Messages in this thread | | | Date | Tue, 02 Jun 2015 22:25:03 +0200 | From | Denys Vlasenko <> | Subject | Re: [PATCH 1/2] x86/asm/entry/32: Open-code CLEAR_RREGS. No code changes. |
| |
On 06/02/2015 09:34 PM, Steven Rostedt wrote: > On Tue, 2 Jun 2015 21:04:01 +0200 > Denys Vlasenko <dvlasenk@redhat.com> wrote: > >> This macro is small, has only four callsites, and one of them is >> slightly different using a conditional parameter. >> >> A few saved lines aren't worth the resulting obfuscation. > > I'm curious, why? Did someone recommend this change?
I'm proposing to do this. Of course, I don't expect that any my patch must be accepted.
> I don't see it as obfuscation at all.
Riddle me this, looking at the current code. What's up with that strange manipulations of %r9 register? Why the code in SYSCALL and SYSENTER entry points is not the same, why "r9 dance" is done only in one of these entry points?
| |