Hi all,
I am working onto to porting MAD to TX3903 MIPS processor on a custom board. As through the mail archive and from my experience with MAD, the configuration for the cross compilation of the package has some problems but that can be easily played around. In a nutshell I am able to compile the package with few modifications to my custom board.
Now if I use standard Makefile's linking, it aborts with an error of relocation truncation which is due to the global data table limit for the mips processor. So changing the value to linker option G helps in that case (default is 8). I have also compiled with the -msoft-floats.
When I run the program on the target, it gives undefined symbols "dpmul", "dpdiv", "dpadd", "dptoi". These symbols are from the floating point library. I do have Math/C/GCC library loaded. Can anyone out there has some experience of dealing with this kind of error or some other workaround. Any pointers will be greatly appreciated.
Thanks in advance -Raj
On Saturday, January 12, 2002, at 04:44 PM, Rajneesh Kumar wrote:
When I run the program on the target, it gives undefined symbols "dpmul", "dpdiv", "dpadd", "dptoi". These symbols are from the floating point library. I do have Math/C/GCC library loaded. Can anyone out there has some experience of dealing with this kind of error or some other workaround. Any pointers will be greatly appreciated.
Where are you getting your cross-compiler?
If you're using a commercial one, or one that's well-supported, you should already have dpmul et al in libgcc.a; you may need to choose the correct libgcc.a by *linking* with -msoft-float, as that will pick the proper version of libgcc.a.
If you're building gcc by hand, well, things are a little more complicated. If you are, could you tell us how your configuring gcc?
Jay
-
Jay Carlson -
Rajneesh Kumar