【经验分享】AMD Embedded R2000 处理器在极限负载下的Linux实时性能测试结果
之前使用FFT运算、C程序编译，网络通信作为负载，测试了AMD Embedded R2000 在Linux下的实时性能。
最近使用Stress-ng的任务作为负载，再次测试了AMD Embedded R2000 在Linux下的实时性能。Stress-ng能产生更极限的负载，比如DDR带宽使用可以达到10GBps。另外还运行了3个glxgears任务，模拟GPU操作对系统的负载。这种情况，对系统的挑战更大。
测试时，使用一个CPU作为普通CPU，运行常规任务和负载任务；其它CPU作为实时CPU，运行实时任务。在测试中，每个CPU运行一个cyclictest的线程来测量实时性能，以实时CPU的cyclictest latency数据作为实时性能测试结果，忽略普通CPU的cyclictest latency数据。
测试18天latency
测试18天后，实时CPU的cyclictest latency最大是14us。
cyclictest的记录如下：
r2000:/home/xilinx/cyclictest-0612b# tail -n 30 cyclictest_load.sh*-glx.3-log.txt
policy: fifo: loadavg: 18.74 18.98 19.03 20/236 7019
T: 0 ( 3564) P:99 I:125 C:13816897496 Min: 1 Act: 3 Avg: 2 Max: 361
T: 1 ( 3565) P:99 I:125 C:13816897493 Min: 2 Act: 5 Avg: 5 Max: 13
T: 2 ( 3566) P:99 I:125 C:13816857500 Min: 2 Act: 5 Avg: 4 Max: 14
T: 3 ( 3567) P:99 I:125 C:13816857497 Min: 4 Act: 5 Avg: 4 Max: 14
测试55天latency
测试55天后，实时CPU的cyclictest latency最大是14us。
# cat /proc/uptime
4831964.16 14496391.50
# uptime -p
up 7 weeks, 6 days, 22 hours, 12 minutes
# tail -n 30 cyclictest_load.sh*-glx.3-log.txt
policy: fifo: loadavg: 19.40 19.25 19.21 21/236 17137
T: 0 ( 3564) P:99 I:125 C:38651161228 Min: 1 Act: 3 Avg: 2 Max: 361
T: 1 ( 3565) P:99 I:125 C:38651121226 Min: 2 Act: 5 Avg: 5 Max: 13
T: 2 ( 3566) P:99 I:125 C:38651121234 Min: 2 Act: 4 Avg: 4 Max: 14
T: 3 ( 3567) P:99 I:125 C:38651121231 Min: 4 Act: 5 Avg: 4 Max: 14
系统信息
系统信息如下：
r2000:/home/xilinx# uname -a
Linux r2000 5.0.19-rt11-yocto-preempt-rt #1 SMP PREEMPT RT Tue Jun 11 07:34:36 UTC
2024 x86_64 x86_64 x86_64 GNU/Linux
r2000:/home/xilinx/cyclictest-0612b# ifconfig
enp1s0f3u3u1 Link encap:Ethernet HWaddr 20:7b:d5:2c:55:04
inet addr:10.164.16.122 Bcast:10.164.16.255 Mask:255.255.255.0
inet6 addr: fe80::227b:d5ff:fe2c:5504/64 Scope:Link
UP BROADCAST RUNNING MULTICAST DYNAMIC MTU:1500 Metric:1
RX packets:1325288 errors:0 dropped:175485 overruns:0 frame:0
TX packets:586889 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:131977408 (125.8 MiB) TX bytes:532982541 (508.2 MiB)
r2000:/home/xilinx# lscpu
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
CPU(s): 4
On-line CPU(s) list: 0-3
Thread(s) per core: 1
Core(s) per socket: 4
Socket(s): 1
NUMA node(s): 1
Vendor ID: AuthenticAMD
CPU family: 23
Model: 24
Model name: AMD Ryzen Embedded R2544 with Radeon Graphics
Stepping: 1
CPU MHz: 3325.598
CPU max MHz: 3350.0000
CPU min MHz: 1600.0000
BogoMIPS: 6688.12
Virtualization: AMD-V
L1d cache: 32K
L1i cache: 64K
L2 cache: 512K
L3 cache: 4096K
NUMA node0 CPU(s): 0-3
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge
mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt
pdpe1gb rdtscp lm constant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid
aperfmperf pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt
aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm
sse4a misalignsse 3dnowprefetch osvw skinit wdt tce topoext perfctr_core
perfctr_nb bpext perfctr_llc mwaitx cpb hw_pstate sme ssbd sev ibpb vmmcall
fsgsbase bmi1 avx2 smep bmi2 rdseed adx smap clflushopt sha_ni xsaveopt
xsavec xgetbv1 xsaves clzero irperf xsaveerptr arat npt lbrv svm_lock
nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter
pfthreshold avic v_vmsave_vmload vgif overflow_recov succor smca
r2000:/home/xilinx# cat /proc/cmdline
BOOT_IMAGE=/bzImage root=PARTUUID=5f0f822f-4633-4410-b63d-8854abe31035
rw rootwait quiet console=tty0 console=ttyS0,115200n8
rcupdate.rcu_cpu_stall_suppress=1 mce=off hpet=disable
numa_balancing=disable igb.blacklist=no rcu_nocb_poll=1024 rootwait
clocksource=tsc tsc=reliable art=virtallow no_ipi_broadcast=1
nosoftlockup audit=0 nmi_watchdog=0 irqaffinity=0 isolcpus=1-15
rcu_nocbs=1-15 nohz_full=1-15 processor.max_cstate=0
processor_idle.max_cstate=0 console=ttyS0,115200n8 console=tty0
rootfstype=ext4,ext3,ext2
r2000:/home/xilinx# cat /proc/uptime
1727677.04 5183530.30
r2000:/home/xilinx# uptime -p
up 2 weeks, 5 days, 23 hours, 54 minutes
r2000:/home/xilinx# uptime
09:39:10 up 19 days, 23:54, 2 users, load average: 18.98, 19.03, 19.05
r2000:/home/xilinx# last -x reboot
wtmp begins Mon Jul 1 00:00:06 2024
r2000:/home/xilinx# cd /home/xilinx/cyclictest*
r2000:/home/xilinx/cyclictest-0612b# ps -eLF | head -n 3 && ps -eLF | grep stress
UID PID PPID LWP C NLWP SZ RSS PSR STIME TTY TIME CMD
root 1 0 1 0 1 22676 6996 0 Jun13 ? 00:10:28 /sbin/init nosoftlockup
root 2 0 2 0 1 0 0 0 Jun13 ? 00:00:00 [kthreadd]
root 2441 851 2441 0 1 4613 5412 0 Jun13 pts/2 00:00:00
stress-ng -t 30d --vm 2 --vm-bytes 128M
root 2448 2441 2448 4 1 4615 1664 0 Jun13 pts/2 21:33:27
stress-ng -t 30d --vm 2 --vm-bytes 128M
r2000:/home/xilinx/cyclictest-0612b# ps -eLF | grep glx
root 2442 851 2442 0 5 105517 81096 0 Jun13 pts/2 03:25:40 glxgears
root 2442 851 2474 4 5 105517 81096 0 Jun13 pts/2 21:33:27 glxgears
root 2442 851 2475 4 5 105517 81096 0 Jun13 pts/2 21:33:27 glxgears
root 2442 851 2476 4 5 105517 81096 0 Jun13 pts/2 21:33:27 glxgears
root 2442 851 2477 4 5 105517 81096 0 Jun13 pts/2 21:33:27 glxgears
root 2443 851 2443 0 5 105486 80488 0 Jun13 pts/2 04:06:03 glxgears
root 2443 851 2466 4 5 105486 80488 0 Jun13 pts/2 21:33:27 glxgears
root 2443 851 2467 4 5 105486 80488 0 Jun13 pts/2 21:33:27 glxgears
root 2443 851 2468 4 5 105486 80488 0 Jun13 pts/2 21:33:27 glxgears
root 2443 851 2469 4 5 105486 80488 0 Jun13 pts/2 21:33:27 glxgears
root 2444 851 2444 0 5 105514 80672 0 Jun13 pts/2 03:34:05 glxgears
root 2444 851 2470 4 5 105514 80672 0 Jun13 pts/2 21:33:27 glxgears
root 2444 851 2471 4 5 105514 80672 0 Jun13 pts/2 21:33:27 glxgears
root 2444 851 2472 4 5 105514 80672 0 Jun13 pts/2 21:33:27 glxgears
root 2444 851 2473 4 5 105514 80672 0 Jun13 pts/2 21:33:27 glxgears
Wed Jul 3 09:39:04 UTC 2024
/proc/uptime:
1727673.02 5183518.24
可以见到，AMD Embedded X86处理器拥有优秀的实时性能。
测试结果，与BIOS设置、Linux编译选型和设置、系统负载有关。请联系对应的FAE、或者hank.fu@amd.com，获取更详细的信息。
付汉杰 hankf@amd.com
