[v2] staging: most: dim2: read done_buffers count locally from HDM channel

Message ID Y05TNQBXLMJMgQ2r@debian-BULLSEYE-live-builder-AMD64
State New
Headers
Series [v2] staging: most: dim2: read done_buffers count locally from HDM channel |

Commit Message

Deepak R Varma Oct. 18, 2022, 7:18 a.m. UTC
  The done_buffer count is already available in the hdm_channel struct.
Calling dim_get_channel_state function to source this value out of
the same structure is unnecessary.
Further, the second parameter struct dim_ch_state_t to this function
is filled by using the hdm_channel inside the function. This filled in
variable is never used in the caller and can be altogether removed.
So, a call to dim_get_channel_state function in this context also
deems expensive.

Signed-off-by: Deepak R Varma <drv@mailo.com>
---
Changes in v2:
   1. Update patch log message to be more descriptive about the reason for change.
      Feedback provided by julia.lawall@inria.fr

PLEASE NOTE: I have only built the module on my machine, but have not tested it.
I am not sure how to test this change. I am willing to test it with appropriate
guidance provided I have the necessary hardware.


 drivers/staging/most/dim2/dim2.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

--
2.30.2
  

Comments

Julia Lawall Oct. 18, 2022, 7:39 a.m. UTC | #1
On Tue, 18 Oct 2022, Deepak R Varma wrote:

> The done_buffer count is already available in the hdm_channel struct.
> Calling dim_get_channel_state function to source this value out of
> the same structure is unnecessary.
> Further, the second parameter struct dim_ch_state_t to this function
> is filled by using the hdm_channel inside the function. This filled in
> variable is never used in the caller and can be altogether removed.
> So, a call to dim_get_channel_state function in this context also
> deems expensive.

Thanks for the rewrite.

I find "source this value out of" hard to understand.

I would have written something like the following:

The function dim_get_channel_state only serves to initialize the ready and
done_buffers fields of the structure passed as its second argument.  In
service_done_flag, this structure is never used again and the only purpose
of the call is to get the value that is put in the done_buffers field.
But that value is just the done_sw_buffers_number field of the call's
first argument.  So the whole call is useless, and we can just replace it
with an access to this field.

This change implies that the variable st is no longer used, so drop it as
well.

---

If you want to work some more in this code, the name of the type
dim_ch_state_t should not have an _t at the end of it.  It looks like it
used to be a typedef and someone eliminated the typedef but didn't drop
the _t.

julia


>
> Signed-off-by: Deepak R Varma <drv@mailo.com>
> ---
> Changes in v2:
>    1. Update patch log message to be more descriptive about the reason for change.
>       Feedback provided by julia.lawall@inria.fr
>
> PLEASE NOTE: I have only built the module on my machine, but have not tested it.
> I am not sure how to test this change. I am willing to test it with appropriate
> guidance provided I have the necessary hardware.
>
>
>  drivers/staging/most/dim2/dim2.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
>
> diff --git a/drivers/staging/most/dim2/dim2.c b/drivers/staging/most/dim2/dim2.c
> index ab72e11ac5ab..4c1f27898a29 100644
> --- a/drivers/staging/most/dim2/dim2.c
> +++ b/drivers/staging/most/dim2/dim2.c
> @@ -259,7 +259,6 @@ static void retrieve_netinfo(struct dim2_hdm *dev, struct mbo *mbo)
>  static void service_done_flag(struct dim2_hdm *dev, int ch_idx)
>  {
>  	struct hdm_channel *hdm_ch = dev->hch + ch_idx;
> -	struct dim_ch_state_t st;
>  	struct list_head *head;
>  	struct mbo *mbo;
>  	int done_buffers;
> @@ -271,7 +270,7 @@ static void service_done_flag(struct dim2_hdm *dev, int ch_idx)
>
>  	spin_lock_irqsave(&dim_lock, flags);
>
> -	done_buffers = dim_get_channel_state(&hdm_ch->ch, &st)->done_buffers;
> +	done_buffers = hdm_ch->ch.done_sw_buffers_number;
>  	if (!done_buffers) {
>  		spin_unlock_irqrestore(&dim_lock, flags);
>  		return;
> --
> 2.30.2
>
>
>
>
>
  
Deepak R Varma Oct. 18, 2022, 12:49 p.m. UTC | #2
On Tue, Oct 18, 2022 at 09:39:08AM +0200, Julia Lawall wrote:
>
>
> On Tue, 18 Oct 2022, Deepak R Varma wrote:
>
> > The done_buffer count is already available in the hdm_channel struct.
> > Calling dim_get_channel_state function to source this value out of
> > the same structure is unnecessary.
> > Further, the second parameter struct dim_ch_state_t to this function
> > is filled by using the hdm_channel inside the function. This filled in
> > variable is never used in the caller and can be altogether removed.
> > So, a call to dim_get_channel_state function in this context also
> > deems expensive.
>
> Thanks for the rewrite.
>
> I find "source this value out of" hard to understand.
>
> I would have written something like the following:
>
> The function dim_get_channel_state only serves to initialize the ready and
> done_buffers fields of the structure passed as its second argument.  In
> service_done_flag, this structure is never used again and the only purpose
> of the call is to get the value that is put in the done_buffers field.
> But that value is just the done_sw_buffers_number field of the call's
> first argument.  So the whole call is useless, and we can just replace it
> with an access to this field.
>
> This change implies that the variable st is no longer used, so drop it as
> well.

This is really well written. Sounds much structured. Now my own log message
sounds a little random :)

Is it okay for me to use your verbiage as is in my patch log?

Thank you,
./drv


>
> ---
>
> If you want to work some more in this code, the name of the type
> dim_ch_state_t should not have an _t at the end of it.  It looks like it
> used to be a typedef and someone eliminated the typedef but didn't drop
> the _t.
>
> julia
>
>
> >
> > Signed-off-by: Deepak R Varma <drv@mailo.com>
> > ---
> > Changes in v2:
> >    1. Update patch log message to be more descriptive about the reason for change.
> >       Feedback provided by julia.lawall@inria.fr
> >
> > PLEASE NOTE: I have only built the module on my machine, but have not tested it.
> > I am not sure how to test this change. I am willing to test it with appropriate
> > guidance provided I have the necessary hardware.
> >
> >
> >  drivers/staging/most/dim2/dim2.c | 3 +--
> >  1 file changed, 1 insertion(+), 2 deletions(-)
> >
> > diff --git a/drivers/staging/most/dim2/dim2.c b/drivers/staging/most/dim2/dim2.c
> > index ab72e11ac5ab..4c1f27898a29 100644
> > --- a/drivers/staging/most/dim2/dim2.c
> > +++ b/drivers/staging/most/dim2/dim2.c
> > @@ -259,7 +259,6 @@ static void retrieve_netinfo(struct dim2_hdm *dev, struct mbo *mbo)
> >  static void service_done_flag(struct dim2_hdm *dev, int ch_idx)
> >  {
> >  	struct hdm_channel *hdm_ch = dev->hch + ch_idx;
> > -	struct dim_ch_state_t st;
> >  	struct list_head *head;
> >  	struct mbo *mbo;
> >  	int done_buffers;
> > @@ -271,7 +270,7 @@ static void service_done_flag(struct dim2_hdm *dev, int ch_idx)
> >
> >  	spin_lock_irqsave(&dim_lock, flags);
> >
> > -	done_buffers = dim_get_channel_state(&hdm_ch->ch, &st)->done_buffers;
> > +	done_buffers = hdm_ch->ch.done_sw_buffers_number;
> >  	if (!done_buffers) {
> >  		spin_unlock_irqrestore(&dim_lock, flags);
> >  		return;
> > --
> > 2.30.2
> >
> >
> >
> >
> >
  
Julia Lawall Oct. 18, 2022, 12:52 p.m. UTC | #3
On Tue, 18 Oct 2022, Deepak R Varma wrote:

> On Tue, Oct 18, 2022 at 09:39:08AM +0200, Julia Lawall wrote:
> >
> >
> > On Tue, 18 Oct 2022, Deepak R Varma wrote:
> >
> > > The done_buffer count is already available in the hdm_channel struct.
> > > Calling dim_get_channel_state function to source this value out of
> > > the same structure is unnecessary.
> > > Further, the second parameter struct dim_ch_state_t to this function
> > > is filled by using the hdm_channel inside the function. This filled in
> > > variable is never used in the caller and can be altogether removed.
> > > So, a call to dim_get_channel_state function in this context also
> > > deems expensive.
> >
> > Thanks for the rewrite.
> >
> > I find "source this value out of" hard to understand.
> >
> > I would have written something like the following:
> >
> > The function dim_get_channel_state only serves to initialize the ready and
> > done_buffers fields of the structure passed as its second argument.  In
> > service_done_flag, this structure is never used again and the only purpose
> > of the call is to get the value that is put in the done_buffers field.
> > But that value is just the done_sw_buffers_number field of the call's
> > first argument.  So the whole call is useless, and we can just replace it
> > with an access to this field.
> >
> > This change implies that the variable st is no longer used, so drop it as
> > well.
>
> This is really well written. Sounds much structured. Now my own log message
> sounds a little random :)
>
> Is it okay for me to use your verbiage as is in my patch log?

Yes.

julia
  
Deepak R Varma Oct. 18, 2022, 1:48 p.m. UTC | #4
On Tue, Oct 18, 2022 at 02:52:21PM +0200, Julia Lawall wrote:
>
>
> On Tue, 18 Oct 2022, Deepak R Varma wrote:
>
> > On Tue, Oct 18, 2022 at 09:39:08AM +0200, Julia Lawall wrote:
> > >
> > >
> > > On Tue, 18 Oct 2022, Deepak R Varma wrote:
> > >
> > > > The done_buffer count is already available in the hdm_channel struct.
> > > > Calling dim_get_channel_state function to source this value out of
> > > > the same structure is unnecessary.
> > > > Further, the second parameter struct dim_ch_state_t to this function
> > > > is filled by using the hdm_channel inside the function. This filled in
> > > > variable is never used in the caller and can be altogether removed.
> > > > So, a call to dim_get_channel_state function in this context also
> > > > deems expensive.
> > >
> > > Thanks for the rewrite.
> > >
> > > I find "source this value out of" hard to understand.
> > >
> > > I would have written something like the following:
> > >
> > > The function dim_get_channel_state only serves to initialize the ready and
> > > done_buffers fields of the structure passed as its second argument.  In
> > > service_done_flag, this structure is never used again and the only purpose
> > > of the call is to get the value that is put in the done_buffers field.
> > > But that value is just the done_sw_buffers_number field of the call's
> > > first argument.  So the whole call is useless, and we can just replace it
> > > with an access to this field.
> > >
> > > This change implies that the variable st is no longer used, so drop it as
> > > well.
> >
> > This is really well written. Sounds much structured. Now my own log message
> > sounds a little random :)
> >
> > Is it okay for me to use your verbiage as is in my patch log?
>
> Yes.

Thank you. Can I convert this into a patch set and include the other suggestion
from you to correct the dim_ch_state_t struct name? Or should these be separate
patches now?

./drv



>
> julia
>
  
Julia Lawall Oct. 18, 2022, 2:19 p.m. UTC | #5
On Tue, 18 Oct 2022, Deepak R Varma wrote:

> On Tue, Oct 18, 2022 at 02:52:21PM +0200, Julia Lawall wrote:
> >
> >
> > On Tue, 18 Oct 2022, Deepak R Varma wrote:
> >
> > > On Tue, Oct 18, 2022 at 09:39:08AM +0200, Julia Lawall wrote:
> > > >
> > > >
> > > > On Tue, 18 Oct 2022, Deepak R Varma wrote:
> > > >
> > > > > The done_buffer count is already available in the hdm_channel struct.
> > > > > Calling dim_get_channel_state function to source this value out of
> > > > > the same structure is unnecessary.
> > > > > Further, the second parameter struct dim_ch_state_t to this function
> > > > > is filled by using the hdm_channel inside the function. This filled in
> > > > > variable is never used in the caller and can be altogether removed.
> > > > > So, a call to dim_get_channel_state function in this context also
> > > > > deems expensive.
> > > >
> > > > Thanks for the rewrite.
> > > >
> > > > I find "source this value out of" hard to understand.
> > > >
> > > > I would have written something like the following:
> > > >
> > > > The function dim_get_channel_state only serves to initialize the ready and
> > > > done_buffers fields of the structure passed as its second argument.  In
> > > > service_done_flag, this structure is never used again and the only purpose
> > > > of the call is to get the value that is put in the done_buffers field.
> > > > But that value is just the done_sw_buffers_number field of the call's
> > > > first argument.  So the whole call is useless, and we can just replace it
> > > > with an access to this field.
> > > >
> > > > This change implies that the variable st is no longer used, so drop it as
> > > > well.
> > >
> > > This is really well written. Sounds much structured. Now my own log message
> > > sounds a little random :)
> > >
> > > Is it okay for me to use your verbiage as is in my patch log?
> >
> > Yes.
>
> Thank you. Can I convert this into a patch set and include the other suggestion
> from you to correct the dim_ch_state_t struct name? Or should these be separate
> patches now?

You can make a patch set.

julia
  

Patch

diff --git a/drivers/staging/most/dim2/dim2.c b/drivers/staging/most/dim2/dim2.c
index ab72e11ac5ab..4c1f27898a29 100644
--- a/drivers/staging/most/dim2/dim2.c
+++ b/drivers/staging/most/dim2/dim2.c
@@ -259,7 +259,6 @@  static void retrieve_netinfo(struct dim2_hdm *dev, struct mbo *mbo)
 static void service_done_flag(struct dim2_hdm *dev, int ch_idx)
 {
 	struct hdm_channel *hdm_ch = dev->hch + ch_idx;
-	struct dim_ch_state_t st;
 	struct list_head *head;
 	struct mbo *mbo;
 	int done_buffers;
@@ -271,7 +270,7 @@  static void service_done_flag(struct dim2_hdm *dev, int ch_idx)

 	spin_lock_irqsave(&dim_lock, flags);

-	done_buffers = dim_get_channel_state(&hdm_ch->ch, &st)->done_buffers;
+	done_buffers = hdm_ch->ch.done_sw_buffers_number;
 	if (!done_buffers) {
 		spin_unlock_irqrestore(&dim_lock, flags);
 		return;