Skip to content

NPU2 support for tiling_exploration per_tile and tile_group examples #3107

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

Merged
jgmelber merged 2 commits into from
May 29, 2026
Merged

NPU2 support for tiling_exploration per_tile and tile_group examples #3107

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
2534cbd
tiling_exploration: add NPU2 support to per_tile and tile_group
hunhoffe May 22, 2026
ea3b6c2
Merge branch 'main' into npu2-tiling-exploration
jgmelber May 29, 2026
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
3 changes: 2 additions & 1 deletion programming_examples/basic/tiling_exploration/per_tile/Makefile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,6 +12,7 @@ srcdir := $(shell dirname $(realpath $(firstword $(MAKEFILE_LIST))))

include ${srcdir}/../../../makefile-common

devicename ?= $(if $(filter 1,$(NPU2)),npu2,npu)
tensor_height = 8
tensor_width = 8
tile_height = 2
Expand All @@ -25,7 +26,7 @@ all: build/final_${data_str}.xclbin

build/aie_${data_str}.mlir: ${srcdir}/${aie_py_src}
mkdir -p ${@D}
python3 $< --tensor-height ${tensor_height} --tensor-width ${tensor_width} --tile-height ${tile_height} --tile-width ${tile_width} > $@
python3 $< --tensor-height ${tensor_height} --tensor-width ${tensor_width} --tile-height ${tile_height} --tile-width ${tile_width} --device ${devicename} > $@

build/final_${data_str}.xclbin: build/aie_${data_str}.mlir
mkdir -p ${@D}
Expand Down
26 changes: 23 additions & 3 deletions programming_examples/basic/tiling_exploration/per_tile/per_tile.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,13 +9,18 @@
import numpy as np

from aie.iron import Buffer, ObjectFifo, Program, Runtime, Worker
from aie.iron.device import NPU1Col1
from aie.iron.device import NPU1Col1, NPU2
from aie.iron.controlflow import range_
from aie.helpers.taplib import TensorTiler2D


def generate_module(
tensor_height, tensor_width, tile_height, tile_width, generate_access_map=False
tensor_height,
tensor_width,
tile_height,
tile_width,
generate_access_map=False,
device="npu",
):
tensor_size = tensor_height * tensor_width
tile_size = tile_height * tile_width
Expand Down Expand Up @@ -57,8 +62,15 @@ def access_order(of_out, counter_buf):
for t in tiler:
rt.drain(of_out.cons(), tensor_out, t, wait=True)

if device == "npu":
dev = NPU1Col1()
elif device == "npu2":
dev = NPU2()
else:
raise ValueError(f"[ERROR] Device name {device} is unknown")

# Create the program from the device type and runtime
my_program = Program(NPU1Col1(), rt)
my_program = Program(dev, rt)

# Place components (assign them resources on the device) and generate an MLIR module
return my_program.resolve_program()
Expand All @@ -71,6 +83,7 @@ def main(opts):
opts.tile_height,
opts.tile_width,
opts.generate_access_map,
opts.device,
)
if not opts.generate_access_map:
print(module)
Expand All @@ -87,6 +100,13 @@ def get_arg_parser():
action="store_true",
help="Produce a file showing data access order",
)
p.add_argument(
"-d",
"--device",
default="npu",
choices=["npu", "npu2"],
help="Target NPU device",
)
return p


Expand Down
11 changes: 11 additions & 0 deletions programming_examples/basic/tiling_exploration/per_tile/run_strix_makefile.lit
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
// (c) Copyright 2026 Advanced Micro Devices, Inc.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
// REQUIRES: ryzen_ai_npu2, peano
//
// RUN: mkdir -p test_stx
// RUN: cd test_stx
// RUN: make -f %S/Makefile clean
// RUN: make -f %S/Makefile devicename=npu2
// RUN: %run_on_npu2% make -f %S/Makefile run_py devicename=npu2
// CHECK: Running...
3 changes: 2 additions & 1 deletion programming_examples/basic/tiling_exploration/tile_group/Makefile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,6 +12,7 @@ srcdir := $(shell dirname $(realpath $(firstword $(MAKEFILE_LIST))))

include ${srcdir}/../../../makefile-common

devicename ?= $(if $(filter 1,$(NPU2)),npu2,npu)
tensor_height = 8
tensor_width = 8
tile_height = 2
Expand All @@ -25,7 +26,7 @@ all: build/final_${data_str}.xclbin

build/aie_${data_str}.mlir: ${srcdir}/${aie_py_src}
mkdir -p ${@D}
python3 $< --tensor-height ${tensor_height} --tensor-width ${tensor_width} --tile-height ${tile_height} --tile-width ${tile_width} > $@
python3 $< --tensor-height ${tensor_height} --tensor-width ${tensor_width} --tile-height ${tile_height} --tile-width ${tile_width} --device ${devicename} > $@

build/final_${data_str}.xclbin: build/aie_${data_str}.mlir
mkdir -p ${@D}
Expand Down
11 changes: 11 additions & 0 deletions programming_examples/basic/tiling_exploration/tile_group/run_strix_makefile.lit
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
// (c) Copyright 2026 Advanced Micro Devices, Inc.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
// REQUIRES: ryzen_ai_npu2, peano
//
// RUN: mkdir -p test_stx
// RUN: cd test_stx
// RUN: make -f %S/Makefile clean
// RUN: make -f %S/Makefile devicename=npu2
// RUN: %run_on_npu2% make -f %S/Makefile run_py devicename=npu2
// CHECK: Running...
26 changes: 23 additions & 3 deletions programming_examples/basic/tiling_exploration/tile_group/tile_group.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,15 +9,20 @@
import numpy as np

from aie.iron import ObjectFifo, Program, Runtime, Worker
from aie.iron.device import NPU1Col1
from aie.iron.device import NPU1Col1, NPU2
from aie.helpers.taplib import TensorTiler2D
from aie.iron.controlflow import range_
import aie.extras.dialects.arith as arith
from aie.helpers.util import np_dtype_to_mlir_type


def generate_module(
tensor_height, tensor_width, tile_height, tile_width, generate_access_map=False
tensor_height,
tensor_width,
tile_height,
tile_width,
generate_access_map=False,
device="npu",
):
# define types
dtype = np.int32
Expand Down Expand Up @@ -57,7 +62,14 @@ def access_order(of_out):
rt.start(worker)
rt.drain(of_out.cons(), tensor_out, t, wait=True)

my_program = Program(NPU1Col1(), rt)
if device == "npu":
dev = NPU1Col1()
elif device == "npu2":
dev = NPU2()
else:
raise ValueError(f"[ERROR] Device name {device} is unknown")

my_program = Program(dev, rt)

# Place components (assign them resources on the device) and generate an MLIR module
return my_program.resolve_program()
Expand All @@ -70,6 +82,7 @@ def main(opts):
opts.tile_height,
opts.tile_width,
opts.generate_access_map,
opts.device,
)
if not opts.generate_access_map:
print(module)
Expand All @@ -86,6 +99,13 @@ def get_arg_parser():
action="store_true",
help="Produce a file showing data access order",
)
p.add_argument(
"-d",
"--device",
default="npu",
choices=["npu", "npu2"],
help="Target NPU device",
)
return p


Expand Down
Loading