I'd like to ask a favour of anyone with a CPU capable of single thread operation at benchmark speeds above 3000 or FP of 40+MIPS here:
I'd like to throw it at the Intel C++ compiler too but their website has defeated me. I once got as far as compiling "Hello World" but have never since been able to find where it was I previously managed feat that or how to upload my C sourcecode and then compile it with their C compiler.
Anyone have a working copy of Intel C++ and be willing to try and compile it please? MS C++ isn't exactly standard so I'm expecting a few errors when it hits any new compiler that follows the standards.
It does some floating point operations and benchmarks to determine how quickly the CPU can do various core numerical operations to create a set of Lego building blocks to be reassembled for maximum speed. This is a sample of its output for an i3770K (results file is about 60k).
# 3 (null loop overhead)
+ 5- 5
* 5 / 29 abs 9 fchsgn 9 chgsign 27 0-x 9 1/(1+x)* 20 1/(1+x)** 18 1/(1+x) 13 sqrt 20 atan 49 atan2 140 log 39 exp 40 sin 57 cos 57 tan 68 tan87 121Timings in nominal CPU cycles. Full list much longer...
I'm am interested in how much faster the latest generation of Intel chips are at handling this problem. I have a wide range of CPUs available but nothing quite cutting edge enough to push the envelope.
In return I can only offer the sourcecode for a couple of state of the art cuberoot algorithms and a cubic solver. The code being optimised is ultimately destined for exoplanet hunting in case you are curious.
My email reply-to address in this email is valid provided that you make no changes to it.