[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <fef7c633-5577-4cdf-803a-a1fe10787186@app.fastmail.com>
Date: Tue, 21 Jan 2025 08:41:38 +0100
From: "Arnd Bergmann" <arnd@...db.de>
To: "Masahiro Yamada" <masahiroy@...nel.org>,
"Arnd Bergmann" <arnd@...nel.org>
Cc: linux-kbuild@...r.kernel.org, linux-kernel@...r.kernel.org,
regressions@...ts.linux.dev, "Han Shen" <shenhan@...gle.com>,
"Nathan Chancellor" <nathan@...nel.org>, "Kees Cook" <kees@...nel.org>,
"Rong Xu" <xur@...gle.com>, "Jann Horn" <jannh@...gle.com>,
"Ard Biesheuvel" <ardb@...nel.org>, Linux-Arch <linux-arch@...r.kernel.org>
Subject: Re: [PATCH] [RFC, DO NOT APPLY] vmlinux.lds: revert link speed regression
On Tue, Jan 21, 2025, at 01:19, Masahiro Yamada wrote:
> On Tue, Jan 21, 2025 at 6:29 AM Arnd Bergmann <arnd@...nel.org> wrote:
>>
>> linux-6.12 linux-6.13
>> ld.lld v20 1.2s 1.2s
>> ld.bfd v2.36 3.2s 5.2s
>> ld.bfd v2.39 59s 388s
>>
>
> Is this problem specific to the BFD linker from binutils?
I only tried the bfd and lld linkers, but I assume it's limited
to the bfd one.
> Did you observe a link speed regression with LLVM=1 build?
No, the ld.lld line above is what I see with LLVM=1, it's
very fast (1.2s) both before and after the change. New
ld.bfd versions were much slower before the regression
for this particular config and got even slower (seven minutes
for each of the three vmlinux link stages).
Arnd
Powered by blists - more mailing lists