Resize

title: 
style: nestedList # TOC style (nestedList|nestedOrderedList|inlineFirstLevel)
minLevel: 0 # Include headings from the specified level
maxLevel: 0 # Include headings up to the specified level
includeLinks: true # Make headings clickable
debugInConsole: false # Print debug info in Obsidian console

CPU

Physical cores: 20
Physical + Logical cores: 40

Architecture:                    x86_64
CPU op-mode(s):                  32-bit, 64-bit
Byte Order:                      Little Endian
Address sizes:                   46 bits physical, 48 bits virtual
CPU(s):                          40
On-line CPU(s) list:             0-39
Thread(s) per core:              2
Core(s) per socket:              10
Socket(s):                       2
NUMA node(s):                    2
Vendor ID:                       GenuineIntel
CPU family:                      6
Model:                           85
Model name:                      Intel(R) Xeon(R) Silver 4210R CPU @ 2.40GHz
Stepping:                        7
CPU MHz:                         2400.000
CPU max MHz:                     3200.0000
CPU min MHz:                     1000.0000

Bilinear

Bilinear Resize Benchmark Results

Method	Resize (640x640x3 to 256x256x3)	Speedup over Python	Speedup over OpenCV	Resize (2160x3840x3 to 480x854)	Speedup over Python	Speedup over OpenCV
Python	1.7087 s	N/A	N/A	10.4762 s	N/A	N/A
OpenCV	0.0003 s	5425.3690	N/A	0.0005 s	21883.2490	N/A
Mojo (For Loops)	0.0077 s	221.8717	0.0410	0.0492 s	217.8379	0.0097
Mojo (SIMD)	0.0010 s	1755.0607	0.323	0.0061 s	1763.9810	0.0789
Mojo (SIMD+ Parallel)	0.0001 s	13858.1098	2.5543	0.0006 s	18995.7696	0.8492

Python

Bilinear resize from (640, 640, 3) to (256, 256):
Error rate: 0.4068756103515625
Opencv: 0.0003149523865431547 s
Python: 1.7087329153902828 s
cv2 resize is 5425.369003057714 times faster than python

Bilinear resize from (2160, 3840, 3) to (480, 854):
Error rate: 0.37517483086130626
Opencv: 0.00047873249957337973 s
Python: 10.476222490519286 s
cv2 resize is 21883.248995744227 times faster than python

Mojo

Error

Mojo numpy Error checking:

Mojo Org Tensor shape:  640x640x3
Mojo Result Tensor shape:  256x256x3
Numpy Org Tensor shape:  (640, 640, 3)
Numpy Result Tensor shape:  (256, 256, 3)
Average error:  0.4065704345703125

Mojo Org Tensor shape:  2160x3840x3
Mojo Result Tensor shape:  480x854x3
Numpy Org Tensor shape:  (2160, 3840, 3)
Numpy Result Tensor shape:  (480, 854, 3
Average error:  0.3744877049180328

Conclusion: Algorithm implementation looks goods, error rates similar to python

Latency

using the best latencies from python:

for 640x640x3 -> 256x256x3:

    Opencv: 0.0003149523865431547 s
    Python: 1.7087329153902828 s

for 2160x3840x3 -> 480x854x3:

    Opencv: 0.00047873249957337973 s
    Python: 10.476222490519286 s

1. Pure For loops:

640x640x3 -> 256x256x3:

    ---------------------
    Benchmark Report (s)
    ---------------------
    Mean: 0.0077014453141175223
    Total: 145.21845284299999
    Iters: 18856
    Warmup Mean: 0.0112494864
    Warmup Total: 0.056247432
    Warmup Iters: 5
    Fastest Mean: 0.0077014453141175223
    Slowest Mean: 0.0077014453141175223
  
    resize:  640x640x3 -> 256x256x3 :  0.0077014453141175223 s
    speedup over python:  221.87171961839474
    speedup over opencv:  0.040895231180284479
    speedup in opencv:  24.452728866883096
    worst speedup in opencv:  24.452728866883096

2160x3840x3 -> 480x854x3:

  ---------------------
  Benchmark Report (s)
  ---------------------
  Mean: 0.049159172993067587
  Total: 113.459371268
  Iters: 2308
  Warmup Mean: 0.062143362000000001
  Warmup Total: 0.31071681000000001
  Warmup Iters: 5
  Fastest Mean: 0.049159172993067594
  Slowest Mean: 0.049159172993067594
  
  resize:  2160x3840x3 -> 480x854x3 :  0.049159172993067587 s
  speedup over python:  217.83793874849002
  speedup over opencv:  0.0097384164628011635
  speedup in opencv:  102.68609930780876
  worst speedup in opencv:  102.68609930780877

2. SIMD Vectorization

640x640x3 -> 256x256x3:

  Mojo Org Tensor shape:  640x640x3
  Mojo Result Tensor shape:  256x256x3
  Numpy Org Tensor shape:  (640, 640, 3)
  Numpy Result Tensor shape:  (256, 256, 3)
  Average error:  0.906005859375
  ---------------------
  Benchmark Report (s)
  ---------------------
  Mean: 0.00097360333796067309
  Total: 120.06963165400001
  Iters: 123325
  Warmup Mean: 0.0011944061999999999
  Warmup Total: 0.005972031
  Warmup Iters: 5
  Fastest Mean: 0.00097360333796067298
  Slowest Mean: 0.00097360333796067298
  
  resize:  640x640x3 -> 256x256x3 :  0.00097360333796067309 s
  speedup over python:  1755.0606584499035
  speedup over opencv:  0.32349148186247961
  speedup in opencv:  3.0912715050256341
  worst speedup in opencv:  3.0912715050256336

2160x3840x3 -> 480x854x3

  Mojo Org Tensor shape:  2160x3840x3
  Mojo Result Tensor shape:  480x854x3
  Numpy Org Tensor shape:  (2160, 3840, 3)
  Numpy Result Tensor shape:  (480, 854, 3)
  Average error:  0.3744877049180328
  ---------------------
  Benchmark Report (s)
  ---------------------
  Mean: 0.0060707756540178568
  Total: 5.4394149860000001
  Iters: 896
  Warmup Mean: 0.0075168404000000005
  Warmup Total: 0.037584201999999997
  Warmup Iters: 5
  Fastest Mean: 0.0060707756540178577
  Slowest Mean: 0.0060707756540178577
  
  resize:  2160x3840x3 -> 480x854x3 :  0.0060707756540178568 s
  speedup over python:  1763.9810010608544
  speedup over opencv:  0.078858539148376761
  speedup in opencv:  12.680934884153052
  worst speedup in opencv:  12.680934884153054

3. SIMD Vectorization + Parallelization

640x640x3 -> 256x256x3:

  Mojo Org Tensor shape:  640x640x3
  Mojo Result Tensor shape:  256x256x3
  Numpy Org Tensor shape:  (640, 640, 3)
  Numpy Result Tensor shape:  (256, 256, 3)
  Average error:  0.906005859375
  ---------------------
  Benchmark Report (s)
  ---------------------
  Mean: 0.00012330201849217637
  Total: 5.6342857349999997
  Iters: 45695
  Warmup Mean: 0.00029481879999999997
  Warmup Total: 0.001474094
  Warmup Iters: 5
  Fastest Mean: 0.0001233020184921764
  Slowest Mean: 0.0001233020184921764
  
  resize:  640x640x3 -> 256x256x3 :  0.00012330201849217637 s
  speedup over python:  13858.10983701539
  speedup over opencv:  2.5543165504880907
  speedup in opencv:  0.39149415518171204
  worst speedup in opencv:  0.39149415518171216

2160x3840x3 -> 480x854x3

  Mojo Org Tensor shape:  2160x3840x3
  Mojo Result Tensor shape:  480x854x3
  Numpy Org Tensor shape:  (2160, 3840, 3)
  Numpy Result Tensor shape:  (480, 854, 3)
  Average error:  0.5209252211813687
  ---------------------
  Benchmark Report (s)
  ---------------------
  Mean: 0.00056374303959517047
  Total: 6.3500015980000004
  Iters: 11264
  Warmup Mean: 0.0017351432
  Warmup Total: 0.008675716
  Warmup Iters: 5
  Fastest Mean: 0.00056374303959517036
  Slowest Mean: 0.00056374303959517036
  
  resize:  2160x3840x3 -> 480x854x3 :  0.00056374303959517047 s
  speedup over python:  18995.769638380512
  speedup over opencv:  0.84920338868780065
  speedup in opencv:  1.1775741987384343
  worst speedup in opencv:  1.1775741987384341

MVPavan's Notes

Explorer

Resize

CPU

Bilinear

Bilinear Resize Benchmark Results

Python

Mojo

Error

Latency

1. Pure For loops:

2. SIMD Vectorization

3. SIMD Vectorization + Parallelization

Graph View

Table of Contents