Hi all, I would be interested in partaking in the development of the fixed-point version of speex. I want to compile it on a TI DSP TMS320C6416. See http://focus.ti.com/docs/prod/folders/print/tms320c6416.html . There is only a mention of this work in progress on the web site but no other details. I use WinXP & debian 2.6 as dev platforms. well versed in cvs. C
Hi Colin, A TI DSP port of Speex would definitely be cool. Right now, most of Speex has been converted to fixed-point so that the few floating point operations that remain don't make a difference in the computation time. So far, it has only been tested on ARM, but it works well. The current fixed-point code also makes little assumptions about the CPU/DSP: no saturation instructions are required and only a 16x16 multiplier and 32- bit accumulator is assumed. Anything else you need to know about the current status? Jean-Marc On sam, 2004-09-11 at 10:52 -0300, Colin MacKenzie wrote:> Hi all, > > I would be interested in partaking in the development of the fixed-point > version of speex. I want to compile it on a TI DSP TMS320C6416. See > http://focus.ti.com/docs/prod/folders/print/tms320c6416.html . There is > only a mention of this work in progress on the web site but no other > details. > > I use WinXP & debian 2.6 as dev platforms. well versed in cvs. > > C > > _______________________________________________ > Speex-dev mailing list > Speex-dev@xiph.org > http://lists.xiph.org/mailman/listinfo/speex-dev
Jean-Marc Valin wrote:> Speex has been converted to fixed-point so that the few floating point > operations that remain don't make a difference in the computation time. > So far, it has only been tested on ARM, but it works well. The current > fixed-point code also makes little assumptions about the CPU/DSP: no > saturation instructions are required and only a 16x16 multiplier and 32- > bit accumulator is assumed.I had a related question. I'm working with a VLIW/SIMD DSP (http://equator.com/productsservices/pdf/BSP-15.pdf), and for some time have been considering a port of Speex (as an alternative to our current 64 kbps wideband audio codec). What could you say about portability of Speex to such an architecture (several execution units, a lot of wide registers, many SIMD instructions)? Vlad
On Sat, Sep 11, 2004 at 10:52:56AM -0300, Colin MacKenzie wrote:> Hi all, > > I would be interested in partaking in the development of the fixed-point > version of speex. I want to compile it on a TI DSP TMS320C6416. See > http://focus.ti.com/docs/prod/folders/print/tms320c6416.html . There is > only a mention of this work in progress on the web site but no other > details. > > I use WinXP & debian 2.6 as dev platforms. well versed in cvs.Hi Colin, you'll want to grab speex 1.1.x -- either grab the 1.1.6 release tarball to play with, or go straight to the subversion trunk. build with FIXED_POINT defined to 1 (./configure --enable-fixed-point on linux or cygwin, or hardcode it in win32/config.h for other builds on win32). Conrad.