search

buddy-compiler / buddy-benchmark

Benchmark Framework for Buddy Projects
Apache License 2.0
45 stars 36 forks source link

[WIP] Experiment with mlir return format and pointwise conv #9

Open Joejiong opened 2 years ago

Joejiong commented 2 years ago

Experiment with mlir return format and pointwise conv optimization setup

problem/issue solved:

> ./bin/pointwise-conv-2d-nhwc-hwcf-benchmark                           pointwise_benchmark [39d7e0c] modified
2021-12-20T18:42:07+08:00
Running ./bin/pointwise-conv-2d-nhwc-hwcf-benchmark
Run on (52 X 2500 MHz CPU s)
CPU Caches:
  L1 Data 32 KiB (x26)
  L1 Instruction 32 KiB (x26)
  L2 Unified 1024 KiB (x26)
  L3 Unified 36608 KiB (x1)
Load Average: 8.09, 11.43, 12.63
***WARNING*** Library was built as DEBUG. Timings may be affected.
-------------------------------------------------------------------------------------
Benchmark                                           Time             CPU   Iterations
-------------------------------------------------------------------------------------
BM_PointwiseConv2DNhwcHwcf/100                  0.124 ms        0.124 ms         5655
BM_PointwiseConv2DNhwcHwcfReturn/100            0.104 ms        0.104 ms         6728
BM_PointwiseConv2DNhwcHwcfReturnOrigin/100      0.150 ms        0.150 ms         4665
inputMemRef: [ 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 ] of shape: [ 1 4 5 2 ]
filterMemRef: [ 3 3 3 3 3 3 3 3 3 3 3 3 3 3 ] of shape: [ 1 1 2 7 ]
outputMemRef: [ 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 ] of shape: [ 1 4 5 7 ]
inputMemReturn: [ 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 ] of shape: [ 1 4 5 2 ]
filterMemReturn: [ 3 3 3 3 3 3 3 3 3 3 3 3 3 3 ] of shape: [ 1 1 2 7 ]
outputMemReturn: [ 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 ] of shape: [ 1 4 5 7 ]
inputMemReturn: [ 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 ] of shape: [ 1 4 5 2 ]
filterMemReturn: [ 3 3 3 3 3 3 3 3 3 3 3 3 3 3 ] of shape: [ 1 1 2 7 ]
outputMemReturnOrigin: [ 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 ] of shape: [ 1 4 5 7 ]