Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <a1815428-36bb-780f-313f-c4dcb22887f3@loongson.cn>
Date:   Wed, 26 Apr 2023 10:23:28 +0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.8.0
Subject: Re: Loongson (and other $ARCHs?) idle VS timer enqueue
Content-Language: en-US
To:     WANG Xuerui <kernel@xen0n.name>,
        Peter Zijlstra <peterz@infradead.org>
Cc:     Frederic Weisbecker <frederic@kernel.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Huacai Chen <chenhuacai@kernel.org>,
        "Rafael J. Wysocki" <rafael.j.wysocki@intel.com>,
        Anna-Maria Behnsen <anna-maria@linutronix.de>,
        LKML <linux-kernel@vger.kernel.org>
References: <ZEKDZEQmKExv0O7Q@lothringen> <87leil2r7v.ffs@tglx>
 <20230422081700.GB1214746@hirez.programming.kicks-ass.net>
 <ZEPteS82TbIhMQxe@lothringen>
 <20230422150409.GL1214746@hirez.programming.kicks-ass.net>
 <7d91fa2a-57c5-6c78-8e2d-7fbdd6a11cba@loongson.cn>
 <20230425114937.GC1335080@hirez.programming.kicks-ass.net>
 <5ba79220-683f-a78a-8c3b-bc0b118226f8@loongson.cn>
 <d9c9b2cb-dd5e-03dd-9a7e-27938af96aaf@xen0n.name>
 <f0f029c5-b095-d713-114f-dc7d5c092d8a@loongson.cn>
 <b882eb61-d783-985e-7687-386f9e03235d@xen0n.name>
From:   maobibo <maobibo@loongson.cn>
In-Reply-To: <b882eb61-d783-985e-7687-386f9e03235d@xen0n.name>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Precedence: bulk


在 2023/4/26 10:10, WANG Xuerui 写道:
> On 2023/4/26 08:46, maobibo wrote:
>>
>>
>> 在 2023/4/25 21:28, WANG Xuerui 写道:
>>> On 2023/4/25 21:25, maobibo wrote:
>>>>
>>>>
>>>> 在 2023/4/25 19:49, Peter Zijlstra 写道:
>>>
>>> <snip>
>>>
>>>>>
>>>>> diff --git a/arch/loongarch/kernel/genex.S b/arch/loongarch/kernel/genex.S
>>>>> index 44ff1ff64260..3c8a6bab98fe 100644
>>>>> --- a/arch/loongarch/kernel/genex.S
>>>>> +++ b/arch/loongarch/kernel/genex.S
>>>>> @@ -18,27 +18,31 @@
>>>>>          .align    5
>>>>>    SYM_FUNC_START(__arch_cpu_idle)
>>>>> -    /* start of rollback region */
>>>>> -    LONG_L    t0, tp, TI_FLAGS
>>>>> -    nop
>>>>> -    andi    t0, t0, _TIF_NEED_RESCHED
>>>>> -    bnez    t0, 1f
>>>>> -    nop
>>>>> -    nop
>>>>> -    nop
>>>>> +    /* start of idle interrupt region */
>>>>> +    move    t0, CSR_CRMD_IE
>>>> addi.d  t0, zero, CSR_CRMD_IE can be used here, move is used for reg to reg
>>>
>>> Or better: li.d t0, CSR_CRMD_IE (prefer pseudo-instruction over concrete ones whenever it helps readability). We don't need to support ancient in-house toolchains without support for even li. ;-)
>>    I am not familiar with compiler:(, how many actual instructions does
>> pseudo-instr li.d takes? It will be ok if it uses only one intr, else
>> there will be problem.
> 
> It's just `ori $t0, $zero, 4` no matter which of li.w or li.d is used. It only matters when the immediate to load is bigger. Given CSR_CRMD_IE is just 4 (1<<2) you can definitely say `li.w` if you want to be extra cautious and it won't hurt. ;-)

That is good to me, you are compiler expert:)

>