163. stream run on arm¶
- 查询GNU
compiler
得知
-O和-O1优化级别相同。stream测试结果也相同。 -O和-O1测试结果较好；-O2和-O3可以使add 和 Triad结果变好，但是整体结果可能变差。stream官方例子带-o- 不指定编译优化选项，或者指定编译优化选项为
-O0时，测试结果最差 - stream官方要求数组大小使每个数组大于cache的4倍。数组大小满足这个需求之后，增大数组，cache-misses增加，但是性能测试结果不变
163.1. 硬件信息¶
163.1.1. CPU¶
root@ubuntu:~# lscpu
Architecture: aarch64
Byte Order: Little Endian
CPU(s): 64
On-line CPU(s) list: 0-63
Thread(s) per core: 1
Core(s) per socket: 32
Socket(s): 2
NUMA node(s): 4
Vendor ID: ARM
Model: 2
Model name: Cortex-A72
Stepping: r0p2
BogoMIPS: 100.00
L1d cache: 32K
L1i cache: 48K
L2 cache: 1024K
L3 cache: 16384K
NUMA node0 CPU(s): 0-15
NUMA node1 CPU(s): 16-31
NUMA node2 CPU(s): 32-47
NUMA node3 CPU(s): 48-63
Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 cpuid
163.1.2. 内存¶
Handle 0x0009, DMI type 17, 40 bytes
Memory Device
Array Handle: 0x0007
Error Information Handle: Not Provided
Total Width: 72 bits
Data Width: 64 bits
Size: 32 GB
Form Factor: DIMM
Set: None
Locator: DIMM000 J11
Bank Locator: SOCKET 0 CHANNEL 0 DIMM 0
Type: DDR4
Type Detail: Synchronous Registered (Buffered)
Speed: 2400 MT/s
Manufacturer: Samsung
Serial Number: 0x351254BC
Asset Tag: 1709
Part Number: M393A4K40BB1-CRC
Rank: 2
Configured Clock Speed: 2400 MT/s
Minimum Voltage: 1.2 V
Maximum Voltage: 1.2 V
Configured Voltage: 1.2 V
root@ubuntu:~# free -h
total used free shared buff/cache available
Mem: 125G 1.8G 123G 17M 786M 123G
Swap: 2.0G 0B 2.0G
163.2. 软件信息¶
root@ubuntu:~# cat/tec /etc/os-release
NAME="Ubuntu"
VERSION="18.04.1 LTS (Bionic Beaver)"
ID=ubuntu
ID_LIKE=debian
PRETTY_NAME="Ubuntu 18.04.1 LTS"
VERSION_ID="18.04"
HOME_URL="https://www.ubuntu.com/"
SUPPORT_URL="https://help.ubuntu.com/"
BUG_REPORT_URL="https://bugs.launchpad.net/ubuntu/"
PRIVACY_POLICY_URL="https://www.ubuntu.com/legal/terms-and-policies/privacy-policy"
VERSION_CODENAME=bionic
UBUNTU_CODENAME=bionic
163.2.1. perf¶
root@ubuntu:~# perf -v
perf version 4.15.18
root@ubuntu:~# dpkg -s linux-tools-common
Package: linux-tools-common
Status: install ok installed
Priority: optional
Section: kernel
Installed-Size: 330
Maintainer: Ubuntu Kernel Team <kernel-team@lists.ubuntu.com>
Architecture: all
Multi-Arch: foreign
Source: linux
Version: 4.15.0-46.49
Depends: lsb-release
Description: Linux kernel version specific tools for version 4.15.0
This package provides the architecture independent parts for kernel
version locked tools (such as perf and x86_energy_perf_policy) for
version PGKVER.
163.2.2. gcc¶
root@ubuntu:~# gcc -v
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/lib/gcc/aarch64-linux-gnu/7/lto-wrapper
Target: aarch64-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Ubuntu/Linaro 7.3.0-27ubuntu1~18.04' --with-bugurl=file:///usr/share/doc/gcc-7/README.Bugs --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++ --prefix=/usr --with-gcc-major-version-only --program-suffix=-7 --program-prefix=aarch64-linux-gnu- --enable-shared --enable-linker-build-id --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --libdir=/usr/lib --enable-nls --with-sysroot=/ --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --with-default-libstdcxx-abi=new --enable-gnu-unique-object --disable-libquadmath --disable-libquadmath-support --enable-plugin --enable-default-pie --with-system-zlib --enable-multiarch --enable-fix-cortex-a53-843419 --disable-werror --enable-checking=release --build=aarch64-linux-gnu --host=aarch64-linux-gnu --target=aarch64-linux-gnu
Thread model: posix
gcc version 7.3.0 (Ubuntu/Linaro 7.3.0-27ubuntu1~18.04)
163.3. 执行结果¶
163.4. 数组10000000，选项无¶
root@ubuntu:~/app/stream# gcc stream.c -o stream
root@ubuntu:~/app/stream# perf stat -e cache-misses ./stream
-------------------------------------------------------------
STREAM version $Revision: 5.10 $
-------------------------------------------------------------
This system uses 8 bytes per array element.
-------------------------------------------------------------
Array size = 10000000 (elements), Offset = 0 (elements)
Memory per array = 76.3 MiB (= 0.1 GiB).
Total memory required = 228.9 MiB (= 0.2 GiB).
Each kernel will be executed 10 times.
The *best* time for each kernel (excluding the first iteration)
will be used to compute the reported bandwidth.
-------------------------------------------------------------
Your clock granularity/precision appears to be 2 microseconds.
Each test below will take on the order of 62633 microseconds.
(= 31316 clock ticks)
Increase the size of the arrays if this shows that
you are not getting at least 20 clock ticks per test.
-------------------------------------------------------------
WARNING -- The above is only a rough guideline.
For best results, please be sure you know the
precision of your system timer.
-------------------------------------------------------------
Function Best Rate MB/s Avg time Min time Max time
Copy: 2549.2 0.062770 0.062765 0.062775
Scale: 3186.0 0.050415 0.050220 0.051743
Add: 4065.9 0.059105 0.059028 0.059161
Triad: 4217.8 0.056916 0.056902 0.056935
-------------------------------------------------------------
Solution Validates: avg error less than 1.000000e-13 on all three arrays
-------------------------------------------------------------
Performance counter stats for './stream':
60,424,173 cache-misses
2.718200988 seconds time elapsed
163.5. 数组10000000，选项-O1¶
root@ubuntu:~/app/stream# gcc -O1 stream.c -o stream
perf stat -e cache-misses ./stream
-------------------------------------------------------------
STREAM version $Revision: 5.10 $
-------------------------------------------------------------
This system uses 8 bytes per array element.
-------------------------------------------------------------
Array size = 10000000 (elements), Offset = 0 (elements)
Memory per array = 76.3 MiB (= 0.1 GiB).
Total memory required = 228.9 MiB (= 0.2 GiB).
Each kernel will be executed 10 times.
The *best* time for each kernel (excluding the first iteration)
will be used to compute the reported bandwidth.
-------------------------------------------------------------
Your clock granularity/precision appears to be 2 microseconds.
Each test below will take on the order of 15230 microseconds.
(= 7615 clock ticks)
Increase the size of the arrays if this shows that
you are not getting at least 20 clock ticks per test.
-------------------------------------------------------------
WARNING -- The above is only a rough guideline.
For best results, please be sure you know the
precision of your system timer.
-------------------------------------------------------------
Function Best Rate MB/s Avg time Min time Max time
Copy: 10603.1 0.015104 0.015090 0.015135
Scale: 11113.3 0.014412 0.014397 0.014426
Add: 11757.3 0.020444 0.020413 0.020470
Triad: 11739.4 0.020467 0.020444 0.020485
-------------------------------------------------------------
Solution Validates: avg error less than 1.000000e-13 on all three arrays
-------------------------------------------------------------
Performance counter stats for './stream':
8,937,017 cache-misses
0.935925494 seconds time elapsed
163.6. 数组10000000，选项-O2¶
root@ubuntu:~/app/stream# gcc -O2 stream.c -o streamperf stat -e cache-misses ./stream
-------------------------------------------------------------
STREAM version $Revision: 5.10 $
-------------------------------------------------------------
This system uses 8 bytes per array element.
-------------------------------------------------------------
Array size = 10000000 (elements), Offset = 0 (elements)
Memory per array = 76.3 MiB (= 0.1 GiB).
Total memory required = 228.9 MiB (= 0.2 GiB).
Each kernel will be executed 10 times.
The *best* time for each kernel (excluding the first iteration)
will be used to compute the reported bandwidth.
-------------------------------------------------------------
Your clock granularity/precision appears to be 1 microseconds.
Each test below will take on the order of 14916 microseconds.
(= 14916 clock ticks)
Increase the size of the arrays if this shows that
you are not getting at least 20 clock ticks per test.
-------------------------------------------------------------
WARNING -- The above is only a rough guideline.
For best results, please be sure you know the
precision of your system timer.
-------------------------------------------------------------
Function Best Rate MB/s Avg time Min time Max time
Copy: 10847.5 0.014777 0.014750 0.014815
Scale: 11175.5 0.014349 0.014317 0.014374
Add: 11782.7 0.020399 0.020369 0.020430
Triad: 11778.0 0.020391 0.020377 0.020417
-------------------------------------------------------------
Solution Validates: avg error less than 1.000000e-13 on all three arrays
-------------------------------------------------------------
Performance counter stats for './stream':
8,511,736 cache-misses
0.916443067 seconds time elapsed
163.7. 数组10000000，选项-O3¶
root@ubuntu:~/app/stream# gcc -O3 stream.c -o stream
perf stat -e cache-misses ./stream
-------------------------------------------------------------
STREAM version $Revision: 5.10 $
-------------------------------------------------------------
This system uses 8 bytes per array element.
-------------------------------------------------------------
Array size = 10000000 (elements), Offset = 0 (elements)
Memory per array = 76.3 MiB (= 0.1 GiB).
Total memory required = 228.9 MiB (= 0.2 GiB).
Each kernel will be executed 10 times.
The *best* time for each kernel (excluding the first iteration)
will be used to compute the reported bandwidth.
-------------------------------------------------------------
Your clock granularity/precision appears to be 1 microseconds.
Each test below will take on the order of 15007 microseconds.
(= 15007 clock ticks)
Increase the size of the arrays if this shows that
you are not getting at least 20 clock ticks per test.
-------------------------------------------------------------
WARNING -- The above is only a rough guideline.
For best results, please be sure you know the
precision of your system timer.
-------------------------------------------------------------
Function Best Rate MB/s Avg time Min time Max time
Copy: 11190.4 0.014314 0.014298 0.014326
Scale: 11327.3 0.014139 0.014125 0.014156
Add: 11374.4 0.021113 0.021100 0.021124
Triad: 11753.8 0.020434 0.020419 0.020447
-------------------------------------------------------------
Solution Validates: avg error less than 1.000000e-13 on all three arrays
-------------------------------------------------------------
Performance counter stats for './stream':
14,925,428 cache-misses
0.911908645 seconds time elapsed
163.8. 数组10000000，选项-O0¶
root@ubuntu:~/app/stream# gcc -O0 stream.c -o stream
perf stat -e cache-misses ./stream
gcc -O0 stream.c -o streamperf stat -e cache-misses ./stream
-
163.9. 数组20000000，选项无¶
root@ubuntu:~/app/stream# gcc -DSTREAM_ARRAY_SIZE 20000000 stream.c= -o stream
root@ubuntu:~/app/stream# perf stat -e cache-misses ./stream
-------------------------------------------------------------
STREAM version $Revision: 5.10 $
-------------------------------------------------------------
This system uses 8 bytes per array element.
-------------------------------------------------------------
Array size = 20000000 (elements), Offset = 0 (elements)
Memory per array = 152.6 MiB (= 0.1 GiB).
Total memory required = 457.8 MiB (= 0.4 GiB).
Each kernel will be executed 10 times.
The *best* time for each kernel (excluding the first iteration)
will be used to compute the reported bandwidth.
-------------------------------------------------------------
Your clock granularity/precision appears to be 2 microseconds.
Each test below will take on the order of 125238 microseconds.
(= 62619 clock ticks)
Increase the size of the arrays if this shows that
you are not getting at least 20 clock ticks per test.
-------------------------------------------------------------
WARNING -- The above is only a rough guideline.
For best results, please be sure you know the
precision of your system timer.
-------------------------------------------------------------
Function Best Rate MB/s Avg time Min time Max time
Copy: 2549.6 0.125998 0.125508 0.128678
Scale: 3345.5 0.097012 0.095650 0.101765
Add: 4172.2 0.117473 0.115047 0.120862
Triad: 4232.1 0.190047 0.113418 0.794803
-------------------------------------------------------------
Solution Validates: avg error less than 1.000000e-13 on all three arrays
-------------------------------------------------------------
Performance counter stats for './stream':
135,075,891 cache-misses
6.586831272 seconds time elapsed
163.10. 数组20000000，选项-O¶
root@ubuntu:~/app/stream# gcc -O -DSTREAM_ARRAY_SIZE=20000000 stream.c -o streamperf stat -e cache-misses ./stream
-------------------------------------------------------------
STREAM version $Revision: 5.10 $
-------------------------------------------------------------
This system uses 8 bytes per array element.
-------------------------------------------------------------
Array size = 20000000 (elements), Offset = 0 (elements)
Memory per array = 152.6 MiB (= 0.1 GiB).
Total memory required = 457.8 MiB (= 0.4 GiB).
Each kernel will be executed 10 times.
The *best* time for each kernel (excluding the first iteration)
will be used to compute the reported bandwidth.
-------------------------------------------------------------
Your clock granularity/precision appears to be 2 microseconds.
Each test below will take on the order of 29887 microseconds.
(= 14943 clock ticks)
Increase the size of the arrays if this shows that
you are not getting at least 20 clock ticks per test.
-------------------------------------------------------------
WARNING -- The above is only a rough guideline.
For best results, please be sure you know the
precision of your system timer.
-------------------------------------------------------------
Function Best Rate MB/s Avg time Min time Max time
Copy: 11145.6 0.028769 0.028711 0.028829
Scale: 11149.8 0.028731 0.028700 0.028757
Add: 12317.8 0.039278 0.038968 0.039673
Triad: 12387.1 0.038914 0.038750 0.039191
-------------------------------------------------------------
Solution Validates: avg error less than 1.000000e-13 on all three arrays
-------------------------------------------------------------
Performance counter stats for './stream':
19,804,344 cache-misses
1.803501622 seconds time elapsed
163.11. 数组20000000，选项-O1¶
root@ubuntu:~/app/stream# gcc -O1 -DSTREAM_ARRAY_SIZE=20000000 stream.c -o streamperf stat -e cache-misses ./stream
-------------------------------------------------------------
STREAM version $Revision: 5.10 $
-------------------------------------------------------------
This system uses 8 bytes per array element.
-------------------------------------------------------------
Array size = 20000000 (elements), Offset = 0 (elements)
Memory per array = 152.6 MiB (= 0.1 GiB).
Total memory required = 457.8 MiB (= 0.4 GiB).
Each kernel will be executed 10 times.
The *best* time for each kernel (excluding the first iteration)
will be used to compute the reported bandwidth.
-------------------------------------------------------------
Your clock granularity/precision appears to be 2 microseconds.
Each test below will take on the order of 32049 microseconds.
(= 16024 clock ticks)
Increase the size of the arrays if this shows that
you are not getting at least 20 clock ticks per test.
-------------------------------------------------------------
WARNING -- The above is only a rough guideline.
For best results, please be sure you know the
precision of your system timer.
-------------------------------------------------------------
Function Best Rate MB/s Avg time Min time Max time
Copy: 9760.2 0.032807 0.032786 0.032823
Scale: 9978.5 0.032094 0.032069 0.032113
Add: 11772.8 0.040799 0.040772 0.040848
Triad: 11914.5 0.040312 0.040287 0.040324
-------------------------------------------------------------
Solution Validates: avg error less than 1.000000e-13 on all three arrays
-------------------------------------------------------------
Performance counter stats for './stream':
21,508,150 cache-misses
1.925709392 seconds time elapsed
163.12. 数组20000000，选项-O2¶
root@ubuntu:~/app/stream# gcc -O2 -DSTREAM_ARRAY_SIZE=20000000 stream.c -o streamperf stat -e cache-misses ./stream
-------------------------------------------------------------
STREAM version $Revision: 5.10 $
-------------------------------------------------------------
This system uses 8 bytes per array element.
-------------------------------------------------------------
Array size = 20000000 (elements), Offset = 0 (elements)
Memory per array = 152.6 MiB (= 0.1 GiB).
Total memory required = 457.8 MiB (= 0.4 GiB).
Each kernel will be executed 10 times.
The *best* time for each kernel (excluding the first iteration)
will be used to compute the reported bandwidth.
-------------------------------------------------------------
Your clock granularity/precision appears to be 2 microseconds.
Each test below will take on the order of 31427 microseconds.
(= 15713 clock ticks)
Increase the size of the arrays if this shows that
you are not getting at least 20 clock ticks per test.
-------------------------------------------------------------
WARNING -- The above is only a rough guideline.
For best results, please be sure you know the
precision of your system timer.
-------------------------------------------------------------
Function Best Rate MB/s Avg time Min time Max time
Copy: 9762.9 0.032804 0.032777 0.032827
Scale: 9688.2 0.033068 0.033030 0.033112
Add: 12236.8 0.039240 0.039226 0.039267
Triad: 12132.6 0.039607 0.039563 0.039621
-------------------------------------------------------------
Solution Validates: avg error less than 1.000000e-13 on all three arrays
-------------------------------------------------------------
Performance counter stats for './stream':
19,257,528 cache-misses
1.883242611 seconds time elapsed
163.13. 数组20000000，选项-O3¶
root@ubuntu:~/app/stream# gcc -O3 -DSTREAM_ARRAY_SIZE=20000000 stream.c -o streamperf stat -e cache-misses ./stream
-------------------------------------------------------------
STREAM version $Revision: 5.10 $
-------------------------------------------------------------
This system uses 8 bytes per array element.
-------------------------------------------------------------
Array size = 20000000 (elements), Offset = 0 (elements)
Memory per array = 152.6 MiB (= 0.1 GiB).
Total memory required = 457.8 MiB (= 0.4 GiB).
Each kernel will be executed 10 times.
The *best* time for each kernel (excluding the first iteration)
will be used to compute the reported bandwidth.
-------------------------------------------------------------
Your clock granularity/precision appears to be 2 microseconds.
Each test below will take on the order of 31440 microseconds.
(= 15720 clock ticks)
Increase the size of the arrays if this shows that
you are not getting at least 20 clock ticks per test.
-------------------------------------------------------------
WARNING -- The above is only a rough guideline.
For best results, please be sure you know the
precision of your system timer.
-------------------------------------------------------------
Function Best Rate MB/s Avg time Min time Max time
Copy: 10445.6 0.030642 0.030635 0.030653
Scale: 10470.5 0.030577 0.030562 0.030598
Add: 11711.1 0.040998 0.040987 0.041024
Triad: 11779.7 0.040759 0.040748 0.040772
-------------------------------------------------------------
Solution Validates: avg error less than 1.000000e-13 on all three arrays
-------------------------------------------------------------
Performance counter stats for './stream':
30,113,752 cache-misses
1.845901002 seconds time elapsed
163.14. 数组20000000，选项-O0¶
root@ubuntu:~/app/stream# gcc -O0 -DSTREAM_ARRAY_SIZE=20000000 stream.c -o streamperf stat -e cache-misses ./stream
-------------------------------------------------------------
STREAM version $Revision: 5.10 $
-------------------------------------------------------------
This system uses 8 bytes per array element.
-------------------------------------------------------------
Array size = 20000000 (elements), Offset = 0 (elements)
Memory per array = 152.6 MiB (= 0.1 GiB).
Total memory required = 457.8 MiB (= 0.4 GiB).
Each kernel will be executed 10 times.
The *best* time for each kernel (excluding the first iteration)
will be used to compute the reported bandwidth.
-------------------------------------------------------------
Your clock granularity/precision appears to be 2 microseconds.
Each test below will take on the order of 125272 microseconds.
(= 62636 clock ticks)
Increase the size of the arrays if this shows that
you are not getting at least 20 clock ticks per test.
-------------------------------------------------------------
WARNING -- The above is only a rough guideline.
For best results, please be sure you know the
precision of your system timer.
-------------------------------------------------------------
Function Best Rate MB/s Avg time Min time Max time
Copy: 2549.0 0.126023 0.125538 0.128636
Scale: 3220.9 0.099850 0.099352 0.101575
Add: 4206.3 0.117327 0.114115 0.120934
Triad: 4233.4 0.114978 0.113385 0.118181
-------------------------------------------------------------
Solution Validates: avg error less than 1.000000e-13 on all three arrays
-------------------------------------------------------------
Performance counter stats for './stream':
124,664,340 cache-misses
5.506577423 seconds time elapsed