This thread has been locked.
If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.
Hi all,
I'm experiencing really long compile/linking times.
Processor: TMS320DM8147, DM8127 Eval Kit, 512 MB RAM.
Following this instructions: http://processors.wiki.ti.com/index.php/Linker_Runs_Too_Long
Results the error:
fatal error #10000: invalid option: --symdebug:none
Using only the -b option for the sysbios cores (DSP, M3VIDEO, M3VPSS):
error #10016-D: --no_sym_merge not supported when producing ELF output files
Does someone know what can be done to reduce compile/link time for this platform?
Weber
Aditional info:
Sometimes it can take +- 1 hour to a full compilation/linking of the entire source code.
I presume you are having this problem when using the TI C6000 compiler, and not the TI ARM compiler. Is that correct?
weber de souza calixto said:fatal error #10000: invalid option: --symdebug:none
One guess is you are linking by directly invoking the linker lnk6x. The option --symdebug:none is a compiler option, not a linker option. Moreover, you should avoid invoking the linker directly. It is better to invoke it through the compiler shell cl6x.
weber de souza calixto said:Does someone know what can be done to reduce compile/link time for this platform?
I think you have uncovered a situation we have not seen. Thus I need a test case from you which allows us to reproduce this behavior. I realize this is particularly difficult for linker issues like this one, but there is no other way to move forward. Is your build organized into a CCS project? If so, I'd appreciate if you would send me the entire project. And please include the compiler version (not the CCS version) you are using.
Thanks and regards,
-George
Hi George,
My bad. I misread the wiki part about the option --symdebug.
For the M3 cores, we are using:
ARM C/C++ CODE GENERATION TOOLS 4.9.7
And for the DSP core:
TMS320C6000 C/C++ CODE GENERATION TOOLS 7.4.2
We are compiling thru make commands, not from the CCS. In fact we are not using CCS for anything more than editing/searching code. Everything else is done by shell.
Our base code is the one provided with the eval kit (DM8127 IPNC RDK 3.5). I'm note sure if we can share the code or not. Can't you get the demo code that comes with DM8127? Or are we allowed share our copy (even you being TI's staff)?
You are right about us calling directly lnk6x for the DSP phase. However the slow linking phase is related with the M3VPSS and M3Video processors. The compile/linking times for A8 and DSP processors are just fine, even with a full clean.
For the M3, the make scripts call cl470 for compiling/linking and ar470 for archiving. I wonder if this is the best way for the M3 cores.
Waiting your reply,
Weber
Another info:
The problem is also present with the default DM8127 source code.
Weber