diff options
| author | Finn Thain | 2018-10-13 11:47:55 +1100 |
|---|---|---|
| committer | Geert Uytterhoeven | 2018-12-03 13:05:42 +0100 |
| commit | b6cf523c16e148e1ece5fc4ff4657d79323a4635 (patch) | |
| tree | a0268d6a6f4dd3651ef3f505d0acf22b5c1fe8b6 /arch | |
| parent | 651022382c7f8da46cb4872a545ee1da6d097d2a (diff) | |
m68k: Unroll raw_outsb() loop
Unroll the raw_outsb() loop using the optimized assembler code from
raw_outsw(). That code is copied and pasted, with movew changed to moveb.
This improves the performance of sequential write transfers using mac_esp
in PIO mode by 5% or 10%. (The DMA controller on the 840av/660av models is
still unsupported so PIO transfers are used.)
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Diffstat (limited to 'arch')
| -rw-r--r-- | arch/m68k/include/asm/raw_io.h | 39 |
1 files changed, 35 insertions, 4 deletions
diff --git a/arch/m68k/include/asm/raw_io.h b/arch/m68k/include/asm/raw_io.h index 85761255dde5..8a6dc6e5a279 100644 --- a/arch/m68k/include/asm/raw_io.h +++ b/arch/m68k/include/asm/raw_io.h @@ -107,12 +107,43 @@ static inline void raw_insb(volatile u8 __iomem *port, u8 *buf, unsigned int len } static inline void raw_outsb(volatile u8 __iomem *port, const u8 *buf, - unsigned int len) + unsigned int nr) { - unsigned int i; + unsigned int tmp; - for (i = 0; i < len; i++) - out_8(port, *buf++); + if (nr & 15) { + tmp = (nr & 15) - 1; + asm volatile ( + "1: moveb %0@+,%2@; dbra %1,1b" + : "=a" (buf), "=d" (tmp) + : "a" (port), "0" (buf), + "1" (tmp)); + } + if (nr >> 4) { + tmp = (nr >> 4) - 1; + asm volatile ( + "1: " + "moveb %0@+,%2@; " + "moveb %0@+,%2@; " + "moveb %0@+,%2@; " + "moveb %0@+,%2@; " + "moveb %0@+,%2@; " + "moveb %0@+,%2@; " + "moveb %0@+,%2@; " + "moveb %0@+,%2@; " + "moveb %0@+,%2@; " + "moveb %0@+,%2@; " + "moveb %0@+,%2@; " + "moveb %0@+,%2@; " + "moveb %0@+,%2@; " + "moveb %0@+,%2@; " + "moveb %0@+,%2@; " + "moveb %0@+,%2@; " + "dbra %1,1b" + : "=a" (buf), "=d" (tmp) + : "a" (port), "0" (buf), + "1" (tmp)); + } } static inline void raw_insw(volatile u16 __iomem *port, u16 *buf, unsigned int nr) |