Skip to content

Commit 34608e3

Browse files
authored
Fix Zephyr FVP memory layout and correct the tutorial (pytorch#22936)
### Summary The Corstone-320 MobileNetV2 overlay assigns the same ISRAM origin to XIP code and writable data, producing overlapping ELF sections and preventing startup. Split the 4 MiB region into 512 KiB for code and 3.5 MiB for writable data, preserving the boot vectors in ITCM and the model in DDR. Update the tutorial and sample README to use ExecuTorch's standard FVP environment, configure simulator options before CMake, and keep Alif and FVP builds in separate directories. Correct the Alif dependencies and export commands, and clarify CPU boundary operators, calibration, memory reservations, and timing. The operator diagnostic now names the namespaces it checks, since other CPU operators can still run. Authored with AI assistance using OpenAI Codex. ### Test plan Validation reused a prepared workspace with Zephyr 4.4.0, Zephyr SDK 1.0.1, and Corstone-320 FVP 11.27.25. After preparing the tutorial workspace and exporting `mv2_u85_256.pte` for `ethos-u85-256`, the FVP check can be repeated with: ```bash source modules/lib/executorch/examples/arm/arm-scratch/setup_path.sh export ARMFVP_EXTRA_FLAGS="--simlimit 10 -C mps4_board.subsystem.ethosu.num_macs=256" west build -d build-fvp -b mps4/corstone320/fvp \ modules/lib/executorch/zephyr/samples/mv2-ethosu -- \ -DET_PTE_FILE_PATH=mv2_u85_256.pte west build -d build-fvp -t run ``` The fresh sample build and FVP run passed, printing `MobileNetV2 Demo Complete` and top-five class IDs `92, 21, 127, 22, 18`, matching PyTorch for the same input. ELF inspection confirmed that code at `0x31000000` and writable data at `0x31080000` no longer overlap. Fresh CMake configuration also verified FVP discovery through `setup_path.sh`, the captured MAC-count/simulation-limit options, and Zephyr's boot-vector configuration. CMake configuration with the U55 model on `ensemble_e8_dk/ae822fa0e5597ls0/rtss_hp` verified the corrected operator diagnostic. All 29 documentation shell blocks passed `bash -n`; relative links, manifest/ATOC syntax, applicable lintrunner checks, and `git diff --check` passed. Alif flashing, inference, and hardware latency validation are pending.
1 parent 8426a13 commit 34608e3

5 files changed

Lines changed: 111 additions & 61 deletions

File tree

‎docs/source/zephyr_alif_tutorial.md‎

Lines changed: 71 additions & 37 deletions
Original file line numberDiff line numberDiff line change
@@ -7,8 +7,7 @@ works on the Arm Corstone-320 FVP for development without hardware.
77

88
## What You'll Build
99

10-
- A quantized INT8 MobileNetV2 model fully delegated to the Ethos-U55 NPU
11-
(110 ops, ~19 ms inference on Alif E8)
10+
- A quantized INT8 MobileNetV2 model with Ethos-U NPU delegation
1211
- A Zephyr RTOS application that loads the `.pte` model, runs inference on a
1312
static test image, and prints the top-5 ImageNet predictions over UART
1413

@@ -18,13 +17,12 @@ works on the Arm Corstone-320 FVP for development without hardware.
1817

1918
| Target | Description |
2019
|--------|-------------|
21-
| **Alif Ensemble E8 DevKit** | Cortex-M55 HP core + Ethos-U55 (256 MACs), 4.5 MB HP SRAM, MRAM |
20+
| **Alif Ensemble E8 DevKit** | Cortex-M55 HP core + Ethos-U55 (256 MACs), MRAM, 8 MiB SRAM |
2221
| **Corstone-320 FVP** | Virtual platform simulating Cortex-M85 + Ethos-U85 (no hardware needed, Linux only) |
2322

2423
### Software
2524

26-
- Linux x86_64 (FVP and Arm toolchain are Linux-only; macOS can export models
27-
but cannot run the FVP or flash)
25+
- Linux x86_64 for the workflow below
2826
- Python 3.12+
2927
- Alif SE Tools for flashing (Alif hardware only)
3028

@@ -41,8 +39,9 @@ west init --manifest-rev v4.4.0
4139

4240
## Step 2: Add ExecuTorch as a Zephyr Module
4341

44-
Copy the submanifest, configure `west` to pull only the modules we need, and
45-
update:
42+
Create the submanifest, configure `west` to pull the modules for this demo, and
43+
update. The project filter below is for the new workspace created in Step 1;
44+
when using an existing application, retain the modules that application needs.
4645

4746
```bash
4847
mkdir -p zephyr/submanifests
@@ -55,18 +54,13 @@ manifest:
5554
path: modules/lib/executorch
5655
EOF
5756

58-
west config manifest.project-filter -- -.*,+zephyr,+executorch,+cmsis,+cmsis_6,+cmsis-nn,+hal_ethos_u
57+
west config manifest.project-filter -- '-.*,+zephyr,+executorch,+cmsis,+cmsis_6,+cmsis-nn,+hal_ethos_u'
5958
west update
6059
west packages pip --install
6160
```
6261

63-
For Alif boards, also add the Alif HAL:
64-
65-
```bash
66-
west config manifest.project-filter -- -.*,+zephyr,+executorch,+cmsis,+cmsis_6,+cmsis-nn,+hal_ethos_u,+hal_alif
67-
west update
68-
west packages pip --install
69-
```
62+
Zephyr 4.4.0 includes the Alif SoC and peripheral support used by this sample
63+
in its own source tree.
7064

7165
Install the Zephyr SDK (compiler toolchain):
7266

@@ -98,7 +92,7 @@ that matches your hardware:
9892
**For Alif E8 (HP Ethos-U55 with 256 MACs):**
9993

10094
```bash
101-
python -m modules.lib.executorch.backends.arm.scripts.aot_arm_compiler \
95+
python -m executorch.backends.arm.scripts.aot_arm_compiler \
10296
--model_name=mv2 \
10397
--quantize --delegate \
10498
--target=ethos-u55-256 \
@@ -110,7 +104,7 @@ If rtss_he is used instead of rtss_hp below use `--target=ethos-u55-128` to matc
110104
**For Corstone-320 FVP (Ethos-U85 with 256 MACs):**
111105

112106
```bash
113-
python -m modules.lib.executorch.backends.arm.scripts.aot_arm_compiler \
107+
python -m executorch.backends.arm.scripts.aot_arm_compiler \
114108
--model_name=mv2 \
115109
--quantize --delegate \
116110
--target=ethos-u85-256 \
@@ -122,39 +116,66 @@ The Vela compiler converts the TOSA intermediate representation into an
122116
optimized command stream for the NPU. Use `mv2_untrained` instead of `mv2` if
123117
you do not want to depend on torchvision pretrained weights.
124118

119+
These export commands retain floating-point model inputs and outputs. The
120+
Cortex-M runs the runtime, input/output processing, and boundary quantization
121+
and dequantization.
122+
An NPU operator count from Vela describes the compiled partition; it does not
123+
count all operations in the ExecuTorch program.
124+
125+
By default, this command calibrates using the model's example input, which is
126+
random for `mv2`. For classification accuracy evaluation, supply representative
127+
calibration data and compare against PyTorch with matching input preprocessing.
128+
125129
## Step 5: Build the Zephyr Application
126130

131+
The separate build directories keep the Alif and FVP board configurations and
132+
firmware images independent. The `-d` option is optional when building only one
133+
target; West otherwise uses `build`.
134+
127135
**For Alif E8:**
128136

129137
```bash
130-
west build -d build -b ensemble_e8_dk/ae822fa0e5597ls0/rtss_hp \
138+
west build -d build-alif -b ensemble_e8_dk/ae822fa0e5597ls0/rtss_hp \
131139
modules/lib/executorch/zephyr/samples/mv2-ethosu -- \
132140
-DET_PTE_FILE_PATH=mv2_ethosu.pte
133141
```
134142

135143
**For Corstone-320 FVP:**
136144

145+
Use the environment from Step 3: `setup_path.sh` supplies the FVP executable
146+
and shared-library paths. Zephyr finds `FVP_Corstone_SSE-320` on `PATH` and
147+
uses the board's simulator configuration. Set the sample's NPU MAC count and
148+
simulation limit before configuring the build; Zephyr captures
149+
`ARMFVP_EXTRA_FLAGS` when CMake runs.
150+
137151
```bash
138-
west build -b mps4/corstone320/fvp \
152+
export ARMFVP_EXTRA_FLAGS="--simlimit 10 -C mps4_board.subsystem.ethosu.num_macs=256"
153+
154+
west build -d build-fvp -b mps4/corstone320/fvp \
139155
modules/lib/executorch/zephyr/samples/mv2-ethosu -- \
140156
-DET_PTE_FILE_PATH=mv2_u85_256.pte
141157
```
142158

143159
## Step 6a: Run on Corstone-320 FVP
144160

145-
Set up the FVP paths and run:
161+
Run the image built in Step 5:
146162

147163
```bash
148-
export FVP_ROOT=$PWD/modules/lib/executorch/examples/arm/arm-scratch/FVP-corstone320
149-
export ARMFVP_BIN_PATH=${FVP_ROOT}/models/Linux64_GCC-9.3
150-
export LD_LIBRARY_PATH=${FVP_ROOT}/python/lib:${ARMFVP_BIN_PATH}:${LD_LIBRARY_PATH}
151-
export ARMFVP_EXTRA_FLAGS="-C mps4_board.uart0.shutdown_on_eot=1 -C mps4_board.subsystem.ethosu.num_macs=256"
152-
153-
west build -t run
164+
west build -d build-fvp -t run
154165
```
155166

156-
MV2 inference is cycle-accurate on the FVP and takes 10-20 minutes of wall
157-
clock. You should see output like:
167+
The Ethos-U model can report NPU cycle counts, while CPU instruction timing
168+
and memory-system behavior are simplified. The `<N> ms` line below measures
169+
`method->execute()` with Zephyr's simulated clock, rather than reading the
170+
NPU cycle counter. It does not establish Alif E8 hardware latency. Host wall
171+
time depends on the host and simulator configuration.
172+
173+
The sample leaves Zephyr running after printing `MobileNetV2 Demo Complete`.
174+
`--simlimit 10` stops the FVP after ten seconds of simulated time. Confirm that
175+
the completion message appears before it exits. If you change the simulation
176+
limit or other FVP flags, rerun `west build -d build-fvp -c` before running again.
177+
178+
You should see output like:
158179

159180
```
160181
========================================
@@ -176,7 +197,7 @@ MobileNetV2 Demo Complete
176197
### Flash with west
177198

178199
```bash
179-
west flash
200+
west flash -d build-alif
180201
```
181202

182203

@@ -192,13 +213,13 @@ SE Tools directory.
192213

193214
The Alif E8 sample build generates the binary consumed by SE Tools:
194215

195-
- `build/zephyr/zephyr.bin`
216+
- `build-alif/zephyr/zephyr.bin`
196217

197218
Create `build/images/zephyr.json` and `build/images/app-device-config.json` in
198219
the SE Tools directory:
199220

200221
```bash
201-
ZEPHYR_BUILD=~/zephyr_workspace/build/zephyr
222+
ZEPHYR_BUILD=~/zephyr_workspace/build-alif/zephyr
202223

203224
cd <PATH_TO_ALIF_SETOOLS>/app-release-exec-linux_FW_1.109.00_DEV
204225

@@ -347,7 +368,9 @@ open the serial terminal, and press reset:
347368
picocom -b 115200 /dev/ttyACM0
348369
```
349370

350-
Press the reset button on the E8 DevKit. You should see:
371+
Press the reset button on the E8 DevKit. A representative log is shown below;
372+
timing, model size, and scores vary with the model and software/hardware
373+
configuration.
351374

352375
```
353376
*** Booting Zephyr OS build v4.4.0 ***
@@ -392,13 +415,19 @@ Use `mv2_untrained` instead of `mv2` if you want to avoid downloading
392415
pretrained weights. In that case the class scores are arbitrary and may all be
393416
close to `0.0000`.
394417

418+
The reported inference time covers one `method->execute()` call, measured with
419+
`k_uptime_get_32()`. Model loading, input preparation, and result reporting are
420+
outside that interval. For hardware comparisons, record the software versions,
421+
model, core/NPU configuration and clocks, and measurements over repeated runs.
422+
395423
## Troubleshooting
396424

397425
| Symptom | Cause | Fix |
398426
|---------|-------|-----|
399-
| Linker: `region 'FLASH' overflowed` | Model PTE too large for ITCM | Use the DDR overlay (FVP) or verify mramAddress (Alif) |
427+
| Linker: `region 'FLASH' overflowed` | Code and model exceed the linked image region | Check the linker map and model placement: DDR on FVP, MRAM on Alif. Changing the ATOC address does not increase linker capacity. |
400428
| Linker: `region 'RAM' overflowed` | Pools + model copy exceed SRAM | Set `CONFIG_ET_ARM_MODEL_PTE_DMA_ACCESSIBLE=y` to skip the SRAM copy |
401-
| FVP hangs after "Ethos-U backend registered" | Cycle-accurate MV2 simulation is slow | Wait 10-20 min, or use Corstone-320 (faster than 300) |
429+
| No Zephyr startup banner on FVP | Incorrect boot or memory layout | Check the ELF addresses and use the Corstone-320 sample overlay with separate regions for code and writable data. |
430+
| FVP exits before the demo completes | Simulation limit too short, configuration mismatch, or runtime failure | Inspect the console, check the exported NPU target and MAC count, and increase `--simlimit` if needed. Reconfigure after changing FVP flags. |
402431
| No Zephyr serial output on Alif | `SW4` still routes VCOM to SE UART | Move `SW4` to `U4` and use 115200 baud |
403432
| No SE Tools response | Serial terminal is open, wrong `SW4` setting, or app blocks ISP | Close terminal, set `SW4=SE`, or enter hard maintenance mode |
404433
| `app-write-mram` fails or app does not boot | `mramAddress` and link address differ | Use `0x80008000`, not `0x80200000` with `CONFIG_FLASH_LOAD_OFFSET=0x8000` |
@@ -408,11 +437,16 @@ close to `0.0000`.
408437

409438
| Region | Corstone-320 FVP | Alif E8 |
410439
|--------|-----------------|---------|
411-
| Code + .rodata | ITCM (512 KB) | MRAM |
412-
| .data + .bss + pools | ISRAM (4 MB) | HP SRAM (4.5 MB) |
413-
| Model PTE (~3.5 MB) | DDR (16 MB, via overlay) | MRAM (DMA-accessible) |
440+
| Code + .rodata | First 512 KiB of ISRAM at `0x31000000`; boot vectors in ITCM | MRAM |
441+
| .data + .bss + pools | Remaining 3.5 MiB of ISRAM at `0x31080000` | Combined SRAM0/SRAM1, 8 MiB at `0x02000000` |
442+
| Model PTE | DDR region at `0x70000000` (16 MiB, via overlay) | MRAM (DMA-accessible) |
414443
| NPU delegation | Ethos-U85 (256 MACs) | Ethos-U55 (256 MACs) |
415444

445+
These are memory-region capacities. The method and temporary allocator pools
446+
reserve 1.5 MiB each; the linker's RAM usage also includes other data and stacks.
447+
Use the build's memory summary and linker map to inspect reserved space. Neither
448+
region capacity nor statically reserved RAM measures peak live allocation.
449+
416450
## Using Claude Code with Zephyr
417451

418452
If you use [Claude Code](https://docs.anthropic.com/en/docs/claude-code), the

‎zephyr/samples/mv2-ethosu/CMakeLists.txt‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -92,7 +92,7 @@ else()
9292
set(EXECUTORCH_SELECT_OPS_MODEL "")
9393
set(_EXECUTORCH_GEN_ZEPHYR_PORTABLE_OPS OFF)
9494
message(
95-
"gen_oplist: No non-delegated ops were found in ${ET_PTE_FILE_PATH}; no ops added to build"
95+
"gen_oplist: No aten:: or dim_order_ops:: operators were found in ${ET_PTE_FILE_PATH}; skipping portable operator selection. Other CPU operators may still run."
9696
)
9797
endif()
9898

‎zephyr/samples/mv2-ethosu/README.md‎

Lines changed: 24 additions & 17 deletions
Original file line numberDiff line numberDiff line change
@@ -20,15 +20,14 @@ The model classifies a static RGB test input tensor with shape `[1, 3, 224, 224]
2020
2121
### Setup FVP paths
2222

23-
Set up FVP paths and macs used, this will also set `shutdown_on_eot` so the FVP auto stops after it has run the example.
23+
Source the environment generated by `examples/arm/setup.sh` and set the MAC
24+
count before configuring the build.
2425

2526
Config Zephyr Corstone300 FVP
2627
<!-- RUN setup_corstone300 -->
2728
```
28-
export FVP_ROOT=$PWD/modules/lib/executorch/examples/arm/arm-scratch/FVP-corstone300
29-
export ARMFVP_BIN_PATH=${FVP_ROOT}/models/Linux64_GCC-9.3
30-
export LD_LIBRARY_PATH=${FVP_ROOT}/python/lib:${ARMFVP_BIN_PATH}:${LD_LIBRARY_PATH:-}
31-
export ARMFVP_EXTRA_FLAGS="-C mps3_board.uart0.shutdown_on_eot=1 -C ethosu.num_macs=128"
29+
source modules/lib/executorch/examples/arm/arm-scratch/setup_path.sh
30+
export ARMFVP_EXTRA_FLAGS="-C ethosu.num_macs=128"
3231
```
3332

3433
### Export the model
@@ -37,7 +36,7 @@ Export a quantized INT8 MobileNetV2 model with Ethos-U55 delegation:
3736

3837
<!-- RUN test_ethos-u55_generate_pte -->
3938
```
40-
python -m modules.lib.executorch.backends.arm.scripts.aot_arm_compiler --model_name=mv2_untrained --quantize --delegate --target=ethos-u55-128 --output=mv2_u55_128.pte
39+
python -m executorch.backends.arm.scripts.aot_arm_compiler --model_name=mv2_untrained --quantize --delegate --target=ethos-u55-128 --output=mv2_u55_128.pte
4140
```
4241

4342
### Build (link-check only)
@@ -51,20 +50,20 @@ west build -b mps3/corstone300/fvp modules/lib/executorch/zephyr/samples/mv2-eth
5150

5251
### Setup FVP paths
5352

54-
Set up FVP paths, libs and macs used, this will also set `shutdown_on_eot` so the FVP auto stops after it has run the example.
53+
Source the environment generated by `examples/arm/setup.sh` and set the MAC
54+
count before configuring the build.
5555

5656
These FVP command-line options are passed through the `ARMFVP_EXTRA_FLAGS`
5757
environment variable. The sample does not set `ARMFVP_FLAGS` in its
5858
`CMakeLists.txt`; the base `ARMFVP_FLAGS` come from the selected Zephyr board's
59-
`board.cmake`.
59+
`board.cmake`. Zephyr captures these options when CMake runs; after changing
60+
them, use `west build -c` to reconfigure.
6061

6162
Config Zephyr Corstone320 FVP
6263
<!-- RUN setup_corstone320 -->
6364
```
64-
export FVP_ROOT=$PWD/modules/lib/executorch/examples/arm/arm-scratch/FVP-corstone320
65-
export ARMFVP_BIN_PATH=${FVP_ROOT}/models/Linux64_GCC-9.3
66-
export LD_LIBRARY_PATH=${FVP_ROOT}/python/lib:${ARMFVP_BIN_PATH}:${LD_LIBRARY_PATH:-}
67-
export ARMFVP_EXTRA_FLAGS="-C mps4_board.uart0.shutdown_on_eot=1 -C mps4_board.subsystem.ethosu.num_macs=256"
65+
source modules/lib/executorch/examples/arm/arm-scratch/setup_path.sh
66+
export ARMFVP_EXTRA_FLAGS="--simlimit 10 -C mps4_board.subsystem.ethosu.num_macs=256"
6867
```
6968

7069
### Export the model
@@ -73,7 +72,7 @@ Export a quantized INT8 MobileNetV2 model with Ethos-U85 delegation:
7372

7473
<!-- RUN test_ethos-u85_generate_pte -->
7574
```
76-
python -m modules.lib.executorch.backends.arm.scripts.aot_arm_compiler --model_name=mv2_untrained --quantize --delegate --target=ethos-u85-256 --output=mv2_u85_256.pte
75+
python -m executorch.backends.arm.scripts.aot_arm_compiler --model_name=mv2_untrained --quantize --delegate --target=ethos-u85-256 --output=mv2_u85_256.pte
7776
```
7877

7978
### Build
@@ -89,12 +88,19 @@ west build -b mps4/corstone320/fvp modules/lib/executorch/zephyr/samples/mv2-eth
8988
west build -t run
9089
```
9190

91+
The sample leaves Zephyr running after `MobileNetV2 Demo Complete`.
92+
`--simlimit 10` stops the FVP after ten seconds of simulated time; confirm that
93+
the completion message appears before it exits. The Ethos-U model can report
94+
NPU cycle counts, while CPU and memory-system timing are simplified. This
95+
sample reports elapsed Zephyr time rather than NPU cycles; use hardware to
96+
measure deployment latency.
97+
9298
## Alif Ensemble E8 DevKit
9399

94100
For boards with Ethos-U55-256 (e.g., Alif E8 HP core), use `--target=ethos-u55-256`:
95101

96102
```bash
97-
python -m modules.lib.executorch.backends.arm.scripts.aot_arm_compiler \
103+
python -m executorch.backends.arm.scripts.aot_arm_compiler \
98104
--model_name=mv2_untrained \
99105
--quantize \
100106
--delegate \
@@ -152,9 +158,10 @@ weights) for meaningful predictions.
152158

153159
## Memory requirements
154160

155-
The default configuration allocates 1.5 MB each for the method and temporary
156-
allocator pools. These defaults are sufficient for a fully NPU-delegated INT8
157-
MobileNetV2. Adjust `CONFIG_EXECUTORCH_METHOD_ALLOCATOR_POOL_SIZE` and
161+
The default configuration reserves 1.5 MiB each for the method and temporary
162+
allocator pools. These defaults support this INT8 MobileNetV2 example with
163+
Ethos-U delegation and CPU quantization/dequantization at the model boundary.
164+
Adjust `CONFIG_EXECUTORCH_METHOD_ALLOCATOR_POOL_SIZE` and
158165
`CONFIG_EXECUTORCH_TEMP_ALLOCATOR_POOL_SIZE` in `prj.conf`, in board-specific
159166
`*.conf` files (for example, `boards/<board>.conf`), or via Zephyr's
160167
`OVERLAY_CONFIG` mechanism for different model configurations.

‎zephyr/samples/mv2-ethosu/boards/mps4_corstone320_fvp.conf‎

Lines changed: 1 addition & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -12,7 +12,6 @@ CONFIG_ETHOS_U=y
1212
# DDR on the FVP so no SRAM copy is needed.
1313
CONFIG_ET_ARM_MODEL_PTE_DMA_ACCESSIBLE=y
1414

15-
# With zephyr,sram redirected to the 4 MiB ISRAM (shared with code),
16-
# the default 1.5 MiB pools fit alongside ~460 KiB of .text and overhead.
15+
# The overlay reserves 3.5 MiB of ISRAM for writable data, including these pools.
1716
CONFIG_EXECUTORCH_METHOD_ALLOCATOR_POOL_SIZE=1572864
1817
CONFIG_EXECUTORCH_TEMP_ALLOCATOR_POOL_SIZE=1572864

‎zephyr/samples/mv2-ethosu/boards/mps4_corstone320_fvp.overlay‎

Lines changed: 14 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -5,16 +5,26 @@
55
* SPDX-License-Identifier: Apache-2.0
66
*/
77

8-
/* Override zephyr,sram to the 4 MiB ISRAM (0x3100_0000) so the allocator
9-
* pools (~3 MiB total) fit alongside code in the same region. The default
10-
* sram@12000000 is only 2 MiB which is too small for MV2.
8+
/* The board uses ISRAM for XIP code. Reserve its first 512 KiB for code
9+
* and the remaining 3.5 MiB for writable data and allocator pools. Using
10+
* the same region for zephyr,flash and zephyr,sram overlaps code and data.
1111
*
1212
* The model PTE blob is placed in DDR (0x7000_0000) via a linker snippet.
1313
* The Ethos-U can DMA from DDR on the Corstone-320 FVP.
1414
*/
15+
&isram {
16+
reg = <0x31000000 DT_SIZE_K(512)>;
17+
};
18+
1519
/ {
1620
chosen {
17-
zephyr,sram = &isram;
21+
zephyr,sram = &executorch_sram;
22+
};
23+
24+
executorch_sram: memory@31080000 {
25+
compatible = "zephyr,memory-region", "mmio-sram";
26+
reg = <0x31080000 DT_SIZE_K(3584)>;
27+
zephyr,memory-region = "EXECUTORCH_SRAM";
1828
};
1929

2030
model_ddr: memory@70000000 {

0 commit comments

Comments
 (0)