[SOLVED] Impact of "2>&1 > /dev/null" VS Output

erfanjoker · 2020-12-23 05:40:24

Hi,
My question is fundamentally about speed difference between these two
In programming, printing the output is relatively a resource-hugging action in comparison to non-outputing, Just as a loop can run milions of times within seconds or even less, But when you try to print something in each cycle of that loop, it can take about minutes and can lead to crashes

Now my question is, how is this in linux ? What exactly happens when we redirect the output of a command to /dev/null instead of display ? Do any processes get executed on it ? Can it have performance improvements in some cases ?

There are some sources that when you start building them, they just generate output of every single line of code they are building, I have seen that these builds usually take about 10-20mins based on the source code, I always assume that they would take fewer minutes if they just STOP Printing every single line of code they are processing , What is your idea ?

Last edited by erfanjoker (2020-12-23 13:39:10)

schard · 2020-12-23 08:50:01

Redirecting the output to a(nother) file can have impact on the program's execution time, yes.
It depends on how fast data can be written to the respective file.
Since /dev/null ist merely discarding all data written to it, it is insanely fast:

#include <stdio.h>


int main () {
	unsigned int i;

	for (i = 0; i <= 1000000; i++) {
		fprintf( stdout, "Garbage to stdout.\n" );
		fprintf( stderr, "Garbage to stderr.\n" );
	}
}

Results:

time ./test

real	0m6,518s
user	0m0,380s
sys	0m3,399s


time ./test > /dev/null

real	0m3,064s
user	0m0,205s
sys	0m1,635s


time ./test 2> /dev/null

real	0m3,586s
user	0m0,269s
sys	0m1,885s


time ./test 2> /dev/null >&2

real	0m0,173s
user	0m0,070s
sys	0m0,102s

Last edited by schard (2020-12-23 08:52:08)

seth · 2020-12-23 09:00:40

First off all:

> /dev/null 2>&1

The order matters.
https://www.gnu.org/software/bash/manua … tions.html
https://catonmat.net/bash-one-liners-ex … part-three

try to print something in each cycle of that loop, it can take about minutes and can lead to crashes

No. That's your code. Or a crap output (terminal)

To answer your question, you save the CPU on the terminal printing stuff and for compiler output that doesn't matter (unless, again, very crappy VTE)
Redirecting stdout and esp. stderr of a compiler to /dev/null is not the best of all ideas because you'll miss warnings and, more importantly, comments why yor code is broken and won't compile.

If there's really sth. that

generate(s) output of every single line of code they are building

there's probably a quiet parameter to restrict that behavior what would likely provide more performance than just redirecting the output.

erfanjoker · 2020-12-23 13:38:55

Thanks all, I got the answer

Arch Linux

#1 2020-12-23 05:40:24

[SOLVED] Impact of "2>&1 > /dev/null" VS Output

#2 2020-12-23 08:50:01

Re: [SOLVED] Impact of "2>&1 > /dev/null" VS Output

#3 2020-12-23 09:00:40

Re: [SOLVED] Impact of "2>&1 > /dev/null" VS Output

#4 2020-12-23 13:38:55

Re: [SOLVED] Impact of "2>&1 > /dev/null" VS Output

Board footer