[Shiunbird]

Hello everyone,

I've always been very frustrated that some it's hard to find cross-platform benchmark tools that survived the test of time.
So, since I've been learning how to program and OpenGL, I decided that it will be my challenge.

My first test is pure C, and it divides the largest 32 bit prime integer by all odd numbers.
It is made to be slow and inefficient on purpose. Here are my results, all single-threaded:

MacBook Pro (employer-issued):
4.35 seconds
Intel Core i7-7567 3.5 GHz

Mac Pro:
8.95 seconds
Xeon X5657 3.06 GHz

Odroid-XU4:
16.56 seconds
Samsung Exynos5422 ARM® Cortex™-A15 Quad 2.0GHz/Cortex™-A7 Quad 1.4GHz

Raspberry Pi 3 Model B Rev 1.2
22.54 seconds
BCM2835

Mac G4:
31.25 seconds
G4 867

Raspberry Pi Zero:
87.21 seconds
1GHz ARMv6 (BCM2835)

HP 9000 785 C8000:
125.04 seconds (GCC)
108 seconds (HP's aCC)
PA8900 1 GHz

WinCE StrongARM SA1110 ARMv4 206 MHz:
240 seconds

WinCE SH4 SS7750 128 MHz:
870 seconds


I'm very surprised by how poorly the PA8900 performs. I'll try HP's compiler once I get HP-UX up.

What other tests do you suggest I could write?
I'm interested in assessing the advantages of having large L2 caches and the performance penalty when going to the RAM, and other tests that could measure the strengths of our UNIX workstations compared to what the market had available.

I'm also half-way through to writing my OpenGL benchmark. It generates different kinds of 2D and 3D scenes, pushes textures, spheres and 2D polygons around. I'm aiming for OpenGL 1.3 and 3.1, in C, with -lGL, lX11, -lm and -lGLU, so you could build it and try.
It already runs quite ok, I just have a broken FPS counter that I'm trying to fix.