BW_PIPE(8)

💖 bw_pipe — bestie we're timing data through pipes 💖

okay so this is bw_pipe from the LMBENCH manual, section 8, and it's genuinely one of the cutest little benchmarks ever. let's get into it 🫶

---

✨ NAME ✨

bw_pipe — time data movement through pipes 🌸

---

✨ SYNOPSIS ✨

bw_pipe [ -m <message size> ] [ -M <total bytes> ] [ -P <parallelism> ]
[ -W <warmups> ] [ -N <repetitions> ]

like, look at her. five whole options. she's serving ✨

---

✨ DESCRIPTION ✨

so what she does is: bw_pipe makes a Unix pipe between two processes and moves total bytes through that pipe in message size chunks. and bestie, hear me on this — pipes are typically sized smaller than that, so don't come for me when the chunks don't fit, that's literally just how pipes are 💅

the default total bytes is 10MB and the default message size is 64KB. iconic defaults, no notes.

---

✨ OUTPUT ✨

the output format is literally this, character for character:

"Pipe bandwidth: %0.2f MB/sec\n", megabytes_per_second

which in real life looks like:

Pipe bandwidth: 4.87 MB/sec

okay but that number? that's the whole point. that's the tea 🍓

---

✨ MEMORY UTILIZATION ✨

this is the part where it gets lowkey unhinged. this benchmark can move up to six times the requested memory per process. SIX. 😭

there are two processes, babes: the sender and the receiver.

most Unix systems implement the read/write system calls as a bcopy from/to kernel space to/from user space. and bcopy will use 2-3 times as much memory bandwidth: there's one read from the source and a write to the destination. the write usually results in a cache line read and then a write back of the cache line at some later point. so yeah, she's eating memory for breakfast 💀

the good news? memory utilization might be reduced by 1/3 if the processor architecture implemented "load cache line" and "store cache line" instructions (as well as getcachelinesize). so like... theoretically slay, if your CPU was built different 🧚

---

✨ ACKNOWLEDGEMENT ✨

funding for the development of this tool was provided by Sun Microsystems Computer Corporation. we love a sponsor, periodt 👑

---

✨ SEE ALSO ✨

• lmbench(8).

go read that one too, iykyk 💜

---

✨ AUTHOR ✨

Carl Staelin and Larry McVoy wrote this and honestly? kings 👑

Comments, suggestions, and bug reports are always welcome. fr fr, they want your feedback, don't be shy 🌈

---

and that's the whole page bestie!! this is (c)1994 Larry McVoy, the date on this version is $Date$, and the source is the LMBENCH manual, section 8. mwah 💋 ✨💖🎀🦋🌸