FasterTransformer 002: cuda调试env

VSCODE ENV

cmake

NVIDIA Nsight Systems

• 当我们装好了CUDA的时候，其实在图形界面下已经装好了一个叫“nsight”的编译器，我们可以直接用终端打开这个编译器，然后写好程序直接编译然后debug就可以了。
• WINDOWS NVIDIA Nsight Systems 入门及使用
• 在Ubuntu20.04上安装Nsight Systems
• https://docs.nvidia.com/nsight-systems/InstallationGuide/index.html
• https://developer.nvidia.com/tools-overview
• https://developer.nvidia.com/nsight-systems/get-started
• [每日一氵] Nsight Systems (nsys) 使用记录以及cuda程序优化
• https://musetransfer.com/s/khonvnids 请点击链接获取《无主题 - NsightSystems-linux-public-2023.2.1.122-3…》, 有效期至6月17日
  

cuda-gdb

• CUDA 番外篇 | Visual Studio Code的CUDA环境

• error while loading shared libraries: libtinfo.so.5: cannot open shared object file: No such file or directory apt-get install libtinfo5

• launch.json

{
    // Use IntelliSense to learn about possible attributes.
    // Hover to view descriptions of existing attributes.
    // For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
    "version": "0.2.0",
    "configurations": [
        {
            "name": "CUDA C++: Launch",
            "type": "cuda-gdb",
            "request": "launch",
            "program": "/home/pdd/test/gemm_test/test",
            "preLaunchTask": "mynvcc"
        },
        {
            "name": "CUDA C++: Attach",
            "type": "cuda-gdb",
            "request": "attach"
        }
    ]
}

• tasks.json

{
    "version": "2.0.0",
    "tasks": [
        {
            "label": "mynvcc",
            "type": "shell",
            "command": "nvcc",
            "args": ["-g","-G","-o","test","gemm_fp16.cu","-lcublas"] //https://stackoverflow.com/questions/64113574/undefined-reference-to-cublascreate-v2-in-tmp-tmpxft-0000120b-0000000-10-my
        }
    ]
}

libtinfo.so.5

(base) pdd@pdd-Dell-G15-5511:~/Downloads$ sudo apt install libncurses5
[sudo] pdd 的密码： 
正在读取软件包列表... 完成
正在分析软件包的依赖关系树... 完成
正在读取状态信息... 完成                 
下列软件包是自动安装的并且现在不需要了：
  fcitx-config-common fcitx-config-gtk fcitx-frontend-all fcitx-frontend-gtk2 fcitx-frontend-gtk3 fcitx-frontend-qt5 fcitx-module-dbus
  fcitx-module-kimpanel fcitx-module-lua fcitx-module-quickphrase-editor5 fcitx-module-x11 fcitx-modules fcitx-ui-classic g++-11 gir1.2-appindicator3-0.1
  gir1.2-gst-plugins-base-1.0 gir1.2-gstreamer-1.0 gir1.2-keybinder-3.0 gir1.2-wnck-3.0 gnome-session-canberra libfcitx-config4 libfcitx-core0
  libfcitx-gclient1 libfcitx-qt5-1 libfcitx-qt5-data libfcitx-utils0 libgettextpo0 libkeybinder-3.0-0 libpresage-data libpresage1v5 libtinyxml2.6.2v5
  libwnck-3-0 libwnck-3-common presage python3-gi-cairo
使用'sudo apt autoremove'来卸载它(它们)。
将会同时安装下列软件：
  libtinfo5
下列【新】软件包将被安装：
  libncurses5 libtinfo5
升级了 0 个软件包，新安装了 2 个软件包，要卸载 0 个软件包，有 432 个软件包未被升级。
需要下载 107 kB/207 kB 的归档。
解压缩后会消耗 883 kB 的额外空间。
您希望继续执行吗？ [Y/n] y
获取:1 https://mirrors.ustc.edu.cn/ubuntu jammy-security/universe amd64 libncurses5 amd64 6.3-2ubuntu0.1 [107 kB]
已下载 107 kB，耗时 0秒 (334 kB/s)    
正在选中未选择的软件包 libtinfo5:amd64。
(正在读取数据库 ... 系统当前共安装有 283690 个文件和目录。)
准备解压 .../libtinfo5_6.3-2ubuntu0.1_amd64.deb  ...
正在解压 libtinfo5:amd64 (6.3-2ubuntu0.1) ...
正在选中未选择的软件包 libncurses5:amd64。
准备解压 .../libncurses5_6.3-2ubuntu0.1_amd64.deb  ...
正在解压 libncurses5:amd64 (6.3-2ubuntu0.1) ...
正在设置 libtinfo5:amd64 (6.3-2ubuntu0.1) ...
正在设置 libncurses5:amd64 (6.3-2ubuntu0.1) ...
正在处理用于 libc-bin (2.35-0ubuntu3.1) 的触发器 ...
/sbin/ldconfig.real: /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_cnn_infer.so.8 is not a symbolic link

/sbin/ldconfig.real: /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_adv_train.so.8 is not a symbolic link

/sbin/ldconfig.real: /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_ops_infer.so.8 is not a symbolic link

/sbin/ldconfig.real: /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_cnn_train.so.8 is not a symbolic link

/sbin/ldconfig.real: /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn.so.8 is not a symbolic link

/sbin/ldconfig.real: /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_ops_train.so.8 is not a symbolic link

/sbin/ldconfig.real: /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_adv_infer.so.8 is not a symbolic link

(base) pdd@pdd-Dell-G15-5511:~/Downloads$ sudo ln -sf /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_cnn_infer.so.8 /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_cnn_infer.so.8
ln: '/usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_cnn_infer.so.8' 与'/usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_cnn_infer.so.8' 为同一文件
(base) pdd@pdd-Dell-G15-5511:~/Downloads$ sudo ln -sf /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_cnn_infer.so.8.8.0 /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_cnn_infer.so.8
(base) pdd@pdd-Dell-G15-5511:~/Downloads$ ^C
(base) pdd@pdd-Dell-G15-5511:~/Downloads$ sudo ln -sf /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_adv_train.so.8.8.0 /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_adv_train.so.8
(base) pdd@pdd-Dell-G15-5511:~/Downloads$ sudo ln -sf /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_ops_infer.so.8.8.0 /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_ops_infer.so.8
(base) pdd@pdd-Dell-G15-5511:~/Downloads$ sudo ln -sf /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_cnn_train.so.8.8.0 /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_cnn_train.so.8
(base) pdd@pdd-Dell-G15-5511:~/Downloads$ sudo ln -sf /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn.so.8.8.0 /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn.so.8 
(base) pdd@pdd-Dell-G15-5511:~/Downloads$ sudo ln -sf /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_ops_train.so.8.8.0 /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_ops_train.so.8
(base) pdd@pdd-Dell-G15-5511:~/Downloads$ sudo ln -sf /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_adv_infer.so.8.8.0 /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_adv_infer.so.8
(base) pdd@pdd-Dell-G15-5511:~/Downloads$ 

How can I get the libncursesw.so.5 file? Install is “up to date” but files do not exist

• sudo apt-get install libncursesw5

debug by gdb after build use cmke

- /home/pdd/test/FastT/FasterTransformer/.vscode/launch.json
{
    "version": "0.2.0",
    "configurations": [
        {
            "name": "Debug",
            "type": "cppdbg",
            "request": "launch",
            "program": "/home/pdd/test/FastT/FasterTransformer/build/bin/transformer_fp32", // target
            "cwd": "${workspaceRoot}",
            //"preLaunchTask": "build",
            "miDebuggerPath": "/usr/bin/gdb"
        }
    ]
}