[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <d9584a56-41a5-3e52-55d0-a02c5d59b823@linaro.org>
Date: Wed, 17 Jul 2019 18:43:31 +0300
From: Stanimir Varbanov <stanimir.varbanov@...aro.org>
To: Aniket Masule <amasule@...eaurora.org>, linux-media@...r.kernel.org
Cc: linux-kernel@...r.kernel.org, linux-arm-msm@...r.kernel.org,
vgarodia@...eaurora.org
Subject: Re: [PATCH v5 4/4] media: venus: Update core selection
Hi,
On 7/16/19 8:09 AM, Aniket Masule wrote:
> Present core assignment is static. Introduced load balancing
> across the cores. Load on earch core is calculated and core
> with minimum load is assigned to given instance.
>
> Signed-off-by: Aniket Masule <amasule@...eaurora.org>
> ---
> drivers/media/platform/qcom/venus/helpers.c | 69 +++++++++++++++++++++++---
> drivers/media/platform/qcom/venus/helpers.h | 2 +-
> drivers/media/platform/qcom/venus/hfi_helper.h | 1 +
> drivers/media/platform/qcom/venus/hfi_parser.h | 5 ++
> drivers/media/platform/qcom/venus/vdec.c | 2 +-
> drivers/media/platform/qcom/venus/venc.c | 2 +-
> 6 files changed, 72 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/media/platform/qcom/venus/helpers.c b/drivers/media/platform/qcom/venus/helpers.c
> index edf403d..d479793 100644
> --- a/drivers/media/platform/qcom/venus/helpers.c
> +++ b/drivers/media/platform/qcom/venus/helpers.c
> @@ -26,6 +26,7 @@
> #include "helpers.h"
> #include "hfi_helper.h"
> #include "hfi_venus_io.h"
> +#include "hfi_parser.h"
>
> struct intbuf {
> struct list_head list;
> @@ -331,6 +332,24 @@ static u32 load_per_instance(struct venus_inst *inst)
> return mbs * inst->fps;
> }
>
> +static u32 load_per_core(struct venus_core *core, u32 core_id)
> +{
> + struct venus_inst *inst = NULL;
> + u32 mbs_per_sec = 0, load = 0;
> +
> + mutex_lock(&core->lock);
> + list_for_each_entry(inst, &core->instances, list) {
> + if (inst->clk_data.core_id != core_id)
> + continue;
> +
> + mbs_per_sec = load_per_instance(inst);
> + load = mbs_per_sec * inst->clk_data.codec_freq_data->vpp_freq;
shouldn't this be load += mbs_per_sec * vpp_freq ? We need to calculate
load for every instance on this core_id.
> + }
> + mutex_unlock(&core->lock);
> +
> + return load;
> +}
> +
<cut>
--
regards,
Stan
Powered by blists - more mailing lists