ump

workqueue in user space

description

Based on multi-thread in user space. You can put your function into ump, so the function is running parallelly. Thus, you can take advantage of multi-core CPUs.

performance

1. without ump

The result of test is "delay is 47921"

2. with ump

The result of test is "delay is 9799"

3. result

The performance is improved by 389%

build

for so

gcc -fpic -shared -o libump.so ump.c

for test

gcc main.c -L. -lump -lpthread

run test

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$PWD

without ump

./a.out

with ump

./a.out 1