Kevin,
Assume you're using the TI ARM compiler and not GCC or another compiler like IAR or Keil?? This is a compiler specific question...
TI's compiler doesn't implement all of the GCC extensions. The compiler manual (google SPNU151 to find it) lists the extensions supported/not supported in Table 5-5 section 5.15.1. "Extended asm / Assembler instructions with C Operands" is one of the unsupported extensions.
I think you'll need to write an assembly routine as an ABI compilaint C function, which would mean moving LR into R0 and then returning immediately. Then try to get that function to inline to take out the overhead. That's what I'd suggest trying anyway.
Best Regards,