#!/usr/bin/python
import sys
import os
import Blender as B

def normpath(path):
    return os.path.abspath(os.path.normpath(path))

Import ('env')

kernel_binaries = []

#Bitness
if B.bitness == 32:
    bits = 32
else:
    bits = 64

if env['WITH_BF_CYCLES_CUDA_BINARIES']:
    kernel = env.Clone()

    # cuda info
    nvcc = env['BF_CYCLES_CUDA_NVCC']
    cuda_archs = env['BF_CYCLES_CUDA_BINARIES_ARCH']

    # build directory
    root_build_dir = normpath(env['BF_BUILDDIR'])
    build_dir = os.path.join(root_build_dir, 'intern/cycles/kernel')

    # source directories and files
    source_dir = Dir('.').srcnode().path
    kernel_file = os.path.join(source_dir, "kernel.cu")
    util_dir = os.path.join(source_dir, "../util")
    svm_dir = os.path.join(source_dir, "../svm")

    # nvcc flags
    nvcc_flags = "-m%s" % (bits)
    nvcc_flags += " --cubin --ptxas-options=\"-v\" --maxrregcount=24"
    nvcc_flags += " --opencc-options -OPT:Olimit=0"
    nvcc_flags += " -DCCL_NAMESPACE_BEGIN= -DCCL_NAMESPACE_END= -DNVCC"
    nvcc_flags += " -I \"%s\" -I \"%s\"" % (util_dir, svm_dir)

    # dependencies
    dependencies = ['kernel.cu'] + kernel.Glob('*.h') + kernel.Glob('../util/*.h') + kernel.Glob('svm/*.h')
    last_cubin_file = None

    # add command for each cuda architecture
    for arch in cuda_archs:
        cubin_file = os.path.join(build_dir, "kernel_%s.cubin" % arch)

        command = "\"%s\" -arch=%s %s \"%s\" -o \"%s\"" % (nvcc, arch, nvcc_flags, kernel_file, cubin_file)

        kernel.Command(cubin_file, 'kernel.cu', command)
        kernel.Depends(cubin_file, dependencies)

        kernel_binaries.append(cubin_file)
        
        if not env['WITH_BF_CYCLES_CUDA_THREADED_COMPILE']:
            # trick to compile one kernel at a time to reduce memory usage
            if last_cubin_file:
                kernel.Depends(cubin_file, last_cubin_file)
            last_cubin_file = cubin_file

Return('kernel_binaries')