On Sun, Dec 16, 2018 at 02:33:39AM +0000, Nadav Amit wrote:
> In general, I think that from the start it was clear that the motivation for
> the patch-set is not just performance and also better code. For example, I
> see no reason to revert the PV-changes or the lock-prefix changes that
> improved the code readability.
One thing that has caught my eye with the asm macros, which actually
decreases readability, is that I can't see the macro properly expanded
when I do
make <filename>.s
For example, I get
#APP
# 164 "./arch/x86/include/asm/cpufeature.h" 1
STATIC_CPU_HAS bitnum=$8 cap_byte="boot_cpu_data+35(%rip)" feature=123 t_yes=.L75 t_no=.L78 always=117 #, MEM[(const char *)&boot_cpu_data + 35B],,,,
# 0 "" 2
.loc 11 164 2 view .LVU480
#NO_APP
but I'd like to see the actual asm as it is really helpful when hacking
on inline asm stuff. And I haven't found a way to make gcc expand asm
macros in .s output.
Now, assuming the gcc inline patch will be backported to gcc8, I think
we should be covered on all modern distros going forward. So I think we
should revert at least the more complex macros.
--
Regards/Gruss,
Boris.
Good mailing practices for 400: avoid top-posting and trim the reply.
