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.

Sign up for GitHub

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

[UKernel] Add bf16/bfp16 ukernel for peano and move tests to peano #1114

Merged
merged 2 commits into from
Feb 17, 2025
Merged

[UKernel] Add bf16/bfp16 ukernel for peano and move tests to peano #1114

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
44da4cb
[UKernel] Add bf16 ukernel for peano and move tests to peano
jtuyls Feb 17, 2025
3b3199e
Clean up
jtuyls Feb 17, 2025
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
85 changes: 11 additions & 74 deletions build_tools/ci/cpu_comparison/run.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1581,74 +1581,6 @@ def __init__(self):
self.existing_names = []
self.tests = []

# Tests Matmul + Trunci.
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Removing the matmul + trunci tests to reduce CI time as there are still matmul+scaling+trunci tests

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We should then cleanup the matmul_trunci* template test and the base class ?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done

# Phoenix : Ukernel + Peano.
self.register(
MatmulTrunci(
256,
128,
32,
"i8",
"i32",
1 * np.ones([256, 32], dtype=np.int8),
1 * np.ones([32, 128], dtype=np.int8),
32 * np.ones([256, 128], dtype=np.int8),
test_params=TestParams(
tile_pipeline="pack-peel-4-level-tiling",
run_on_target=["npu1_4col"],
aie_compilation_flags=[
"--iree-amdaie-num-rows=4",
"--iree-amdaie-num-cols=4",
],
use_ukernel=True,
),
)
)
# Phoenix : Vectorization + Peano.
self.register(
MatmulTrunci(
256,
128,
32,
"i8",
"i32",
1 * np.ones([256, 32], dtype=np.int8),
1 * np.ones([32, 128], dtype=np.int8),
32 * np.ones([256, 128], dtype=np.int8),
test_params=TestParams(
tile_pipeline="pack-peel-4-level-tiling",
run_on_target=["npu1_4col"],
aie_compilation_flags=[
"--iree-amdaie-num-rows=4",
"--iree-amdaie-num-cols=4",
],
),
)
)
# Strix : Ukernel + Chess.
self.register(
MatmulTrunci(
256,
128,
32,
"i8",
"i32",
1 * np.ones([256, 32], dtype=np.int8),
1 * np.ones([32, 128], dtype=np.int8),
32 * np.ones([256, 128], dtype=np.int8),
test_params=TestParams(
tile_pipeline="pack-peel-4-level-tiling",
run_on_target=["npu4"],
aie_compilation_flags=[
"--iree-amdaie-num-rows=4",
"--iree-amdaie-num-cols=8",
],
use_chess=True,
use_ukernel=True,
),
)
)

# Tests Matmul + Trunci with Scaling.
# Phoenix : Ukernel + Peano.
self.register(
Expand Down Expand Up @@ -1696,7 +1628,7 @@ def __init__(self):
use_scaling=True,
)
)
# Strix : Ukernel + Chess.
# Strix : Ukernel + Peano.
self.register(
MatmulTrunci(
256,
Expand All @@ -1714,8 +1646,9 @@ def __init__(self):
"--iree-amdaie-num-rows=4",
"--iree-amdaie-num-cols=8",
],
use_chess=True,
use_chess=False,
use_ukernel=True,
use_chess_for_ukernel=False,
),
use_scaling=True,
)
Expand Down Expand Up @@ -1943,7 +1876,8 @@ def __init__(self):
"f32",
test_params=TestParams(
use_ukernel=True,
use_chess=True,
use_chess=False,
use_chess_for_ukernel=False,
run_on_target=["npu4"],
),
)
Expand All @@ -1958,11 +1892,12 @@ def __init__(self):
test_params=TestParams(
name_suffix="npu4_4x8",
use_ukernel=True,
use_chess=False,
use_chess_for_ukernel=False,
aie_compilation_flags=[
"--iree-amdaie-num-rows=4",
"--iree-amdaie-num-cols=8",
],
use_chess=True,
run_on_target=["npu4"],
),
)
Expand Down Expand Up @@ -2004,7 +1939,8 @@ def __init__(self):
"--iree-amdaie-num-rows=4",
"--iree-amdaie-num-cols=8",
],
use_chess=True,
use_chess=False,
use_chess_for_ukernel=False,
),
)
)
Expand All @@ -2024,7 +1960,8 @@ def __init__(self):
"--iree-amdaie-num-rows=4",
"--iree-amdaie-num-cols=8",
],
use_chess=True,
use_chess=False,
use_chess_for_ukernel=False,
),
)
)
Expand Down
Loading
Loading