MIME-Version: 1.0
Date:   Wed, 01 Dec 2021 07:56:29 +0000
From:   Marc Zyngier <maz@kernel.org>
To:     Vladimir Murzin <vladimir.murzin@arm.com>
Cc:     Mark Rutland <mark.rutland@arm.com>, linux-kernel@vger.kernel.org,
        aou@eecs.berkeley.edu, catalin.marinas@arm.com,
        deanbo422@gmail.com, green.hu@gmail.com, guoren@kernel.org,
        jonas@southpole.se, kernelfans@gmail.com,
        linux-arm-kernel@lists.infradead.org, linux@armlinux.org.uk,
        nickhu@andestech.com, palmer@dabbelt.com, paulmck@kernel.org,
        paul.walmsley@sifive.com, peterz@infradead.org, shorne@gmail.com,
        stefan.kristiansson@saunalahti.fi, tglx@linutronix.de,
        torvalds@linux-foundation.org, tsbogend@alpha.franken.de,
        vgupta@kernel.org, will@kernel.org
Subject: Re: [PATCH 09/15] irq: arm: perform irqentry in entry code
In-Reply-To: <59cb8da0-7dbd-6e70-56be-a40615fd694b@arm.com>
References: <20211021180236.37428-1-mark.rutland@arm.com>
 <20211021180236.37428-10-mark.rutland@arm.com>
 <0efc4465-12b5-a568-0228-c744ec0509a3@arm.com>
 <20211022153602.GE86184@C02TD0UTHF1T.local>
 <1dc39ac9-1a05-cf8d-2aef-633903a6338d@arm.com>
 <20211022175854.GK86184@C02TD0UTHF1T.local> <87tuh8uchn.wl-maz@kernel.org>
 <cada0034-7427-f4ae-0f88-7bbb1be2b1a5@arm.com>
 <87ilxnc21r.wl-maz@kernel.org>
 <c43c8c0e-946e-93c6-00d0-f015f948833a@arm.com>
 <59cb8da0-7dbd-6e70-56be-a40615fd694b@arm.com>
User-Agent: Roundcube Webmail/1.4.11
Message-ID: <78cc83d10cb6d251e53f5947ddc52735@kernel.org>
Content-Type: text/plain; charset=US-ASCII;
 format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 2021-11-30 08:49, Vladimir Murzin wrote:

> One month and a week later...
> 
> I observe that in terms of performance
> 
>    MRS       r0, ipsr
> 
> is equivalent to readl_relaxed(BASEADDR_V7M_SCB + V7M_SCB_ICSR)
> 
>    MOV.W   r3, #3758153728
>    LDR.W   r0, [r3, #3332]
> 
> Old compilers can produce less performant sequence like
> 
>    LDR      r3,0xbcc0
>    ADD.W    r3,r3,#0xaf00
>    LDR      r0,[r3,#0]
> 
> So, what would be your preference?

If there is no significant overhead to reading the MMIO register
and that you see a benefit in enabling COMPILE_TEST, then this
probably is the way to go.

Thanks,

         M.
-- 
Jazz is not dead. It just smells funny...