Skip to content

Instantly share code, notes, and snippets.

@emallson

emallson/Benchmark - h5py vs np.memmap.ipynb

Last active January 7, 2016 17:40
Show Gist options

  • Select an option

    Learn more about clone URLs
  • Save emallson/0e9ff54c14c85ba486e3 to your computer and use it in GitHub Desktop.

Select an option

Learn more about clone URLs
Save emallson/0e9ff54c14c85ba486e3 to your computer and use it in GitHub Desktop.
Download ZIP
Display the source blob
Display the rendered blob
Raw
Benchmark - h5py vs np.memmap.ipynb
{
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"import h5py\n",
"import numpy as np"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Generate Some Data ###\n",
"\n",
"A 10,000 x 10,000 random float matrix. Approximately 763MB when stored in the `test.h5`"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"with h5py.File(\"test.h5\") as f:\n",
" f['/test'] = np.random.rand(10000, 10000)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Define numpy mmap ###"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"def _mmap_h5(path, h5path):\n",
" with h5py.File(path) as f:\n",
" ds = f[h5path]\n",
" # We get the dataset address in the HDF5 fiel.\n",
" offset = ds.id.get_offset()\n",
" # We ensure we have a non-compressed contiguous array.\n",
" assert ds.chunks is None\n",
" assert ds.compression is None\n",
" assert offset > 0\n",
" dtype = ds.dtype\n",
" shape = ds.shape\n",
" arr = np.memmap(path, mode='r', shape=shape, offset=offset, dtype=dtype)\n",
" return arr"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Experiment 1: Load Data, Slice rectangle [:2, :3]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### h5py"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {
"collapsed": false,
"scrolled": true
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"10 loops, best of 3: 2.01 s per loop\n"
]
}
],
"source": [
"%%timeit -n 10 -r 3\n",
"f = h5py.File(\"test.h5\", 'r')\n",
"x = f['/test'][...]\n",
"print(x[:2, :3])\n",
"f.close()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### numpy"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {
"collapsed": false,
"scrolled": true
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"10 loops, best of 3: 3.42 ms per loop\n"
]
}
],
"source": [
"%%timeit -n 10 -r 3\n",
"x = _mmap_h5('test.h5', '/test')[...]\n",
"print(x[:2, :3])"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Experiment 2: Don't Load, Just Slice [:2, :3]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### h5py"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {
"collapsed": false,
"scrolled": true
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"10 loops, best of 3: 4.36 ms per loop\n"
]
}
],
"source": [
"%%timeit -n 10 -r 3\n",
"f = h5py.File(\"test.h5\", 'r')\n",
"x = f['/test']\n",
"print(x[:2, :3])\n",
"f.close()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### numpy"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {
"collapsed": false,
"scrolled": true
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"[[ 0.35610554 0.88638298 0.61585581]\n",
" [ 0.52239096 0.26190995 0.02339244]]\n",
"10 loops, best of 3: 3.67 ms per loop\n"
]
}
],
"source": [
"%%timeit -n 10 -r 3\n",
"x = _mmap_h5('test.h5', '/test')\n",
"print(x[:2, :3])"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Experiment 3: Aggregate over all data"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### h5py"
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {
"collapsed": false,
"scrolled": true
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"50000732.3183\n",
"10 loops, best of 3: 4.68 s per loop\n"
]
}
],
"source": [
"%%timeit -n 10 -r 3\n",
"f = h5py.File(\"test.h5\", 'r')\n",
"x = f['/test']\n",
"# h5py.Dataset does not have a sum() function. \n",
"# The options are to use [...] to load all, or use np.sum on the dataset itself\n",
"print(np.sum(x))\n",
"f.close()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### numpy"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {
"collapsed": false,
"scrolled": true
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"50000732.3183242\n",
"10 loops, best of 3: 276 ms per loop\n"
]
}
],
"source": [
"%%timeit -n 10 -r 3\n",
"x = _mmap_h5('test.h5', '/test')\n",
"print(x.sum())"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Conclusion\n",
"\n",
"`np.memmap` and `h5py` perform equally well when you don't load the entire dataset (`[...]`). However, when the entire dataset is loaded or aggregation is performed, then `np.memmap` significantly outperforms `h5py`"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Final Timings\n",
"\n",
"| Method | Experiment 1 | Experiment 2 | Experiment 3 |\n",
"|-------------|--------------|--------------|--------------|\n",
"| `h5py` | 2.01s | 4.36ms | 4.68s |\n",
"| `np.memmap` | **3.42ms** | **3.67ms** | **276ms** |"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.4.3"
}
},
"nbformat": 4,
"nbformat_minor": 0
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment