When trying to install from source from the current main branch, I get failures due to "-Wno-switch-bool" not being recognised by nvcc - see error message snippet below.
Is it possible that the flag was supposed to be added to the cxx extra compile args, rather than the nvcc args? If I add it there (line 395, like this: "cxx": ["-O3", "-Wno-switch-bool"],), compilation also seems to work fine.
I've opened a PR that moves the flag to cxx: #244
Error message snippet:
...
nvcc fatal : Unknown option '-Wno-switch-bool'
ninja: build stopped: subcommand failed.
Traceback (most recent call last):
File "...miniconda3/envs/sglang/lib/python3.11/site-packages/torch/utils/cpp_extension.py", line 2107, in _run_ninja_build
subprocess.run(
File ".../miniconda3/envs/sglang/lib/python3.11/subprocess.py", line 571, in run
raise CalledProcessError(retcode, process.args,
subprocess.CalledProcessError: Command '['ninja', '-v', '-j', '127']' returned non-zero exit status 1.
When trying to install from source from the current main branch, I get failures due to "-Wno-switch-bool" not being recognised by
nvcc
- see error message snippet below.It seems this is caused by this commit: https://github.com/flashinfer-ai/flashinfer/commit/84a433a6d5f6cebcd535276b76531fdd49781e64 which adds this flag to the
nvcc
extra compile arguments insetup.py
. If I comment out the"-Wno-switch-bool"
line insetup.py
I can compile and install just fine.Is it possible that the flag was supposed to be added to the
cxx
extra compile args, rather than thenvcc
args? If I add it there (line 395, like this:"cxx": ["-O3", "-Wno-switch-bool"],
), compilation also seems to work fine.I've opened a PR that moves the flag to
cxx
: #244Error message snippet: