logcumsumexp_inductor_static.py

# 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

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_dberard/i5/ci5pipzer3zksupmrmrjquugfjpdi4554fbou3bdeayfbyshpqda.py
# Topologically Sorted Source Nodes: [logcumsumexp], Original ATen: [aten.logcumsumexp]
# Source node to ATen node mapping:
#   logcumsumexp => logcumsumexp
# Graph fragment:
#   %logcumsumexp : [num_users=1] = call_function[target=torch.ops.aten.logcumsumexp.default](args = (%arg0_1, 0), kwargs = {})
triton_per_fused_logcumsumexp_0 = async_compile.triton('triton_per_fused_logcumsumexp_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()
@triton.jit
def _triton_helper_fn_minimum_maximum_ne_isinf_bitwise_not_bitwise_or_sub_exp_log1p_add_where0(arg0_0, arg1_0):
    tmp0 = triton_helpers.minimum(arg0_0, arg1_0)
    tmp1 = triton_helpers.maximum(arg0_0, arg1_0)
    tmp2 = tmp0 != tmp1
    tmp3 = libdevice.isinf(tmp0).to(tl.int1)
    tmp4 = ~tmp3
    tmp5 = tmp2 | tmp4
    tmp6 = tmp0 - tmp1
    tmp7 = tl_math.exp(tmp6)
    tmp8 = libdevice.log1p(tmp7)
    tmp9 = tmp8 + tmp1
    tmp10 = tl.where(tmp5, tmp9, arg0_0)
    return tmp10
@triton_heuristics.persistent_reduction(
    size_hints=[32, 16],
    reduction_hint=ReductionHint.DEFAULT,
    filename=__file__,
    triton_meta={'signature': {'in_ptr0': '*fp32', 'out_ptr0': '*fp32', 'xnumel': 'i32', 'rnumel': 'i32'}, 'device': DeviceProperties(type='cuda', index=0, cc=90, major=9, regs_per_multiprocessor=65536, max_threads_per_multi_processor=2048, multi_processor_count=132, warp_size=32), 'constants': {}, 'configs': [AttrsDescriptor.from_dict({'arg_properties': {'tt.divisibility': (0, 1, 2, 3), 'tt.equal_to': ()}, 'cls': 'AttrsDescriptor'})]},
    inductor_meta={'autotune_hints': set(), 'kernel_name': 'triton_per_fused_logcumsumexp_0', 'mutated_arg_names': [], 'optimize_mem': True, 'no_x_dim': False, 'num_load': 1, 'num_reduction': 0, 'backend_hash': '63A118008A8D9F5C10AB08E75F67129677D36B011F4277B312BCA9383911253E', '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}
)
@triton.jit
def triton_per_fused_logcumsumexp_0(in_ptr0, out_ptr0, xnumel, rnumel, XBLOCK : tl.constexpr):
    xnumel = 32
    rnumel = 16
    RBLOCK: tl.constexpr = 16
    xoffset = tl.program_id(0) * XBLOCK
    xindex = xoffset + tl.arange(0, XBLOCK)[:, None]
    xmask = xindex < xnumel
    rindex = tl.arange(0, RBLOCK)[None, :]
    roffset = 0
    rmask = tl.full([XBLOCK, RBLOCK], True, tl.int1)
    r1 = rindex
    x0 = xindex
    tmp0 = tl.load(in_ptr0 + (x0 + (32*r1)), xmask, other=0.0)
    tmp1 = tmp0.to(tl.float32)
    tmp2 = tl.broadcast_to(tmp1, [XBLOCK, RBLOCK])
    tmp3, = tl.associative_scan((tmp2,), 1, _triton_helper_fn_minimum_maximum_ne_isinf_bitwise_not_bitwise_or_sub_exp_log1p_add_where0)
    tl.store(out_ptr0 + (x0 + (32*r1)), tmp3, xmask)
''', device_str='cuda')


async_compile.wait(globals())
del async_compile

def call(args):
    arg0_1, = args
    args.clear()
    assert_size_stride(arg0_1, (16, 32), (32, 1))
    with torch.cuda._DeviceGuard(0):
        torch.cuda.set_device(0)
        buf0 = empty_strided_cuda((16, 32), (32, 1), torch.float32)
        # Topologically Sorted Source Nodes: [logcumsumexp], Original ATen: [aten.logcumsumexp]
        stream0 = get_raw_stream(0)
        triton_per_fused_logcumsumexp_0.run(arg0_1, buf0, 32, 16, grid=grid(32), stream=stream0)
        del arg0_1
    return (buf0, )


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((16, 32), (32, 1), device='cuda:0', dtype=torch.float32)
    fn = lambda: call([arg0_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)