'--device-debug (-G)' overrides '--generate-line-info (-lineinfo)' in nvcc

[Aurorabili]

Xmake 版本

2.9.6

操作系统版本和架构

Ubuntu 22.04, Linux Kernel 5.14.0-427.31.1.el9_4.x86_64

描述问题

使用 xmake create -l cuda 创建默认 cuda 项目并在 mode.debug 下构建得到 nvcc 警告:

nvcc warning : '--device-debug (-G)' overrides '--generate-line-info (-lineinfo)'
ptxas warning : Conflicting options --device-debug and --generate-line-info specified, ignoring --generate-line-info option

期待的结果

在 mode.debug 下仅包含 --device-debug (-G) 选项。

工程配置

add_rules("mode.debug", "mode.release")

target("hello-cuda")
    set_kind("binary")
    add_files("src/*.cu")
    set_toolchains("cuda")
    add_cugencodes("native")

附加信息和错误日志

console log

devcontainer ➜ /workspaces/hello-cuda $ xmake f -m debug
checking for platform ... linux
checking for architecture ... x86_64
checking for Cuda SDK directory ... /usr/local/cuda
devcontainer ➜ /workspaces/hello-cuda $ xmake -rvD      
checking for /usr/local/cuda/bin/nvcc ... ok
checking for flags (-O0) ... ok
> nvcc "-O0"
checking for flags (-rdc=true) ... ok
> nvcc "-rdc=true"
checking for flags (-gencode arch=compute_61,code=sm_61) ... ok
> nvcc "-gencode" "arch=compute_61,code=sm_61"
[ 50%]: compiling.debug src/main.cu
/usr/local/cuda/bin/nvcc -c -Xcompiler -fPIE -G -g -lineinfo -O0 -I/usr/local/cuda/include -m64 -rdc=true -gencode arch=compute_61,code=sm_61 -o build/.objs/hello-cuda/linux/x86_64/debug/src/main.cu.o src/main.cu
checking for flags (-MMD -MF) ... ok
> nvcc "-MMD" "-MF" "/dev/null"
nvcc warning : '--device-debug (-G)' overrides '--generate-line-info (-lineinfo)'
ptxas warning : Conflicting options --device-debug and --generate-line-info specified, ignoring --generate-line-info option
checking for nvcc ... /usr/local/cuda/bin/nvcc
checking for the cuda linker (culd) ... nvcc
checking for flags (-gencode arch=compute_61,code=sm_61) ... ok
> nvcc "-gencode" "arch=compute_61,code=sm_61"
[ 75%]: devlinking.debug hello-cuda_gpucode.cu.o
/usr/local/cuda/bin/nvcc -o build/.objs/hello-cuda/linux/x86_64/debug/rules/cuda/devlink/hello-cuda_gpucode.cu.o build/.objs/hello-cuda/linux/x86_64/debug/src/main.cu.o -L/usr/local/cuda/lib64 -lcudadevrt -lcudart_static -lrt -lpthread -ldl -m64 -gencode arch=compute_61,code=sm_61 -dlink
checking for g++ ... /usr/bin/g++
checking for the linker (ld) ... g++
checking for /usr/bin/g++ ... ok
checking for flags (-fPIC) ... ok
> g++ "-fPIC" "-m64" "-m64"
checking for flags (-Wl,-rpath=/usr/local/cuda/lib64) ... ok
> g++ "-Wl,-rpath=/usr/local/cuda/lib64" "-m64" "-m64"
[ 75%]: linking.debug hello-cuda
/usr/bin/g++ -o build/linux/x86_64/debug/hello-cuda build/.objs/hello-cuda/linux/x86_64/debug/src/main.cu.o build/.objs/hello-cuda/linux/x86_64/debug/rules/cuda/devlink/hello-cuda_gpucode.cu.o -m64 -L/usr/local/cuda/lib64 -Wl,-rpath=/usr/local/cuda/lib64 -lcudadevrt -lcudart_static -lrt -lpthread -ldl

build cache stats:
cache directory: build/.build_cache
cache hit rate: 0%
cache hit: 0
cache hit total time: 0.000s
cache miss: 0
cache miss total time: 0.000s
new cached files: 0
remote cache hit: 0
remote new cached files: 0
preprocess failed: 0
compile fallback count: 0
compile total time: 0.000s

[100%]: build ok, spent 7.894s

compile_commands.json

[
  {
    "directory": "/workspaces/hello-cuda",
    "arguments": [
      "/usr/local/cuda/bin/nvcc",
      "-c",
      "-Xcompiler",
      "-fPIE",
      "-G",
      "-g",
      "-lineinfo",
      "-O0",
      "-I/usr/local/cuda/include",
      "-m64",
      "-rdc=true",
      "-gencode",
      "arch=compute_61,code=sm_61",
      "-o",
      "build/.objs/hello-cuda/linux/x86_64/debug/src/main.cu.o",
      "src/main.cu"
    ],
    "file": "src/main.cu"
  }
]

[waruqi]

@xq114 Any idea? 6523b8e

[xq114]

Can we turn on --device-debug for debug mode and --generate-line-info for releasedbg mode? The first option will turn off optimization while the second won't. https://nanxiao.me/en/dont-use-g-compile-option-for-profiling-cuda-programs/ seems -lineinfo is made for profiling needs.

[waruqi]

Can we turn on --device-debug for debug mode and --generate-line-info for releasedbg mode? The first option will turn off optimization while the second won't. https://nanxiao.me/en/dont-use-g-compile-option-for-profiling-cuda-programs/ seems -lineinfo is made for profiling needs.

那就只能 symbols 里面同时去掉 -G/-lineinfo，只保留 -g 。。然后 在 module/rules 里面分别对 debug/releasedbg 单独加 add_cuflags("-G") 和 lineinfo

[Aurorabili]

好主意。我注意到 module/rules 中存在 mode.profile，将 -lineinfo 加在这里是否会比加在 mode.releasedbg 更符合语义？

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically.

---

Good idea. I noticed that mode.profile exists in module/rules, would adding -lineinfo there be more semantic than adding mode.releasedbg?

[waruqi]

好主意。我注意到 module/rules 中存在 mode.profile，将 -lineinfo 加在这里是否会比加在 mode.releasedbg 更符合语义？

也可以

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically.

---

Good idea. I noticed that mode.profile exists in module/rules, would adding -lineinfo there be more semantic than adding mode.releasedbg?

OK