import triton
import triton.language as tl

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()

@triton_heuristics.reduction(
    size_hints={'x': 1, 'r0_': 2048},
    reduction_hint=ReductionHint.INNER,
    filename=__file__,
    triton_meta={'signature': {'in_ptr0': '*bf16', 'in_ptr1': '*bf16', 'out_ptr1': '*bf16', 'xnumel': 'constexpr', 'r0_numel': 'i32', 'XBLOCK': 'constexpr', 'R0_BLOCK': 'constexpr'}, 'device': DeviceProperties(type='cuda', index=0, multi_processor_count=48, cc=121, major=12, regs_per_multiprocessor=65536, max_threads_per_multi_processor=1536, max_threads_per_block=1024, warp_size=32), 'constants': {'xnumel': 1}, 'native_matmul': False, 'enable_fp_fusion': True, 'launch_pdl': False, 'disable_ftz': False, 'configs': [{(0,): [['tt.divisibility', 16]], (1,): [['tt.divisibility', 16]], (2,): [['tt.divisibility', 16]], (4,): [['tt.divisibility', 16]]}]},
    inductor_meta={'grid_type': 'Grid1D', 'autotune_hints': set(), 'kernel_name': 'triton_red_fused__to_copy_add_mean_mul_pow_rsqrt_0', 'mutated_arg_names': [], 'optimize_mem': True, 'no_x_dim': False, 'atomic_add_found': False, 'num_load': 3, 'num_store': 1, 'num_reduction': 1, 'backend_hash': 'BE9F3F68E84A48F2C239366BB10BBF3F0E1498DBECF7F747524E7B7961E82579', '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, 'deterministic': False, 'force_filter_reduction_configs': False, 'mix_order_reduction_allow_multi_stages': False, 'are_deterministic_algorithms_enabled': False, 'tiling_scores': {'r0_': 16384}}
)
@triton.jit
def triton_red_fused__to_copy_add_mean_mul_pow_rsqrt_0(in_ptr0, in_ptr1, out_ptr1, xnumel, r0_numel, XBLOCK : tl.constexpr, R0_BLOCK : tl.constexpr):
    xnumel = 1
    r0_numel = 2048
    rnumel = r0_numel
    RBLOCK: tl.constexpr = R0_BLOCK
    xoffset = tl.program_id(0) * XBLOCK
    xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
    xmask = tl.full([XBLOCK], True, tl.int1)[:, None]
    r0_base = tl.arange(0, R0_BLOCK)[None, :]
    rbase = r0_base
    _tmp4 = tl.full([XBLOCK, R0_BLOCK], 0, tl.float32)
    for r0_offset in tl.range(0, r0_numel, R0_BLOCK):
        r0_index = r0_offset + r0_base
        r0_mask = r0_index < r0_numel
        roffset = r0_offset
        rindex = r0_index
        r0_0 = r0_index
        tmp0 = tl.load(in_ptr0 + (r0_0), r0_mask, eviction_policy='evict_last', other=0.0).to(tl.float32)
        tmp1 = tmp0.to(tl.float32)
        tmp2 = tmp1 * tmp1
        tmp3 = tl.broadcast_to(tmp2, [XBLOCK, R0_BLOCK])
        tmp5 = _tmp4 + tmp3
        _tmp4 = tl.where(r0_mask, tmp5, _tmp4)
    tmp4 = tl.sum(_tmp4, 1)[:, None]
    for r0_offset in tl.range(0, r0_numel, R0_BLOCK):
        r0_index = r0_offset + r0_base
        r0_mask = r0_index < r0_numel
        roffset = r0_offset
        rindex = r0_index
        r0_0 = r0_index
        tmp6 = tl.load(in_ptr0 + (r0_0), r0_mask, eviction_policy='evict_first', other=0.0).to(tl.float32)
        tmp15 = tl.load(in_ptr1 + (r0_0), r0_mask, eviction_policy='evict_first', other=0.0).to(tl.float32)
        tmp7 = tmp6.to(tl.float32)
        tmp8 = tl.full([1, 1], 2048.0, tl.float32)
        tmp9 = (tmp4 / tmp8)
        tmp10 = tl.full([1, 1], 1e-05, tl.float32)
        tmp11 = tmp9 + tmp10
        tmp12 = libdevice.rsqrt(tmp11)
        tmp13 = tmp7 * tmp12
        tmp14 = tmp13.to(tl.float32)
        tmp16 = tmp14 * tmp15
        tl.store(out_ptr1 + (tl.broadcast_to(r0_0, [XBLOCK, R0_BLOCK])), tmp16, r0_mask)