warpcore 1.1.3

Streamlined multi-threaded process acceleration

warpcore

Streamlined multi-threaded process acceleration

When working with software that needs to be performant, it’s challenging to deal with all the pitfalls of multi-threading while balancing code stability.

Smoothing out the bumps in the road to multi-threading is the primary goal of the project. It’s just that simple.

Installation

OS X, Linux & Windows:

pip install warpcore

Usage Examples

List Operations

1. Build a list of arguments that will be passed to a designated function.

jobs = []
jobs.append("Picard")
jobs.append("Janeway")
jobs.append("Kirk")
jobs.append("Sisko")
jobs.append("Archer")

2. Create a function that will iterate over the list:

def do_the_thing(name):
    print(f"Star Fleet Captain {name}")

3. Create a single-threaded version to test:

for name in jobs:
    do_the_thing(name)

4. Once that works, convert the for-loop into a warpcore call

warpcore.list_engage(jobs, do_the_thing)

Full example:

from warpcore.engineering import WarpCore

def do_the_thing(name):
    print(f"Star Fleet Captain {name}")

jobs = []
jobs.append("Picard")
jobs.append("Janeway")
jobs.append("Kirk")
jobs.append("Sisko")
jobs.append("Archer")

# Single-threaded operation (for testing/debug)
# for name in jobs:
#     do_the_thing(name)

# Multi-threaded operation (for normal operation)
warpcore = WarpCore()
warpcore.list_engage(jobs, do_the_thing)

Please refer to example0.py and example1.py for basic and more advanced usage examples respectively.

Dictionary Operations

1. Build a dict of arguments that will be passed to a designated function.

database = {
    "Picard": "USS Enterprise-D",
    "Janeway": "USS Voyager",
    "Kirk": "USS Enterprise-A",
    "Sisko": "Deep Space 9",
    "Archer": "Enterprise NX-01"
}

2. Create a function that will iterate over the dictionary:

*Note when using dicts, make sure your worker function accepts the key and value as arguments. (See below)

def do_the_thing(key, value):
    print(f"Star Fleet Captain {key} is/was in command of {value}")

3. Create a single-threaded version to test:

for key, value in database.items():
    do_the_thing(key, value)

4. Once that works, convert the for-loop into a warpcore call

warpcore.dict_engage(database, do_the_thing)

Full example:

from warpcore.engineering import WarpCore

def do_the_thing(key, value):
    print(f"Star Fleet Captain {key} is/was in command of {value}")

database = {
    "Picard": "USS Enterprise-D",
    "Janeway": "USS Voyager",
    "Kirk": "USS Enterprise-A",
    "Sisko": "Deep Space 9",
    "Archer": "Enterprise NX-01"
}

# Single-threaded operation (for testing/debug)
# for key, value in database.items():
#     do_the_thing(key, value)

# Multi-threaded operation (for normal operation)
warpcore = WarpCore()
warpcore.dict_engage(jobs, do_the_thing)

Fine Tuning for Performance

TL;DR: example2.py Is a working sample of the profiling system.

Your workload and processor architecture will dictate which settings work best for any situation.

You can leave things at default, but if you want to squeeze even more performance out, consider using the profiling feature.

# Regular operation
warpcore.list_engage(tasks_list, do_the_thing)

# Performance Profiling mode of same function as above
warpcore.list_profile(tasks_list, do_the_thing)

Profiling simply runs your code, but benchmarks execution time of the full job list. Then tweaks the settings and re-runs the jobs again.

Each time it re-runs, it displays the performance metrics of the last run on console.

Once complete, it will display the suggested combination of settings

Example 1

RESULTS: Best performance (85.8% gain) using * compute:True * with max_parallel: 51

This translates to the following setup:

warpcore = WarpCore(51)
warpcore.list_engage(tasks_list, do_the_thing, compute=True)

Example 2

RESULTS: Best performance (91.4% gain) using * compute:False (Default)* with max_parallel: 32

This translates to the following setup:

warpcore = WarpCore(32)
warpcore.list_engage(tasks_list, do_the_thing, compute=False)
# or just leave out 'compute' keyword to assume False
warpcore.list_engage(tasks_list, do_the_thing)

Meta

Brandon Blackburn – PGP Encrypted Chat @ Keybase

Distributed under the Apache 2.0 license. See LICENSE for more information.

TL;DR: For a human-readable & fast explanation of the Apache 2.0 license visit: http://www.tldrlegal.com/l/apache2

https://github.com/BlackburnHax/warpcore