From 0837f52463583f76670ab2350e0f1541cb0351f5 Mon Sep 17 00:00:00 2001 From: Paul Mundt Date: Tue, 4 Aug 2009 18:09:54 +0900 Subject: sh: Partially unroll the SH-4 __flush_xxx_region() flushers. This does a bit of unrolling for the SH-4 region flushers. Based on an earlier patch by SUGIOKA Toshinobu. Signed-off-by: Paul Mundt --- arch/sh/mm/flush-sh4.c | 104 +++++++++++++++++++++++++++++++++++++++---------- 1 file changed, 83 insertions(+), 21 deletions(-) diff --git a/arch/sh/mm/flush-sh4.c b/arch/sh/mm/flush-sh4.c index e6d918f6ec0..edefc53891a 100644 --- a/arch/sh/mm/flush-sh4.c +++ b/arch/sh/mm/flush-sh4.c @@ -10,16 +10,37 @@ */ void __weak __flush_wback_region(void *start, int size) { - unsigned long v; - unsigned long begin, end; + unsigned long v, cnt, end; - begin = (unsigned long)start & ~(L1_CACHE_BYTES-1); + v = (unsigned long)start & ~(L1_CACHE_BYTES-1); end = ((unsigned long)start + size + L1_CACHE_BYTES-1) & ~(L1_CACHE_BYTES-1); - for (v = begin; v < end; v+=L1_CACHE_BYTES) { - asm volatile("ocbwb %0" - : /* no output */ - : "m" (__m(v))); + cnt = (end - v) / L1_CACHE_BYTES; + + while (cnt >= 8) { + asm volatile("ocbwb @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbwb @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbwb @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbwb @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbwb @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbwb @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbwb @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbwb @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + cnt -= 8; + } + + while (cnt) { + asm volatile("ocbwb @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + cnt--; } } @@ -31,16 +52,36 @@ void __weak __flush_wback_region(void *start, int size) */ void __weak __flush_purge_region(void *start, int size) { - unsigned long v; - unsigned long begin, end; + unsigned long v, cnt, end; - begin = (unsigned long)start & ~(L1_CACHE_BYTES-1); + v = (unsigned long)start & ~(L1_CACHE_BYTES-1); end = ((unsigned long)start + size + L1_CACHE_BYTES-1) & ~(L1_CACHE_BYTES-1); - for (v = begin; v < end; v+=L1_CACHE_BYTES) { - asm volatile("ocbp %0" - : /* no output */ - : "m" (__m(v))); + cnt = (end - v) / L1_CACHE_BYTES; + + while (cnt >= 8) { + asm volatile("ocbp @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbp @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbp @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbp @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbp @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbp @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbp @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbp @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + cnt -= 8; + } + while (cnt) { + asm volatile("ocbp @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + cnt--; } } @@ -49,15 +90,36 @@ void __weak __flush_purge_region(void *start, int size) */ void __weak __flush_invalidate_region(void *start, int size) { - unsigned long v; - unsigned long begin, end; + unsigned long v, cnt, end; - begin = (unsigned long)start & ~(L1_CACHE_BYTES-1); + v = (unsigned long)start & ~(L1_CACHE_BYTES-1); end = ((unsigned long)start + size + L1_CACHE_BYTES-1) & ~(L1_CACHE_BYTES-1); - for (v = begin; v < end; v+=L1_CACHE_BYTES) { - asm volatile("ocbi %0" - : /* no output */ - : "m" (__m(v))); + cnt = (end - v) / L1_CACHE_BYTES; + + while (cnt >= 8) { + asm volatile("ocbi @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbi @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbi @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbi @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbi @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbi @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbi @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + asm volatile("ocbi @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + cnt -= 8; + } + + while (cnt) { + asm volatile("ocbi @%0" : : "r" (v)); + v += L1_CACHE_BYTES; + cnt--; } } -- cgit v1.2.3