Simple Yocto benchmark

Use this benchmark to measure your machine's performance in building a small Yocto image. The download and build is deterministic and should build equally everywhere.

Install git and download this from github:

   sudo apt install git
   git clone https://github.com/sveinse/yocto-benchmark.git
   cd yocto-benchmark

If the machine haven't been used for Yocto building before, the necessary host-tools for Yocto must be installed. To do this on Ubuntu, execute

   sudo ./host-install

Clone and download the necessary sources. Will download 5.3 Gb from the net.

   ./clone

Run the timed compilation. The results will be logged in time-compile.*. It will consume around 8 Gb disk when RM_WORK="1", and around 18 Gb if RM_WORK is disabled (see the compile file for settings).

   ./compile

It can be worth running the test (step 4) multiple times to see if system caching makes any difference. To rerun, please run ./clean and then rerun the compilation with ./compile.

Finally, ./distclean can be used to remove everything.

Results

Here are some results from my own testing:

DL380 Virtual

System: HP DL380, E5-2637 v2 @3.5GHz (2x4 cores, 16 threads), 128 GB RAM. Running MS Server 2012 with HyperV. Guest running Ubuntu, 80GB RAM, 16 CPU. Dedicated PHY access to SSD storage device.

# Standard compilation
15185.56user 1659.29system 29:37.33elapsed 947%CPU  (2017-12-15)
15119.29user 1640.76system 29:23.00elapsed 950%CPU  (2017-12-15)

# rm_work disabled
14684.79user 1578.54system 28:56.96elapsed 936%CPU  (2017-12-14)
14709.04user 1526.15system 30:24.63elapsed 889%CPU  (2017-12-14)

# Using tmpfs
15447.03user 1677.42system 26:17.88elapsed 1085%CPU  (2018-01-03)
15720.32user 1765.11system 32:19.13elapsed 901%CPU   (2018-01-03)

# Using tmpfs, rw_work disabled
14878.02user 1575.68system 25:35.92elapsed 1071%CPU  (2017-12-15)
14877.92user 1584.26system 25:39.91elapsed 1069%CPU  (2017-12-15)

Lenovo W530

System: Lenovo W530 laptop. i7-3820QM CPU @2.7 GHz (4 cores, 8 threads), 8 GB memory. Intel 520 SSD.

# Standard compilation
16426.48user 1210.09system 45:43.44elapsed 642%CPU  (2018-01-03)
16219.27user 1193.68system 44:24.72elapsed 653%CPU  (2018-01-03)

# rm_work disabled
15370.77user 1107.08system 41:22.29elapsed 663%CPU
15792.09user 1142.44system 42:55.50elapsed 657%CPU

# Enforced limitation of 4 CPUs
15425.47user 1165.22system 46:03.35elapsed 600%CPU  (2018-01-03)

# VirtualBox Ubuntu guest on Ubuntu host. 4096Mb, 4 CPU
10724.01user 1482.69system 1:03:15elapsed 321%CPU  (2018-01-03)
10888.51user 1500.76system 1:04:04elapsed 322%CPU  (2018-01-03)

# VMWare Player on Ubuntu host. 4096Mb, 8 CPU
19583.42user 2952.69system 1:00:04elapsed 625%CPU  (2018-01-03)

# VirtalBox on Windows 8 host. 4096Mb, 4 CPU
 9815.62user 953.91system 56:54.96elapsed 315%CPU  (2018-01-03)
10007.63user 982.05system 58:07.10elapsed 315%CPU  (2018-01-03)

# VirtalBox on Windows 8 host. 4096Mb, 4 CPU, PHY disk
9552.24user 1038.75system 53:11.13elapsed 331%CPU  (2018-01-03)
9604.82user 1025.48system 53:24.65elapsed 331%CPU  (2018-01-03)

Workstation

System: Workstation. 2x Xeon E5-2670 v2 @ 2.5GHz (2x10cores, 40 threads). 64 GB Memory. Linux host.

15854.25user 1476.46system 19:18.75elapsed 1495%CPU  (2018-01-03)

Lenovo P51

System: Lenovo P51 laptop. i7-7820HQ CPU @2.9 GHz (4 cores, 8 threads), 32 GB memory. 2x Intel nvme SSD.

13216.55user 1355.77system 33:59.39elapsed 714%CPU  (2018-01-17)

ASUS ROG G750JH

System: ASUS ROG G750JH. i7-4700HQ CPU @2.40GHz (4 cores, 8 threads), 32 GB memory, Samsung 850 PRO SSD

15938.39user 1695.40system 41:49.66elapsed 702%CPU  (2018-02-08)

2018-02-08 Svein Seldal sveinse@seldal.com