Re: RFA: Inserting NOPs for profiling.

[Richard Earnshaw <rearnsha at arm dot com>]

> You probably should ask people at ARM if any of their tools rely on this
> symbol being present in gcc output.   If they do, then you have a problem.

Not to my knowledge, since at present our debugger doesn't support gcc 
output; but I'll ask around.

> It might be easier to solve the problem by adding a "nop" macro to the
> assembler, so that the compiler doesn't have to worry about emitting an
> special arm or thumb instruction.  In fact, it looks like the assembler
> already supports nop and handles it correctly.  So all the compiler needs
> to do is emit a simple "nop" string.  You don't need any define_expand for
> that.

The reason for avoiding the pseudo-op "nop" was to try and avoid assembler 
dependencies (just because gas supports nop, doesn't mean that other 
assemblers do).  However, gcc does rely on the adr pseudo-op, so there is 
precedent there if we really need it.

R.