Re: [PATCH v3] arm64: lib: accelerate do_csum with NEON instruction

From: "huanglingyan (A)" <huanglingyan2@huawei.com>
To: Will Deacon <will.deacon@arm.com>
Cc: Zhangshaokun <zhangshaokun@hisilicon.com>,
	Catalin Marinas <catalin.marinas@arm.com>,
	linux-arm-kernel@lists.infradead.org, ard.biesheuvel@linaro.org
Subject: Re: [PATCH v3] arm64: lib: accelerate do_csum with NEON instruction
Date: Wed, 9 Jan 2019 10:03:05 +0800
Message-ID: <cd5bb83e-bb0e-b348-5365-095c5fcd9648@huawei.com> (raw)
In-Reply-To: <20190108135444.GB14476@fuggles.cambridge.arm.com>

On 2019/1/8 21:54, Will Deacon wrote:
> [re-adding Ard and LAKML -- not sure why the headers are so munged]
>
> On Mon, Jan 07, 2019 at 10:38:55AM +0800, huanglingyan (A) wrote:
>> On 2019/1/6 16:26, Ard Biesheuvel wrote:
>>     Please change this into
>>
>>     if (IS_ENABLED(CONFIG_KERNEL_MODE_NEON) &&
>>         len >= CSUM_NEON_THRESHOLD &&
>>         may_use_simd()) {
>>             kernel_neon_begin();
>>             res = do_csum_neon(buff, len);
>>             kernel_neon_end();
>>         }
>>
>>     and drop the intermediate do_csum_arm()
>>
>>
>>         +               return do_csum_arm(buff, len);
>>         +#endif  /* CONFIG_KERNEL_MODE_NEON */
>>
>>     No else? What happens if len < CSUM_NEON_THRESHOLD ?
>>
>>
>>         +#undef do_csum
>>
>>     Can we drop this?
>>
>> Using NEON instructions will bring some costs. The spending maybe introduced
>> when reservering/restoring
>> neon registers with kernel_neon_begin()/kernel_neon_end(). Therefore NEON code
>> is Only used when
>> the length exceeds CSUM_NEON_THRESHOLD. General do csum() codes in lib/
>> checksum.c will be used in
>> shorter length. To achieve this goal, I use the "#undef do_csum" in else clause
>> to have the oppotunity to
>> utilize the general codes.
> I don't think that's how it works :/
>
> Before we get deeper into the implementation, please could you justify the
> need for a CPU-optimised checksum implementation at all? I thought this was
> usually offloaded to the NIC?
>
> Will
>
> .
This problem is introduced when testing Intel x710 network card on my ARM server.
Ip forward is set for ease of testing. Then send lots of packages to server by Tesgine
machine and then receive.

The bandwidth in Intel 8180 is 9.5 Gbps while only 5.8 Gbps in ARM. It shows that
do_csum() costs 36% in ARM and only 6% in Intel with perf tools. That's why I decide
to modify do_csum() function in ARM.

As a newbee of linux kernel, I has little knowledge of the implementation of such
situation. Looking forward to get your help of improving this patch.

Lingyan Huang
.

**************************

**************************
>

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel