Re: [PATCH] x86: enable dead code and data elimination

Re: [PATCH] x86: enable dead code and data elimination

[Rui Salvaterra]

Hi, Tobias,

On Thu, 1 Jul 2021 at 10:22, <tobias.karnat@remondis.de> wrote:
>
> Hello,
>
> I have found your patch to enable x86 dead code and data elimination on the openwrt-devel mailing list and I wonder if you can sent it upstream to the kernel mailing list?

As I quoted, mine is an adaptation of Nicholas Piggin's original
patch, basically a forward-port to 5.4 (5.10 in my OpenWrt tree [1]).
Since the original proposal has been sent upstream and hasn't been
merged, there are probably good reasons for it (even though it works
perfectly for my use case). Nicholas, any idea on why your original
patch [2] hasn't been merged?

[1] https://github.com/rsalvaterra/openwrt/commit/0082c486b5637dc8e1771e2b9b38b728cb2fa01d
[can be rebased at any time, caveat lector]
[2] https://lore.kernel.org/lkml/20170709031333.29443-1-npiggin@gmail.com/]

> It works for me even combined with the new Clang ThinLTO kernel build. @Sedat you might be interested as well, as I saw your attempt here https://lkml.org/lkml/2021/2/27/181

I haven't tested with Clang, only GCC, but it's nice to know.

Thanks,
Rui

Re: [PATCH] x86: enable dead code and data elimination

[Nicholas Piggin]

Excerpts from Rui Salvaterra's message of July 1, 2021 8:00 pm:
> Hi, Tobias,
> 
> On Thu, 1 Jul 2021 at 10:22, <tobias.karnat@remondis.de> wrote:
>>
>> Hello,
>>
>> I have found your patch to enable x86 dead code and data elimination on the openwrt-devel mailing list and I wonder if you can sent it upstream to the kernel mailing list?
> 
> As I quoted, mine is an adaptation of Nicholas Piggin's original
> patch, basically a forward-port to 5.4 (5.10 in my OpenWrt tree [1]).
> Since the original proposal has been sent upstream and hasn't been
> merged, there are probably good reasons for it (even though it works
> perfectly for my use case). Nicholas, any idea on why your original
> patch [2] hasn't been merged?

Not sure, I didn't do much testing or try too hard to get it upstream. 
If you have found it works and is helpful I don't see why it couldn't
be merged. Feel free to submit it.

Thanks,
Nick