Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add debug log statements to optimizer to see why nodes are fused #388

Merged
merged 2 commits into from
Feb 19, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions .github/workflows/docs.yml
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,8 @@ on:
push:
branches:
- main
permissions:
contents: write
jobs:
build:
runs-on: ubuntu-20.04
Expand Down
17 changes: 17 additions & 0 deletions cubed/core/optimization.py
Original file line number Diff line number Diff line change
@@ -1,3 +1,5 @@
import logging

import networkx as nx

from cubed.primitive.blockwise import (
Expand All @@ -7,6 +9,8 @@
fuse_multiple,
)

logger = logging.getLogger(__name__)


def simple_optimize_dag(dag):
"""Apply map blocks fusion."""
Expand Down Expand Up @@ -108,16 +112,20 @@ def can_fuse_predecessors(

# if node itself can't be fused then there is nothing to fuse
if not is_fusable(nodes[name]):
logger.debug("can't fuse %s since it is not fusable", name)
return False

# if no predecessor ops can be fused then there is nothing to fuse
if all(not is_fusable(nodes[pre]) for pre in predecessor_ops(dag, name)):
logger.debug("can't fuse %s since no predecessor ops can be fused", name)
return False

# if node is in never_fuse or always_fuse list then it overrides logic below
if never_fuse is not None and name in never_fuse:
logger.debug("can't fuse %s since it is in 'never_fuse'", name)
return False
if always_fuse is not None and name in always_fuse:
logger.debug("can fuse %s since it is in 'always_fuse'", name)
return True

# if there is more than a single predecessor op, and the total number of source arrays to
Expand All @@ -128,6 +136,12 @@ def can_fuse_predecessors(
for pre in predecessor_ops(dag, name)
)
if total_source_arrays > max_total_source_arrays:
logger.debug(
"can't fuse %s since total number of source arrays (%s) exceeds max (%s)",
name,
total_source_arrays,
max_total_source_arrays,
)
return False

predecessor_primitive_ops = [
Expand All @@ -136,6 +150,7 @@ def can_fuse_predecessors(
if is_fusable(nodes[pre])
]
return can_fuse_multiple_primitive_ops(
name,
nodes[name]["primitive_op"],
predecessor_primitive_ops,
max_total_num_input_blocks=max_total_num_input_blocks,
Expand Down Expand Up @@ -219,6 +234,8 @@ def multiple_inputs_optimize_dag(
):
"""Fuse multiple inputs."""
for name in list(nx.topological_sort(dag)):
if name.startswith("array-"):
continue
dag = fuse_predecessors(
dag,
name,
Expand Down
51 changes: 48 additions & 3 deletions cubed/primitive/blockwise.py
Original file line number Diff line number Diff line change
@@ -1,4 +1,5 @@
import itertools
import logging
import math
from collections.abc import Iterator
from dataclasses import dataclass
Expand All @@ -23,6 +24,9 @@

from .types import CubedArrayProxy, MemoryModeller, PrimitiveOperation

logger = logging.getLogger(__name__)


sym_counter = 0


Expand Down Expand Up @@ -352,6 +356,7 @@ def can_fuse_primitive_ops(


def can_fuse_multiple_primitive_ops(
name: str,
primitive_op: PrimitiveOperation,
predecessor_primitive_ops: List[PrimitiveOperation],
*,
Expand All @@ -362,27 +367,67 @@ def can_fuse_multiple_primitive_ops(
):
# If the peak projected memory for running all the predecessor ops in
# order is larger than allowed_mem then we can't fuse.
if peak_projected_mem(predecessor_primitive_ops) > primitive_op.allowed_mem:
peak_projected = peak_projected_mem(predecessor_primitive_ops)
if peak_projected > primitive_op.allowed_mem:
logger.debug(
"can't fuse %s since peak projected memory for predecessor ops (%s) is greater than allowed (%s)",
name,
peak_projected,
primitive_op.allowed_mem,
)
return False
# If the number of input blocks for each input is not uniform, then we
# can't fuse. (This should never happen since all operations are
# currently uniform, and fused operations are too if fuse is applied in
# topological order.)
num_input_blocks = primitive_op.pipeline.config.num_input_blocks
if not all(num_input_blocks[0] == n for n in num_input_blocks):
logger.debug(
"can't fuse %s since number of input blocks for each input is not uniform: %s",
name,
num_input_blocks,
)
return False
if max_total_num_input_blocks is None:
# If max total input blocks not specified, then only fuse if num
# tasks of predecessor ops match.
return all(
ret = all(
primitive_op.num_tasks == p.num_tasks for p in predecessor_primitive_ops
)
if ret:
logger.debug(
"can fuse %s since num tasks of predecessor ops match", name
)
else:
logger.debug(
"can't fuse %s since num tasks of predecessor ops do not match",
name,
)
return ret
else:
total_num_input_blocks = 0
for ni, p in zip(num_input_blocks, predecessor_primitive_ops):
for nj in p.pipeline.config.num_input_blocks:
total_num_input_blocks += ni * nj
return total_num_input_blocks <= max_total_num_input_blocks
ret = total_num_input_blocks <= max_total_num_input_blocks
if ret:
logger.debug(
"can fuse %s since total number of input blocks (%s) does not exceed max (%s)",
name,
total_num_input_blocks,
max_total_num_input_blocks,
)
else:
logger.debug(
"can't fuse %s since total number of input blocks (%s) exceeds max (%s)",
name,
total_num_input_blocks,
max_total_num_input_blocks,
)
return ret
logger.debug(
"can't fuse %s since primitive op and predecessors are not all candidates", name
)
return False


Expand Down
Loading