From mboxrd@z Thu Jan 1 00:00:00 1970 From: yalin wang Date: Thu, 20 Aug 2015 11:30:58 +0000 Subject: Re: [RFC] fbdev/riva:change to use generice function to implement reverse_order() Message-Id: <867D66CD-9A3B-4536-B537-8C065C85E497@gmail.com> List-Id: References: <55D5B3A9.6040901@ti.com> In-Reply-To: <55D5B3A9.6040901@ti.com> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Tomi Valkeinen Cc: adaplas@gmail.com, plagnioj@jcrosoft.com, linux-fbdev@vger.kernel.org, open list > On Aug 20, 2015, at 19:02, Tomi Valkeinen wrote: > > > On 10/08/15 13:12, yalin wang wrote: >> This change to use swab32(bitrev32()) to implement reverse_order() >> function, have better performance on some platforms. > > Which platforms? Presuming you tested this, roughly how much better > performance? If you didn't, how do you know it's faster? i investigate on arm64 platforms: for (j = dsize; j--;) { tmp = data[k++]; tmp = reverse_order(tmp); NVDmaNext(par, tmp); bac: 110006a4 add w4, w21, #0x1 bb0: 5ac000a3 rbit w3, w5 if (dsize) { NVDmaStart(info, par, RECT_EXPAND_TWO_COLOR_DATA(0), dsize); for (j = dsize; j--;) { tmp = data[k++]; bb4: 110006d6 add w22, w22, #0x1 bb8: 5ac00861 rev w1, w3 tmp = reverse_order(tmp); NVDmaNext(par, tmp); bbc: b9041e64 str w4, [x19,#1052] bc0: 8b3548c2 add x2, x6, w21, uxtw #2 bc4: b9000041 str w1, [x2] this is the disassemble code after apply the patch, only need: rbit w3, w5 rev w1, w3 2 instruction to get the reverse_order() result, apparently after than the origianl macro code. > >> Signed-off-by: yalin wang >> --- >> drivers/video/fbdev/riva/fbdev.c | 19 ++++++------------- >> 1 file changed, 6 insertions(+), 13 deletions(-) >> >> diff --git a/drivers/video/fbdev/riva/fbdev.c b/drivers/video/fbdev/riva/fbdev.c >> index f1ad274..4803901 100644 >> --- a/drivers/video/fbdev/riva/fbdev.c >> +++ b/drivers/video/fbdev/riva/fbdev.c >> @@ -40,6 +40,7 @@ >> #include >> #include >> #include >> +#include >> #include >> #ifdef CONFIG_PMAC_BACKLIGHT >> #include >> @@ -84,6 +85,7 @@ >> #define SetBit(n) (1<<(n)) >> #define Set8Bits(value) ((value)&0xff) >> >> +#define reverse_order(v) swab32(bitrev32(v)) >> /* HW cursor parameters */ >> #define MAX_CURS 32 >> >> @@ -451,15 +453,6 @@ static inline unsigned char MISCin(struct riva_par *par) >> return (VGA_RD08(par->riva.PVIO, 0x3cc)); >> } >> >> -static inline void reverse_order(u32 *l) > > I would suggest to do the work in the inline function, instead of a > macro. And if you keep the function prototype the same, then the changes > to each reverse_order call site are not needed. > ok, i will change to a inline function.