On Nov 10, 2007, at 12:10 PM, Lauro Ramos Venancio wrote:
>>
>> Does it make sense to make this work? It seems much better than
>> having to hack arm.h for every subtarget that doesn't use @ as a
>> comment character.
>>
>
> I'm afraid I didn't understand your point. I think the problem is
> - Target can have inline asm extensions (not only comments - ARM
> target has two extensions: '%@' and '%r').
> So we need a way to define this extensions. To do this, I copied the
> idea used by the gcc codegen.
>
> arm.h code for gcc codegen:
>
> /* These are a couple of extensions to the formats accepted
> by asm_fprintf:
> %@ prints out ASM_COMMENT_START
> %r prints out REGISTER_PREFIX reg_names[arg] */
> #define ASM_FPRINTF_EXTENSIONS(FILE, ARGS, P) \
> case '@': \
> fputs (ASM_COMMENT_START, FILE); \
> break; \
> \
> case 'r': \
> fputs (REGISTER_PREFIX, FILE); \
> fputs (reg_names [va_arg (ARGS, int)], FILE); \
> break;
Hi Lauro,
It seems like there is two options: when lowering from GCC asm to
llvm asm, we can either change %@ (for example) into
ASM_COMMENT_START. Alternatively, we can lower it to "${:comment}",
which llvm asm already expands out to the right thing. Does this
make sense? The important point here is that not all arm targets use
@ for their comment character, and we don't want to break that.
-Chris
_______________________________________________
llvm-commits mailing list
llvm-commits@cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
