Semidbm performance

semidbmperf1

Was trying to optimize sequential writes (fill_sequential benchmark).  Showing:

* Current released version
* performance regression with python3 support
* Better performance with binary file format

Benchmark params:

100,000 keys, key size 16 bytes, value size 100 bytes


Current release (v0.3.1):

((v0.3.1) %) $ scripts/benchmark -d semidbm -n 100000
Generating random data.
Benchmarking: <module 'semidbm' from 'semidbm/semidbm/__init__.pyc'>
    num_keys  : 100000
    key_size  : 16
    value_size: 100
fill_sequential     : time:     1.522,   micros/ops:    15.223,   ops/s:  65691.412,  MB/s:      7.267
read_hot            : time:     0.464,   micros/ops:     4.643,   ops/s: 215398.489,  MB/s:     23.829
read_sequential     : time:     0.441,   micros/ops:     4.407,   ops/s: 226909.248,  MB/s:     25.102
read_random         : time:     0.492,   micros/ops:     4.922,   ops/s: 203189.662,  MB/s:     22.478
delete_sequential   : time:     1.062,   micros/ops:    10.619,   ops/s:  94175.000,  MB/s:     10.418


SLOW version, after python3 support and str/bytes awesomeness:

((ee8ccbb...) %) $ scripts/benchmark -d semidbm -n 100000
Generating random data.
('Benchmarking:', <module 'semidbm' from 'semidbm/semidbm/__init__.pyc'>)
    num_keys  : 100000
    key_size  : 16
    value_size: 100
fill_sequential     : time:     2.025,   micros/ops:    20.247,   ops/s:  49389.226,  MB/s:      5.464
read_cold           : time:     0.436,   micros/ops:     4.358,   ops/s: 229479.965,  MB/s:     25.387
read_sequential     : time:     0.429,   micros/ops:     4.291,   ops/s: 233051.329,  MB/s:     25.782
read_hot            : time:     0.437,   micros/ops:     4.366,   ops/s: 229017.855,  MB/s:     25.335
read_random         : time:     0.486,   micros/ops:     4.858,   ops/s: 205866.323,  MB/s:     22.774
delete_sequential   : time:     1.111,   micros/ops:    11.114,   ops/s:  89976.190,  MB/s:      9.954


Binary file format (master @ 0dca44b)

(master % u=) $ scripts/benchmark -d semidbm -n 100000
Generating random data.
('Benchmarking:', <module 'semidbm' from 'semidbm/semidbm/__init__.pyc'>)
    num_keys  : 100000
    key_size  : 16
    value_size: 100
fill_sequential     : time:     1.260,   micros/ops:    12.604,   ops/s:  79340.204,  MB/s:      8.777
read_cold           : time:     0.403,   micros/ops:     4.031,   ops/s: 248049.663,  MB/s:     27.441
read_sequential     : time:     0.408,   micros/ops:     4.081,   ops/s: 245043.917,  MB/s:     27.108
read_hot            : time:     0.400,   micros/ops:     3.998,   ops/s: 250131.885,  MB/s:     27.671
read_random         : time:     0.456,   micros/ops:     4.558,   ops/s: 219395.018,  MB/s:     24.271
delete_sequential   : time:     1.015,   micros/ops:    10.147,   ops/s:  98549.056,  MB/s:     10.902