import os
import torch

from torch import Tensor

os.environ["TORCHINDUCTOR_FX_GRAPH_CACHE"] = "0"
os.environ["TORCHINDUCTOR_UNIQUE_KERNEL_NAMES"] = "1"
os.environ["TORCHINDUCTOR_BENCHMARK_KERNEL"] = "1"


@torch.compile
def view_copy(target: Tensor, source: Tensor):
    assert target.dtype == torch.bfloat16
    assert source.dtype == torch.uint16
    target.view(torch.uint16).copy_(source)


target: Tensor = torch.ones(65536 * 1024, dtype=torch.bfloat16, device="cuda")
source = torch.full_like(target, 4, dtype=torch.uint16)

target.view(torch.uint16).copy_(source)
print(target[0])  # 3.6734e-40
view_copy(target, source)
print(target[0])  # 4.

@triton_heuristics.pointwise(
    size_hints={'x': 67108864}, 
    filename=__file__,
    triton_meta={'signature': {'in_ptr0': '*u16', 'out_ptr0': '*bf16', 'xnumel': 'i32'}, 'device': DeviceProperties(type='cuda', index=0, multi_processor_count=132, cc=90, major=9, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, warp_size=32), 'constants': {}, 'configs': [AttrsDescriptor.from_dict({'arg_properties': {'tt.divisibility': (0, 1, 2), 'tt.equal_to': ()}, 'cls': 'AttrsDescriptor'})]},
    inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_0', 'mutated_arg_names': ['out_ptr0'], 'optimize_mem': True, 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A0D3A2B50857E9501D843044B01F725922648D76E6D26323B14F8A4EA4473D1B', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False, 'kernel_num_gb': 0.268435456},
    min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
    xnumel = 67108864
    xoffset = tl.program_id(0) * XBLOCK
    xindex = xoffset + tl.arange(0, XBLOCK)[:]
    xmask = tl.full([XBLOCK], True, tl.int1)
    x0 = xindex
    tmp0 = tl.load(in_ptr0 + (x0), None)
    tmp1 = tmp0.to(tl.float32, bitcast=False)
    tl.store(out_ptr0 + (x0), tmp1, None)

# AOT ID: ['0_inference']
from ctypes import c_void_p, c_long, c_int
import torch
import math
import random
import os
import tempfile
from math import inf, nan
from torch._inductor.hooks import run_intermediate_hooks
from torch._inductor.utils import maybe_profile
from torch._inductor.codegen.memory_planning import _align as align
from torch import device, empty_strided
from torch._inductor.async_compile import AsyncCompile
from torch._inductor.select_algorithm import extern_kernels
from torch._inductor.codegen.multi_kernel import MultiKernelCall
import triton
import triton.language as tl
from torch._inductor.runtime.triton_heuristics import (
    grid,
    split_scan_grid,
    grid_combo_kernels,
    start_graph,
    end_graph,
    cooperative_reduction_grid,
)
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
from torch._C import _cuda_getCurrentRawStream as get_raw_stream

aten = torch.ops.aten
inductor_ops = torch.ops.inductor
_quantized = torch.ops._quantized
assert_size_stride = torch._C._dynamo.guards.assert_size_stride
empty_strided_cpu = torch._C._dynamo.guards._empty_strided_cpu
empty_strided_cuda = torch._C._dynamo.guards._empty_strided_cuda
empty_strided_xpu = torch._C._dynamo.guards._empty_strided_xpu
reinterpret_tensor = torch._C._dynamo.guards._reinterpret_tensor
alloc_from_pool = torch.ops.inductor._alloc_from_pool
async_compile = AsyncCompile()
empty_strided_p2p = torch._C._distributed_c10d._SymmetricMemory.empty_strided_p2p


# kernel path: /tmp/torchinductor_root/5m/c5mmszhvnmnuimm6a3j7emw2wh7vx6mwt6uar6eqk5mygz5jqgw4.py
# Topologically Sorted Source Nodes: [], Original ATen: []
# Source node to ATen node mapping:
# Graph fragment:
#   %copy_ : [num_users=0] = call_function[target=torch.ops.aten.copy_.default](args = (%arg0_1, %view_2), kwargs = {})
triton_poi_fused_0 = async_compile.triton('triton_poi_fused_0', '''
import triton
import triton.language as tl
from triton.compiler.compiler import AttrsDescriptor

from torch._inductor.runtime import triton_helpers, triton_heuristics
from torch._inductor.runtime.triton_helpers import libdevice, math as tl_math
from torch._inductor.runtime.hints import AutotuneHint, ReductionHint, TileHint, DeviceProperties
triton_helpers.set_driver_to_gpu()

from torch._dynamo.testing import rand_strided
from torch._C import _cuda_getCurrentRawStream as get_raw_stream
import torch
from torch._inductor.runtime.triton_heuristics import grid, split_scan_grid

@triton_heuristics.pointwise(
    size_hints={'x': 67108864}, 
    filename=__file__,
    triton_meta={'signature': {'in_ptr0': '*u16', 'out_ptr0': '*bf16', 'xnumel': 'i32'}, 'device': DeviceProperties(type='cuda', index=0, multi_processor_count=132, cc=90, major=9, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, warp_size=32), 'constants': {}, 'configs': [AttrsDescriptor.from_dict({'arg_properties': {'tt.divisibility': (0, 1, 2), 'tt.equal_to': ()}, 'cls': 'AttrsDescriptor'})]},
    inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_poi_fused_0', 'mutated_arg_names': ['out_ptr0'], 'optimize_mem': True, 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': 'A0D3A2B50857E9501D843044B01F725922648D76E6D26323B14F8A4EA4473D1B', 'are_deterministic_algorithms_enabled': False, 'assert_indirect_indexing': True, 'autotune_local_cache': True, 'autotune_pointwise': True, 'autotune_remote_cache': None, 'force_disable_caches': False, 'dynamic_scale_rblock': True, 'max_autotune': False, 'max_autotune_pointwise': False, 'min_split_scan_rblock': 256, 'spill_threshold': 16, 'store_cubin': False, 'kernel_num_gb': 0.268435456},
    min_elem_per_thread=0
)
@triton.jit
def triton_poi_fused_0(in_ptr0, out_ptr0, xnumel, XBLOCK : tl.constexpr):
    xnumel = 67108864
    xoffset = tl.program_id(0) * XBLOCK
    xindex = xoffset + tl.arange(0, XBLOCK)[:]
    xmask = tl.full([XBLOCK], True, tl.int1)
    x0 = xindex
    tmp0 = tl.load(in_ptr0 + (x0), None)
    tmp1 = tmp0.to(tl.float32, bitcast=False)
    tl.store(out_ptr0 + (x0), tmp1, None)


def get_args():
    arg_0 = rand_strided((67108864,), (1,), device='cuda:0', dtype=torch.uint16)
    arg_1 = rand_strided((67108864,), (1,), device='cuda:0', dtype=torch.bfloat16)
    return arg_0, arg_1,


def call(args):
    with torch.cuda._DeviceGuard(0):
        torch.cuda.set_device(0)
        stream0 = get_raw_stream(0)
        triton_poi_fused_0.run(*args, 67108864, grid=grid(67108864), stream=stream0)


def benchmark_all_configs(args):
    with torch.cuda._DeviceGuard(0):
        torch.cuda.set_device(0)
        return triton_poi_fused_0.benchmark_all_configs(*args, 67108864, grid=grid(67108864))


if __name__ == '__main__':
    from torch._inductor.runtime.benchmarking import benchmarker

    args = get_args()
    ms = benchmarker.benchmark_gpu(lambda: call(args), rep=40)
    num_gb = 0.268435456
    gb_per_s = num_gb / (ms / 1e3)
    print(f"{ms:.3f}ms    {num_gb:.3f}GB    {gb_per_s:.2f}GB/s")
''', device_str='cuda')


async_compile.wait(globals())
del async_compile

def call(args):
    arg0_1, arg1_1 = args
    args.clear()
    assert_size_stride(arg0_1, (67108864, ), (1, ))
    assert_size_stride(arg1_1, (67108864, ), (1, ))
    with torch.cuda._DeviceGuard(0):
        torch.cuda.set_device(0)
        # Topologically Sorted Source Nodes: [], Original ATen: []
        stream0 = get_raw_stream(0)
        triton_poi_fused_0.run(arg1_1, arg0_1, 67108864, grid=grid(67108864), stream=stream0)
        del arg0_1
        del arg1_1
    return ()


def benchmark_compiled_module(times=10, repeat=10):
    from torch._dynamo.testing import rand_strided
    from torch._inductor.utils import print_performance
    arg0_1 = rand_strided((67108864, ), (1, ), device='cuda:0', dtype=torch.bfloat16)
    arg1_1 = rand_strided((67108864, ), (1, ), device='cuda:0', dtype=torch.uint16)
    fn = lambda: call([arg0_1, arg1_1])
    return print_performance(fn, times=times, repeat=repeat)


if __name__ == "__main__":
    from torch._inductor.wrapper_benchmark import compiled_module_main
    compiled_module_main('None', benchmark_compiled_module)