UDP client CPU utilization has increased on Jetpack 5

Robotics & Edge Computing Jetson Systems Jetson AGX Xavier
,
1

I have several UDP senders/receivers running on my AGX Xavier that are now consuming more than double the expected CPU after upgrading to Jetpack 5 from Jetpack 4. Please confirm if this issue is reproducible on your end by running the minimal example below.

Test Commands
Sender:
iperf -s

Receiver:
./topp iperf -c 127.0.0.1 -u -b 30M -t 600

topp is a profiling script that measures CPU utilization of a program(originally posted on stackoverflow)

#!/bin/bash
set -e
topp() (
  if [ -n "$O" ]; then
    $* &
  else
    $* &>/dev/null &
  fi
  pid="$!"
  trap "kill $pid" SIGINT
  o='%cpu,%mem,vsz,rss'
  printf '%s\n' "$o"
  i=0
  while s="$(ps --no-headers -o "$o" -p "$pid")"; do
    printf "$i $s\n"
    i=$(($i + 1))
    sleep "${T:-0.1}"
  done
)

topp $*

CPU Utilization:
Jetpack 4: 3.7%
Jetpack 5: 8.7%

3

Hi,
Could you run sudo jetson_clocks and compare sudo tegrastats of Jetpack 4 and Jetpack 5? And the latest versions are 4.6.4 and 5.1.3. It would be great if you can compare the latest releases.

4

tegrastats isn’t very helpful when you have to profile a single program, that is why I used the script I shared above.

However, here’s the output of tegrastats while I had the iperf client running on Jetpack 5.1.1. I am using a carrier board from Connect Tech that currently does not have a 5.1.3-based BSP.

01-30-2024 05:45:28 RAM 3421/31012MB (lfb 151x4MB) SWAP 169/36138MB (cached 1MB) CPU [2%@2263,0%@2263,20%@2263,0%@2263,0%@2262,1%@2263,0%@2262,0%@2261] EMC_FREQ 0%@2133 GR3D_FREQ 0%@1377 VIC_FREQ 115 APE 150 AUX@45C CPU@44.5C thermal@44.7C Tboard@44C AO@44C GPU@44.5C iwlwifi@59C Tdiode@47.25C PMIC@50C GPU 936mW/936mW CPU 1218mW/1218mW SOC 2436mW/2436mW CV 0mW/0mW VDDRQ 280mW/280mW SYS5V 4737mW/4737mW
01-30-2024 05:45:29 RAM 3421/31012MB (lfb 151x4MB) SWAP 169/36138MB (cached 1MB) CPU [4%@2262,0%@2262,0%@2263,17%@2261,1%@2263,0%@2263,0%@2261,0%@2261] EMC_FREQ 0%@2133 GR3D_FREQ 0%@1377 VIC_FREQ 115 APE 150 AUX@45C CPU@44.5C thermal@44.7C Tboard@44C AO@44C GPU@44.5C iwlwifi@59C Tdiode@47C PMIC@50C GPU 937mW/936mW CPU 1218mW/1218mW SOC 2437mW/2436mW CV 0mW/0mW VDDRQ 281mW/280mW SYS5V 4737mW/4737mW
01-30-2024 05:45:30 RAM 3421/31012MB (lfb 151x4MB) SWAP 169/36138MB (cached 1MB) CPU [3%@2263,0%@2259,0%@2262,20%@2262,0%@2262,0%@2262,0%@2262,0%@2262] EMC_FREQ 0%@2133 GR3D_FREQ 0%@1377 VIC_FREQ 115 APE 150 AUX@45C CPU@44.5C thermal@44.7C Tboard@44C AO@44C GPU@44.5C iwlwifi@59C Tdiode@47C PMIC@50C GPU 937mW/936mW CPU 1218mW/1218mW SOC 2437mW/2436mW CV 0mW/0mW VDDRQ 281mW/280mW SYS5V 4737mW/4737mW
01-30-2024 05:45:31 RAM 3421/31012MB (lfb 151x4MB) SWAP 169/36138MB (cached 1MB) CPU [2%@2262,0%@2263,1%@2262,19%@2263,1%@2263,0%@2258,0%@2263,0%@2256] EMC_FREQ 0%@2133 GR3D_FREQ 0%@1377 VIC_FREQ 115 APE 150 AUX@45C CPU@44.5C thermal@44.7C Tboard@44C AO@44C GPU@44.5C iwlwifi@59C Tdiode@47C PMIC@50C GPU 937mW/936mW CPU 1218mW/1218mW SOC 2437mW/2436mW CV 0mW/0mW VDDRQ 281mW/280mW SYS5V 4737mW/4737mW
01-30-2024 05:45:32 RAM 3420/31012MB (lfb 151x4MB) SWAP 169/36138MB (cached 1MB) CPU [4%@2265,1%@2262,0%@2263,19%@2262,0%@2263,0%@2261,0%@2263,0%@2260] EMC_FREQ 0%@2133 GR3D_FREQ 0%@1377 VIC_FREQ 115 APE 150 AUX@45C CPU@44.5C thermal@44.7C Tboard@44C AO@44C GPU@44.5C iwlwifi@59C Tdiode@46.75C PMIC@50C GPU 937mW/936mW CPU 1218mW/1218mW SOC 2437mW/2436mW CV 0mW/0mW VDDRQ 281mW/280mW SYS5V 4737mW/4737mW
01-30-2024 05:45:33 RAM 3420/31012MB (lfb 151x4MB) SWAP 169/36138MB (cached 1MB) CPU [3%@2263,0%@2261,0%@2265,21%@2263,0%@2263,0%@2262,0%@2263,0%@2262] EMC_FREQ 0%@2133 GR3D_FREQ 0%@1377 VIC_FREQ 115 APE 150 AUX@45C CPU@44.5C thermal@44.7C Tboard@44C AO@44C GPU@44.5C iwlwifi@59C Tdiode@47C PMIC@50C GPU 937mW/936mW CPU 1218mW/1218mW SOC 2437mW/2436mW CV 0mW/0mW VDDRQ 281mW/280mW SYS5V 4776mW/4743mW
01-30-2024 05:45:34 RAM 3430/31012MB (lfb 151x4MB) SWAP 169/36138MB (cached 1MB) CPU [5%@2263,15%@2263,3%@2262,15%@2262,1%@2267,4%@2262,5%@2258,2%@2263] EMC_FREQ 0%@2133 GR3D_FREQ 0%@1377 VIC_FREQ 115 APE 150 AUX@45C CPU@45.5C thermal@44.7C Tboard@44C AO@44C GPU@44.5C iwlwifi@59C Tdiode@47C PMIC@50C GPU 936mW/936mW CPU 1967mW/1325mW SOC 2436mW/2436mW CV 0mW/0mW VDDRQ 280mW/280mW SYS5V 4776mW/4748mW
01-30-2024 05:45:35 RAM 3446/31012MB (lfb 151x4MB) SWAP 169/36138MB (cached 1MB) CPU [92%@2263,5%@2262,10%@2262,9%@2261,8%@2263,2%@2263,4%@2266,9%@2263] EMC_FREQ 0%@2133 GR3D_FREQ 0%@1377 VIC_FREQ 115 APE 150 AUX@45C CPU@45.5C thermal@44.85C Tboard@44C AO@44C GPU@44.5C iwlwifi@59C Tdiode@47.25C PMIC@50C GPU 936mW/936mW CPU 2715mW/1498mW SOC 2528mW/2448mW CV 0mW/0mW VDDRQ 374mW/292mW SYS5V 4776mW/4751mW
6

CPU usage has improved considerably after disabling the KPTI mitigation for Meltdown, by adding kpti=0 to the kernel boot parameters. CPU usage for the example posted earlier in the thread is now around 6% of a CPU core.