On 01/11/2013 05:36 PM, Adam Sampson wrote: > Kristoff Bonne <[email protected]> writes: > >> if I use the codec2 library to decode a codec2 stream, I get slightly >> different result on i386 and on ARM. > That's not unusual -- the i387 FPU has an 80-bit wide stack, so > intermediate floating-point results are stored with more precision on > i386 than on other platforms, and the resulting rounding errors will be > different. > > You may find that you get something closer to your expected results if > you specify -msse -mfpmath=sse as compiler flags. See the description in > the GCC manual here: > http://gcc.gnu.org/onlinedocs/gcc-4.7.2/gcc/i386-and-x86_002d64-Options.html > That's just the tip of the iceberg of things which affect the outcome of floating point calculations. Try -O0, -O1 and -O2 for the compilation of most floating point codecs and the answers can change quite a lot.
Steve ------------------------------------------------------------------------------ Master Visual Studio, SharePoint, SQL, ASP.NET, C# 2012, HTML5, CSS, MVC, Windows 8 Apps, JavaScript and much more. Keep your skills current with LearnDevNow - 3,200 step-by-step video tutorials by Microsoft MVPs and experts. SALE $99.99 this month only -- learn more at: http://p.sf.net/sfu/learnmore_122912 _______________________________________________ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
