kernel_optimize_test/arch/mn10300/mm/cache-inv-by-reg.S
Akira Takeuchi 9731d23710 MN10300: AM34: Add cacheflushing by using the AM34 purge registers
The AM34 CPU core provides an automated way of purging the cache rather than
manually iterating over all the tags in the cache.  Make it possible to use
these.

Signed-off-by: Akira Takeuchi <takeuchi.akr@jp.panasonic.com>
Signed-off-by: Kiyoshi Owada <owada.kiyoshi@jp.panasonic.com>
Signed-off-by: David Howells <dhowells@redhat.com>
2010-10-27 17:28:45 +01:00

357 lines
8.4 KiB
ArmAsm

/* MN10300 CPU cache invalidation routines, using automatic purge registers
*
* Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
* Written by David Howells (dhowells@redhat.com)
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public Licence
* as published by the Free Software Foundation; either version
* 2 of the Licence, or (at your option) any later version.
*/
#include <linux/sys.h>
#include <linux/linkage.h>
#include <asm/smp.h>
#include <asm/page.h>
#include <asm/cache.h>
#include <asm/irqflags.h>
#include <asm/cacheflush.h>
#define mn10300_local_dcache_inv_range_intr_interval \
+((1 << MN10300_DCACHE_INV_RANGE_INTR_LOG2_INTERVAL) - 1)
#if mn10300_local_dcache_inv_range_intr_interval > 0xff
#error MN10300_DCACHE_INV_RANGE_INTR_LOG2_INTERVAL must be 8 or less
#endif
.am33_2
#ifndef CONFIG_SMP
.globl mn10300_icache_inv
.globl mn10300_icache_inv_page
.globl mn10300_icache_inv_range
.globl mn10300_icache_inv_range2
.globl mn10300_dcache_inv
.globl mn10300_dcache_inv_page
.globl mn10300_dcache_inv_range
.globl mn10300_dcache_inv_range2
mn10300_icache_inv = mn10300_local_icache_inv
mn10300_icache_inv_page = mn10300_local_icache_inv_page
mn10300_icache_inv_range = mn10300_local_icache_inv_range
mn10300_icache_inv_range2 = mn10300_local_icache_inv_range2
mn10300_dcache_inv = mn10300_local_dcache_inv
mn10300_dcache_inv_page = mn10300_local_dcache_inv_page
mn10300_dcache_inv_range = mn10300_local_dcache_inv_range
mn10300_dcache_inv_range2 = mn10300_local_dcache_inv_range2
#endif /* !CONFIG_SMP */
###############################################################################
#
# void mn10300_local_icache_inv(void)
# Invalidate the entire icache
#
###############################################################################
ALIGN
.globl mn10300_local_icache_inv
.type mn10300_local_icache_inv,@function
mn10300_local_icache_inv:
mov CHCTR,a0
movhu (a0),d0
btst CHCTR_ICEN,d0
beq mn10300_local_icache_inv_end
# invalidate
or CHCTR_ICINV,d0
movhu d0,(a0)
movhu (a0),d0
mn10300_local_icache_inv_end:
ret [],0
.size mn10300_local_icache_inv,.-mn10300_local_icache_inv
###############################################################################
#
# void mn10300_local_dcache_inv(void)
# Invalidate the entire dcache
#
###############################################################################
ALIGN
.globl mn10300_local_dcache_inv
.type mn10300_local_dcache_inv,@function
mn10300_local_dcache_inv:
mov CHCTR,a0
movhu (a0),d0
btst CHCTR_DCEN,d0
beq mn10300_local_dcache_inv_end
# invalidate
or CHCTR_DCINV,d0
movhu d0,(a0)
movhu (a0),d0
mn10300_local_dcache_inv_end:
ret [],0
.size mn10300_local_dcache_inv,.-mn10300_local_dcache_inv
###############################################################################
#
# void mn10300_local_dcache_inv_range(unsigned long start, unsigned long end)
# void mn10300_local_dcache_inv_range2(unsigned long start, unsigned long size)
# void mn10300_local_dcache_inv_page(unsigned long start)
# Invalidate a range of addresses on a page in the dcache
#
###############################################################################
ALIGN
.globl mn10300_local_dcache_inv_page
.globl mn10300_local_dcache_inv_range
.globl mn10300_local_dcache_inv_range2
.type mn10300_local_dcache_inv_page,@function
.type mn10300_local_dcache_inv_range,@function
.type mn10300_local_dcache_inv_range2,@function
mn10300_local_dcache_inv_page:
and ~(PAGE_SIZE-1),d0
mov PAGE_SIZE,d1
mn10300_local_dcache_inv_range2:
add d0,d1
mn10300_local_dcache_inv_range:
# If we are in writeback mode we check the start and end alignments,
# and if they're not cacheline-aligned, we must flush any bits outside
# the range that share cachelines with stuff inside the range
#ifdef CONFIG_MN10300_CACHE_WBACK
btst ~(L1_CACHE_BYTES-1),d0
bne 1f
btst ~(L1_CACHE_BYTES-1),d1
beq 2f
1:
bra mn10300_local_dcache_flush_inv_range
2:
#endif /* CONFIG_MN10300_CACHE_WBACK */
movm [d2,d3,a2],(sp)
mov CHCTR,a0
movhu (a0),d2
btst CHCTR_DCEN,d2
beq mn10300_local_dcache_inv_range_end
# round the addresses out to be full cachelines, unless we're in
# writeback mode, in which case we would be in flush and invalidate by
# now
#ifndef CONFIG_MN10300_CACHE_WBACK
and L1_CACHE_TAG_ADDRESS|L1_CACHE_TAG_ENTRY,d0 # round start
# addr down
mov L1_CACHE_BYTES-1,d2
add d2,d1
and L1_CACHE_TAG_ADDRESS|L1_CACHE_TAG_ENTRY,d1 # round end addr up
#endif /* !CONFIG_MN10300_CACHE_WBACK */
sub d0,d1,d2 # calculate the total size
mov d0,a2 # A2 = start address
mov d1,a1 # A1 = end address
LOCAL_CLI_SAVE(d3)
mov DCPGCR,a0 # make sure the purger isn't busy
setlb
mov (a0),d0
btst DCPGCR_DCPGBSY,d0
lne
# skip initial address alignment calculation if address is zero
mov d2,d1
cmp 0,a2
beq 1f
dcivloop:
/* calculate alignsize
*
* alignsize = L1_CACHE_BYTES;
* while (! start & alignsize) {
* alignsize <<=1;
* }
* d1 = alignsize;
*/
mov L1_CACHE_BYTES,d1
lsr 1,d1
setlb
add d1,d1
mov d1,d0
and a2,d0
leq
1:
/* calculate invsize
*
* if (totalsize > alignsize) {
* invsize = alignsize;
* } else {
* invsize = totalsize;
* tmp = 0x80000000;
* while (! invsize & tmp) {
* tmp >>= 1;
* }
* invsize = tmp;
* }
* d1 = invsize
*/
cmp d2,d1
bns 2f
mov d2,d1
mov 0x80000000,d0 # start from 31bit=1
setlb
lsr 1,d0
mov d0,e0
and d1,e0
leq
mov d0,d1
2:
/* set mask
*
* mask = ~(invsize-1);
* DCPGMR = mask;
*/
mov d1,d0
add -1,d0
not d0
mov d0,(DCPGMR)
# invalidate area
mov a2,d0
or DCPGCR_DCI,d0
mov d0,(a0) # DCPGCR = (mask & start) | DCPGCR_DCI
setlb # wait for the purge to complete
mov (a0),d0
btst DCPGCR_DCPGBSY,d0
lne
sub d1,d2 # decrease size remaining
add d1,a2 # increase next start address
/* check invalidating of end address
*
* a2 = a2 + invsize
* if (a2 < end) {
* goto dcivloop;
* } */
cmp a1,a2
bns dcivloop
LOCAL_IRQ_RESTORE(d3)
mn10300_local_dcache_inv_range_end:
ret [d2,d3,a2],12
.size mn10300_local_dcache_inv_page,.-mn10300_local_dcache_inv_page
.size mn10300_local_dcache_inv_range,.-mn10300_local_dcache_inv_range
.size mn10300_local_dcache_inv_range2,.-mn10300_local_dcache_inv_range2
###############################################################################
#
# void mn10300_local_icache_inv_page(unsigned long start)
# void mn10300_local_icache_inv_range2(unsigned long start, unsigned long size)
# void mn10300_local_icache_inv_range(unsigned long start, unsigned long end)
# Invalidate a range of addresses on a page in the icache
#
###############################################################################
ALIGN
.globl mn10300_local_icache_inv_page
.globl mn10300_local_icache_inv_range
.globl mn10300_local_icache_inv_range2
.type mn10300_local_icache_inv_page,@function
.type mn10300_local_icache_inv_range,@function
.type mn10300_local_icache_inv_range2,@function
mn10300_local_icache_inv_page:
and ~(PAGE_SIZE-1),d0
mov PAGE_SIZE,d1
mn10300_local_icache_inv_range2:
add d0,d1
mn10300_local_icache_inv_range:
movm [d2,d3,a2],(sp)
mov CHCTR,a0
movhu (a0),d2
btst CHCTR_ICEN,d2
beq mn10300_local_icache_inv_range_reg_end
/* calculate alignsize
*
* alignsize = L1_CACHE_BYTES;
* for (i = (end - start - 1) / L1_CACHE_BYTES ; i > 0; i >>= 1) {
* alignsize <<= 1;
* }
* d2 = alignsize;
*/
mov L1_CACHE_BYTES,d2
sub d0,d1,d3
add -1,d3
lsr L1_CACHE_SHIFT,d3
beq 2f
1:
add d2,d2
lsr 1,d3
bne 1b
2:
/* a1 = end */
mov d1,a1
LOCAL_CLI_SAVE(d3)
mov ICIVCR,a0
/* wait for busy bit of area invalidation */
setlb
mov (a0),d1
btst ICIVCR_ICIVBSY,d1
lne
/* set mask
*
* mask = ~(alignsize-1);
* ICIVMR = mask;
*/
mov d2,d1
add -1,d1
not d1
mov d1,(ICIVMR)
/* a2 = mask & start */
and d1,d0,a2
icivloop:
/* area invalidate
*
* ICIVCR = (mask & start) | ICIVCR_ICI
*/
mov a2,d0
or ICIVCR_ICI,d0
mov d0,(a0)
/* wait for busy bit of area invalidation */
setlb
mov (a0),d1
btst ICIVCR_ICIVBSY,d1
lne
/* check invalidating of end address
*
* a2 = a2 + alignsize
* if (a2 < end) {
* goto icivloop;
* } */
add d2,a2
cmp a1,a2
bns icivloop
LOCAL_IRQ_RESTORE(d3)
mn10300_local_icache_inv_range_reg_end:
ret [d2,d3,a2],12
.size mn10300_local_icache_inv_page,.-mn10300_local_icache_inv_page
.size mn10300_local_icache_inv_range,.-mn10300_local_icache_inv_range
.size mn10300_local_icache_inv_range2,.-mn10300_local_icache_inv_range2