enable arm neon intrinsics for msvc build

[nihui]

• compiler check
• guard all gnu inline assembly
• enable windows arm ci
• implement runtime cpu feature
• neon intrinsics coverage

[codecov-commenter]

Codecov Report

Attention: 7 lines in your changes are missing coverage. Please review.

Comparison is base (465debe) 94.72% compared to head (f505902) 94.41%.
Report is 3 commits behind head on master.

Files	Patch %	Lines
src/mat.cpp	0.00%	6 Missing ⚠️
src/cpu.cpp	50.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #5151      +/-   ##
==========================================
- Coverage   94.72%   94.41%   -0.31%     
==========================================
  Files         774      774              
  Lines      242342   243079     +737     
==========================================
- Hits       229551   229504      -47     
- Misses      12791    13575     +784     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[nihui]

pi@raspberrypi:~/ncnn/build/benchmark $ ./benchncnn 10 4 0 -1 1
loop_count = 10
num_threads = 4
powersave = 0
gpu_device = -1
cooling_down = 1
          squeezenet  min =    8.49  max =    8.78  avg =    8.69
     squeezenet_int8  min =    8.61  max =    9.05  avg =    8.81
           mobilenet  min =   10.99  max =   11.14  avg =   11.06
      mobilenet_int8  min =   11.22  max =   11.48  avg =   11.31
        mobilenet_v2  min =   13.40  max =   13.75  avg =   13.61
        mobilenet_v3  min =    9.05  max =   12.14  avg =    9.50
          shufflenet  min =    4.43  max =    4.46  avg =    4.44
       shufflenet_v2  min =    3.63  max =    3.71  avg =    3.66
             mnasnet  min =    8.02  max =    8.14  avg =    8.08
     proxylessnasnet  min =    9.43  max =    9.68  avg =    9.52
     efficientnet_b0  min =   14.89  max =   15.14  avg =   15.03
   efficientnetv2_b0  min =   16.56  max =   20.13  avg =   17.12
        regnety_400m  min =   12.22  max =   12.38  avg =   12.30
           blazeface  min =    1.61  max =    1.66  avg =    1.64
           googlenet  min =   31.02  max =   31.48  avg =   31.21
      googlenet_int8  min =   29.68  max =   30.22  avg =   29.96
            resnet18  min =   23.85  max =   24.17  avg =   24.03
       resnet18_int8  min =   21.39  max =   21.60  avg =   21.47
             alexnet  min =   24.61  max =   24.84  avg =   24.74
               vgg16  min =  166.10  max =  168.17  avg =  166.83
          vgg16_int8  min =  129.48  max =  131.56  avg =  130.40
            resnet50  min =   55.17  max =   55.79  avg =   55.60
       resnet50_int8  min =   50.97  max =   53.51  avg =   51.45
      squeezenet_ssd  min =   39.42  max =   39.99  avg =   39.72
 squeezenet_ssd_int8  min =   36.80  max =   38.57  avg =   37.31
       mobilenet_ssd  min =   30.53  max =   30.94  avg =   30.74
  mobilenet_ssd_int8  min =   28.72  max =   29.63  avg =   29.25
      mobilenet_yolo  min =   71.02  max =   71.81  avg =   71.34
  mobilenetv2_yolov3  min =   48.18  max =   48.69  avg =   48.36
         yolov4-tiny  min =   55.63  max =   56.20  avg =   55.85
           nanodet_m  min =   13.68  max =   14.00  avg =   13.81
    yolo-fastest-1.1  min =    6.44  max =    6.57  avg =    6.50
      yolo-fastestv2  min =    5.73  max =    5.82  avg =    5.76
  vision_transformer  min =  616.99  max =  640.29  avg =  627.76
          FastestDet  min =    5.60  max =    5.80  avg =    5.70
pi@raspberrypi:~/ncnn/build/benchmark $ ./benchncnn 10 1 0 -1 1
loop_count = 10
num_threads = 1
powersave = 0
gpu_device = -1
cooling_down = 1
          squeezenet  min =   13.22  max =   13.38  avg =   13.31
     squeezenet_int8  min =   12.89  max =   13.17  avg =   13.05
           mobilenet  min =   21.46  max =   21.60  avg =   21.51
      mobilenet_int8  min =   15.41  max =   15.72  avg =   15.58
        mobilenet_v2  min =   18.76  max =   19.56  avg =   19.01
        mobilenet_v3  min =   12.88  max =   13.25  avg =   13.07
          shufflenet  min =    7.26  max =    7.30  avg =    7.28
       shufflenet_v2  min =    7.22  max =    7.32  avg =    7.26
             mnasnet  min =   13.42  max =   13.71  avg =   13.57
     proxylessnasnet  min =   16.61  max =   16.69  avg =   16.66
     efficientnet_b0  min =   25.88  max =   26.04  avg =   25.95
   efficientnetv2_b0  min =   28.97  max =   29.17  avg =   29.06
        regnety_400m  min =   17.47  max =   17.54  avg =   17.51
           blazeface  min =    3.08  max =    3.13  avg =    3.11
           googlenet  min =   51.52  max =   52.01  avg =   51.71
      googlenet_int8  min =   52.89  max =   53.08  avg =   52.98
            resnet18  min =   31.67  max =   31.99  avg =   31.83
       resnet18_int8  min =   39.00  max =   39.21  avg =   39.05
             alexnet  min =   36.66  max =   37.07  avg =   36.88
               vgg16  min =  205.90  max =  209.63  avg =  207.62
          vgg16_int8  min =  287.79  max =  289.39  avg =  288.78
            resnet50  min =   95.31  max =   96.43  avg =   95.96
       resnet50_int8  min =   85.52  max =   85.89  avg =   85.76
      squeezenet_ssd  min =   43.67  max =   44.13  avg =   43.88
 squeezenet_ssd_int8  min =   47.20  max =   48.27  avg =   47.77
       mobilenet_ssd  min =   50.50  max =   51.04  avg =   50.81
  mobilenet_ssd_int8  min =   40.89  max =   42.05  avg =   41.54
      mobilenet_yolo  min =  114.10  max =  116.32  avg =  115.16
  mobilenetv2_yolov3  min =   68.92  max =   69.05  avg =   68.98
         yolov4-tiny  min =   73.24  max =   73.61  avg =   73.45
           nanodet_m  min =   21.32  max =   21.59  avg =   21.45
    yolo-fastest-1.1  min =    8.72  max =    8.84  avg =    8.78
      yolo-fastestv2  min =    7.89  max =    8.00  avg =    7.92
  vision_transformer  min = 1267.07  max = 1268.49  avg = 1267.67
          FastestDet  min =    7.88  max =    8.06  avg =    7.97

[nihui]

after apt upgrade and reboot

pi@raspberrypi:~/ncnn/build/benchmark $ ./benchncnn 10 4 0 -1 0
loop_count = 10
num_threads = 4
powersave = 0
gpu_device = -1
cooling_down = 0
          squeezenet  min =    8.31  max =    8.54  avg =    8.45
     squeezenet_int8  min =    8.46  max =    8.93  avg =    8.70
           mobilenet  min =   10.98  max =   11.19  avg =   11.07
      mobilenet_int8  min =   10.98  max =   11.26  avg =   11.14
        mobilenet_v2  min =   13.50  max =   13.80  avg =   13.67
        mobilenet_v3  min =    8.72  max =    8.92  avg =    8.85
          shufflenet  min =    4.43  max =    4.52  avg =    4.47
       shufflenet_v2  min =    3.58  max =    3.71  avg =    3.63
             mnasnet  min =    7.98  max =    8.16  avg =    8.05
     proxylessnasnet  min =    9.43  max =    9.52  avg =    9.48
     efficientnet_b0  min =   14.78  max =   15.04  avg =   14.93
   efficientnetv2_b0  min =   16.38  max =   19.24  avg =   16.90
        regnety_400m  min =   11.72  max =   11.86  avg =   11.80
           blazeface  min =    1.60  max =    1.64  avg =    1.62
           googlenet  min =   30.47  max =   31.14  avg =   30.84
      googlenet_int8  min =   29.60  max =   30.04  avg =   29.82
            resnet18  min =   23.24  max =   26.26  avg =   23.78
       resnet18_int8  min =   21.06  max =   21.58  avg =   21.31
             alexnet  min =   24.64  max =   28.61  avg =   25.22
               vgg16  min =  164.78  max =  168.18  avg =  166.34
          vgg16_int8  min =  129.48  max =  130.91  avg =  130.19
            resnet50  min =   54.83  max =   55.58  avg =   55.31
       resnet50_int8  min =   50.47  max =   50.78  avg =   50.64
      squeezenet_ssd  min =   39.51  max =   40.68  avg =   39.78
 squeezenet_ssd_int8  min =   36.46  max =   38.04  avg =   37.17
       mobilenet_ssd  min =   30.16  max =   30.82  avg =   30.49
  mobilenet_ssd_int8  min =   28.51  max =   29.46  avg =   29.08
      mobilenet_yolo  min =   70.41  max =   71.12  avg =   70.79
  mobilenetv2_yolov3  min =   47.81  max =   48.28  avg =   48.01
         yolov4-tiny  min =   55.16  max =   55.58  avg =   55.39
           nanodet_m  min =   13.31  max =   13.75  avg =   13.57
    yolo-fastest-1.1  min =    6.32  max =    6.45  avg =    6.38
      yolo-fastestv2  min =    5.94  max =    6.05  avg =    6.01
  vision_transformer  min =  584.05  max =  610.38  avg =  596.09
          FastestDet  min =    5.79  max =    5.95  avg =    5.89
pi@raspberrypi:~/ncnn/build/benchmark $ ./benchncnn 10 1 0 -1 0
loop_count = 10
num_threads = 1
powersave = 0
gpu_device = -1
cooling_down = 0
          squeezenet  min =   12.79  max =   13.10  avg =   12.91
     squeezenet_int8  min =   12.71  max =   13.06  avg =   12.91
           mobilenet  min =   21.39  max =   21.50  avg =   21.43
      mobilenet_int8  min =   15.40  max =   15.68  avg =   15.52
        mobilenet_v2  min =   18.66  max =   18.96  avg =   18.80
        mobilenet_v3  min =   12.82  max =   13.24  avg =   13.02
          shufflenet  min =    7.27  max =    7.33  avg =    7.30
       shufflenet_v2  min =    7.20  max =    7.27  avg =    7.24
             mnasnet  min =   13.42  max =   13.67  avg =   13.54
     proxylessnasnet  min =   16.52  max =   16.61  avg =   16.57
     efficientnet_b0  min =   25.25  max =   25.74  avg =   25.56
   efficientnetv2_b0  min =   28.75  max =   28.90  avg =   28.82
        regnety_400m  min =   17.38  max =   17.52  avg =   17.46
           blazeface  min =    3.09  max =    3.12  avg =    3.10
           googlenet  min =   50.80  max =   51.47  avg =   51.15
      googlenet_int8  min =   52.25  max =   52.74  avg =   52.52
            resnet18  min =   31.69  max =   32.07  avg =   31.89
       resnet18_int8  min =   38.26  max =   38.60  avg =   38.44
             alexnet  min =   36.83  max =   37.42  avg =   37.07
               vgg16  min =  206.50  max =  209.37  avg =  207.51
          vgg16_int8  min =  286.72  max =  287.99  avg =  287.15
            resnet50  min =   95.03  max =   96.05  avg =   95.60
       resnet50_int8  min =   84.42  max =   86.22  avg =   85.23
      squeezenet_ssd  min =   43.46  max =   43.96  avg =   43.65
 squeezenet_ssd_int8  min =   47.01  max =   47.90  avg =   47.49
       mobilenet_ssd  min =   50.54  max =   51.04  avg =   50.75
  mobilenet_ssd_int8  min =   41.10  max =   42.67  avg =   41.86
      mobilenet_yolo  min =  112.72  max =  114.50  avg =  113.65
  mobilenetv2_yolov3  min =   68.36  max =   68.94  avg =   68.56
         yolov4-tiny  min =   72.89  max =   73.27  avg =   73.07
           nanodet_m  min =   21.08  max =   21.39  avg =   21.25
    yolo-fastest-1.1  min =    8.64  max =    8.79  avg =    8.68
      yolo-fastestv2  min =    7.78  max =    7.90  avg =    7.82
  vision_transformer  min = 1263.35  max = 1265.39  avg = 1264.57
          FastestDet  min =    7.85  max =    7.96  avg =    7.91