ncnn-20230517-android-shared 32位Android APK 上 执行矩阵乘法forward时崩溃

[garbe-github-support]

复现条件
开源库这个位置的代码
https://github.com/Digital2Slave/ncnn-android-yolov8-seg/blob/7f76b69b936b1edd34456dc11279e51ecdb28084/app/src/main/jni/yolo.cpp#L162C39-L162C39

把这个代码复制到自己的项目中（没有测试这个项目的代码会不会崩溃）
然后，用
ncnn-20230517-android-shared 包， 打包32位的APK
荣耀honor80 pro VIVO IQoo 测试，运行到这个位置APP就会直接崩溃
使用同版本的 非shared的不会出错，使用更低版本的shared 非shared 都也不会出错，打包成64位也不会出错

输入矩阵尺寸信息
a 尺寸 = w 32 h 2 c 1 cstep = 64, elemSize= 4 elemPack = 1
b 尺寸 = w 15360 h 32 c 1 cstep = 491520, elemSize= 4 elemPack = 1

报错信息
2023-06-28 20:34:23.462 12454-14030/xxx E/DEBUG: #00 pc 0004c9cc /data/app/~~mHLNNM_lu_r4qz_zkueSpA==/xxx -zciNAeg6Ht3OMNprAXiqtQ==/lib/arm/libcrashsdk.so
2023-06-28 20:34:23.462 12454-14030/xxx u E/DEBUG: #1 pc 0004ac5c /data/app/~~mHLNNM_lu_r4qz_zkueSpA==/xxx -zciNAeg6Ht3OMNprAXiqtQ==/lib/arm/libcrashsdk.so
2023-06-28 20:34:23.462 12454-14030/xxx E/DEBUG: #2 pc 00007bac /data/app/~~mHLNNM_lu_r4qz_zkueSpA==/xxx -zciNAeg6Ht3OMNprAXiqtQ==/lib/arm/libcrashsdk.so
2023-06-28 20:34:23.462 12454-14030/xxx E/DEBUG: #3 pc 0002d23c /data/app/~~mHLNNM_lu_r4qz_zkueSpA==/xxx -zciNAeg6Ht3OMNprAXiqtQ==/lib/arm/libcrashsdk.so
2023-06-28 20:34:23.462 12454-14030/xxx E/DEBUG: #4 pc 0005c083 /apex/com.android.runtime/lib/bionic/libc.so (__cxa_finalize+190)
2023-06-28 20:34:23.462 12454-14030/xxx E/DEBUG: #5 pc 0005147b /apex/com.android.runtime/lib/bionic/libc.so (exit+10)
2023-06-28 20:34:23.462 12454-14030/xxx E/DEBUG: #6 pc 002d4bf0 /data/app/~~mHLNNM_lu_r4qz_zkueSpA==/xxxxx-zciNAeg6Ht3OMNprAXiqtQ==/lib/arm/libncnn.so (__kmp_invoke_microtask+1064)

[nihui]

应该是这个commit修复了，可以测试下 7883f4d

[garbe-github-support]

好的 ，谢谢啦

[glenn-jocher]

@garbe-github-support I have great news 😃! I've recently added official support for Ultralytics YOLOv8 NCNN export ✅ in PR ultralytics/ultralytics#3529 with the help of @nihui which is part of ultralytics==8.0.129. NCNN works for all tasks including Detect, Segment, Pose and Classify.

You can now export with CLI:

yolo export model=yolov8n.pt format=ncnn

or Python:

from ultralytics import YOLO

# Create a model
model = YOLO('yolov8n.pt')

# Export the model to NCNN with arguments
model.export(format='ncnn', half=True, imgsz=640)

Output is a yolov8n_ncnn_model/ directory containing model.bin, model.param and metadata.yaml, along with extra PNNX files. For details see https://github.com/pnnx/pnnx README.

To get this update:

• Git – Run git pull from within your ultralytics/ directory or run git clone https://github.com/ultralytics/ultralytics again
• Pip – Update with pip install -U ultralytics
• Notebooks – Check out the updated notebooks
• Docker – Run sudo docker pull ultralytics/ultralytics:latest to update your image

Please let us know if NCNN export is working correctly for you, and don't hesitate to report any other issues you find or feature requests you may have. Happy training with YOLOv8 🚀!

[garbe-github-support]

Thank you, it's a great job！