如果硬碟沒有分割區，可以用以下指令擴大硬碟區域:
https://unix.stackexchange.com/questions/643026/how-to-resize-a-disk-without-partition

1
2

# resize2fs /dev/[DISK_ID][PARTITION_NUMBER]
resize2fs /dev/sdc

https://tech.sars.tw/posts/gcp-linux-enlarge-disk-size/

https://www.tecmint.com/parted-command-create-linux-partitions/

https://godleon.github.io/blog/Linux/Linux-extend-lvm-from-unused-space/

LVM說明
https://linux.vbird.org/linux_basic_train/centos8/unit14.php

可以用symble link的方式，將服務檔放到/etc/systemd/system/目錄下，讓systemd管理服務。

更改service檔後，需執行以下指令，讓systemd重新載入設定檔:

1
2
3

sudo systemctl daemon-reload
sudo systemctl enable appname
sudo systemctl start appname

移除服務:

1
2
3
4
5
6
7

systemctl stop [servicename]
systemctl disable [servicename]
rm /your/service/locations/[servicename]
# and symlinks that might be related
rm /your/service/locations/[servicename] 
systemctl daemon-reload
systemctl reset-failed

查看服務整的log

1

journalctl -u appname

參考:
製作fastapi服務
設定檔重載
symbolic link
移除服務

載入Region_TRT plugin

必須要載入libnvinfer_plugin，可以用python的trt.init_libnvinfer_plugins載入，

1
2

TRT_LOGGER = trt.Logger(trt.Logger.ERROR)
trt.init_libnvinfer_plugins(TRT_LOGGER,"")

https://docs.nvidia.com/deeplearning/tensorrt/api/python_api/infer/Plugin/IPluginRegistry.html#tensorrt.init_libnvinfer_plugins

印出目前已經載入的plugin

PLUGIN_CREATORS = trt.get_plugin_registry().plugin_creator_list
for plugin_creator in PLUGIN_CREATORS:
print(plugin_creator.name)

參考並修改:https://developer.nvidia.com/zh-cn/blog/tensorrt-custom-layer-cn/

tensorrt Region layer 說明書

https://docs.nvidia.com/deeplearning/tensorrt/api/c_api/structnvinfer1_1_1plugin_1_1_region_parameters.html#ad2c6bba4f07221add9b6d9abf0a4e312

inference範例參考:
https://leimao.github.io/blog/TensorRT-Python-Inference/

VScode

(float[5])v_anchors
查看大小為5的float array

參考:
https://github.com/microsoft/vscode-cpptools/issues/688#issuecomment-299262784

https://stackoverflow.com/questions/52721440/how-to-expand-an-array-while-debugging-in-visual-studio-code

GDB

列印位址

1
2

//變數c的位址
p&c

https://stackoverflow.com/questions/43861731/gdb-print-the-value-of-memory-address

列印位址的內容

1

(gdb) p *(struct node *) 0x00000000004004fc

https://stackoverflow.com/questions/43861731/gdb-print-the-value-of-memory-address

segmentation fault 除錯

1. gdb run 並且帶入執行參數
   gdb –args executablename arg1 arg2 arg3
   (gdb) run

https://stackoverflow.com/questions/6121094/how-do-i-run-a-program-with-commandline-arguments-using-gdb-within-a-bash-script

2. https://www.cnblogs.com/realjimmy/p/12850884.html

參考:

https://hackernoon.com/how-to-set-up-c-debugging-in-vscode-using-a-makefile

參考:
https://lloydrochester.com/post/c/unix-daemon-example/

https://www.freedesktop.org/software/systemd/man/latest/daemon.html

https://stackoverflow.com/questions/17954432/creating-a-daemon-in-linux/17955149#17955149

clangd cuda 設定
https://www.youtube.com/watch?v=gN3XeFwZ4ng

範例下載
https://github.com/NVIDIA/cuda-samples/tree/v12.4

非常單的cuda 程式撰寫
https://developer.nvidia.com/blog/even-easier-introduction-cuda/

Nsight Visual Studio Code Edition
https://marketplace.visualstudio.com/items?itemName=NVIDIA.nsight-vscode-edition

程式撰寫

https://www.olcf.ornl.gov/cuda-training-series/
https://learnopencv.com/demystifying-gpu-architectures-for-deep-learning/
https://paulrichmond.shef.ac.uk/teaching/COM4521/

http://www.cs.nccu.edu.tw/~jang/teaching/CompNet_files/Wireshark-%E5%9F%BA%E7%A4%8E%E6%95%99%E5%AD%B8.pdf

w32tm有兩個地方可以設定，一個是registry，一個是policy(GPO)
registry可以用regedit開啟，位於
HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\W32Time

policy用gpedit.msc可以開啟，位於
Computer Configuration\Policies\Administrative Templates\System\Windows Time Service

下面指令測試是否可以連線到NTP server

1

w32tm /stripchart /computer:time.windows.com /samples:5 

下面指令可以設定NTP server

1

w32tm /config /manualpeerlist:NTP.iead.local /syncfromflags:manual /update

w32tm /query /configuration
確認看到以下訊息，(本機)代表是從registry讀取的設定，(原則)代表是從 policy(GPO) 讀取的設定

1
2
3
4
5

NtpServer (本機)
DllName: C:\Windows\SYSTEM32\w32time.DLL (本機)
Enabled: 1 (原則)
InputProvider: 0 (本機)
AllowNonstandardModeCombinations: 1 (本機)

w32tm /query /configuration /Verbose
可以詳細看到設定的NTP伺服器

設定w32tm log
https://learn.microsoft.com/en-us/troubleshoot/windows-server/active-directory/turn-on-debug-logging-in-windows-time-service

To turn on debug logging in the Windows Time service:

Start Registry Editor.

Locate and then click the registry key: HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\W32Time\Config.

On the Edit menu, click New Value, and then add the following registry values:

Value Name: FileLogSize
Data Type: DWORD
Value data: 10000000
This registry value specifies the size of the log file in bytes. A value of 10000000 bytes will limit the log file to approximately 10 MB.

Value name: FileLogName
Data Type: String
Value data: C:\Windows\Temp\w32time.log
This registry value specifies the location of the log file. The path is not fixed. You can use a different path.

Value name: FileLogEntries
Data Type: String
Value: 0-116
This registry value specifies the level of detail of the information in the debug log. If you must have more detailed logging information, contact a Microsoft Support Professional.

執行 regedit
執行

IBM重點參考文章
https://www.dell.com/support/kbdoc/zh-cn/000134430/windows-time-service-%E6%95%85%E9%9A%9C-%E5%A4%84%E7%90%86

https://www.ravenswoodtechnology.com/network-time-protocol-configurations-a-deeper-dive/

https://www.ravenswoodtechnology.com/in-sync-proper-time-configuration-in-ad/

https://superuser.com/questions/451018/how-can-i-query-an-ntp-server-under-windows

https://www.renanrodrigues.com/post/how-to-configure-ntp-server-in-active-directory-step-by-step

Windows作為NTP server
https://support.hanwhavision.com/hc/en-us/articles/26570683589529-How-to-Setup-an-NTP-Server-on-Windows-10

• 取得compute capability
  編譯CUDA提供的example並且執行deviceQuery程式，注意example要下載正確的CUDA版本

• Cuda Error (209): cudaLaunchKernel returned cudaErrorNoKernelImageForDevice

https://stackoverflow.com/questions/62901027/cuda-error-209-cudalaunchkernel-returned-cudaerrornokernelimagefordevice

• 關於compute capability更進一步的說明
  https://stackoverflow.com/questions/35656294/cuda-how-to-use-arch-and-code-and-sm-vs-compute/35657430#35657430

• 正確取得CUDA error
  https://stackoverflow.com/questions/14038589/what-is-the-canonical-way-to-check-for-errors-using-the-cuda-runtime-api

參考

https://stackoverflow.com/questions/62901027/cuda-error-209-cudalaunchkernel-returned-cudaerrornokernelimagefordevice