Hi Dave, <br><br>I tried defining <span>ARM5E_ASM and ARM4_AS, and you are right, it doesn't compile, <br></span><span></span>But I can't say if the problem come from the Microsoft compiler, because if it doesn't know<br>
<span>ARM assembly optimizations, who can know ? <br>We may need to include/install something else to use </span><span>ARM5E_ASM or ARM4_ASM, <br>or maybe this part of speex's code is not finished.<br></span><br>I tried to compare both l
<span>ibspeex_armce.lib and </span><span></span><span>hand-built FIXED_POINT build as well.<br>I encoded 30sec and it tooks around 6sec with </span>l<span>ibspeex_armce.lib (1.1.16)<br>and around 4sec with </span><span>1.2beta1
. <br><br><br>rgds, <br>Patrick<br><br></span><br><br><span>>Thanks Patrick. I have done this too. However, it was my impression that
<br></span><span>></span><span>the libspeex_armce.lib was cross-compiled in GCC. Building in VS2005<br></span><span>></span><span>doesn't allow you to take advantage of the ARM assembly optimizations,<br></span><span>
></span><span>since the Microsoft compiler doesn't know how to do that (try defining<br></span><span>></span><span>ARM5E_ASM or ARM4_ASM and you will see what I mean). At least back in<br></span><span>></span><span>
the 1.1.16 days this seemed to make a big difference for me: the time<br></span><span>></span><span>taken during a decent sized encode() call was much greater with a<br></span><span>></span><span>hand-built FIXED_POINT build than it was with libspeex_armce.lib, but
<br></span><span>></span><span>perhaps I was doing something else wrong. I haven't had a chance to see<br></span><span>></span><span>if that's still the case with 1.2beta1.<br></span><br>