diff mbox series

[v3] usb: gadget: aspeed: improve vhub port irq handling

Message ID 20200315191430.12379-1-rentao.bupt@gmail.com (mailing list archive)
State Superseded
Headers show
Series [v3] usb: gadget: aspeed: improve vhub port irq handling | expand

Commit Message

Tao Ren March 15, 2020, 7:14 p.m. UTC
From: Tao Ren <rentao.bupt@gmail.com>

This patch evaluates vhub ports' irq mask before going through per-port
irq handling one by one, which helps to speed up irq handling in case
there is no port interrupt.

Signed-off-by: Tao Ren <rentao.bupt@gmail.com>
---
 Changes in v3:
   - assign istat to (unsigned long) bitmap before calling
     "for_each_set_bit_from".
 Changes in v2:
   - use "for_each_set_bit" to speed up port irq handling.

 drivers/usb/gadget/udc/aspeed-vhub/core.c | 12 +++++++++---
 drivers/usb/gadget/udc/aspeed-vhub/vhub.h |  8 +++-----
 2 files changed, 12 insertions(+), 8 deletions(-)

Comments

Tao Ren April 1, 2020, 9:58 p.m. UTC | #1
Hi Ben,

Any further comments on the patch?


Cheers,

Tao

On Sun, Mar 15, 2020 at 12:14:30PM -0700, rentao.bupt@gmail.com wrote:
> From: Tao Ren <rentao.bupt@gmail.com>
> 
> This patch evaluates vhub ports' irq mask before going through per-port
> irq handling one by one, which helps to speed up irq handling in case
> there is no port interrupt.
> 
> Signed-off-by: Tao Ren <rentao.bupt@gmail.com>
> ---
>  Changes in v3:
>    - assign istat to (unsigned long) bitmap before calling
>      "for_each_set_bit_from".
>  Changes in v2:
>    - use "for_each_set_bit" to speed up port irq handling.
> 
>  drivers/usb/gadget/udc/aspeed-vhub/core.c | 12 +++++++++---
>  drivers/usb/gadget/udc/aspeed-vhub/vhub.h |  8 +++-----
>  2 files changed, 12 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/usb/gadget/udc/aspeed-vhub/core.c b/drivers/usb/gadget/udc/aspeed-vhub/core.c
> index f8d35dd60c34..555e8645fb1e 100644
> --- a/drivers/usb/gadget/udc/aspeed-vhub/core.c
> +++ b/drivers/usb/gadget/udc/aspeed-vhub/core.c
> @@ -134,11 +134,15 @@ static irqreturn_t ast_vhub_irq(int irq, void *data)
>  	}
>  
>  	/* Handle device interrupts */
> -	for (i = 0; i < vhub->max_ports; i++) {
> -		u32 dev_mask = VHUB_IRQ_DEVICE1 << i;
> +	if (istat & vhub->port_irq_mask) {
> +		unsigned long bitmap = istat;
> +		int offset = VHUB_IRQ_DEV1_BIT;
> +		int size = VHUB_IRQ_DEV1_BIT + vhub->max_ports;
>  
> -		if (istat & dev_mask)
> +		for_each_set_bit_from(offset, &bitmap, size) {
> +			i = offset - VHUB_IRQ_DEV1_BIT;
>  			ast_vhub_dev_irq(&vhub->ports[i].dev);
> +		}
>  	}
>  
>  	/* Handle top-level vHub EP0 interrupts */
> @@ -332,6 +336,8 @@ static int ast_vhub_probe(struct platform_device *pdev)
>  
>  	spin_lock_init(&vhub->lock);
>  	vhub->pdev = pdev;
> +	vhub->port_irq_mask = GENMASK(VHUB_IRQ_DEV1_BIT + vhub->max_ports - 1,
> +				      VHUB_IRQ_DEV1_BIT);
>  
>  	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
>  	vhub->regs = devm_ioremap_resource(&pdev->dev, res);
> diff --git a/drivers/usb/gadget/udc/aspeed-vhub/vhub.h b/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> index fac79ef6d669..23a1ac91f8d2 100644
> --- a/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> +++ b/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> @@ -51,14 +51,11 @@
>  #define VHUB_CTRL_UPSTREAM_CONNECT		(1 << 0)
>  
>  /* IER & ISR */
> +#define VHUB_IRQ_DEV1_BIT			9
>  #define VHUB_IRQ_USB_CMD_DEADLOCK		(1 << 18)
>  #define VHUB_IRQ_EP_POOL_NAK			(1 << 17)
>  #define VHUB_IRQ_EP_POOL_ACK_STALL		(1 << 16)
> -#define VHUB_IRQ_DEVICE5			(1 << 13)
> -#define VHUB_IRQ_DEVICE4			(1 << 12)
> -#define VHUB_IRQ_DEVICE3			(1 << 11)
> -#define VHUB_IRQ_DEVICE2			(1 << 10)
> -#define VHUB_IRQ_DEVICE1			(1 << 9)
> +#define VHUB_IRQ_DEVICE1			(1 << (VHUB_IRQ_DEV1_BIT))
>  #define VHUB_IRQ_BUS_RESUME			(1 << 8)
>  #define VHUB_IRQ_BUS_SUSPEND 			(1 << 7)
>  #define VHUB_IRQ_BUS_RESET 			(1 << 6)
> @@ -402,6 +399,7 @@ struct ast_vhub {
>  	/* Per-port info */
>  	struct ast_vhub_port		*ports;
>  	u32				max_ports;
> +	u32				port_irq_mask;
>  
>  	/* Generic EP data structures */
>  	struct ast_vhub_ep		*epns;
> -- 
> 2.17.1
>
Benjamin Herrenschmidt April 2, 2020, 10:45 a.m. UTC | #2
On Wed, 2020-04-01 at 14:58 -0700, Tao Ren wrote:
> Hi Ben,
> 
> Any further comments on the patch?

Ah sorry, nope. Did you check the generated assembly to see if it
looked any better ? :-)

Otherwise,

Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

> 
> Cheers,
> 
> Tao
> 
> On Sun, Mar 15, 2020 at 12:14:30PM -0700, rentao.bupt@gmail.com
> wrote:
> > From: Tao Ren <rentao.bupt@gmail.com>
> > 
> > This patch evaluates vhub ports' irq mask before going through per-
> > port
> > irq handling one by one, which helps to speed up irq handling in
> > case
> > there is no port interrupt.
> > 
> > Signed-off-by: Tao Ren <rentao.bupt@gmail.com>
> > ---
> >  Changes in v3:
> >    - assign istat to (unsigned long) bitmap before calling
> >      "for_each_set_bit_from".
> >  Changes in v2:
> >    - use "for_each_set_bit" to speed up port irq handling.
> > 
> >  drivers/usb/gadget/udc/aspeed-vhub/core.c | 12 +++++++++---
> >  drivers/usb/gadget/udc/aspeed-vhub/vhub.h |  8 +++-----
> >  2 files changed, 12 insertions(+), 8 deletions(-)
> > 
> > diff --git a/drivers/usb/gadget/udc/aspeed-vhub/core.c
> > b/drivers/usb/gadget/udc/aspeed-vhub/core.c
> > index f8d35dd60c34..555e8645fb1e 100644
> > --- a/drivers/usb/gadget/udc/aspeed-vhub/core.c
> > +++ b/drivers/usb/gadget/udc/aspeed-vhub/core.c
> > @@ -134,11 +134,15 @@ static irqreturn_t ast_vhub_irq(int irq, void
> > *data)
> >  	}
> >  
> >  	/* Handle device interrupts */
> > -	for (i = 0; i < vhub->max_ports; i++) {
> > -		u32 dev_mask = VHUB_IRQ_DEVICE1 << i;
> > +	if (istat & vhub->port_irq_mask) {
> > +		unsigned long bitmap = istat;
> > +		int offset = VHUB_IRQ_DEV1_BIT;
> > +		int size = VHUB_IRQ_DEV1_BIT + vhub->max_ports;
> >  
> > -		if (istat & dev_mask)
> > +		for_each_set_bit_from(offset, &bitmap, size) {
> > +			i = offset - VHUB_IRQ_DEV1_BIT;
> >  			ast_vhub_dev_irq(&vhub->ports[i].dev);
> > +		}
> >  	}
> >  
> >  	/* Handle top-level vHub EP0 interrupts */
> > @@ -332,6 +336,8 @@ static int ast_vhub_probe(struct
> > platform_device *pdev)
> >  
> >  	spin_lock_init(&vhub->lock);
> >  	vhub->pdev = pdev;
> > +	vhub->port_irq_mask = GENMASK(VHUB_IRQ_DEV1_BIT + vhub-
> > >max_ports - 1,
> > +				      VHUB_IRQ_DEV1_BIT);
> >  
> >  	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> >  	vhub->regs = devm_ioremap_resource(&pdev->dev, res);
> > diff --git a/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> > b/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> > index fac79ef6d669..23a1ac91f8d2 100644
> > --- a/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> > +++ b/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> > @@ -51,14 +51,11 @@
> >  #define VHUB_CTRL_UPSTREAM_CONNECT		(1 << 0)
> >  
> >  /* IER & ISR */
> > +#define VHUB_IRQ_DEV1_BIT			9
> >  #define VHUB_IRQ_USB_CMD_DEADLOCK		(1 << 18)
> >  #define VHUB_IRQ_EP_POOL_NAK			(1 << 17)
> >  #define VHUB_IRQ_EP_POOL_ACK_STALL		(1 << 16)
> > -#define VHUB_IRQ_DEVICE5			(1 << 13)
> > -#define VHUB_IRQ_DEVICE4			(1 << 12)
> > -#define VHUB_IRQ_DEVICE3			(1 << 11)
> > -#define VHUB_IRQ_DEVICE2			(1 << 10)
> > -#define VHUB_IRQ_DEVICE1			(1 << 9)
> > +#define VHUB_IRQ_DEVICE1			(1 <<
> > (VHUB_IRQ_DEV1_BIT))
> >  #define VHUB_IRQ_BUS_RESUME			(1 << 8)
> >  #define VHUB_IRQ_BUS_SUSPEND 			(1 << 7)
> >  #define VHUB_IRQ_BUS_RESET 			(1 << 6)
> > @@ -402,6 +399,7 @@ struct ast_vhub {
> >  	/* Per-port info */
> >  	struct ast_vhub_port		*ports;
> >  	u32				max_ports;
> > +	u32				port_irq_mask;
> >  
> >  	/* Generic EP data structures */
> >  	struct ast_vhub_ep		*epns;
> > -- 
> > 2.17.1
> >
Tao Ren April 3, 2020, 6:48 a.m. UTC | #3
On Thu, Apr 02, 2020 at 09:45:38PM +1100, Benjamin Herrenschmidt wrote:
> On Wed, 2020-04-01 at 14:58 -0700, Tao Ren wrote:
> > Hi Ben,
> > 
> > Any further comments on the patch?
> 
> Ah sorry, nope. Did you check the generated assembly to see if it
> looked any better ? :-)
> 
> Otherwise,
> 
> Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

Thanks Ben for the review. Let me see if I can collect some runtime data
on my ast2400 BMC platform. Will update back later.


Cheers,

Tao
> 
> > 
> > Cheers,
> > 
> > Tao
> > 
> > On Sun, Mar 15, 2020 at 12:14:30PM -0700, rentao.bupt@gmail.com
> > wrote:
> > > From: Tao Ren <rentao.bupt@gmail.com>
> > > 
> > > This patch evaluates vhub ports' irq mask before going through per-
> > > port
> > > irq handling one by one, which helps to speed up irq handling in
> > > case
> > > there is no port interrupt.
> > > 
> > > Signed-off-by: Tao Ren <rentao.bupt@gmail.com>
> > > ---
> > >  Changes in v3:
> > >    - assign istat to (unsigned long) bitmap before calling
> > >      "for_each_set_bit_from".
> > >  Changes in v2:
> > >    - use "for_each_set_bit" to speed up port irq handling.
> > > 
> > >  drivers/usb/gadget/udc/aspeed-vhub/core.c | 12 +++++++++---
> > >  drivers/usb/gadget/udc/aspeed-vhub/vhub.h |  8 +++-----
> > >  2 files changed, 12 insertions(+), 8 deletions(-)
> > > 
> > > diff --git a/drivers/usb/gadget/udc/aspeed-vhub/core.c
> > > b/drivers/usb/gadget/udc/aspeed-vhub/core.c
> > > index f8d35dd60c34..555e8645fb1e 100644
> > > --- a/drivers/usb/gadget/udc/aspeed-vhub/core.c
> > > +++ b/drivers/usb/gadget/udc/aspeed-vhub/core.c
> > > @@ -134,11 +134,15 @@ static irqreturn_t ast_vhub_irq(int irq, void
> > > *data)
> > >  	}
> > >  
> > >  	/* Handle device interrupts */
> > > -	for (i = 0; i < vhub->max_ports; i++) {
> > > -		u32 dev_mask = VHUB_IRQ_DEVICE1 << i;
> > > +	if (istat & vhub->port_irq_mask) {
> > > +		unsigned long bitmap = istat;
> > > +		int offset = VHUB_IRQ_DEV1_BIT;
> > > +		int size = VHUB_IRQ_DEV1_BIT + vhub->max_ports;
> > >  
> > > -		if (istat & dev_mask)
> > > +		for_each_set_bit_from(offset, &bitmap, size) {
> > > +			i = offset - VHUB_IRQ_DEV1_BIT;
> > >  			ast_vhub_dev_irq(&vhub->ports[i].dev);
> > > +		}
> > >  	}
> > >  
> > >  	/* Handle top-level vHub EP0 interrupts */
> > > @@ -332,6 +336,8 @@ static int ast_vhub_probe(struct
> > > platform_device *pdev)
> > >  
> > >  	spin_lock_init(&vhub->lock);
> > >  	vhub->pdev = pdev;
> > > +	vhub->port_irq_mask = GENMASK(VHUB_IRQ_DEV1_BIT + vhub-
> > > >max_ports - 1,
> > > +				      VHUB_IRQ_DEV1_BIT);
> > >  
> > >  	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> > >  	vhub->regs = devm_ioremap_resource(&pdev->dev, res);
> > > diff --git a/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> > > b/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> > > index fac79ef6d669..23a1ac91f8d2 100644
> > > --- a/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> > > +++ b/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> > > @@ -51,14 +51,11 @@
> > >  #define VHUB_CTRL_UPSTREAM_CONNECT		(1 << 0)
> > >  
> > >  /* IER & ISR */
> > > +#define VHUB_IRQ_DEV1_BIT			9
> > >  #define VHUB_IRQ_USB_CMD_DEADLOCK		(1 << 18)
> > >  #define VHUB_IRQ_EP_POOL_NAK			(1 << 17)
> > >  #define VHUB_IRQ_EP_POOL_ACK_STALL		(1 << 16)
> > > -#define VHUB_IRQ_DEVICE5			(1 << 13)
> > > -#define VHUB_IRQ_DEVICE4			(1 << 12)
> > > -#define VHUB_IRQ_DEVICE3			(1 << 11)
> > > -#define VHUB_IRQ_DEVICE2			(1 << 10)
> > > -#define VHUB_IRQ_DEVICE1			(1 << 9)
> > > +#define VHUB_IRQ_DEVICE1			(1 <<
> > > (VHUB_IRQ_DEV1_BIT))
> > >  #define VHUB_IRQ_BUS_RESUME			(1 << 8)
> > >  #define VHUB_IRQ_BUS_SUSPEND 			(1 << 7)
> > >  #define VHUB_IRQ_BUS_RESET 			(1 << 6)
> > > @@ -402,6 +399,7 @@ struct ast_vhub {
> > >  	/* Per-port info */
> > >  	struct ast_vhub_port		*ports;
> > >  	u32				max_ports;
> > > +	u32				port_irq_mask;
> > >  
> > >  	/* Generic EP data structures */
> > >  	struct ast_vhub_ep		*epns;
> > > -- 
> > > 2.17.1
> > > 
>
Tao Ren April 7, 2020, 6:02 a.m. UTC | #4
Hi Ben,

On Thu, Apr 02, 2020 at 11:48:27PM -0700, Tao Ren wrote:
> On Thu, Apr 02, 2020 at 09:45:38PM +1100, Benjamin Herrenschmidt wrote:
> > On Wed, 2020-04-01 at 14:58 -0700, Tao Ren wrote:
> > > Hi Ben,
> > > 
> > > Any further comments on the patch?
> > 
> > Ah sorry, nope. Did you check the generated assembly to see if it
> > looked any better ? :-)
> > 
> > Otherwise,
> > 
> > Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> 
> Thanks Ben for the review. Let me see if I can collect some runtime data
> on my ast2400 BMC platform. Will update back later.

I ran some testing on my ast2400 and ast2500 BMC and looks like the
for() loop runs faster than for_each_set_bit_from() loop in my
environment. I'm not sure if something needs to be revised in my test
code, but please kindly share your suggestions:

I use get_cycles() to calculate execution time of 2 different loops, and
ast_vhub_dev_irq() is replaced with barrier() to avoid "noise"; below
are the results:

  - when downstream port number is 5 and only 1 irq bit is set, it takes
    ~30 cycles to finish for_each_set_bit() loop, and 20-25 cycles to
    finish the for() loop.

  - if downstream port number is 5 and all 5 bits are set, then
    for_each_set_bit() loop takes ~50 cycles and for() loop takes ~25
    cycles.

  - when I increase downsteam port number to 16 and set 1 irq bit, the
    for_each_set_bit() loop takes ~30 cycles and for() loop takes 25
    cycles. It's a little surprise to me because I thought for() loop
    would cost 60+ cycles (3 times of the value when port number is 5).

  - if downstream port number is 16 and all irq status bits are set,
    then for_each_set_bit() loop takes 60-70 cycles and for() loop takes
    30+ cycles.


Cheers,

Tao

> 
> > > 
> > > On Sun, Mar 15, 2020 at 12:14:30PM -0700, rentao.bupt@gmail.com
> > > wrote:
> > > > From: Tao Ren <rentao.bupt@gmail.com>
> > > > 
> > > > This patch evaluates vhub ports' irq mask before going through per-
> > > > port
> > > > irq handling one by one, which helps to speed up irq handling in
> > > > case
> > > > there is no port interrupt.
> > > > 
> > > > Signed-off-by: Tao Ren <rentao.bupt@gmail.com>
> > > > ---
> > > >  Changes in v3:
> > > >    - assign istat to (unsigned long) bitmap before calling
> > > >      "for_each_set_bit_from".
> > > >  Changes in v2:
> > > >    - use "for_each_set_bit" to speed up port irq handling.
> > > > 
> > > >  drivers/usb/gadget/udc/aspeed-vhub/core.c | 12 +++++++++---
> > > >  drivers/usb/gadget/udc/aspeed-vhub/vhub.h |  8 +++-----
> > > >  2 files changed, 12 insertions(+), 8 deletions(-)
> > > > 
> > > > diff --git a/drivers/usb/gadget/udc/aspeed-vhub/core.c
> > > > b/drivers/usb/gadget/udc/aspeed-vhub/core.c
> > > > index f8d35dd60c34..555e8645fb1e 100644
> > > > --- a/drivers/usb/gadget/udc/aspeed-vhub/core.c
> > > > +++ b/drivers/usb/gadget/udc/aspeed-vhub/core.c
> > > > @@ -134,11 +134,15 @@ static irqreturn_t ast_vhub_irq(int irq, void
> > > > *data)
> > > >  	}
> > > >  
> > > >  	/* Handle device interrupts */
> > > > -	for (i = 0; i < vhub->max_ports; i++) {
> > > > -		u32 dev_mask = VHUB_IRQ_DEVICE1 << i;
> > > > +	if (istat & vhub->port_irq_mask) {
> > > > +		unsigned long bitmap = istat;
> > > > +		int offset = VHUB_IRQ_DEV1_BIT;
> > > > +		int size = VHUB_IRQ_DEV1_BIT + vhub->max_ports;
> > > >  
> > > > -		if (istat & dev_mask)
> > > > +		for_each_set_bit_from(offset, &bitmap, size) {
> > > > +			i = offset - VHUB_IRQ_DEV1_BIT;
> > > >  			ast_vhub_dev_irq(&vhub->ports[i].dev);
> > > > +		}
> > > >  	}
> > > >  
> > > >  	/* Handle top-level vHub EP0 interrupts */
> > > > @@ -332,6 +336,8 @@ static int ast_vhub_probe(struct
> > > > platform_device *pdev)
> > > >  
> > > >  	spin_lock_init(&vhub->lock);
> > > >  	vhub->pdev = pdev;
> > > > +	vhub->port_irq_mask = GENMASK(VHUB_IRQ_DEV1_BIT + vhub-
> > > > >max_ports - 1,
> > > > +				      VHUB_IRQ_DEV1_BIT);
> > > >  
> > > >  	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> > > >  	vhub->regs = devm_ioremap_resource(&pdev->dev, res);
> > > > diff --git a/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> > > > b/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> > > > index fac79ef6d669..23a1ac91f8d2 100644
> > > > --- a/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> > > > +++ b/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> > > > @@ -51,14 +51,11 @@
> > > >  #define VHUB_CTRL_UPSTREAM_CONNECT		(1 << 0)
> > > >  
> > > >  /* IER & ISR */
> > > > +#define VHUB_IRQ_DEV1_BIT			9
> > > >  #define VHUB_IRQ_USB_CMD_DEADLOCK		(1 << 18)
> > > >  #define VHUB_IRQ_EP_POOL_NAK			(1 << 17)
> > > >  #define VHUB_IRQ_EP_POOL_ACK_STALL		(1 << 16)
> > > > -#define VHUB_IRQ_DEVICE5			(1 << 13)
> > > > -#define VHUB_IRQ_DEVICE4			(1 << 12)
> > > > -#define VHUB_IRQ_DEVICE3			(1 << 11)
> > > > -#define VHUB_IRQ_DEVICE2			(1 << 10)
> > > > -#define VHUB_IRQ_DEVICE1			(1 << 9)
> > > > +#define VHUB_IRQ_DEVICE1			(1 <<
> > > > (VHUB_IRQ_DEV1_BIT))
> > > >  #define VHUB_IRQ_BUS_RESUME			(1 << 8)
> > > >  #define VHUB_IRQ_BUS_SUSPEND 			(1 << 7)
> > > >  #define VHUB_IRQ_BUS_RESET 			(1 << 6)
> > > > @@ -402,6 +399,7 @@ struct ast_vhub {
> > > >  	/* Per-port info */
> > > >  	struct ast_vhub_port		*ports;
> > > >  	u32				max_ports;
> > > > +	u32				port_irq_mask;
> > > >  
> > > >  	/* Generic EP data structures */
> > > >  	struct ast_vhub_ep		*epns;
> > > > -- 
> > > > 2.17.1
> > > > 
> >
Benjamin Herrenschmidt April 7, 2020, 11:36 p.m. UTC | #5
On Mon, 2020-04-06 at 23:02 -0700, Tao Ren wrote:
> I ran some testing on my ast2400 and ast2500 BMC and looks like the
> for() loop runs faster than for_each_set_bit_from() loop in my
> environment. I'm not sure if something needs to be revised in my test
> code, but please kindly share your suggestions:
> 
> I use get_cycles() to calculate execution time of 2 different loops, and
> ast_vhub_dev_irq() is replaced with barrier() to avoid "noise"; below
> are the results:
> 
>   - when downstream port number is 5 and only 1 irq bit is set, it takes
>     ~30 cycles to finish for_each_set_bit() loop, and 20-25 cycles to
>     finish the for() loop.
> 
>   - if downstream port number is 5 and all 5 bits are set, then
>     for_each_set_bit() loop takes ~50 cycles and for() loop takes ~25
>     cycles.
> 
>   - when I increase downsteam port number to 16 and set 1 irq bit, the
>     for_each_set_bit() loop takes ~30 cycles and for() loop takes 25
>     cycles. It's a little surprise to me because I thought for() loop
>     would cost 60+ cycles (3 times of the value when port number is 5).
> 
>   - if downstream port number is 16 and all irq status bits are set,
>     then for_each_set_bit() loop takes 60-70 cycles and for() loop takes
>     30+ cycles.

I suspect the CPU doesn't have an efficient find-zero-bit primitive,
check the generated asm. In that case I would go back to the simple for
loop.

Cheers,
Ben.
Tao Ren April 8, 2020, 5:40 a.m. UTC | #6
On Wed, Apr 08, 2020 at 09:36:16AM +1000, Benjamin Herrenschmidt wrote:
> On Mon, 2020-04-06 at 23:02 -0700, Tao Ren wrote:
> > I ran some testing on my ast2400 and ast2500 BMC and looks like the
> > for() loop runs faster than for_each_set_bit_from() loop in my
> > environment. I'm not sure if something needs to be revised in my test
> > code, but please kindly share your suggestions:
> > 
> > I use get_cycles() to calculate execution time of 2 different loops, and
> > ast_vhub_dev_irq() is replaced with barrier() to avoid "noise"; below
> > are the results:
> > 
> >   - when downstream port number is 5 and only 1 irq bit is set, it takes
> >     ~30 cycles to finish for_each_set_bit() loop, and 20-25 cycles to
> >     finish the for() loop.
> > 
> >   - if downstream port number is 5 and all 5 bits are set, then
> >     for_each_set_bit() loop takes ~50 cycles and for() loop takes ~25
> >     cycles.
> > 
> >   - when I increase downsteam port number to 16 and set 1 irq bit, the
> >     for_each_set_bit() loop takes ~30 cycles and for() loop takes 25
> >     cycles. It's a little surprise to me because I thought for() loop
> >     would cost 60+ cycles (3 times of the value when port number is 5).
> > 
> >   - if downstream port number is 16 and all irq status bits are set,
> >     then for_each_set_bit() loop takes 60-70 cycles and for() loop takes
> >     30+ cycles.
> 
> I suspect the CPU doesn't have an efficient find-zero-bit primitive,
> check the generated asm. In that case I would go back to the simple for
> loop.
> 
> Cheers,
> Ben.

_find_next_bit_le() function is defined in arch/arm/lib/findbit.S. I'm
looking at the code: will run more tests and send out patch v4 with
simple for loop later.


Cheers,

Tao
diff mbox series

Patch

diff --git a/drivers/usb/gadget/udc/aspeed-vhub/core.c b/drivers/usb/gadget/udc/aspeed-vhub/core.c
index f8d35dd60c34..555e8645fb1e 100644
--- a/drivers/usb/gadget/udc/aspeed-vhub/core.c
+++ b/drivers/usb/gadget/udc/aspeed-vhub/core.c
@@ -134,11 +134,15 @@  static irqreturn_t ast_vhub_irq(int irq, void *data)
 	}
 
 	/* Handle device interrupts */
-	for (i = 0; i < vhub->max_ports; i++) {
-		u32 dev_mask = VHUB_IRQ_DEVICE1 << i;
+	if (istat & vhub->port_irq_mask) {
+		unsigned long bitmap = istat;
+		int offset = VHUB_IRQ_DEV1_BIT;
+		int size = VHUB_IRQ_DEV1_BIT + vhub->max_ports;
 
-		if (istat & dev_mask)
+		for_each_set_bit_from(offset, &bitmap, size) {
+			i = offset - VHUB_IRQ_DEV1_BIT;
 			ast_vhub_dev_irq(&vhub->ports[i].dev);
+		}
 	}
 
 	/* Handle top-level vHub EP0 interrupts */
@@ -332,6 +336,8 @@  static int ast_vhub_probe(struct platform_device *pdev)
 
 	spin_lock_init(&vhub->lock);
 	vhub->pdev = pdev;
+	vhub->port_irq_mask = GENMASK(VHUB_IRQ_DEV1_BIT + vhub->max_ports - 1,
+				      VHUB_IRQ_DEV1_BIT);
 
 	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
 	vhub->regs = devm_ioremap_resource(&pdev->dev, res);
diff --git a/drivers/usb/gadget/udc/aspeed-vhub/vhub.h b/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
index fac79ef6d669..23a1ac91f8d2 100644
--- a/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
+++ b/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
@@ -51,14 +51,11 @@ 
 #define VHUB_CTRL_UPSTREAM_CONNECT		(1 << 0)
 
 /* IER & ISR */
+#define VHUB_IRQ_DEV1_BIT			9
 #define VHUB_IRQ_USB_CMD_DEADLOCK		(1 << 18)
 #define VHUB_IRQ_EP_POOL_NAK			(1 << 17)
 #define VHUB_IRQ_EP_POOL_ACK_STALL		(1 << 16)
-#define VHUB_IRQ_DEVICE5			(1 << 13)
-#define VHUB_IRQ_DEVICE4			(1 << 12)
-#define VHUB_IRQ_DEVICE3			(1 << 11)
-#define VHUB_IRQ_DEVICE2			(1 << 10)
-#define VHUB_IRQ_DEVICE1			(1 << 9)
+#define VHUB_IRQ_DEVICE1			(1 << (VHUB_IRQ_DEV1_BIT))
 #define VHUB_IRQ_BUS_RESUME			(1 << 8)
 #define VHUB_IRQ_BUS_SUSPEND 			(1 << 7)
 #define VHUB_IRQ_BUS_RESET 			(1 << 6)
@@ -402,6 +399,7 @@  struct ast_vhub {
 	/* Per-port info */
 	struct ast_vhub_port		*ports;
 	u32				max_ports;
+	u32				port_irq_mask;
 
 	/* Generic EP data structures */
 	struct ast_vhub_ep		*epns;