MIME-Version: 1.0
References: <1637104560-37432-1-git-send-email-quic_c_gdjako@quicinc.com> <1637104560-37432-2-git-send-email-quic_c_gdjako@quicinc.com>
In-Reply-To: <1637104560-37432-2-git-send-email-quic_c_gdjako@quicinc.com>
From:   Arnd Bergmann <arnd@arndb.de>
Date:   Wed, 17 Nov 2021 08:37:58 +0100
Message-ID: <CAK8P3a3W13wb5B_FMSCEp0JTUFrPk=Fp8N59oa58wxn+WeTroA@mail.gmail.com>
Subject: Re: [RFC 2/2] drm/msm/gem: Make use of the system cache
To:     Georgi Djakov <quic_c_gdjako@quicinc.com>
Cc:     Catalin Marinas <catalin.marinas@arm.com>,
        Will Deacon <will@kernel.org>, Rob Clark <robdclark@gmail.com>,
        Sean Paul <sean@poorly.run>,
        Robin Murphy <robin.murphy@arm.com>,
        Marc Zyngier <maz@kernel.org>, Arnd Bergmann <arnd@arndb.de>,
        Linux ARM <linux-arm-kernel@lists.infradead.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        "open list:IOMMU DRIVERS" <iommu@lists.linux-foundation.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, Nov 17, 2021 at 12:16 AM Georgi Djakov
<quic_c_gdjako@quicinc.com> wrote:
>
> Instead of writing to WC cmdstream buffers that go all the way to the main
> memory, let's use the system cache to improve the performance.
>
> Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
> ---
>  drivers/gpu/drm/msm/msm_gem.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/msm/msm_gem.c b/drivers/gpu/drm/msm/msm_gem.c
> index 104fdfc14027..921a1c24721e 100644
> --- a/drivers/gpu/drm/msm/msm_gem.c
> +++ b/drivers/gpu/drm/msm/msm_gem.c
> @@ -214,7 +214,7 @@ void msm_gem_put_pages(struct drm_gem_object *obj)
>  static pgprot_t msm_gem_pgprot(struct msm_gem_object *msm_obj, pgprot_t prot)
>  {
>         if (msm_obj->flags & (MSM_BO_WC|MSM_BO_UNCACHED))
> -               return pgprot_writecombine(prot);
> +               return pgprot_syscached(prot);
>         return prot;
>  }

Based on the definition in patch 1, doesn't this mean that 32-bit
kernels degrade
from writecombined to uncached, making them a lot slower?

My feeling about this series is that there should be a clearer
definition of what
exactly happens on systems with and without system cache.

         Arnd