[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAEEQ3w=mXXTwxph+_ttBLQhvNb66bq9b6Wo-Ka35qpp+3iq2JQ@mail.gmail.com>
Date: Mon, 25 Mar 2024 11:59:43 +0800
From: yunhui cui <cuiyunhui@...edance.com>
To: Alexandre Ghiti <alexghiti@...osinc.com>
Cc: Qingfang Deng <dqfext@...il.com>, aou@...s.berkeley.edu, linux-kernel@...r.kernel.org,
linux-riscv@...ts.infradead.org, palmer@...belt.com, paul.walmsley@...ive.com,
samuel.holland@...ive.com
Subject: Re: [External] Re: [PATCH] riscv: Improve sbi_ecall() code generation
by reordering arguments
Hi Alex,
On Mon, Mar 25, 2024 at 2:20 AM Alexandre Ghiti <alexghiti@...osinccom> wrote:
>
> Hi Qingfang,
>
> On Sat, Mar 23, 2024 at 7:26 AM Qingfang Deng <dqfext@...il.com> wrote:
> >
> > Hi Alexandre,
> >
> > You can simply make sbi_ecall `__always_inline`, so the C function call
> > overhead can be fully avoided.
>
> I understand your point, though I don't think we need to mark
> sbi_ecall() as inline, it's not in any hot path so it's not worth it.
> This patch simply gets rid of a really useless overhead, but it does
> not visibly accelerate anything.
>
> I hope it makes sense,
>
> Thanks,
>
> Alex
>
The compiler will have inline size restrictions, so I agree with the
modification of this patch instead of inline.
Please refer to:
https://github.com/gcc-mirror/gcc/blob/master/gcc/common.opt,
"finline-limit-"
So:
Reviewed-by: Yunhui Cui <cuiyunhui@...edance.com>
Thanks,
Yunhui
Powered by blists - more mailing lists