On 06/24/2018 11:59 PM, Jan Hubicka wrote:
> Hi,
> searching for other occurences I see:
> jan@skylake:~/trunk/gcc/config/i386> grep ASM_BYTE *md *.c
> i386.md: return ASM_BYTE "0x9e";
> i386.md: fputs (ASM_BYTE "0x66\n", asm_out_file);
> i386.md: fputs (ASM_BYTE "0x66\n", asm_out_file);
> i386.c: fputs (ASM_BYTE "0x48, 0x8d, 0xa4, 0x24, 0x00, 0x00, 0x00, 0x00\n",
> i386.c: fputs (ASM_BYTE "0x8b, 0xff, 0x55, 0x8b, 0xec\n", asm_out_file);
> i386.c: fputs ("\n" ASM_BYTE "0xf2\n\t", file);
> i386.c: fputs ("\n" ASM_BYTE "0xf3\n\t", file);
> i386.c: fprintf (file, "1:" ASM_BYTE "0x0f, 0x1f, 0x44, 0x00, 0x00\n");
> i386.c:#undef TARGET_ASM_BYTE_OP
> i386.c:#define TARGET_ASM_BYTE_OP ASM_BYTE
>
I've just reviewed all these occurances, as well as the bigger-sized
ASM_<size> ones, and indeed in a couple of places we get a breakpoint at
an incorrect location, but AFAIU not in the middle of an insn as is the
case here.
> Perhaps we want to add new macro like ASM_INSN_BYTE which is used to output
> such prefixes...
Maybe. And I suspect that with the .insn/.endinsn construct I mentioned
earlier in this thread, this new macro would be easier to implement.
Thanks,
- Tom
