Subject: [linux-audio-dev] Floating point math speed on pentium
From: Karl MacMillan (karlmac_AT_peabody.jhu.edu)
Date: Sun Oct 01 2000 - 02:13:11 EEST
Hi,
I have been trying to understand the performance of a simple test
programs using both floats and doubles and was hoping that someone on this
list could help give some insight. Basically I have a small program that
does some simple floating point math using random numbers using floats or
doubles in a loop. I use clock() to time the loop so that I can compare
the speed with floats and doubles. With no optimization things are as
expected - doubles take quite a bit longer. With '-O3 -march=i686' the
time is exactly the same. This is on a celeron 366. Any ideas?
Karl
-- _____________________________________________________ | Karl W. MacMillan | | Computer Music Department | | Peabody Institute of the Johns Hopkins University | | karlmac_AT_peabody.jhu.edu | | www.peabody.jhu.edu/~karlmac | -----------------------------------------------------
This archive was generated by hypermail 2b28 : Sun Oct 01 2000 - 02:34:19 EEST