40bdac7dbc
This is an implementation of a suggestion made by Chris Torek:
--------------------
Something else I noticed in passing: the EX and EX_LD/EX_ST macros
scattered throughout the various .S files make a fair bit of .fixup
code, all of which does the same thing. At the cost of one symbol
in copy_in_user.S, you could just have one common two-instruction
retl-and-mov-1 fixup that they all share.
--------------------
The following is with a defconfig build:
text data bss dec hex filename
3972767 344024 584449 4901240 4ac978 vmlinux.orig
3968887 344024 584449 4897360 4aba50 vmlinux
Signed-off-by: David S. Miller <davem@davemloft.net>
155 lines
3.5 KiB
ArmAsm
155 lines
3.5 KiB
ArmAsm
/* bzero.S: Simple prefetching memset, bzero, and clear_user
|
|
* implementations.
|
|
*
|
|
* Copyright (C) 2005 David S. Miller <davem@davemloft.net>
|
|
*/
|
|
|
|
.text
|
|
|
|
.globl __memset
|
|
.type __memset, #function
|
|
__memset: /* %o0=buf, %o1=pat, %o2=len */
|
|
|
|
.globl memset
|
|
.type memset, #function
|
|
memset: /* %o0=buf, %o1=pat, %o2=len */
|
|
and %o1, 0xff, %o3
|
|
mov %o2, %o1
|
|
sllx %o3, 8, %g1
|
|
or %g1, %o3, %o2
|
|
sllx %o2, 16, %g1
|
|
or %g1, %o2, %o2
|
|
sllx %o2, 32, %g1
|
|
ba,pt %xcc, 1f
|
|
or %g1, %o2, %o2
|
|
|
|
.globl __bzero
|
|
.type __bzero, #function
|
|
__bzero: /* %o0=buf, %o1=len */
|
|
clr %o2
|
|
1: mov %o0, %o3
|
|
brz,pn %o1, __bzero_done
|
|
cmp %o1, 16
|
|
bl,pn %icc, __bzero_tiny
|
|
prefetch [%o0 + 0x000], #n_writes
|
|
andcc %o0, 0x3, %g0
|
|
be,pt %icc, 2f
|
|
1: stb %o2, [%o0 + 0x00]
|
|
add %o0, 1, %o0
|
|
andcc %o0, 0x3, %g0
|
|
bne,pn %icc, 1b
|
|
sub %o1, 1, %o1
|
|
2: andcc %o0, 0x7, %g0
|
|
be,pt %icc, 3f
|
|
stw %o2, [%o0 + 0x00]
|
|
sub %o1, 4, %o1
|
|
add %o0, 4, %o0
|
|
3: and %o1, 0x38, %g1
|
|
cmp %o1, 0x40
|
|
andn %o1, 0x3f, %o4
|
|
bl,pn %icc, 5f
|
|
and %o1, 0x7, %o1
|
|
prefetch [%o0 + 0x040], #n_writes
|
|
prefetch [%o0 + 0x080], #n_writes
|
|
prefetch [%o0 + 0x0c0], #n_writes
|
|
prefetch [%o0 + 0x100], #n_writes
|
|
prefetch [%o0 + 0x140], #n_writes
|
|
4: prefetch [%o0 + 0x180], #n_writes
|
|
stx %o2, [%o0 + 0x00]
|
|
stx %o2, [%o0 + 0x08]
|
|
stx %o2, [%o0 + 0x10]
|
|
stx %o2, [%o0 + 0x18]
|
|
stx %o2, [%o0 + 0x20]
|
|
stx %o2, [%o0 + 0x28]
|
|
stx %o2, [%o0 + 0x30]
|
|
stx %o2, [%o0 + 0x38]
|
|
subcc %o4, 0x40, %o4
|
|
bne,pt %icc, 4b
|
|
add %o0, 0x40, %o0
|
|
brz,pn %g1, 6f
|
|
nop
|
|
5: stx %o2, [%o0 + 0x00]
|
|
subcc %g1, 8, %g1
|
|
bne,pt %icc, 5b
|
|
add %o0, 0x8, %o0
|
|
6: brz,pt %o1, __bzero_done
|
|
nop
|
|
__bzero_tiny:
|
|
1: stb %o2, [%o0 + 0x00]
|
|
subcc %o1, 1, %o1
|
|
bne,pt %icc, 1b
|
|
add %o0, 1, %o0
|
|
__bzero_done:
|
|
retl
|
|
mov %o3, %o0
|
|
.size __bzero, .-__bzero
|
|
.size __memset, .-__memset
|
|
.size memset, .-memset
|
|
|
|
#define EX_ST(x,y) \
|
|
98: x,y; \
|
|
.section __ex_table,"a";\
|
|
.align 4; \
|
|
.word 98b, __retl_o1; \
|
|
.text; \
|
|
.align 4;
|
|
|
|
.globl __clear_user
|
|
.type __clear_user, #function
|
|
__clear_user: /* %o0=buf, %o1=len */
|
|
brz,pn %o1, __clear_user_done
|
|
cmp %o1, 16
|
|
bl,pn %icc, __clear_user_tiny
|
|
EX_ST(prefetcha [%o0 + 0x00] %asi, #n_writes)
|
|
andcc %o0, 0x3, %g0
|
|
be,pt %icc, 2f
|
|
1: EX_ST(stba %g0, [%o0 + 0x00] %asi)
|
|
add %o0, 1, %o0
|
|
andcc %o0, 0x3, %g0
|
|
bne,pn %icc, 1b
|
|
sub %o1, 1, %o1
|
|
2: andcc %o0, 0x7, %g0
|
|
be,pt %icc, 3f
|
|
EX_ST(stwa %g0, [%o0 + 0x00] %asi)
|
|
sub %o1, 4, %o1
|
|
add %o0, 4, %o0
|
|
3: and %o1, 0x38, %g1
|
|
cmp %o1, 0x40
|
|
andn %o1, 0x3f, %o4
|
|
bl,pn %icc, 5f
|
|
and %o1, 0x7, %o1
|
|
EX_ST(prefetcha [%o0 + 0x040] %asi, #n_writes)
|
|
EX_ST(prefetcha [%o0 + 0x080] %asi, #n_writes)
|
|
EX_ST(prefetcha [%o0 + 0x0c0] %asi, #n_writes)
|
|
EX_ST(prefetcha [%o0 + 0x100] %asi, #n_writes)
|
|
EX_ST(prefetcha [%o0 + 0x140] %asi, #n_writes)
|
|
4: EX_ST(prefetcha [%o0 + 0x180] %asi, #n_writes)
|
|
EX_ST(stxa %g0, [%o0 + 0x00] %asi)
|
|
EX_ST(stxa %g0, [%o0 + 0x08] %asi)
|
|
EX_ST(stxa %g0, [%o0 + 0x10] %asi)
|
|
EX_ST(stxa %g0, [%o0 + 0x18] %asi)
|
|
EX_ST(stxa %g0, [%o0 + 0x20] %asi)
|
|
EX_ST(stxa %g0, [%o0 + 0x28] %asi)
|
|
EX_ST(stxa %g0, [%o0 + 0x30] %asi)
|
|
EX_ST(stxa %g0, [%o0 + 0x38] %asi)
|
|
subcc %o4, 0x40, %o4
|
|
bne,pt %icc, 4b
|
|
add %o0, 0x40, %o0
|
|
brz,pn %g1, 6f
|
|
nop
|
|
5: EX_ST(stxa %g0, [%o0 + 0x00] %asi)
|
|
subcc %g1, 8, %g1
|
|
bne,pt %icc, 5b
|
|
add %o0, 0x8, %o0
|
|
6: brz,pt %o1, __clear_user_done
|
|
nop
|
|
__clear_user_tiny:
|
|
1: EX_ST(stba %g0, [%o0 + 0x00] %asi)
|
|
subcc %o1, 1, %o1
|
|
bne,pt %icc, 1b
|
|
add %o0, 1, %o0
|
|
__clear_user_done:
|
|
retl
|
|
clr %o0
|
|
.size __clear_user, .-__clear_user
|