[v1,tty] 8250: microchip: pci1xxxx: Refactor TX Burst code to use pre-existing APIs

  Updated the TX Burst implementation by changing the circular buffer
processing with the pre-existing APIs in kernel. Also updated conditional
statements and alignment issues for better readability.

Signed-off-by: Rengarajan S <rengarajan.s@microchip.com>
---
 drivers/tty/serial/8250/8250_pci1xxxx.c | 39 ++++++++++++-------------
 1 file changed, 19 insertions(+), 20 deletions(-)
  

On Thu, 22 Feb 2024, Rengarajan S wrote:

> Updated the TX Burst implementation by changing the circular buffer
> processing with the pre-existing APIs in kernel. Also updated conditional
> statements and alignment issues for better readability.
> 
> Signed-off-by: Rengarajan S <rengarajan.s@microchip.com>
> ---

> @@ -434,16 +435,7 @@ static void pci1xxxx_tx_burst(struct uart_port *port, u32 uart_status)
>  
>  	xmit = &port->state->xmit;
>  
> -	if (port->x_char) {
> -		writeb(port->x_char, port->membase + UART_TX);
> -		port->icount.tx++;
> -		port->x_char = 0;
> -		return;
> -	}
> -
> -	if ((uart_tx_stopped(port)) || (uart_circ_empty(xmit))) {
> -		port->ops->stop_tx(port);
> -	} else {
> +	if (!(port->x_char)) {
>  		data_empty_count = (pci1xxxx_read_burst_status(port) &
>  				    UART_BST_STAT_TX_COUNT_MASK) >> 8;
>  		do {
> @@ -453,15 +445,22 @@ static void pci1xxxx_tx_burst(struct uart_port *port, u32 uart_status)
>  						    &data_empty_count,
>  						    &valid_byte_count);
>  
> -			port->icount.tx++;
>  			if (uart_circ_empty(xmit))
>  				break;
>  		} while (data_empty_count && valid_byte_count);
> +	} else {
> +		writeb(port->x_char, port->membase + UART_TX);
> +		port->icount.tx++;
> +		port->x_char = 0;
> +		return;

Why you made this reorganization for x_char handling?? It seems 
entirely wrong thing to do, x_char should have precendence over 
sending normal chars.

This patch would have been some much simpler to review if it would have 
not attempted to n things in one go, please try to split into sensible 
changes.
  

On 22. 02. 24, 14:49, Rengarajan S wrote:
> Updated the TX Burst implementation by changing the circular buffer
> processing with the pre-existing APIs in kernel. Also updated conditional
> statements and alignment issues for better readability.

Hi,

so why are you keeping the nested double loop?
  

On Fri, 2024-02-23 at 07:08 +0100, Jiri Slaby wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you
> know the content is safe
> 
> On 22. 02. 24, 14:49, Rengarajan S wrote:
> > Updated the TX Burst implementation by changing the circular buffer
> > processing with the pre-existing APIs in kernel. Also updated
> > conditional
> > statements and alignment issues for better readability.
> 
> Hi,
> 
> so why are you keeping the nested double loop?
> 

Hi, in order to differentiate Burst mode handling with byte mode had
seperate loops for both. Since, having single while loop also does not
align with rx implementation (where we have seperate handling for burst
and byte) have retained the double loop. 

> --
> js
> suse labs
>
  

On Thu, 2024-02-22 at 18:01 +0200, Ilpo Järvinen wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you
> know the content is safe
> 
> On Thu, 22 Feb 2024, Rengarajan S wrote:
> 
> > Updated the TX Burst implementation by changing the circular buffer
> > processing with the pre-existing APIs in kernel. Also updated
> > conditional
> > statements and alignment issues for better readability.
> > 
> > Signed-off-by: Rengarajan S <rengarajan.s@microchip.com>
> > ---
> 
> > @@ -434,16 +435,7 @@ static void pci1xxxx_tx_burst(struct uart_port
> > *port, u32 uart_status)
> > 
> >       xmit = &port->state->xmit;
> > 
> > -     if (port->x_char) {
> > -             writeb(port->x_char, port->membase + UART_TX);
> > -             port->icount.tx++;
> > -             port->x_char = 0;
> > -             return;
> > -     }
> > -
> > -     if ((uart_tx_stopped(port)) || (uart_circ_empty(xmit))) {
> > -             port->ops->stop_tx(port);
> > -     } else {
> > +     if (!(port->x_char)) {
> >               data_empty_count = (pci1xxxx_read_burst_status(port)
> > &
> >                                   UART_BST_STAT_TX_COUNT_MASK) >>
> > 8;
> >               do {
> > @@ -453,15 +445,22 @@ static void pci1xxxx_tx_burst(struct
> > uart_port *port, u32 uart_status)
> >                                                  
> > &data_empty_count,
> >                                                  
> > &valid_byte_count);
> > 
> > -                     port->icount.tx++;
> >                       if (uart_circ_empty(xmit))
> >                               break;
> >               } while (data_empty_count && valid_byte_count);
> > +     } else {
> > +             writeb(port->x_char, port->membase + UART_TX);
> > +             port->icount.tx++;
> > +             port->x_char = 0;
> > +             return;
> 
> Why you made this reorganization for x_char handling?? It seems
> entirely wrong thing to do, x_char should have precendence over
> sending normal chars.
> 
> This patch would have been some much simpler to review if it would
> have
> not attempted to n things in one go, please try to split into
> sensible
> changes.
> 

Hi, Thanks for reviewing the patch. Will address the comments and share
the updated patch shortly.

> 
> --
>  i.
>
  

On 23. 02. 24, 10:21, Rengarajan.S@microchip.com wrote:
> On Fri, 2024-02-23 at 07:08 +0100, Jiri Slaby wrote:
>> EXTERNAL EMAIL: Do not click links or open attachments unless you
>> know the content is safe
>>
>> On 22. 02. 24, 14:49, Rengarajan S wrote:
>>> Updated the TX Burst implementation by changing the circular buffer
>>> processing with the pre-existing APIs in kernel. Also updated
>>> conditional
>>> statements and alignment issues for better readability.
>>
>> Hi,
>>
>> so why are you keeping the nested double loop?
>>
> 
> Hi, in order to differentiate Burst mode handling with byte mode had
> seperate loops for both. Since, having single while loop also does not
> align with rx implementation (where we have seperate handling for burst
> and byte) have retained the double loop.

So obviously, align RX to a single loop if possible. The current TX code 
is very hard to follow and sort of unmaintainable (and buggy). And IMO 
it's unnecessary as I proposed [1]. And even if RX cannot be one loop, 
you still can make TX easy to read as the two need not be the same.

[1] 
https://lore.kernel.org/all/b8325c3f-bf5b-4c55-8dce-ef395edce251@kernel.org/

thanks,
  

On Fri, 2024-02-23 at 10:26 +0100, Jiri Slaby wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you
> know the content is safe
> 
> On 23. 02. 24, 10:21, Rengarajan.S@microchip.com wrote:
> > On Fri, 2024-02-23 at 07:08 +0100, Jiri Slaby wrote:
> > > EXTERNAL EMAIL: Do not click links or open attachments unless you
> > > know the content is safe
> > > 
> > > On 22. 02. 24, 14:49, Rengarajan S wrote:
> > > > Updated the TX Burst implementation by changing the circular
> > > > buffer
> > > > processing with the pre-existing APIs in kernel. Also updated
> > > > conditional
> > > > statements and alignment issues for better readability.
> > > 
> > > Hi,
> > > 
> > > so why are you keeping the nested double loop?
> > > 
> > 
> > Hi, in order to differentiate Burst mode handling with byte mode
> > had
> > seperate loops for both. Since, having single while loop also does
> > not
> > align with rx implementation (where we have seperate handling for
> > burst
> > and byte) have retained the double loop.
> 
> So obviously, align RX to a single loop if possible. The current TX
> code
> is very hard to follow and sort of unmaintainable (and buggy). And
> IMO
> it's unnecessary as I proposed [1]. And even if RX cannot be one
> loop,
> you still can make TX easy to read as the two need not be the same.
> 
> [1]
> https://lore.kernel.org/all/b8325c3f-bf5b-4c55-8dce-ef395edce251@kernel.org/


Sure. Will update the TX implementation as suggested and will send out
the patch shortly.
> 
> thanks,
> --
> js
> suse labs
>
  

Hi Jiri,

On Fri, 2024-02-23 at 10:26 +0100, Jiri Slaby wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you
> know the content is safe
> 
> On 23. 02. 24, 10:21, Rengarajan.S@microchip.com wrote:
> > On Fri, 2024-02-23 at 07:08 +0100, Jiri Slaby wrote:
> > > EXTERNAL EMAIL: Do not click links or open attachments unless you
> > > know the content is safe
> > > 
> > > On 22. 02. 24, 14:49, Rengarajan S wrote:
> > > > Updated the TX Burst implementation by changing the circular
> > > > buffer
> > > > processing with the pre-existing APIs in kernel. Also updated
> > > > conditional
> > > > statements and alignment issues for better readability.
> > > 
> > > Hi,
> > > 
> > > so why are you keeping the nested double loop?
> > > 
> > 
> > Hi, in order to differentiate Burst mode handling with byte mode
> > had
> > seperate loops for both. Since, having single while loop also does
> > not
> > align with rx implementation (where we have seperate handling for
> > burst
> > and byte) have retained the double loop.
> 
> So obviously, align RX to a single loop if possible. The current TX
> code
> is very hard to follow and sort of unmaintainable (and buggy). And
> IMO
> it's unnecessary as I proposed [1]. And even if RX cannot be one
> loop,
> you still can make TX easy to read as the two need not be the same.
> 
> [1]
> https://lore.kernel.org/all/b8325c3f-bf5b-4c55-8dce-ef395edce251@kernel.org/


while (data_empty_count) {
   cnt = CIRC_CNT_TO_END();
   if (!cnt)
     break;
   if (cnt < UART_BURST_SIZE || (tail & 3)) { // is_unaligned()
     writeb();
     cnt = 1;
   } else {
     writel()
     cnt = UART_BURST_SIZE;
   }
   uart_xmit_advance(cnt);
   data_empty_count -= cnt;
}

With the above implementation we are observing performance drop of 2
Mbps at baud rate of 4 Mbps. The reason for this is the fact that for
each iteration we are checking if the the data need to be processed via
DWORDs or Bytes. The condition check for each iteration is causing the
drop in performance.

With the previous implementation(with nested loops) the performance is
found to be around 4 Mbps at baud rate of 4 Mbps. In that
implementation we handle sending DWORDs continuosly until the transfer
size < 4. Can you let us know any other alternatives for the above
performance drop.


> 
> thanks,
> --
> js
> suse labs
>
  

On 04. 03. 24, 5:37, Rengarajan.S@microchip.com wrote:
> Hi Jiri,
> 
> On Fri, 2024-02-23 at 10:26 +0100, Jiri Slaby wrote:
>> EXTERNAL EMAIL: Do not click links or open attachments unless you
>> know the content is safe
>>
>> On 23. 02. 24, 10:21, Rengarajan.S@microchip.com wrote:
>>> On Fri, 2024-02-23 at 07:08 +0100, Jiri Slaby wrote:
>>>> EXTERNAL EMAIL: Do not click links or open attachments unless you
>>>> know the content is safe
>>>>
>>>> On 22. 02. 24, 14:49, Rengarajan S wrote:
>>>>> Updated the TX Burst implementation by changing the circular
>>>>> buffer
>>>>> processing with the pre-existing APIs in kernel. Also updated
>>>>> conditional
>>>>> statements and alignment issues for better readability.
>>>>
>>>> Hi,
>>>>
>>>> so why are you keeping the nested double loop?
>>>>
>>>
>>> Hi, in order to differentiate Burst mode handling with byte mode
>>> had
>>> seperate loops for both. Since, having single while loop also does
>>> not
>>> align with rx implementation (where we have seperate handling for
>>> burst
>>> and byte) have retained the double loop.
>>
>> So obviously, align RX to a single loop if possible. The current TX
>> code
>> is very hard to follow and sort of unmaintainable (and buggy). And
>> IMO
>> it's unnecessary as I proposed [1]. And even if RX cannot be one
>> loop,
>> you still can make TX easy to read as the two need not be the same.
>>
>> [1]
>> https://lore.kernel.org/all/b8325c3f-bf5b-4c55-8dce-ef395edce251@kernel.org/
> 
> 
> while (data_empty_count) {
>     cnt = CIRC_CNT_TO_END();
>     if (!cnt)
>       break;
>     if (cnt < UART_BURST_SIZE || (tail & 3)) { // is_unaligned()
>       writeb();
>       cnt = 1;
>     } else {
>       writel()
>       cnt = UART_BURST_SIZE;
>     }
>     uart_xmit_advance(cnt);
>     data_empty_count -= cnt;
> }
> 
> With the above implementation we are observing performance drop of 2
> Mbps at baud rate of 4 Mbps. The reason for this is the fact that for
> each iteration we are checking if the the data need to be processed via
> DWORDs or Bytes. The condition check for each iteration is causing the
> drop in performance.

Hi,

the check is by several orders of magnitude faster than the I/O proper. 
So I don't think that's the root cause.

> With the previous implementation(with nested loops) the performance is
> found to be around 4 Mbps at baud rate of 4 Mbps. In that
> implementation we handle sending DWORDs continuosly until the transfer
> size < 4. Can you let us know any other alternatives for the above
> performance drop.

Could you attach the patch you are testing?

thanks,