jittor在Ubuntu20.04.1中运行custom_op例程出现bug

我在windows下运行这段程序没有问题，转到Ubuntu中却报了错，docker版本和Jittor镜像都是完全一致的。不明白具体是什么原因导致的。
docker版本 20.10.7
程序：

import jittor as jt

header ="""
#pragma once
#include "op.h"

namespace jittor {

struct CustomOp : Op {
    Var* output;
    CustomOp(NanoVector shape, NanoString dtype=ns_float);
    
    const char* name() const override { return "custom"; }
    DECLARE_jit_run;
};

} // jittor
"""

src = """
#include "var.h"
#include "custom_op.h"

namespace jittor {
#ifndef JIT
CustomOp::CustomOp(NanoVector shape, NanoString dtype) {
    flags.set(NodeFlags::_cuda, 1);
    flags.set(NodeFlags::_cpu, 1);
    output = create_output(shape, dtype);
}

void CustomOp::jit_prepare() {
    add_jit_define("T", output->dtype());
}

#else // JIT
#ifdef JIT_cpu
void CustomOp::jit_run() {
    index_t num = output->num;
    auto* __restrict__ x = output->ptr<T>();
    for (index_t i=0; i<num; i++)
        x[i] = (T)i;
}
#else
// JIT_cuda
__global__ void kernel(index_t n, T *x) {
    int index = blockIdx.x * blockDim.x + threadIdx.x;
    int stride = blockDim.x * gridDim.x;
    for (int i = index; i < n; i += stride)
        x[i] = (T)-i;
}

void CustomOp::jit_run() {
    index_t num = output->num;
    auto* __restrict__ x = output->ptr<T>();
    int blockSize = 256;
    int numBlocks = (num + blockSize - 1) / blockSize;
    kernel<<<numBlocks, blockSize>>>(num, x);
}
#endif // JIT_cpu
#endif // JIT

} // jittor
"""

my_op = jt.compile_custom_op(header, src, "custom", warp=False)

报错信息：

RemoteTraceback                           Traceback (most recent call last)
RemoteTraceback: 
"""
Traceback (most recent call last):
  File "/usr/lib/python3.7/multiprocessing/pool.py", line 121, in worker
    result = (True, func(*args, **kwds))
  File "/usr/local/lib/python3.7/dist-packages/jittor_utils/__init__.py", line 146, in do_compile
    return cc.cache_compile(cmd, cache_path, jittor_path)
RuntimeError: [f 0716 00:51:51.026669 96 log.cc:329] Check failed ret(256) == 0(0) Run cmd failed: cd /root/.cache/jittor/default/g++ && /usr/bin/g++ /root/.cache/jittor/default/g++/custom_ops/gen_ops_custom.cc   -Wall -Werror -Wno-unknown-pragmas -std=c++14 -fPIC -march=native  -fdiagnostics-color=always  -I/usr/local/lib/python3.7/dist-packages/jittor/src -I/usr/include/python3.7m -I/usr/local/lib/python3.7/dist-packages/pybind11/include -I/root/.cache/jittor/default/g++  -O2  -I'/root/.cache/jittor/default/g++/custom_ops'  -c  -o /root/.cache/jittor/default/g++/obj_files/gen_ops_custom.cc.o


The above exception was the direct cause of the following exception:

RuntimeError                              Traceback (most recent call last)
<ipython-input-1-d2b4a46fbb07> in <module>
     64 """
     65 
---> 66 my_op = jt.compile_custom_op(header, src, "custom", warp=False)

/usr/local/lib/python3.7/dist-packages/jittor/lock.py in inner(*args, **kw)
     30         def inner(*args, **kw):
     31             with self:
---> 32                 ret = func(*args, **kw)
     33             return ret
     34         return inner

/usr/local/lib/python3.7/dist-packages/jittor/compiler.py in compile_custom_op(header, source, op_name, warp)
    584     with open(ccname, 'w') as f:
    585         f.write(source)
--> 586     m = compile_custom_ops([hname, ccname])
    587     return getattr(m, op_name)
    588 

/usr/local/lib/python3.7/dist-packages/jittor/lock.py in inner(*args, **kw)
     30         def inner(*args, **kw):
     31             with self:
---> 32                 ret = func(*args, **kw)
     33             return ret
     34         return inner

/usr/local/lib/python3.7/dist-packages/jittor/compiler.py in compile_custom_ops(filenames, extra_flags, return_module, dlopen_flags, gen_name_)
    673     LOG.vvv(f"Build custum ops lib:{gen_lib}")
    674     LOG.vvvv(f"Build sources:{builds}")
--> 675     compile(cc_path, extra_flags+cc_flags+opt_flags+includes, builds, gen_lib)
    676 
    677     # add python path and import

/usr/local/lib/python3.7/dist-packages/jittor/compiler.py in compile(compiler, flags, inputs, output, combind_build)
     88             cmd = cmd.replace("-Ofast", "-O2")
     89         cmds.append(cmd)
---> 90     jit_utils.run_cmds(cmds, cache_path, jittor_path, "Compiling "+base_output)
     91     cmd = f"{compiler} {' '.join(obj_files)} {flags} {lto_flags} {link} -o {output}"
     92     return do_compile(cmd)

/usr/local/lib/python3.7/dist-packages/jittor_utils/__init__.py in run_cmds(cmds, cache_path, jittor_path, msg)
    176         n = len(cmds)
    177         dp = DelayProgress(msg, n)
--> 178         for i,_ in enumerate(p.imap_unordered(do_compile, cmds)):
    179             dp.update(i)
    180     finally:

/usr/lib/python3.7/multiprocessing/pool.py in next(self, timeout)
    746         if success:
    747             return value
--> 748         raise value
    749 
    750     __next__ = next                    # XXX