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': 8192, 'r0_': 1024},
    reduction_hint=ReductionHint.INNER,
    filename=__file__,
    triton_meta={'signature': {'in_ptr0': '*bf16', 'in_ptr1': '*bf16', 'in_ptr2': '*bf16', 'in_ptr3': '*bf16', 'out_ptr1': '*bf16', 'xnumel': 'i32', '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': {}, '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]], (3,): [['tt.divisibility', 16]], (4,): [['tt.divisibility', 16]], (6,): [['tt.divisibility', 16]]}]},
    inductor_meta={'grid_type': 'Grid1D', 'autotune_hints': set(), 'kernel_name': 'triton_red_fused__to_copy__unsafe_view_add_cat_expand_mean_mul_pow_rsqrt_view_5', 'mutated_arg_names': [], 'optimize_mem': True, 'no_x_dim': False, 'atomic_add_found': False, 'num_load': 7, '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': {'x': 0, 'r0_': 32612352}}
)
@triton.jit
def triton_red_fused__to_copy__unsafe_view_add_cat_expand_mean_mul_pow_rsqrt_view_5(in_ptr0, in_ptr1, in_ptr2, in_ptr3, out_ptr1, xnumel, r0_numel, XBLOCK : tl.constexpr, R0_BLOCK : tl.constexpr):
    r0_numel = 1024
    rnumel = r0_numel
    RBLOCK: tl.constexpr = R0_BLOCK
    xoffset = tl.program_id(0) * XBLOCK
    xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
    xmask = xindex < xnumel
    r0_base = tl.arange(0, R0_BLOCK)[None, :]
    rbase = r0_base
    x0 = (xindex % 5)
    x1 = xindex // 5
    x3 = xindex
    _tmp16 = 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_2 = r0_index
        tmp11 = tl.load(in_ptr2 + (r0_2 + 1024*x3), r0_mask & xmask, eviction_policy='evict_last', other=0.0).to(tl.float32)
        tmp0 = x0
        tmp1 = tl.full([1, 1], 0, tl.int64)
        tmp2 = tmp0 >= tmp1
        tmp3 = tl.full([1, 1], 1, tl.int64)
        tmp4 = tmp0 < tmp3
        tmp5 = tl.load(in_ptr0 + (tl.broadcast_to(r0_2, [XBLOCK, R0_BLOCK])), r0_mask & tmp4 & xmask, eviction_policy='evict_last', other=0.0).to(tl.float32)
        tmp6 = tmp0 >= tmp3
        tmp7 = tl.full([1, 1], 5, tl.int64)
        tmp8 = tmp0 < tmp7
        tmp9 = tl.load(in_ptr1 + (r0_2 + 1024*((-1) + x0) + 4096*x1), r0_mask & tmp6 & xmask, eviction_policy='evict_last', other=0.0).to(tl.float32)
        tmp10 = tl.where(tmp4, tmp5, tmp9)
        tmp12 = tmp10 + tmp11
        tmp13 = tmp12.to(tl.float32)
        tmp14 = tmp13 * tmp13
        tmp15 = tl.broadcast_to(tmp14, [XBLOCK, R0_BLOCK])
        tmp17 = _tmp16 + tmp15
        _tmp16 = tl.where(r0_mask & xmask, tmp17, _tmp16)
    tmp16 = tl.sum(_tmp16, 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_2 = r0_index
        tmp29 = tl.load(in_ptr2 + (r0_2 + 1024*x3), r0_mask & xmask, eviction_policy='evict_first', other=0.0).to(tl.float32)
        tmp39 = tl.load(in_ptr3 + (r0_2), r0_mask, eviction_policy='evict_last', other=0.0).to(tl.float32)
        tmp18 = x0
        tmp19 = tl.full([1, 1], 0, tl.int64)
        tmp20 = tmp18 >= tmp19
        tmp21 = tl.full([1, 1], 1, tl.int64)
        tmp22 = tmp18 < tmp21
        tmp23 = tl.load(in_ptr0 + (tl.broadcast_to(r0_2, [XBLOCK, R0_BLOCK])), r0_mask & tmp22 & xmask, eviction_policy='evict_last', other=0.0).to(tl.float32)
        tmp24 = tmp18 >= tmp21
        tmp25 = tl.full([1, 1], 5, tl.int64)
        tmp26 = tmp18 < tmp25
        tmp27 = tl.load(in_ptr1 + (r0_2 + 1024*((-1) + x0) + 4096*x1), r0_mask & tmp24 & xmask, eviction_policy='evict_first', other=0.0).to(tl.float32)
        tmp28 = tl.where(tmp22, tmp23, tmp27)
        tmp30 = tmp28 + tmp29
        tmp31 = tmp30.to(tl.float32)
        tmp32 = tl.full([1, 1], 1024.0, tl.float32)
        tmp33 = (tmp16 / tmp32)
        tmp34 = tl.full([1, 1], 1e-05, tl.float32)
        tmp35 = tmp33 + tmp34
        tmp36 = libdevice.rsqrt(tmp35)
        tmp37 = tmp31 * tmp36
        tmp38 = tmp37.to(tl.float32)
        tmp40 = tmp38 * tmp39
        tl.store(out_ptr1 + (r0_2 + 1024*x3), tmp40, r0_mask & xmask)