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.

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

[hexagon][testing] add max_pool2d benchmark #11720

Merged
merged 1 commit into from
Jun 17, 2022

Conversation

cconvey
Copy link
Contributor

@cconvey cconvey commented Jun 14, 2022

  • Add benchmarking framework for Hexagon maxpool-2d kernels,
    and one (simple) kernel.

  • Minor refactor hexagon benchmark utilities.

cc @mehrdadh

@cconvey
Copy link
Contributor Author

cconvey commented Jun 14, 2022

Example of the current output:

$ cat /tmp/tmphut9v83z/benchmark-results.csv | column -s $'\t' -t -n | less -SN
      1 basic_kernel  sched_type  N  H    W    C   DTYPE  KERNEL  STRIDE  DILATION  PADDING       IO_TENSOR_MEM_SCOPE  row_status  timings_min_usecs  timings_max_usecs  timings_median_usecs  timings_mean_usecs  timings_st
      2 max_pool2d    1           1  8    8    32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     197.200            197.200            197.200               197.200             0.000
      3 max_pool2d    1           1  8    8    64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     531.300            531.300            531.300               531.300             0.000
      4 max_pool2d    1           1  8    16   32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     459.500            459.500            459.500               459.500             0.000
      5 max_pool2d    1           1  8    16   64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     1641.800           1641.800           1641.800              1641.800            0.000
      6 max_pool2d    1           1  8    32   32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     979.700            979.700            979.700               979.700             0.000
      7 max_pool2d    1           1  8    32   64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     3513.800           3513.800           3513.800              3513.800            0.000
      8 max_pool2d    1           1  8    64   32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     2019.900           2019.900           2019.900              2019.900            0.000
      9 max_pool2d    1           1  8    64   64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     7040.600           7040.600           7040.600              7040.600            0.000
     10 max_pool2d    1           1  8    128  32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     4100.400           4100.400           4100.400              4100.400            0.000
     11 max_pool2d    1           1  8    128  64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     13208.500          13208.500          13208.500             13208.500           0.000
     12 max_pool2d    1           1  16   8    32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     455.400            455.400            455.400               455.400             0.000
     13 max_pool2d    1           1  16   8    64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     1482.100           1482.100           1482.100              1482.100            0.000
     14 max_pool2d    1           1  16   16   32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     1537.500           1537.500           1537.500              1537.500            0.000
     15 max_pool2d    1           1  16   16   64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     3894.300           3894.300           3894.300              3894.300            0.000
     16 max_pool2d    1           1  16   32   32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     3286.100           3286.100           3286.100              3286.100            0.000
     17 max_pool2d    1           1  16   32   64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     7447.000           7447.000           7447.000              7447.000            0.000
     18 max_pool2d    1           1  16   64   32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     6665.300           6665.300           6665.300              6665.300            0.000
     19 max_pool2d    1           1  16   64   64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     15824.300          15824.300          15824.300             15824.300           0.000
     20 max_pool2d    1           1  16   128  32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     12444.100          12444.100          12444.100             12444.100           0.000
     21 max_pool2d    1           1  16   128  64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     32704.500          32704.500          32704.500             32704.500           0.000
     22 max_pool2d    1           1  32   8    32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     971.800            971.800            971.800               971.800             0.000
     23 max_pool2d    1           1  32   8    64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     3171.200           3171.200           3171.200              3171.200            0.000
     24 max_pool2d    1           1  32   16   32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     3289.100           3289.100           3289.100              3289.100            0.000
     25 max_pool2d    1           1  32   16   64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     8009.500           8009.500           8009.500              8009.500            0.000
     26 max_pool2d    1           1  32   32   32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     6866.800           6866.800           6866.800              6866.800            0.000
     27 max_pool2d    1           1  32   32   64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     15006.800          15006.800          15006.800             15006.800           0.000
     28 max_pool2d    1           1  32   64   32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     13205.000          13205.000          13205.000             13205.000           0.000
     29 max_pool2d    1           1  32   64   64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     36247.900          36247.900          36247.900             36247.900           0.000
     30 max_pool2d    1           1  32   128  32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     28164.600          28164.600          28164.600             28164.600           0.000
     31 max_pool2d    1           1  32   128  64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     67673.200          67673.200          67673.200             67673.200           0.000
     32 max_pool2d    1           1  64   8    32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     2004.100           2004.100           2004.100              2004.100            0.000
     33 max_pool2d    1           1  64   8    64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     6370.700           6370.700           6370.700              6370.700            0.000
     34 max_pool2d    1           1  64   16   32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     6640.800           6640.800           6640.800              6640.800            0.000
     35 max_pool2d    1           1  64   16   64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     15813.800          15813.800          15813.800             15813.800           0.000
     36 max_pool2d    1           1  64   32   32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     13045.600          13045.600          13045.600             13045.600           0.000
     37 max_pool2d    1           1  64   32   64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     33073.000          33073.000          33073.000             33073.000           0.000
     38 max_pool2d    1           1  64   64   32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     28791.800          28791.800          28791.800             28791.800           0.000
     39 max_pool2d    1           1  64   64   64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     76532.600          76532.600          76532.600             76532.600           0.000
     40 max_pool2d    1           1  64   128  32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     59452.100          59452.100          59452.100             59452.100           0.000
     41 max_pool2d    1           1  64   128  64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     155364.400         155364.400         155364.400            155364.400          0.000
     42 max_pool2d    1           1  128  8    32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     4068.500           4068.500           4068.500              4068.500            0.000
     43 max_pool2d    1           1  128  8    64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     12074.000          12074.000          12074.000             12074.000           0.000
     44 max_pool2d    1           1  128  16   32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     12445.000          12445.000          12445.000             12445.000           0.000
     45 max_pool2d    1           1  128  16   64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     34239.600          34239.600          34239.600             34239.600           0.000
     46 max_pool2d    1           1  128  32   32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     27057.200          27057.200          27057.200             27057.200           0.000
     47 max_pool2d    1           1  128  32   64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     69079.900          69079.900          69079.900             69079.900           0.000
     48 max_pool2d    1           1  128  64   32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     63014.100          63014.100          63014.100             63014.100           0.000
     49 max_pool2d    1           1  128  64   64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     135556.200         135556.200         135556.200            135556.200          0.000
     50 max_pool2d    1           1  128  128  32  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     131045.600         131045.600         131045.600            131045.600          0.000
     51 max_pool2d    1           1  128  128  64  int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     298585.000         298585.000         298585.000            298585.000          0.000

@cconvey
Copy link
Contributor Author

cconvey commented Jun 14, 2022

CC: @csullivan

@github-actions github-actions bot requested a review from mehrdadh June 15, 2022 16:08
@cconvey
Copy link
Contributor Author

cconvey commented Jun 15, 2022

Updated benchmark output, due to a new revision of the PR code:

basic_kernel  sched_type  input_shape_4d     block_shape  DTYPE  KERNEL  STRIDE  DILATION  PADDING       IO_TENSOR_MEM_SCOPE  row_status  timings_min_usecs  timings_max_usecs  timings_median_usecs  timings_mean_usecs  timings_stddev_usecs  host_files_dir_path                                                                                                                                                                                 comments
max_pool2d    1           [1, 8, 8, 32]      [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     76.200             76.200             76.200                76.200              0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_8_8_32-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm      
max_pool2d    1           [1, 8, 8, 32]      [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     197.300            197.300            197.300               197.300             0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_8_8_32-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm      
max_pool2d    1           [1, 8, 8, 64]      [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     143.600            143.600            143.600               143.600             0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_8_8_64-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm      
max_pool2d    1           [1, 8, 8, 64]      [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     531.100            531.100            531.100               531.100             0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_8_8_64-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm      
max_pool2d    1           [1, 8, 32, 32]     [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     283.500            283.500            283.500               283.500             0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_8_32_32-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm     
max_pool2d    1           [1, 8, 32, 32]     [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     979.800            979.800            979.800               979.800             0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_8_32_32-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm     
max_pool2d    1           [1, 8, 32, 64]     [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     529.200            529.200            529.200               529.200             0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_8_32_64-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm     
max_pool2d    1           [1, 8, 32, 64]     [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     3513.700           3513.700           3513.700              3513.700            0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_8_32_64-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm     
max_pool2d    1           [1, 8, 128, 32]    [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     1122.700           1122.700           1122.700              1122.700            0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_8_128_32-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm    
max_pool2d    1           [1, 8, 128, 32]    [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     4100.300           4100.300           4100.300              4100.300            0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_8_128_32-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm    
max_pool2d    1           [1, 8, 128, 64]    [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     2105.000           2105.000           2105.000              2105.000            0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_8_128_64-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm    
max_pool2d    1           [1, 8, 128, 64]    [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     13259.000          13259.000          13259.000             13259.000           0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_8_128_64-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm    
max_pool2d    1           [1, 32, 8, 32]     [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     4.600              4.600              4.600                 4.600               0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_32_8_32-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm     
max_pool2d    1           [1, 32, 8, 32]     [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     971.900            971.900            971.900               971.900             0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_32_8_32-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm     
max_pool2d    1           [1, 32, 8, 64]     [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     528.700            528.700            528.700               528.700             0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_32_8_64-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm     
max_pool2d    1           [1, 32, 8, 64]     [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     3171.100           3171.100           3171.100              3171.100            0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_32_8_64-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm     
max_pool2d    1           [1, 32, 32, 32]    [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     1107.800           1107.800           1107.800              1107.800            0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_32_32_32-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm    
max_pool2d    1           [1, 32, 32, 32]    [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     6864.800           6864.800           6864.800              6864.800            0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_32_32_32-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm    
max_pool2d    1           [1, 32, 32, 64]    [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     2108.700           2108.700           2108.700              2108.700            0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_32_32_64-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm    
max_pool2d    1           [1, 32, 32, 64]    [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     14937.400          14937.400          14937.400             14937.400           0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_32_32_64-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm    
max_pool2d    1           [1, 32, 128, 32]   [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     4418.500           4418.500           4418.500              4418.500            0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_32_128_32-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm   
max_pool2d    1           [1, 32, 128, 32]   [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     27957.400          27957.400          27957.400             27957.400           0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_32_128_32-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm   
max_pool2d    1           [1, 32, 128, 64]   [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     8104.600           8104.600           8104.600              8104.600            0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_32_128_64-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm   
max_pool2d    1           [1, 32, 128, 64]   [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     73345.100          73345.100          73345.100             73345.100           0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_32_128_64-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm   
max_pool2d    1           [1, 128, 8, 32]    [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     5.800              5.800              5.800                 5.800               0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_128_8_32-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm    
max_pool2d    1           [1, 128, 8, 32]    [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     4068.300           4068.300           4068.300              4068.300            0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_128_8_32-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm    
max_pool2d    1           [1, 128, 8, 64]    [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     2103.200           2103.200           2103.200              2103.200            0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_128_8_64-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm    
max_pool2d    1           [1, 128, 8, 64]    [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     12057.600          12057.600          12057.600             12057.600           0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_128_8_64-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm    
max_pool2d    1           [1, 128, 32, 32]   [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     4419.300           4419.300           4419.300              4419.300            0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_128_32_32-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm   
max_pool2d    1           [1, 128, 32, 32]   [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     28135.200          28135.200          28135.200             28135.200           0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_128_32_32-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm   
max_pool2d    1           [1, 128, 32, 64]   [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     7944.200           7944.200           7944.200              7944.200            0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_128_32_64-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm   
max_pool2d    1           [1, 128, 32, 64]   [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     60932.900          60932.900          60932.900             60932.900           0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_128_32_64-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm   
max_pool2d    1           [1, 128, 128, 32]  [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     16996.200          16996.200          16996.200             16996.200           0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_128_128_32-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm  
max_pool2d    1           [1, 128, 128, 32]  [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     126598.000         126598.000         126598.000            126598.000          0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_128_128_32-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm  
max_pool2d    1           [1, 128, 128, 64]  [8, 8, 32]   int8   (1, 1)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     32163.600          32163.600          32163.600             32163.600           0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_128_128_64-block_shape:8_8_32-DTYPE:int8-KERNEL:1_1-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm  
max_pool2d    1           [1, 128, 128, 64]  [8, 8, 32]   int8   (3, 3)  (1, 1)  (1, 1)    (0, 0, 0, 0)  global.vtcm          SUCCESS     312788.700         312788.700         312788.700            312788.700          0.000                 /tmp/tmpszyuisxa/basic_kernel:max_pool2d-sched_type:1-input_shape_4d:1_128_128_64-block_shape:8_8_32-DTYPE:int8-KERNEL:3_3-STRIDE:1_1-DILATION:1_1-PADDING:0_0_0_0-IO_TENSOR_MEM_SCOPE:global.vtcm  

@cconvey cconvey changed the title [hexagon][testing] add max_pool2d benchmark WIP [hexagon][testing] add max_pool2d benchmark Jun 16, 2022
- Add benchmarking framework for Hexagon maxpool-2d kernels,
  and one (simple) kernel.
@cconvey cconvey changed the title WIP [hexagon][testing] add max_pool2d benchmark [hexagon][testing] add max_pool2d benchmark Jun 17, 2022
@cconvey cconvey requested a review from mehrdadh June 17, 2022 17:27
Copy link
Member

@mehrdadh mehrdadh left a comment

Choose a reason for hiding this comment

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

LGTM, thanks!

@csullivan csullivan merged commit c5465d8 into apache:main Jun 17, 2022
blackkker pushed a commit to blackkker/tvm that referenced this pull request Jul 7, 2022
- Add benchmarking framework for Hexagon maxpool-2d kernels,
  and one (simple) kernel.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants