usb: gadget: configfs: Use memcpy_and_pad()
PATCH 1/1 usb: gadget: configfs: Use memcpy_and_pad()
Instead of zeroing some memory and then copying data in part or all of it, use memcpy_and_pad().
This avoids writing some memory twice and should save a few cycles.
--- drivers/usb/gadget/configfs.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/usb/gadget/configfs.c +1 -2
@@ -909,8 +909,7 @@ static ssize_t webusb_landingPage_store(struct config_item *item, const char *pamutex_lock (&gi->lock);// ensure 0 bytes are set, in case the new landing page is shorter then the old one. -memset (gi->landing_page,0 ,sizeof (gi->landing_page)); -memcpy (gi->landing_page, page, l); +memcpy_and_pad (gi->landing_page,sizeof (gi->landing_page), page, l,0 );mutex_unlock (&gi->lock);return len;
Signature · 1 line
2.39.1
From: Andy Shevchenko
Quoted 5 lines
Sent: 02 February 2023 15:18
Instead of zeroing some memory and then copying data in part or all of it, use
memcpy_and_pad().
This avoids writing some memory twice and should save a few cycles.
Maybe, maybe not.
It rather depends on the lengths involved (the code doesn't seem to be in the main tree).
The cost of the conditionals and the misaligned length/start for the memset() could easily overwhelm any apparent saving.
A memset() of a constant whole number of words is going to be significantly faster than the partial one.
David
Quoted 21 lines
Signed-off-by Andy Shevchenko <[email protected]>--- drivers/usb/gadget/configfs.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)drivers/usb/gadget/configfs.c +1
-3@@ -909,8 +909,7 @@ static ssize_t webusb_landingPage_store(struct config_item *item, const char *pamutex_lock (&gi->lock);// ensure 0 bytes are set, in case the new landing page is shorter then the old one. -memset (gi->landing_page,0 ,sizeof (gi->landing_page)); -memcpy (gi->landing_page, page, l); +memcpy_and_pad (gi->landing_page,sizeof (gi->landing_page), page, l,0 );mutex_unlock (&gi->lock);return len; --
Quoted 1 line
2.39.1
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)
Quoted 11 lines · David Laight
From: Andy Shevchenko
Quoted 5 lines
Sent: 02 February 2023 15:18
Instead of zeroing some memory and then copying data in part or all of it, use
memcpy_and_pad().
This avoids writing some memory twice and should save a few cycles.Maybe, maybe not.
It rather depends on the lengths involved (the code doesn't seem to be in the main tree).
Quoted 5 lines
The cost of the conditionals and the misaligned length/start for the
memset()could easily overwhelm any apparent saving.A
memset()of a constant whole number of words is going to be significantly faster than the partial one.
Then you can put some (little I suppose) efforts in optimizing memcpy_and_pad() once for all, no?
Signature · 2 lines
With Best Regards, Andy Shevchenko