import os
import subprocess

import torch
from setuptools import setup, find_packages
from torch.utils.cpp_extension import BuildExtension, CUDAExtension, ROCM_HOME

from typing import Optional, Union
import subprocess
from pathlib import Path


# ninja build does not work unless include_dirs are abs path
this_dir = os.path.dirname(os.path.abspath(__file__))


def check_cuda_torch_binary_vs_bare_metal(cuda_dir):
    torch_binary_major = torch.version.cuda.split(".")[0]
    torch_binary_minor = torch.version.cuda.split(".")[1]

    print("\nCompiling cuda extensions with")

def append_nvcc_threads(nvcc_extra_args):
    return nvcc_extra_args


if not torch.cuda.is_available():
    # https://github.com/NVIDIA/apex/issues/486
    # Extension builds after https://github.com/pytorch/pytorch/pull/23408 attempt to query torch.cuda.get_device_capability(),
    # which will fail if you are compiling in an environment without visible GPUs (e.g. during an nvidia-docker build command).
    print(
        '\nWarning: Torch did not find available GPUs on this system.\n',
        'If your intention is to cross-compile, this is not an error.\n'
        'By default, FastFold will cross-compile for Pascal (compute capabilities 6.0, 6.1, 6.2),\n'
        'Volta (compute capability 7.0), Turing (compute capability 7.5),\n'
        'and, if the CUDA version is >= 11.0, Ampere (compute capability 8.0).\n'
        'If you wish to cross-compile for a single specific architecture,\n'
        'export TORCH_CUDA_ARCH_LIST="compute capability" before running setup.py.\n')

print("\n\ntorch.__version__  = {}\n\n".format(torch.__version__))
TORCH_MAJOR = int(torch.__version__.split('.')[0])
TORCH_MINOR = int(torch.__version__.split('.')[1])

if TORCH_MAJOR < 1 or (TORCH_MAJOR == 1 and TORCH_MINOR < 10):
    raise RuntimeError("FastFold requires Pytorch 1.10 or newer.\n" +
                       "The latest stable release can be obtained from https://pytorch.org/")

cmdclass = {}
ext_modules = []

# Set up macros for forward/backward compatibility hack around
# https://github.com/pytorch/pytorch/commit/4404762d7dd955383acee92e6f06b48144a0742e
# and
# https://github.com/NVIDIA/apex/issues/456
# https://github.com/pytorch/pytorch/commit/eb7b39e02f7d75c26d8a795ea8c7fd911334da7e#diff-4632522f237f1e4e728cb824300403ac
version_dependent_macros = ['-DVERSION_GE_1_1', '-DVERSION_GE_1_3', '-DVERSION_GE_1_5']

if ROCM_HOME is None:
    raise RuntimeError(
        "Are you sure your environment has nvcc available?  If you're installing within a container from https://hub.docker.com/r/pytorch/pytorch, only images whose names contain 'devel' will provide nvcc."
    )
else:
    # check_cuda_torch_binary_vs_bare_metal(ROCM_HOME)

    def cuda_ext_helper(name, sources, extra_cuda_flags):
        return CUDAExtension(
            name=name,
            sources=[
                os.path.join('fastfold/model/fastnn/kernel/cuda_native/csrc', path) for path in sources
            ],
            include_dirs=[
                os.path.join(this_dir, 'fastfold/model/fastnn/kernel/cuda_native/csrc/include'),
                os.path.join(this_dir, 'fastfold/model/fastnn/kernel/cuda_native/csrc/'),
            ],
            extra_compile_args={
                'cxx': ['-O3'] + version_dependent_macros,
                'hipcc':
                    append_nvcc_threads(['-O3', '--use_fast_math'] + version_dependent_macros +
                                        extra_cuda_flags)
            })



    cc_flag = ['-gencode', 'arch=compute_70,code=sm_70']

    extra_cuda_flags = [
        '-std=c++14', '-maxrregcount=50', '-U__CUDA_NO_HALF_OPERATORS__',
        '-U__CUDA_NO_HALF_CONVERSIONS__', '--expt-relaxed-constexpr', '--expt-extended-lambda'
    ]

    ext_modules.append(
        cuda_ext_helper('fastfold_layer_norm_cuda',
                        ['layer_norm_cuda.cpp', 'layer_norm_cuda_kernel.cu'],
                        extra_cuda_flags + cc_flag))

    ext_modules.append(
        cuda_ext_helper('fastfold_softmax_cuda', ['softmax_cuda.cpp', 'softmax_cuda_kernel.cu'],
                        extra_cuda_flags + cc_flag))


def get_sha(pytorch_root: Union[str, Path]) -> str:
    try:
        return subprocess.check_output(['git', 'rev-parse', 'HEAD'], cwd=pytorch_root).decode('ascii').strip()
    except Exception:
        return 'Unknown'


def get_version_add(sha: Optional[str] = None) -> str:
    add_version_path = "jax/version.py"
    if sha != 'Unknown':
        if sha is None:
            sha_path = os.getenv('FASTFOLD_DOWNLOAD_PATH', "")
            sha = get_sha(sha_path)
        version = 'git' + sha[:7]

    if os.getenv('FASTFOLD_BUILD_VERSION'):
        version_dtk = os.getenv('FASTFOLD_BUILD_VERSION', "")
        version += "." + version_dtk

    with open(add_version_path, encoding="utf-8", mode="w") as file:
        file.write("__version__='0.2.21'+'+{}'\n".format(version))
        file.write("_minimum_jaxlib_version='0.1.69'\n")
    file.close()


def get_version():
    get_version_add()
    version_file = 'version.py'
    with open(version_file, encoding='utf-8') as f:
        exec(compile(f.read(), version_file, 'exec'))
    return locals()['__version__']


setup(
    name='fastfold',
    #version='0.2.0',
    version=get_version(),
    packages=find_packages(exclude=(
        'assets',
        'benchmark',
        '*.egg-info',
    )),
    description=
    'Optimizing Protein Structure Prediction Model Training and Inference on GPU Clusters',
    ext_modules=ext_modules,
    package_data={'fastfold': ['model/fastnn/kernel/cuda_native/csrc/*']},
    cmdclass={'build_ext': BuildExtension} if ext_modules else {},
    #install_requires=['einops', 'colossalai'],
)