/* memset.S: optimised assembly memset
*
* Copyright (C) 2003, 2004 Red Hat, Inc. All Rights Reserved.
* Written by David Howells (dhowells@redhat.com)
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, see
* .
*/
#include
.text
.p2align 4
###############################################################################
#
# void *memset(void *p, char ch, size_t count)
#
# - NOTE: must not use any stack. exception detection performs function return
# to caller's fixup routine, aborting the remainder of the set
# GR4, GR7, GR8, and GR11 must be managed
#
###############################################################################
.globl memset
.type memset,@function
memset:
orcc.p gr10,gr0,gr5,icc3 ; GR5 = count
andi gr9,#0xff,gr9
or.p gr8,gr0,gr4 ; GR4 = address
beqlr icc3,#0
# conditionally write a byte to 2b-align the address
setlos.p #1,gr6
andicc gr4,#1,gr0,icc0
ckne icc0,cc7
cstb.p gr9,@(gr4,gr0) ,cc7,#1
csubcc gr5,gr6,gr5 ,cc7,#1 ; also set ICC3
cadd.p gr4,gr6,gr4 ,cc7,#1
beqlr icc3,#0
# conditionally write a word to 4b-align the address
andicc.p gr4,#2,gr0,icc0
subicc gr5,#2,gr0,icc1
setlos.p #2,gr6
ckne icc0,cc7
slli.p gr9,#8,gr12 ; need to double up the pattern
cknc icc1,cc5
or.p gr9,gr12,gr12
andcr cc7,cc5,cc7
csth.p gr12,@(gr4,gr0) ,cc7,#1
csubcc gr5,gr6,gr5 ,cc7,#1 ; also set ICC3
cadd.p gr4,gr6,gr4 ,cc7,#1
beqlr icc3,#0
# conditionally write a dword to 8b-align the address
andicc.p gr4,#4,gr0,icc0
subicc gr5,#4,gr0,icc1
setlos.p #4,gr6
ckne icc0,cc7
slli.p gr12,#16,gr13 ; need to quadruple-up the pattern
cknc icc1,cc5
or.p gr13,gr12,gr12
andcr cc7,cc5,cc7
cst.p gr12,@(gr4,gr0) ,cc7,#1
csubcc gr5,gr6,gr5 ,cc7,#1 ; also set ICC3
cadd.p gr4,gr6,gr4 ,cc7,#1
beqlr icc3,#0
or.p gr12,gr12,gr13 ; need to octuple-up the pattern
# the address is now 8b-aligned - loop around writing 64b chunks
setlos #8,gr7
subi.p gr4,#8,gr4 ; store with update index does weird stuff
setlos #64,gr6
subicc gr5,#64,gr0,icc0
0: cknc icc0,cc7
cstdu gr12,@(gr4,gr7) ,cc7,#1
cstdu gr12,@(gr4,gr7) ,cc7,#1
cstdu gr12,@(gr4,gr7) ,cc7,#1
cstdu gr12,@(gr4,gr7) ,cc7,#1
cstdu gr12,@(gr4,gr7) ,cc7,#1
cstdu.p gr12,@(gr4,gr7) ,cc7,#1
csubcc gr5,gr6,gr5 ,cc7,#1 ; also set ICC3
cstdu.p gr12,@(gr4,gr7) ,cc7,#1
subicc gr5,#64,gr0,icc0
cstdu.p gr12,@(gr4,gr7) ,cc7,#1
beqlr icc3,#0
bnc icc0,#2,0b
# now do 32-byte remnant
subicc.p gr5,#32,gr0,icc0
setlos #32,gr6
cknc icc0,cc7
cstdu.p gr12,@(gr4,gr7) ,cc7,#1
csubcc gr5,gr6,gr5 ,cc7,#1 ; also set ICC3
cstdu.p gr12,@(gr4,gr7) ,cc7,#1
setlos #16,gr6
cstdu.p gr12,@(gr4,gr7) ,cc7,#1
subicc gr5,#16,gr0,icc0
cstdu.p gr12,@(gr4,gr7) ,cc7,#1
beqlr icc3,#0
# now do 16-byte remnant
cknc icc0,cc7
cstdu.p gr12,@(gr4,gr7) ,cc7,#1
csubcc gr5,gr6,gr5 ,cc7,#1 ; also set ICC3
cstdu.p gr12,@(gr4,gr7) ,cc7,#1
beqlr icc3,#0
# now do 8-byte remnant
subicc gr5,#8,gr0,icc1
cknc icc1,cc7
cstdu.p gr12,@(gr4,gr7) ,cc7,#1
csubcc gr5,gr7,gr5 ,cc7,#1 ; also set ICC3
setlos.p #4,gr7
beqlr icc3,#0
# now do 4-byte remnant
subicc gr5,#4,gr0,icc0
addi.p gr4,#4,gr4
cknc icc0,cc7
cstu.p gr12,@(gr4,gr7) ,cc7,#1
csubcc gr5,gr7,gr5 ,cc7,#1 ; also set ICC3
subicc.p gr5,#2,gr0,icc1
beqlr icc3,#0
# now do 2-byte remnant
setlos #2,gr7
addi.p gr4,#2,gr4
cknc icc1,cc7
csthu.p gr12,@(gr4,gr7) ,cc7,#1
csubcc gr5,gr7,gr5 ,cc7,#1 ; also set ICC3
subicc.p gr5,#1,gr0,icc0
beqlr icc3,#0
# now do 1-byte remnant
setlos #0,gr7
addi.p gr4,#2,gr4
cknc icc0,cc7
cstb.p gr12,@(gr4,gr0) ,cc7,#1
bralr
.size memset, .-memset
libc_hidden_def(memset)