MIME-Version: 1.0
References: <20230208093322.75816-1-hsiangkao@linux.alibaba.com>
 <Y/ewpGQkpWvOf7qh@gmail.com> <ca1e604a-92ba-023b-8896-dcad9413081d@linux.alibaba.com>
 <8e067230-ce1b-1c75-0c23-87b926357f96@linux.alibaba.com> <CAB=BE-SQZA7gETEvxnHmy0FDQ182fUSRoa0bJBNouN33SFx3hQ@mail.gmail.com>
In-Reply-To: <CAB=BE-SQZA7gETEvxnHmy0FDQ182fUSRoa0bJBNouN33SFx3hQ@mail.gmail.com>
From:   Sandeep Dhavale <dhavale@google.com>
Date:   Mon, 27 Feb 2023 20:47:48 -0800
Message-ID: <CAB=BE-Svf7TMPs-eA+sVuGtYjVWfKd1Nd_AkA9im4Op7TCLW3g@mail.gmail.com>
Subject: Re: [PATCH v5] erofs: add per-cpu threads for decompression as an option
To:     Gao Xiang <hsiangkao@linux.alibaba.com>
Cc:     Eric Biggers <ebiggers@kernel.org>, linux-erofs@lists.ozlabs.org,
        LKML <linux-kernel@vger.kernel.org>,
        Nathan Huckleberry <nhuck@google.com>,
        Yue Hu <huyue2@coolpad.com>, kernel-team@android.com
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Hi all,
I completed the tests and the results are consistent with
our previous observation. We can see that removing WQ_UNBOUND
helps but the scheduling latency by using high priority per cpu
kthreads is even lower. Below is the table.

|---------------------+-------+-------+------+-------|
| Table               | avg   | med   | min  | max   |
|---------------------+-------+-------+------+-------|
| Default erofs       | 19323 | 19758 | 3986 | 35051 |
|---------------------+-------+-------+------+-------|
| !WQ_UNBOUND         | 11202 | 10798 | 3493 | 19822 |
|---------------------+-------+-------+------+-------|
| hipri pcpu kthreads | 7182  | 7017  | 2463 | 12300 |
|---------------------+-------+-------+------+-------|


Thanks,
Sandeep.