python/tvm/topi/cuda/softmax.py

# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements.  See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership.  The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License.  You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied.  See the License for the
# specific language governing permissions and limitations
# under the License.
# pylint: disable=invalid-name, unused-variable, trailing-whitespace
"""Schedule for softmax operator"""
from tvm.target import Target
from tvm import te
from tvm.contrib import cudnn
from .. import generic
from .injective import schedule_injective_from_existing
from ..utils import get_const_int, traverse_inline


def _schedule_softmax(softmax_op, s, outs, tgt):
    op_tag = softmax_op.tag
    axis = get_const_int(softmax_op.attrs["axis"])  # reduce axis
    if op_tag == "softmax_output":
        expsum = softmax_op.input_tensors[1]
        exp = softmax_op.input_tensors[0]
        max_elem = s[exp].op.input_tensors[1]
        delta = None
    elif op_tag == "fast_softmax_output":
        expsum = softmax_op.input_tensors[1]
        exp = softmax_op.input_tensors[0]
        delta = s[exp].op.input_tensors[0]
        max_elem = s[delta].op.input_tensors[1]
    elif op_tag == "log_softmax_output":
        exp = None
        delta = None
        max_elem = softmax_op.input_tensors[1]
        expsum = softmax_op.input_tensors[2]
    else:
        raise ValueError(
            f"Tag is expected to be softmax_output or log_softmax_output. Got {op_tag}"
        )

    # The nvptx and rocm backends only supports 32-bits warp shuffle
    # instructions.
    #
    # TODO(tvm-team) Fix nvptx codegen or deprecate nvptx backend.
    def sched_warp_softmax():
        if tgt.kind.name in ["nvptx", "rocm"]:
            dtype = softmax_op.output(0).dtype
            return dtype in ["float32", "int32"]
        if tgt.kind.name != "cuda":
            # this is used as the gpu schedule for other arches which
            # may not have warp reductions
            return False
        return True

    if len(outs[0].shape) != 2:
        ops = [max_elem.op, expsum.op, softmax_op]
        if delta is not None:
            ops.append(delta.op)
        if exp is not None:
            ops.append(exp.op)
        if softmax_op != outs[0].op:
            ops.append(outs[0].op)

        for op in ops:
            s = schedule_injective_from_existing(s, op.output(0))

    elif sched_warp_softmax():
        # A warp of 32 threads performs a row reduction.
        num_thread = tgt.thread_warp_size
        block_x = te.thread_axis("blockIdx.x")
        thread_x = te.thread_axis((0, num_thread), "threadIdx.x")

        # (4) softmax
        output = outs[0]
        xo, xi = s[output].split(output.op.axis[axis], nparts=num_thread)
        xio, xii = s[output].split(xi, factor=4)
        s[output].vectorize(xii)
        s[output].bind(xo, thread_x)
        s[output].bind(output.op.axis[axis ^ 1], block_x)
        s[output].reorder(output.op.axis[axis ^ 1], xo, xio, xii)

        if softmax_op != outs[0].op:
            s[softmax_op].compute_at(s[output], xio)
            s[softmax_op].vectorize(softmax_op.axis[axis])  # vec_len == 4

        # (3) expsum
        k = expsum.op.reduce_axis[0]
        ko, _ = s[expsum].split(k, nparts=num_thread)
        s[expsum].bind(ko, thread_x)
        s[expsum].compute_at(s[output], xo)

        # (2) exp
        if delta is not None:
            s[exp].compute_inline()
            s[delta].compute_inline()
        elif exp is not None:
            xo, xi = s[exp].split(exp.op.axis[axis], nparts=num_thread)
            _, xii = s[exp].split(xi, factor=4)
            s[exp].vectorize(xii)
            s[exp].bind(xo, thread_x)
            s[exp].compute_at(s[expsum], expsum.op.axis[0])
            s[exp].compute_at(s[output], output.op.axis[axis ^ 1])
            s[exp].set_scope("warp")

        # (1) max_elem
        k = max_elem.op.reduce_axis[0]
        ko, _ = s[max_elem].split(k, nparts=num_thread)
        s[max_elem].bind(ko, thread_x)
        if exp is not None and delta is None:
            s[max_elem].compute_at(s[exp], xo)
        else:
            s[max_elem].bind(ko, thread_x)
            s[max_elem].bind(max_elem.op.axis[0], block_x)

    else:
        num_thread = 64
        block_x = te.thread_axis("blockIdx.x")
        thread_x = te.thread_axis((0, num_thread), "threadIdx.x")

        if delta is not None:
            s[exp].compute_inline()
            s[delta].compute_inline()
        elif exp is not None:
            s[exp].bind(exp.op.axis[axis ^ 1], block_x)

        s[max_elem].bind(max_elem.op.axis[0], block_x)
        k = expsum.op.reduce_axis[0]
        ko, ki = s[expsum].split(k, factor=num_thread)
        EF = s.rfactor(expsum, ki)
        s[expsum].bind(s[expsum].op.axis[0], block_x)
        s[expsum].bind(s[expsum].op.reduce_axis[0], thread_x)
        s[EF].compute_at(s[expsum], s[expsum].op.reduce_axis[0])
        s[expsum].set_store_predicate(thread_x.var.equal(0))

        output = outs[0]
        tx, xi = s[output].split(output.op.axis[axis], nparts=num_thread)
        s[output].bind(output.op.axis[axis ^ 1], block_x)
        s[output].bind(tx, thread_x)
        s[output].reorder(output.op.axis[axis ^ 1], tx, xi)

        if softmax_op != outs[0].op:
            s[softmax_op].compute_at(s[output], tx)


def schedule_softmax(outs):
    """Schedule for softmax op.

    Parameters
    ----------
    outs: Array of Tensor
          The computation graph description of softmax in the format
          of an array of tensors.

    Returns
    -------
    sch: Schedule
        The computation schedule for the op.
    """
    outs = [outs] if isinstance(outs, te.tensor.Tensor) else outs
    s = te.create_schedule([x.op for x in outs])
    tgt = Target.current(allow_none=False)

    def _callback(op):
        if "softmax" in op.tag:
            _schedule_softmax(op, s, outs, tgt)

    traverse_inline(s, outs[0].op, _callback)
    return s


def softmax_cudnn(x, axis=-1):
    """Perform softmax on the data using cudnn"""
    return cudnn.softmax(x, axis)


def schedule_softmax_cudnn(outs):
    """Schedule for softmax cudnn op"""
    return generic.schedule_extern(outs)


def log_softmax_cudnn(x, axis=-1):
    """Perform log_softmax on the data using cudnn"""
    return cudnn.log_softmax(x, axis)


def schedule_log_softmax_cudnn(outs):
    """Schedule for log_softmax cudnn op"""
    return generic.schedule_extern(outs)