4 -DUSE_FORK -DPERFORMANCE_RUN=1  -lrt
    
18. 
    
19. Parallel Fork : 4
    
20. 
    
21. Memory location  : Please put data memory location here
    
22. 
    
23. (e.g. code in flash, data on heap etc)
    
24. 
    
25. seedcrc          : 0x18f2
    
26. 
    
27. [0]crclist       : 0xe3c1
    
28. 
    
29. [1]crclist       : 0xe3c1
    
30. 
    
31. [2]crclist       : 0xe3c1
    
32. 
    
33. [3]crclist       : 0xe3c1
    
34. 
    
35. [0]crcmatrix     : 0x0747
    
36. 
    
37. [1]crcmatrix     : 0x0747
    
38. 
    
39. [2]crcmatrix     : 0x0747
    
40. 
    
41. [3]crcmatrix     : 0x0747
    
42. 
    
43. [0]crcstate      : 0x8d84
    
44. 
    
45. [1]crcstate      : 0x8d84
    
46. 
    
47. [2]crcstate      : 0x8d84
    
48. 
    
49. [3]crcstate      : 0x8d84
    
50. 
    
51. [0]crcfinal      : 0x0956
    
52. 
    
53. [1]crcfinal      : 0x0956
    
54. 
    
55. [2]crcfinal      : 0x0956
    
56. 
    
57. [3]crcfinal      : 0x0956
    
58. 
    
59. Correct operation validated. See README.md for run and reporting rules.

复制代码

对比

https://www.eembc.org/coremark/scores.php

下搜索A55没有对应的芯片的跑分,
可以和A53对比下,

我们这里的得分28440比A53的19678还是高很多的,并且还仅是-O2优化。

计算圆周率

time echo "[scale](https://so.csdn.net/so/search?q=scale&spm=1001.2101.3001.7020)=5000; 4*a(1)" | bc -l -q

执行时间如下

1. real    0m47.623s
   
2. 
   
3. user    0m47.596s
   
4. 
   
5. sys     0m0.012s

复制代码

RAM带宽

参考https://www.cs.virginia.edu/stream/ref.html

git clone [https://github.com/qinyunti/STREAM.git](https://github.com/qinyunti/STREAM.git)

cd STREAM/

gcc -O3 stream.c -o stream

打印如下

1. root@firefly:~/coremark/STREAM# ./stream
   
2. 
   
3. ---
   
4. 
   
5. STREAM version $Revision: 5.10 $
   
6. 
   
7. ---
   
8. 
   
9. This system uses 8 bytes per array element.
   
10. 
    
11. ---
    
12. 
    
13. Array size = 10000000 (elements), Offset = 0 (elements)
    
14. 
    
15. Memory per array = 76.3 MiB (= 0.1 GiB).
    
16. 
    
17. Total memory required = 228.9 MiB (= 0.2 GiB).
    
18. 
    
19. Each kernel will be executed 10 times.
    
20. 
    
21. The *best* time for each kernel (excluding the first iteration)
    
22. 
    
23. will be used to compute the reported bandwidth.
    
24. 
    
25. ---
    
26. 
    
27. Your clock granularity/precision appears to be 1 microseconds.
    
28. 
    
29. Each test below will take on the order of 43055 microseconds.
    
30. 
    
31. (= 43055 clock ticks)
    
32. 
    
33. Increase the size of the arrays if this shows that
    
34. 
    
35. you are not getting at least 20 clock ticks per test.
    
36. 
    
37. ---
    
38. 
    
39. WARNING -- The above is only a rough guideline.
    
40. 
    
41. For best results, please be sure you know the
    
42. 
    
43. precision of your system timer.
    
44. 
    
45. ---
    
46. 
    
47. Function    Best Rate MB/s  Avg time     Min time     Max time
    
48. 
    
49. Copy:            6306.2     0.025627     0.025372     0.025743
    
50. 
    
51. Scale:           5647.5     0.028464     0.028331     0.028618
    
52. 
    
53. Add:             5446.5     0.044271     0.044065     0.044582
    
54. 
    
55. Triad:           5169.9     0.046605     0.046423     0.046989
    
56. 
    
57. ---
    
58. 
    
59. Solution Validates: avg error less than 1.000000e-13 on all three arrays

复制代码

压力测试

参考https://pyropus.ca./software/memtester/

wget [https://pyropus.ca./software/mem ... tester-4.5.1.tar.gz](https://pyropus.ca./software/mem ... tester-4.5.1.tar.gz)

tar -xvf memtester-4.5.1.tar.gz

cd memtester-4.5.1/

gcc -O3 memtester.c tests.c -o memtester

./memtester 512M 1

512M表示测试RAM大小

1表示测试一次

打印如下

1. root@firefly:~/memtester-4.5.1# ./memtester 512M 1
   
2. 
   
3. memtester version 4.5.1 (64-bit)
   
4. 
   
5. Copyright (C) 2001-2020 Charles Cazabon.
   
6. 
   
7. Licensed under the GNU General Public License version 2 (only).
   
8. 
   
9. pagesize is 4096
   
10. 
    
11. pagesizemask is 0xfffffffffffff000
    
12. 
    
13. want 512MB (536870912 bytes)
    
14. 
    
15. got  512MB (536870912 bytes), trying mlock ...locked.
    
16. 
    
17. Loop 1/1:
    
18. 
    
19. Stuck Address       : ok
    
20. 
    
21. Random Value        : ok
    
22. 
    
23. Compare XOR         : ok
    
24. 
    
25. Compare SUB         : ok
    
26. 
    
27. Compare MUL         : ok
    
28. 
    
29. Compare DIV         : ok
    
30. 
    
31. Compare OR          : ok
    
32. 
    
33. Compare AND         : ok
    
34. 
    
35. Sequential Increment: ok
    
36. 
    
37. Solid Bits          : ok
    
38. 
    
39. Block Sequential    : ok
    
40. 
    
41. Checkerboard        : ok
    
42. 
    
43. Bit Spread          : ok
    
44. 
    
45. Bit Flip            : ok
    
46. 
    
47. Walking Ones        : ok
    
48. 
    
49. Walking Zeroes      : ok
    
50. 
    
51. Done.

复制代码

EMMC

dmesg | grep mmc

可以看到打印

1. mmc3: new ultra high speed SDR104 SDIO card at address 0001
   
2. 
   
3. [    2.312867] mmc3:mmc host rescan start!

复制代码

其中high speed SDR104表示emmc 设备支持的时钟模式:

Speed Mode	clock (MHz)
Default Speed	26
Hight Speed SDR	52
Hight Speed DDR	52
HS200	200
HS400	200

SDR : 单边沿采样

DDR : 双边沿采样

所以我们这里x8-bit理论最大吞吐量应该是52MB/S。

输入df回车

我们看到EMMC的/dev/mmcblk0p7挂在了目录/userdata

我们就在该目录下读写文件测试

1. root@firefly:~/memtester-4.5.1# df
   
2. 
   
3. Filesystem     1K-blocks    Used Available Use% Mounted on
   
4. 
   
5. udev             1984744       8   1984736   1% /dev
   
6. 
   
7. tmpfs             399616    1168    398448   1% /run
   
8. 
   
9. /dev/mmcblk0p6   2666944 2599912         0 100% /root-ro
   
10. 
    
11. /dev/mmcblk0p7  26999224 6355668  20627172  24% /userdata
    
12. 
    
13. overlayroot     26999224 6355668  20627172  24% /
    
14. 
    
15. tmpfs            1998060       0   1998060   0% /dev/shm
    
16. 
    
17. tmpfs               5120       4      5116   1% /run/lock
    
18. 
    
19. tmpfs            1998060       0   1998060   0% /sys/fs/cgroup
    
20. 
    
21. tmpfs             399612       0    399612   0% /run/user/0
    
22. 
    
23. tmpfs             399612       8    399604   1% /run/user/1000
    
24. 
    
25. root@firefly:~/memtester-4.5.1#

复制代码

读

dd if=/userdata/test.bin of=/dev/null bs=块大小 count=块数量

写

dd if=/dev/zero of=/userdata/test.bin bs=块大小 count=块数量

测试记录如下

bs/count 1GB	bs/count 1GB	指令	结果
读	16k/65536	dd if=/userdata/test.bin of=/dev/null bs=16k count=65536 iflag=direct	36.5 MB/s
4k/262144	dd if=/userdata/test.bin of=/dev/null bs=4k count=262144 iflag=direct	14.7 MB/s
1k/1048576	dd if=/userdata/test.bin of=/dev/null bs=1k count=10485764 iflag=direct	2.2 MB/s
写	16k/65536	dd if=/dev/zero of=/userdata/test.bin bs=16k count=65536 conv=fdatasync	118 MB/s
4k/262144	dd if=/dev/zero of=/userdata/test.bin bs=4k count=262144 conv=fdatasync	112 MB/s
1k/1048576	dd if=/dev/zero of=/userdata/test.bin bs=1k count=1048576 conv=fdatasync	64.6 MB/s

QT

sudo apt-get install qt5-default qtcreator

直接板上使用qtcreator开发,操作也比较流畅

GPU

sudo apt install glmark2

运行

输入glmark2回车

最终得分

1. =======================================================
   
2. 
   
3. glmark2 Score: 49
   
4. 
   
5. =======================================================

复制代码

视频硬件编解码

/usr/local/test.mp4

1080P, 24Fps, H264

播放流畅

总结

综合各方面,该开发板性能都非常不错,特别适合人机交互,AI,边缘计算等高性能要求的场景。