Titan compile » History » Version 2
Version 1 (Henrik Levämäki, 2016-06-21 12:00) → Version 2/4 (Henrik Levämäki, 2016-07-04 14:37)
h1. Titan compile
The CPUs on Titan support so called AVX/AVX2 instructions (https://en.wikipedia.org/wiki/Advanced_Vector_Extensions), which make floating point-intensive calculations faster (depending on the code, ~2X speed boost vs. Pleione are possible).
The following command can be used to access CPU information (make, model, supported instruction sets etc.):
<pre>
less /proc/cpuinfo
</pre>
With Intel compilers, one can therefore use the following compiler flag to take advantage of AVX2:
<pre>
-xCORE-AVX2
</pre>
The CPUs on Titan support so called AVX/AVX2 instructions (https://en.wikipedia.org/wiki/Advanced_Vector_Extensions), which make floating point-intensive calculations faster (depending on the code, ~2X speed boost vs. Pleione are possible).
The following command can be used to access CPU information (make, model, supported instruction sets etc.):
<pre>
less /proc/cpuinfo
</pre>
With Intel compilers, one can therefore use the following compiler flag to take advantage of AVX2:
<pre>
-xCORE-AVX2
</pre>