Rev 1663 | Rev 2071 | Go to most recent revision | Blame | Compare with Previous | Last modification | View Log | Download | RSS feed
#
# Copyright (C) 2005 Jakub Jermar
# All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions
# are met:
#
# - Redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer.
# - Redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in the
# documentation and/or other materials provided with the distribution.
# - The name of the author may not be used to endorse or promote products
# derived from this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
# IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
# OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
# IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
# INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
# NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
# THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#
.text
.global context_save
.global context_restore
context_save:
alloc loc0 = ar.pfs, 1, 8, 0, 0
mov loc1 = ar.unat ;;
/* loc2 */
mov loc3 = ar.rsc
.auto
/*
* Flush dirty registers to backing store.
* After this ar.bsp and ar.bspstore are equal.
*/
flushrs
mov loc4 = ar.bsp
/*
* Put RSE to enforced lazy mode.
* So that ar.rnat can be read.
*/
and loc5 = ~3, loc3
mov ar.rsc = loc5
mov loc5 = ar.rnat
.explicit
mov loc6 = ar.lc
/*
* Save application registers
*/
st8 [in0] = loc0, 8 ;; /* save ar.pfs */
st8 [in0] = loc1, 8 ;; /* save ar.unat (caller) */
mov loc2 = in0 ;;
add in0 = 8, in0 ;; /* skip ar.unat (callee) */
st8 [in0] = loc3, 8 ;; /* save ar.rsc */
st8 [in0] = loc4, 8 ;; /* save ar.bsp */
st8 [in0] = loc5, 8 ;; /* save ar.rnat */
st8 [in0] = loc6, 8 ;; /* save ar.lc */
/*
* Save general registers including NaT bits
*/
st8.spill [in0] = r1, 8 ;;
st8.spill [in0] = r4, 8 ;;
st8.spill [in0] = r5, 8 ;;
st8.spill [in0] = r6, 8 ;;
st8.spill [in0] = r7, 8 ;;
st8.spill [in0] = r12, 8 ;; /* save sp */
st8.spill [in0] = r13, 8 ;; /* save tp */
mov loc3 = ar.unat ;;
st8 [loc2] = loc3 /* save ar.unat (callee) */
/*
* Save branch registers
*/
mov loc2 = b0 ;;
st8 [in0] = loc2, 8 /* save pc */
mov loc3 = b1 ;;
st8 [in0] = loc3, 8
mov loc4 = b2 ;;
st8 [in0] = loc4, 8
mov loc5 = b3 ;;
st8 [in0] = loc5, 8
mov loc6 = b4 ;;
st8 [in0] = loc6, 8
mov loc7 = b5 ;;
st8 [in0] = loc7, 8
/*
* Save predicate registers
*/
mov loc2 = pr ;;
st8 [in0] = loc2, 16;; /* Next fpu registers should be spilled to 16B aligned address */
/*
* Save floating-point registers.
*/
stf.spill [in0] = f2, 16 ;;
stf.spill [in0] = f3, 16 ;;
stf.spill [in0] = f4, 16 ;;
stf.spill [in0] = f5, 16 ;;
stf.spill [in0] = f16, 16 ;;
stf.spill [in0] = f17, 16 ;;
stf.spill [in0] = f18, 16 ;;
stf.spill [in0] = f19, 16 ;;
stf.spill [in0] = f20, 16 ;;
stf.spill [in0] = f21, 16 ;;
stf.spill [in0] = f22, 16 ;;
stf.spill [in0] = f23, 16 ;;
stf.spill [in0] = f24, 16 ;;
stf.spill [in0] = f25, 16 ;;
stf.spill [in0] = f26, 16 ;;
stf.spill [in0] = f27, 16 ;;
stf.spill [in0] = f28, 16 ;;
stf.spill [in0] = f29, 16 ;;
stf.spill [in0] = f30, 16 ;;
stf.spill [in0] = f31, 16 ;;
mov ar.unat = loc1
add r8 = r0, r0, 1 /* context_save returns 1 */
br.ret.sptk.many b0
context_restore:
alloc loc0 = ar.pfs, 1, 9, 0, 0 ;;
ld8 loc0 = [in0], 8 ;; /* load ar.pfs */
ld8 loc1 = [in0], 8 ;; /* load ar.unat (caller) */
ld8 loc2 = [in0], 8 ;; /* load ar.unat (callee) */
ld8 loc3 = [in0], 8 ;; /* load ar.rsc */
ld8 loc4 = [in0], 8 ;; /* load ar.bsp */
ld8 loc5 = [in0], 8 ;; /* load ar.rnat */
ld8 loc6 = [in0], 8 ;; /* load ar.lc */
.auto
/*
* Invalidate the ALAT
*/
invala
/*
* Put RSE to enforced lazy mode.
* So that ar.bspstore and ar.rnat can be written.
*/
movl loc8 = ~3
and loc8 = loc3, loc8
mov ar.rsc = loc8
/*
* Flush dirty registers to backing store.
* We do this because we want the following move
* to ar.bspstore to assign the same value to ar.bsp.
*/
flushrs
/*
* Restore application registers
*/
mov ar.bspstore = loc4 /* rse.bspload = ar.bsp = ar.bspstore = loc4 */
mov ar.rnat = loc5
mov ar.pfs = loc0
mov ar.rsc = loc3
.explicit
mov ar.unat = loc2 ;;
mov ar.lc = loc6
/*
* Restore general registers including NaT bits
*/
ld8.fill r1 = [in0], 8 ;;
ld8.fill r4 = [in0], 8 ;;
ld8.fill r5 = [in0], 8 ;;
ld8.fill r6 = [in0], 8 ;;
ld8.fill r7 = [in0], 8 ;;
ld8.fill r12 = [in0], 8 ;; /* restore sp */
ld8.fill r13 = [in0], 8 ;;
/*
* Restore branch registers
*/
ld8 loc2 = [in0], 8 ;; /* restore pc */
mov b0 = loc2
ld8 loc3 = [in0], 8 ;;
mov b1 = loc3
ld8 loc4 = [in0], 8 ;;
mov b2 = loc4
ld8 loc5 = [in0], 8 ;;
mov b3 = loc5
ld8 loc6 = [in0], 8 ;;
mov b4 = loc6
ld8 loc7 = [in0], 8 ;;
mov b5 = loc7
/*
* Restore predicate registers
*/
ld8 loc2 = [in0], 16 ;;
mov pr = loc2, ~0
/*
* Restore floating-point registers.
*/
ldf.fill f2 = [in0], 16 ;;
ldf.fill f3 = [in0], 16 ;;
ldf.fill f4 = [in0], 16 ;;
ldf.fill f5 = [in0], 16 ;;
ldf.fill f16 = [in0], 16 ;;
ldf.fill f17 = [in0], 16 ;;
ldf.fill f18 = [in0], 16 ;;
ldf.fill f19 = [in0], 16 ;;
ldf.fill f20 = [in0], 16 ;;
ldf.fill f21 = [in0], 16 ;;
ldf.fill f22 = [in0], 16 ;;
ldf.fill f23 = [in0], 16 ;;
ldf.fill f24 = [in0], 16 ;;
ldf.fill f25 = [in0], 16 ;;
ldf.fill f26 = [in0], 16 ;;
ldf.fill f27 = [in0], 16 ;;
ldf.fill f28 = [in0], 16 ;;
ldf.fill f29 = [in0], 16 ;;
ldf.fill f30 = [in0], 16 ;;
ldf.fill f31 = [in0], 16 ;;
mov ar.unat = loc1
mov r8 = r0 /* context_restore returns 0 */
br.ret.sptk.many b0