Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [RFC PATCH v3 2/2] arm64: tlb: Use the TLBI RANGE feature in
 arm64
To:     Catalin Marinas <catalin.marinas@arm.com>
CC:     <linux-arch@vger.kernel.org>, <suzuki.poulose@arm.com>,
        <maz@kernel.org>, <linux-kernel@vger.kernel.org>,
        <xiexiangyou@huawei.com>, <steven.price@arm.com>,
        <zhangshaokun@hisilicon.com>, <linux-mm@kvack.org>,
        <arm@kernel.org>, <prime.zeng@hisilicon.com>,
        <guohanjun@huawei.com>, <olof@lixom.net>,
        <kuhn.chenqun@huawei.com>, <will@kernel.org>,
        <linux-arm-kernel@lists.infradead.org>
References: <20200414112835.1121-1-yezhenyu2@huawei.com>
 <20200414112835.1121-3-yezhenyu2@huawei.com> <20200514152840.GC1907@gaia>
 <54468aae-dbb1-66bd-c633-82fc75936206@huawei.com>
 <20200520170759.GE18302@gaia>
From:   Zhenyu Ye <yezhenyu2@huawei.com>
Message-ID: <a159dd5a-7189-308f-9ebc-98f0eadca852@huawei.com>
Date:   Mon, 1 Jun 2020 22:57:35 +0800
User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:68.0) Gecko/20100101
 Thunderbird/68.3.0
MIME-Version: 1.0
In-Reply-To: <20200520170759.GE18302@gaia>
Content-Type: text/plain; charset="gbk"
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi Catalin,

I have sent the v4 of this series [1] and combine the two function with
a single loop.  See codes for details.

[1] https://lore.kernel.org/linux-arm-kernel/20200601144713.2222-1-yezhenyu2@huawei.com/

On 2020/5/21 1:08, Catalin Marinas wrote:
>> This optimization is only effective when the range is a multiple of 256KB
>> (when the page size is 4KB), and I'm worried about the performance
>> of ilog2().  I traced the __flush_tlb_range() last year and found that in
>> most cases the range is less than 256K (see details in [1]).
> 
> THP or hugetlbfs would exercise bigger strides but I guess it depends on
> the use-case. ilog2() should be reduced to a few instructions on arm64
> AFAICT (haven't tried but it should use the CLZ instruction).
> 

Not bigger than 256K, but the range must be a integer multiple of 256KB,
so I still start from scale 0.

Thanks,
Zhenyu