1
0
mirror of https://git.FreeBSD.org/src.git synced 2024-12-17 10:26:15 +00:00

Implement bzero et memset in support.S

This commit is contained in:
Olivier Houchard 2004-05-14 23:31:10 +00:00
parent bb52bd2f06
commit c8d8367098
Notes: svn2git 2020-12-20 02:59:44 +00:00
svn path=/head/; revision=129250

View File

@ -70,3 +70,164 @@ ENTRY(disable_intr)
ENTRY(enable_intr)
IRQenableALL
/*
* memset: Sets a block of memory to the specified value
*
* On entry:
* r0 - dest address
* r1 - byte to write
* r2 - number of bytes to write
*
* On exit:
* r0 - dest address
*/
/* LINTSTUB: Func: void bzero(void *, size_t) */
ENTRY(bzero)
mov r3, #0x00
b do_memset
/* LINTSTUB: Func: void *memset(void *, int, size_t) */
ENTRY(memset)
and r3, r1, #0xff /* We deal with bytes */
mov r1, r2
do_memset:
cmp r1, #0x04 /* Do we have less than 4 bytes */
mov ip, r0
blt .Lmemset_lessthanfour
/* Ok first we will word align the address */
ands r2, ip, #0x03 /* Get the bottom two bits */
bne .Lmemset_wordunaligned /* The address is not word aligned */
/* We are now word aligned */
.Lmemset_wordaligned:
orr r3, r3, r3, lsl #8 /* Extend value to 16-bits */
#ifdef __XSCALE__
tst ip, #0x04 /* Quad-align for Xscale */
#else
cmp r1, #0x10
#endif
orr r3, r3, r3, lsl #16 /* Extend value to 32-bits */
#ifdef __XSCALE__
subne r1, r1, #0x04 /* Quad-align if necessary */
strne r3, [ip], #0x04
cmp r1, #0x10
#endif
blt .Lmemset_loop4 /* If less than 16 then use words */
mov r2, r3 /* Duplicate data */
cmp r1, #0x80 /* If < 128 then skip the big loop */
blt .Lmemset_loop32
/* Do 128 bytes at a time */
.Lmemset_loop128:
subs r1, r1, #0x80
#ifdef __XSCALE__
strged r2, [ip], #0x08
strged r2, [ip], #0x08
strged r2, [ip], #0x08
strged r2, [ip], #0x08
strged r2, [ip], #0x08
strged r2, [ip], #0x08
strged r2, [ip], #0x08
strged r2, [ip], #0x08
strged r2, [ip], #0x08
strged r2, [ip], #0x08
strged r2, [ip], #0x08
strged r2, [ip], #0x08
strged r2, [ip], #0x08
strged r2, [ip], #0x08
strged r2, [ip], #0x08
strged r2, [ip], #0x08
#else
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
#endif
bgt .Lmemset_loop128
moveq pc, lr /* Zero length so just exit */
add r1, r1, #0x80 /* Adjust for extra sub */
/* Do 32 bytes at a time */
.Lmemset_loop32:
subs r1, r1, #0x20
#ifdef __XSCALE__
strged r2, [ip], #0x08
strged r2, [ip], #0x08
strged r2, [ip], #0x08
strged r2, [ip], #0x08
#else
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
#endif
bgt .Lmemset_loop32
moveq pc, lr /* Zero length so just exit */
adds r1, r1, #0x10 /* Partially adjust for extra sub */
/* Deal with 16 bytes or more */
#ifdef __XSCALE__
strged r2, [ip], #0x08
strged r2, [ip], #0x08
#else
stmgeia ip!, {r2-r3}
stmgeia ip!, {r2-r3}
#endif
moveq pc, lr /* Zero length so just exit */
addlt r1, r1, #0x10 /* Possibly adjust for extra sub */
/* We have at least 4 bytes so copy as words */
.Lmemset_loop4:
subs r1, r1, #0x04
strge r3, [ip], #0x04
bgt .Lmemset_loop4
moveq pc, lr /* Zero length so just exit */
#ifdef __XSCALE__
/* Compensate for 64-bit alignment check */
adds r1, r1, #0x04
moveq pc, lr
cmp r1, #2
#else
cmp r1, #-2
#endif
strb r3, [ip], #0x01 /* Set 1 byte */
strgeb r3, [ip], #0x01 /* Set another byte */
strgtb r3, [ip] /* and a third */
mov pc, lr /* Exit */
.Lmemset_wordunaligned:
rsb r2, r2, #0x004
strb r3, [ip], #0x01 /* Set 1 byte */
cmp r2, #0x02
strgeb r3, [ip], #0x01 /* Set another byte */
sub r1, r1, r2
strgtb r3, [ip], #0x01 /* and a third */
cmp r1, #0x04 /* More than 4 bytes left? */
bge .Lmemset_wordaligned /* Yup */
.Lmemset_lessthanfour:
cmp r1, #0x00
moveq pc, lr /* Zero length so exit */
strb r3, [ip], #0x01 /* Set 1 byte */
cmp r1, #0x02
strgeb r3, [ip], #0x01 /* Set another byte */
strgtb r3, [ip] /* and a third */
mov pc, lr /* Exit */