|
Hello bhaktavatsala The floating point format is not an IEEE format and is described in the users guide. This processor stores the bit values differently to conserve register space and power as shown. |<---- access as a 40b float(register) ---------------->| |<---- access as a 32b float(memory)---->| |<------ access as 32b long --------->| +-----------------+-+ | 8b 2's comp Exp | 32b 2's compliment mantissa | +-----------------+-+ | Sign Basically if you access data from memory, lets say data or coefficient for an FIR, this data is most likely to be in 32b float format since it is stored in a 32b memory. You may note here how odd it would be to build and maintain a 40b (or greater) memory bus and that this would consume extra silicon and therefor also power. However, when the MAC operation occurs, you would like to keep as much precision as possible. Since this occurs as a Rx+Rx operation, this is done with 40b float precision, but unique to this processor, spanning only 1 register. Format conversion functions, TMS_IEEE() and IEEE_TMS(), are found in \DSK3\DSK3\COMMON\TMSFLOAT.CPP. If you want to look at some example code that shows how to use these functions, REGVIEW.EXE (built from REGVIEW.CPP) is a simple DOS Console type of application, so you dont have to learn how to program windows. If you want to dive into Windows programming, these functions are embedded in all of those programs as well. Building applications DOS Console Applications ------------------------ Built using Borland C++ ver 3.1. This is an older tool that only compiles 16b DOS applications. The project files are found in \DSK3\DSK3\BC31.PRJ. Note here that by being a DOS console appliation, these applications dont have the ability to call the protected mode 32b DLL support functions that access the PC printer port. These applications work with DOS, Win31, Win95, Win98, WinMe and I am told WinXP if you can figure out how to get the 'compatability mode' to work. Another option, but I suppose WinXP has slammed this door shut by now, is DIRECTIO (search for this on the web). Windows Applications -------------------- These applications will build using either Borland C++ 4.5, or MSVC ver 6.0. The project files are found in either \DSK3\DSK3\BC45.IDE, or in \DSK3\DSK3\MSVC.WKS\DSK3DW. Note that for MSVC each of the component applications is tree'd off of the main DK3DW workspace. Also, for the applications to work with Windows NT derived OS's (WinNT, 2K and XP) you need to build the windows applcations in 32 bit mode for either BC or MSVC. Hope this helps Keith Larson DSP and Analog Consulting http://home.comcast.net/~klarsondsp bhaktavatsala wrote: > Dear all, > I am giving input data through data file to TMS320vc33 DSP, and need to > perform floating point operations on the received data. > > 1. What type of data does this processor accept ? > 2. I tried to output, the input data file through DSP CC, and gave >various formats of data through PC (data file), but the output was junk >data. The aim of our program was to read the input data in float format >and output the same thing. > > Can i get any sample programs for this particular processor. > I am new to this processor, and help needed desperately. > > regards, > > Bhaktavatsala >===== >"Give to the World the Best of You > The Best in the World will come back to You" >*************************************************************** >S.Bhaktavatsala >*************************************************************** > > >__________________________________ > |
Re: How to give floating point data to DSP using CC
Started by ●May 24, 2004
