qiurui
/
glibc
mirror of git://sourceware.org/git/glibc.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Don't use %r19 (linkage table pointer).

This commit is contained in:
Ulrich Drepper 2000-10-15 03:22:34 +00:00
parent 257c3b5223
commit acc97172a1
8 changed files with 319 additions and 310 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

98
sysdeps/hppa/hppa1.1/addmul_1.s
View File

@ -1,71 +1,72 @@
; HP-PA-1.1 __mpn_addmul_1 -- Multiply a limb vector with a limb and ;! HP-PA-1.1 __mpn_addmul_1 -- Multiply a limb vector with a limb and
; add the result to a second limb vector. ;! add the result to a second limb vector.
; Copyright (C) 1992, 1993, 1994 Free Software Foundation, Inc. ;! Copyright (C) 1992, 1993, 1994 Free Software Foundation, Inc.
; This file is part of the GNU MP Library. ;! This file is part of the GNU MP Library.
; The GNU MP Library is free software; you can redistribute it and/or modify ;! The GNU MP 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 ;! 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 ;! the Free Software Foundation; either version 2 of the License, or (at your
; option) any later version. ;! option) any later version.
; The GNU MP Library is distributed in the hope that it will be useful, but ;! The GNU MP Library is distributed in the hope that it will be useful, but
; WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY ;! WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
; or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public ;! or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
; License for more details. ;! License for more details.
; You should have received a copy of the GNU Library General Public License ;! You should have received a copy of the GNU Library General Public License
; along with the GNU MP Library; see the file COPYING.LIB. If not, write to ;! along with the GNU MP Library; see the file COPYING.LIB. If not, write to
; the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, ;! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
; MA 02111-1307, USA. ;! MA 02111-1307, USA.
; INPUT PARAMETERS ;! INPUT PARAMETERS
; res_ptr r26 ;! res_ptr r26
; s1_ptr r25 ;! s1_ptr r25
; size r24 ;! size r24
; s2_limb r23 ;! s2_limb r23
; This runs at 11 cycles/limb on a PA7000. With the used instructions, it ;! This runs at 11 cycles/limb on a PA7000. With the used instructions, it
; can not become faster due to data cache contention after a store. On the ;! can not become faster due to data cache contention after a store. On the
; PA7100 it runs at 10 cycles/limb, and that can not be improved either, ;! PA7100 it runs at 10 cycles/limb, and that can not be improved either,
; since only the xmpyu does not need the integer pipeline, so the only ;! since only the xmpyu does not need the integer pipeline, so the only
; dual-issue we will get are addc+xmpyu. Unrolling could gain a cycle/limb ;! dual-issue we will get are addc+xmpyu. Unrolling could gain a cycle/limb
; on the PA7100. ;! on the PA7100.
; There are some ideas described in mul_1.s that applies to this code too. ;! There are some ideas described in mul_1.s that applies to this code too.
.code .text
.export __mpn_addmul_1 .export __mpn_addmul_1
__mpn_addmul_1 __mpn_addmul_1:
.proc .proc
.callinfo frame=64,no_calls .callinfo frame=64,no_calls
.entry .entry
ldo 64(%r30),%r30 ldo 64(%r30),%r30
fldws,ma 4(%r25),%fr5 fldws,ma 4(%r25),%fr5
stw %r23,-16(%r30) ; move s2_limb ... stw %r23,-16(%r30) ;! move s2_limb ...
addib,= -1,%r24,L$just_one_limb addib,= -1,%r24,L$just_one_limb
fldws -16(%r30),%fr4 ; ... into fr4 fldws -16(%r30),%fr4 ;! ... into fr4
add %r0,%r0,%r0 ; clear carry add %r0,%r0,%r0 ;! clear carry
xmpyu %fr4,%fr5,%fr6 xmpyu %fr4,%fr5,%fr6
fldws,ma 4(%r25),%fr7 fldws,ma 4(%r25),%fr7
fstds %fr6,-16(%r30) fstds %fr6,-16(%r30)
xmpyu %fr4,%fr7,%fr8 xmpyu %fr4,%fr7,%fr8
ldw -12(%r30),%r19 ; least significant limb in product ldw -12(%r30),%r20 ;! least significant limb in product
ldw -16(%r30),%r28 ldw -16(%r30),%r28
fstds %fr8,-16(%r30) fstds %fr8,-16(%r30)
addib,= -1,%r24,L$end addib,= -1,%r24,L$end
ldw -12(%r30),%r1 ldw -12(%r30),%r1
; Main loop ;! Main loop
L$loop ldws 0(%r26),%r29 L$loop:
ldws 0(%r26),%r29
fldws,ma 4(%r25),%fr5 fldws,ma 4(%r25),%fr5
add %r29,%r19,%r19 add %r29,%r20,%r20
stws,ma %r19,4(%r26) stws,ma %r20,4(%r26)
addc %r28,%r1,%r19 addc %r28,%r1,%r20
xmpyu %fr4,%fr5,%fr6 xmpyu %fr4,%fr5,%fr6
ldw -16(%r30),%r28 ldw -16(%r30),%r28
fstds %fr6,-16(%r30) fstds %fr6,-16(%r30)
@ -73,27 +74,28 @@ L$loop ldws 0(%r26),%r29
addib,<> -1,%r24,L$loop addib,<> -1,%r24,L$loop
ldw -12(%r30),%r1 ldw -12(%r30),%r1
L$end ldw 0(%r26),%r29 L$end:
add %r29,%r19,%r19 ldw 0(%r26),%r29
stws,ma %r19,4(%r26) add %r29,%r20,%r20
addc %r28,%r1,%r19 stws,ma %r20,4(%r26)
addc %r28,%r1,%r20
ldw -16(%r30),%r28 ldw -16(%r30),%r28
ldws 0(%r26),%r29 ldws 0(%r26),%r29
addc %r0,%r28,%r28 addc %r0,%r28,%r28
add %r29,%r19,%r19 add %r29,%r20,%r20
stws,ma %r19,4(%r26) stws,ma %r20,4(%r26)
addc %r0,%r28,%r28 addc %r0,%r28,%r28
bv 0(%r2) bv 0(%r2)
ldo -64(%r30),%r30 ldo -64(%r30),%r30
L$just_one_limb L$just_one_limb:
xmpyu %fr4,%fr5,%fr6 xmpyu %fr4,%fr5,%fr6
ldw 0(%r26),%r29 ldw 0(%r26),%r29
fstds %fr6,-16(%r30) fstds %fr6,-16(%r30)
ldw -12(%r30),%r1 ldw -12(%r30),%r1
ldw -16(%r30),%r28 ldw -16(%r30),%r28
add %r29,%r1,%r19 add %r29,%r1,%r20
stw %r19,0(%r26) stw %r20,0(%r26)
addc %r0,%r28,%r28 addc %r0,%r28,%r28
bv 0(%r2) bv 0(%r2)
ldo -64(%r30),%r30 ldo -64(%r30),%r30

100
sysdeps/hppa/hppa1.1/mul_1.s
View File

@ -1,92 +1,94 @@
; HP-PA-1.1 __mpn_mul_1 -- Multiply a limb vector with a limb and store ;! HP-PA-1.1 __mpn_mul_1 -- Multiply a limb vector with a limb and store
; the result in a second limb vector. ;! the result in a second limb vector.
; Copyright (C) 1992, 1993, 1994 Free Software Foundation, Inc. ;! Copyright (C) 1992, 1993, 1994 Free Software Foundation, Inc.
; This file is part of the GNU MP Library. ;! This file is part of the GNU MP Library.
; The GNU MP Library is free software; you can redistribute it and/or modify ;! The GNU MP 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 ;! 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 ;! the Free Software Foundation; either version 2 of the License, or (at your
; option) any later version. ;! option) any later version.
; The GNU MP Library is distributed in the hope that it will be useful, but ;! The GNU MP Library is distributed in the hope that it will be useful, but
; WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY ;! WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
; or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public ;! or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
; License for more details. ;! License for more details.
; You should have received a copy of the GNU Library General Public License ;! You should have received a copy of the GNU Library General Public License
; along with the GNU MP Library; see the file COPYING.LIB. If not, write to ;! along with the GNU MP Library; see the file COPYING.LIB. If not, write to
; the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, ;! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
; MA 02111-1307, USA. ;! MA 02111-1307, USA.
; INPUT PARAMETERS ;! INPUT PARAMETERS
; res_ptr r26 ;! res_ptr r26
; s1_ptr r25 ;! s1_ptr r25
; size r24 ;! size r24
; s2_limb r23 ;! s2_limb r23
; This runs at 9 cycles/limb on a PA7000. With the used instructions, it can ;! This runs at 9 cycles/limb on a PA7000. With the used instructions, it can
; not become faster due to data cache contention after a store. On the ;! not become faster due to data cache contention after a store. On the
; PA7100 it runs at 7 cycles/limb, and that can not be improved either, since ;! PA7100 it runs at 7 cycles/limb, and that can not be improved either, since
; only the xmpyu does not need the integer pipeline, so the only dual-issue ;! only the xmpyu does not need the integer pipeline, so the only dual-issue
; we will get are addc+xmpyu. Unrolling would not help either CPU. ;! we will get are addc+xmpyu. Unrolling would not help either CPU.
; We could use fldds to read two limbs at a time from the S1 array, and that ;! We could use fldds to read two limbs at a time from the S1 array, and that
; could bring down the times to 8.5 and 6.5 cycles/limb for the PA7000 and ;! could bring down the times to 8.5 and 6.5 cycles/limb for the PA7000 and
; PA7100, respectively. We don't do that since it does not seem worth the ;! PA7100, respectively. We don't do that since it does not seem worth the
; (alignment) troubles... ;! (alignment) troubles...
; At least the PA7100 is rumored to be able to deal with cache-misses ;! At least the PA7100 is rumored to be able to deal with cache-misses
; without stalling instruction issue. If this is true, and the cache is ;! without stalling instruction issue. If this is true, and the cache is
; actually also lockup-free, we should use a deeper software pipeline, and ;! actually also lockup-free, we should use a deeper software pipeline, and
; load from S1 very early! (The loads and stores to -12(sp) will surely be ;! load from S1 very early; (The loads and stores to -12(sp) will surely be
; in the cache.) ;! in the cache.)
.code .text
.export __mpn_mul_1 .export __mpn_mul_1
__mpn_mul_1 __mpn_mul_1:
.proc .proc
.callinfo frame=64,no_calls .callinfo frame=64,no_calls
.entry .entry
ldo 64(%r30),%r30 ldo 64(%r30),%r30
fldws,ma 4(%r25),%fr5 fldws,ma 4(%r25),%fr5
stw %r23,-16(%r30) ; move s2_limb ... stw %r23,-16(%r30) ;! move s2_limb ...
addib,= -1,%r24,L$just_one_limb addib,= -1,%r24,L$just_one_limb
fldws -16(%r30),%fr4 ; ... into fr4 fldws -16(%r30),%fr4 ;! ... into fr4
add %r0,%r0,%r0 ; clear carry add %r0,%r0,%r0 ;! clear carry
xmpyu %fr4,%fr5,%fr6 xmpyu %fr4,%fr5,%fr6
fldws,ma 4(%r25),%fr7 fldws,ma 4(%r25),%fr7
fstds %fr6,-16(%r30) fstds %fr6,-16(%r30)
xmpyu %fr4,%fr7,%fr8 xmpyu %fr4,%fr7,%fr8
ldw -12(%r30),%r19 ; least significant limb in product ldw -12(%r30),%r20 ;! least significant limb in product
ldw -16(%r30),%r28 ldw -16(%r30),%r28
fstds %fr8,-16(%r30) fstds %fr8,-16(%r30)
addib,= -1,%r24,L$end addib,= -1,%r24,L$end
ldw -12(%r30),%r1 ldw -12(%r30),%r1
; Main loop ;! Main loop
L$loop fldws,ma 4(%r25),%fr5 L$loop:
stws,ma %r19,4(%r26) fldws,ma 4(%r25),%fr5
addc %r28,%r1,%r19 stws,ma %r20,4(%r26)
addc %r28,%r1,%r20
xmpyu %fr4,%fr5,%fr6 xmpyu %fr4,%fr5,%fr6
ldw -16(%r30),%r28 ldw -16(%r30),%r28
fstds %fr6,-16(%r30) fstds %fr6,-16(%r30)
addib,<> -1,%r24,L$loop addib,<> -1,%r24,L$loop
ldw -12(%r30),%r1 ldw -12(%r30),%r1
L$end stws,ma %r19,4(%r26) L$end:
addc %r28,%r1,%r19 stws,ma %r20,4(%r26)
addc %r28,%r1,%r20
ldw -16(%r30),%r28 ldw -16(%r30),%r28
stws,ma %r19,4(%r26) stws,ma %r20,4(%r26)
addc %r0,%r28,%r28 addc %r0,%r28,%r28
bv 0(%r2) bv 0(%r2)
ldo -64(%r30),%r30 ldo -64(%r30),%r30
L$just_one_limb L$just_one_limb:
xmpyu %fr4,%fr5,%fr6 xmpyu %fr4,%fr5,%fr6
fstds %fr6,-16(%r30) fstds %fr6,-16(%r30)
ldw -16(%r30),%r28 ldw -16(%r30),%r28

102
sysdeps/hppa/hppa1.1/submul_1.s
View File

@ -1,77 +1,78 @@
; HP-PA-1.1 __mpn_submul_1 -- Multiply a limb vector with a limb and ;! HP-PA-1.1 __mpn_submul_1 -- Multiply a limb vector with a limb and
; subtract the result from a second limb vector. ;! subtract the result from a second limb vector.
; Copyright (C) 1992, 1993, 1994 Free Software Foundation, Inc. ;! Copyright (C) 1992, 1993, 1994 Free Software Foundation, Inc.
; This file is part of the GNU MP Library. ;! This file is part of the GNU MP Library.
; The GNU MP Library is free software; you can redistribute it and/or modify ;! The GNU MP 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 ;! 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 ;! the Free Software Foundation; either version 2 of the License, or (at your
; option) any later version. ;! option) any later version.
; The GNU MP Library is distributed in the hope that it will be useful, but ;! The GNU MP Library is distributed in the hope that it will be useful, but
; WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY ;! WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
; or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public ;! or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
; License for more details. ;! License for more details.
; You should have received a copy of the GNU Library General Public License ;! You should have received a copy of the GNU Library General Public License
; along with the GNU MP Library; see the file COPYING.LIB. If not, write to ;! along with the GNU MP Library; see the file COPYING.LIB. If not, write to
; the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, ;! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
; MA 02111-1307, USA. ;! MA 02111-1307, USA.
; INPUT PARAMETERS ;! INPUT PARAMETERS
; res_ptr r26 ;! res_ptr r26
; s1_ptr r25 ;! s1_ptr r25
; size r24 ;! size r24
; s2_limb r23 ;! s2_limb r23
; This runs at 12 cycles/limb on a PA7000. With the used instructions, it ;! This runs at 12 cycles/limb on a PA7000. With the used instructions, it
; can not become faster due to data cache contention after a store. On the ;! can not become faster due to data cache contention after a store. On the
; PA7100 it runs at 11 cycles/limb, and that can not be improved either, ;! PA7100 it runs at 11 cycles/limb, and that can not be improved either,
; since only the xmpyu does not need the integer pipeline, so the only ;! since only the xmpyu does not need the integer pipeline, so the only
; dual-issue we will get are addc+xmpyu. Unrolling could gain a cycle/limb ;! dual-issue we will get are addc+xmpyu. Unrolling could gain a cycle/limb
; on the PA7100. ;! on the PA7100.
; There are some ideas described in mul_1.s that applies to this code too. ;! There are some ideas described in mul_1.s that applies to this code too.
; It seems possible to make this run as fast as __mpn_addmul_1, if we use ;! It seems possible to make this run as fast as __mpn_addmul_1, if we use
; sub,>>= %r29,%r19,%r22 ;! sub,>>= %r29,%r20,%r22
; addi 1,%r28,%r28 ;! addi 1,%r28,%r28
; but that requires reworking the hairy software pipeline... ;! but that requires reworking the hairy software pipeline...
.code .text
.export __mpn_submul_1 .export __mpn_submul_1
__mpn_submul_1 __mpn_submul_1:
.proc .proc
.callinfo frame=64,no_calls .callinfo frame=64,no_calls
.entry .entry
ldo 64(%r30),%r30 ldo 64(%r30),%r30
fldws,ma 4(%r25),%fr5 fldws,ma 4(%r25),%fr5
stw %r23,-16(%r30) ; move s2_limb ... stw %r23,-16(%r30) ;! move s2_limb ...
addib,= -1,%r24,L$just_one_limb addib,= -1,%r24,L$just_one_limb
fldws -16(%r30),%fr4 ; ... into fr4 fldws -16(%r30),%fr4 ;! ... into fr4
add %r0,%r0,%r0 ; clear carry add %r0,%r0,%r0 ;! clear carry
xmpyu %fr4,%fr5,%fr6 xmpyu %fr4,%fr5,%fr6
fldws,ma 4(%r25),%fr7 fldws,ma 4(%r25),%fr7
fstds %fr6,-16(%r30) fstds %fr6,-16(%r30)
xmpyu %fr4,%fr7,%fr8 xmpyu %fr4,%fr7,%fr8
ldw -12(%r30),%r19 ; least significant limb in product ldw -12(%r30),%r20 ;! least significant limb in product
ldw -16(%r30),%r28 ldw -16(%r30),%r28
fstds %fr8,-16(%r30) fstds %fr8,-16(%r30)
addib,= -1,%r24,L$end addib,= -1,%r24,L$end
ldw -12(%r30),%r1 ldw -12(%r30),%r1
; Main loop ;! Main loop
L$loop ldws 0(%r26),%r29 L$loop:
ldws 0(%r26),%r29
fldws,ma 4(%r25),%fr5 fldws,ma 4(%r25),%fr5
sub %r29,%r19,%r22 sub %r29,%r20,%r22
add %r22,%r19,%r0 add %r22,%r20,%r0
stws,ma %r22,4(%r26) stws,ma %r22,4(%r26)
addc %r28,%r1,%r19 addc %r28,%r1,%r20
xmpyu %fr4,%fr5,%fr6 xmpyu %fr4,%fr5,%fr6
ldw -16(%r30),%r28 ldw -16(%r30),%r28
fstds %fr6,-16(%r30) fstds %fr6,-16(%r30)
@ -79,22 +80,23 @@ L$loop ldws 0(%r26),%r29
addib,<> -1,%r24,L$loop addib,<> -1,%r24,L$loop
ldw -12(%r30),%r1 ldw -12(%r30),%r1
L$end ldw 0(%r26),%r29 L$end:
sub %r29,%r19,%r22 ldw 0(%r26),%r29
add %r22,%r19,%r0 sub %r29,%r20,%r22
add %r22,%r20,%r0
stws,ma %r22,4(%r26) stws,ma %r22,4(%r26)
addc %r28,%r1,%r19 addc %r28,%r1,%r20
ldw -16(%r30),%r28 ldw -16(%r30),%r28
ldws 0(%r26),%r29 ldws 0(%r26),%r29
addc %r0,%r28,%r28 addc %r0,%r28,%r28
sub %r29,%r19,%r22 sub %r29,%r20,%r22
add %r22,%r19,%r0 add %r22,%r20,%r0
stws,ma %r22,4(%r26) stws,ma %r22,4(%r26)
addc %r0,%r28,%r28 addc %r0,%r28,%r28
bv 0(%r2) bv 0(%r2)
ldo -64(%r30),%r30 ldo -64(%r30),%r30
L$just_one_limb L$just_one_limb:
xmpyu %fr4,%fr5,%fr6 xmpyu %fr4,%fr5,%fr6
ldw 0(%r26),%r29 ldw 0(%r26),%r29
fstds %fr6,-16(%r30) fstds %fr6,-16(%r30)

69
sysdeps/hppa/hppa1.1/udiv_qrnnd.s
View File

@ -1,53 +1,55 @@
; HP-PA __udiv_qrnnd division support, used from longlong.h. ;! HP-PA __udiv_qrnnd division support, used from longlong.h.
; This version runs fast on PA 7000 and later. ;! This version runs fast on PA 7000 and later.
; Copyright (C) 1993, 1994 Free Software Foundation, Inc. ;! Copyright (C) 1993, 1994 Free Software Foundation, Inc.
; This file is part of the GNU MP Library. ;! This file is part of the GNU MP Library.
; The GNU MP Library is free software; you can redistribute it and/or modify ;! The GNU MP 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 ;! 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 ;! the Free Software Foundation; either version 2 of the License, or (at your
; option) any later version. ;! option) any later version.
; The GNU MP Library is distributed in the hope that it will be useful, but ;! The GNU MP Library is distributed in the hope that it will be useful, but
; WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY ;! WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
; or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public ;! or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
; License for more details. ;! License for more details.
; You should have received a copy of the GNU Library General Public License ;! You should have received a copy of the GNU Library General Public License
; along with the GNU MP Library; see the file COPYING.LIB. If not, write to ;! along with the GNU MP Library; see the file COPYING.LIB. If not, write to
; the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, ;! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
; MA 02111-1307, USA. ;! MA 02111-1307, USA.
; INPUT PARAMETERS ;! INPUT PARAMETERS
; rem_ptr gr26 ;! rem_ptr gr26
; n1 gr25 ;! n1 gr25
; n0 gr24 ;! n0 gr24
; d gr23 ;! d gr23
.code .text
L$0000 .word 0x43f00000 L$0000:
.word 0x43f00000
.word 0x0 .word 0x0
.export __udiv_qrnnd .export __udiv_qrnnd
__udiv_qrnnd __udiv_qrnnd:
.proc .proc
.callinfo frame=64,no_calls .callinfo frame=64,no_calls
.entry .entry
ldo 64(%r30),%r30 ldo 64(%r30),%r30
stws %r25,-16(0,%r30) ; n_hi stws %r25,-16(0,%r30) ;! n_hi
stws %r24,-12(0,%r30) ; n_lo stws %r24,-12(0,%r30) ;! n_lo
ldil L'L$0000,%r19 b,l L$0,%r1
ldo R'L$0000(%r19),%r19 ldo L$0000-L$0(%r1),%r1
L$0:
fldds -16(0,%r30),%fr5 fldds -16(0,%r30),%fr5
stws %r23,-12(0,%r30) stws %r23,-12(0,%r30)
comib,<= 0,%r25,L$1 comib,<= 0,%r25,L$1
fcnvxf,dbl,dbl %fr5,%fr5 fcnvxf,dbl,dbl %fr5,%fr5
fldds 0(0,%r19),%fr4 fldds 0(0,%r1),%fr4
fadd,dbl %fr4,%fr5,%fr5 fadd,dbl %fr4,%fr5,%fr5
L$1 L$1:
fcpy,sgl %fr0,%fr6L fcpy,sgl %fr0,%fr6L
fldws -12(0,%r30),%fr6R fldws -12(0,%r30),%fr6R
fcnvxf,dbl,dbl %fr6,%fr4 fcnvxf,dbl,dbl %fr6,%fr4
@ -62,13 +64,14 @@ L$1
ldws -12(0,%r30),%r21 ldws -12(0,%r30),%r21
ldws -16(0,%r30),%r20 ldws -16(0,%r30),%r20
sub %r24,%r21,%r22 sub %r24,%r21,%r22
subb %r25,%r20,%r19 subb %r25,%r20,%r1
comib,= 0,%r19,L$2 comib,= 0,%r1,L$2
ldo -64(%r30),%r30 ldo -64(%r30),%r30
add %r22,%r23,%r22 add %r22,%r23,%r22
ldo -1(%r28),%r28 ldo -1(%r28),%r28
L$2 bv 0(%r2) L$2:
bv 0(%r2)
stws %r22,0(0,%r26) stws %r22,0(0,%r26)
.exit .exit

54
sysdeps/hppa/lshift.s
View File

@ -1,34 +1,34 @@
; HP-PA __mpn_lshift -- ;! HP-PA __mpn_lshift --
; Copyright (C) 1992, 1994 Free Software Foundation, Inc. ;! Copyright (C) 1992, 1994 Free Software Foundation, Inc.
; This file is part of the GNU MP Library. ;! This file is part of the GNU MP Library.
; The GNU MP Library is free software; you can redistribute it and/or modify ;! The GNU MP 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 ;! 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 ;! the Free Software Foundation; either version 2 of the License, or (at your
; option) any later version. ;! option) any later version.
; The GNU MP Library is distributed in the hope that it will be useful, but ;! The GNU MP Library is distributed in the hope that it will be useful, but
; WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY ;! WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
; or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public ;! or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
; License for more details. ;! License for more details.
; You should have received a copy of the GNU Library General Public License ;! You should have received a copy of the GNU Library General Public License
; along with the GNU MP Library; see the file COPYING.LIB. If not, write to ;! along with the GNU MP Library; see the file COPYING.LIB. If not, write to
; the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, ;! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
; MA 02111-1307, USA. ;! MA 02111-1307, USA.
; INPUT PARAMETERS ;! INPUT PARAMETERS
; res_ptr gr26 ;! res_ptr gr26
; s_ptr gr25 ;! s_ptr gr25
; size gr24 ;! size gr24
; cnt gr23 ;! cnt gr23
.code .text
.export __mpn_lshift .export __mpn_lshift
__mpn_lshift __mpn_lshift:
.proc .proc
.callinfo frame=64,no_calls .callinfo frame=64,no_calls
.entry .entry
@ -39,12 +39,12 @@ __mpn_lshift
subi 32,%r23,%r1 subi 32,%r23,%r1
mtsar %r1 mtsar %r1
addib,= -1,%r24,L$0004 addib,= -1,%r24,L$0004
vshd %r0,%r22,%r28 ; compute carry out limb vshd %r0,%r22,%r28 ;! compute carry out limb
ldws,mb -4(0,%r25),%r29 ldws,mb -4(0,%r25),%r29
addib,= -1,%r24,L$0002 addib,= -1,%r24,L$0002
vshd %r22,%r29,%r20 vshd %r22,%r29,%r20
L$loop ldws,mb -4(0,%r25),%r22 L$loop: ldws,mb -4(0,%r25),%r22
stws,mb %r20,-4(0,%r26) stws,mb %r20,-4(0,%r26)
addib,= -1,%r24,L$0003 addib,= -1,%r24,L$0003
vshd %r29,%r22,%r20 vshd %r29,%r22,%r20
@ -53,12 +53,12 @@ L$loop ldws,mb -4(0,%r25),%r22
addib,<> -1,%r24,L$loop addib,<> -1,%r24,L$loop
vshd %r22,%r29,%r20 vshd %r22,%r29,%r20
L$0002 stws,mb %r20,-4(0,%r26) L$0002: stws,mb %r20,-4(0,%r26)
vshd %r29,%r0,%r20 vshd %r29,%r0,%r20
bv 0(%r2) bv 0(%r2)
stw %r20,-4(0,%r26) stw %r20,-4(0,%r26)
L$0003 stws,mb %r20,-4(0,%r26) L$0003: stws,mb %r20,-4(0,%r26)
L$0004 vshd %r22,%r0,%r20 L$0004: vshd %r22,%r0,%r20
bv 0(%r2) bv 0(%r2)
stw %r20,-4(0,%r26) stw %r20,-4(0,%r26)

54
sysdeps/hppa/rshift.s
View File

@ -1,34 +1,34 @@
; HP-PA __mpn_rshift -- ;! HP-PA __mpn_rshift --
; Copyright (C) 1992, 1994 Free Software Foundation, Inc. ;! Copyright (C) 1992, 1994 Free Software Foundation, Inc.
; This file is part of the GNU MP Library. ;! This file is part of the GNU MP Library.
; The GNU MP Library is free software; you can redistribute it and/or modify ;! The GNU MP 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 ;! 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 ;! the Free Software Foundation; either version 2 of the License, or (at your
; option) any later version. ;! option) any later version.
; The GNU MP Library is distributed in the hope that it will be useful, but ;! The GNU MP Library is distributed in the hope that it will be useful, but
; WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY ;! WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
; or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public ;! or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
; License for more details. ;! License for more details.
; You should have received a copy of the GNU Library General Public License ;! You should have received a copy of the GNU Library General Public License
; along with the GNU MP Library; see the file COPYING.LIB. If not, write to ;! along with the GNU MP Library; see the file COPYING.LIB. If not, write to
; the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, ;! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
; MA 02111-1307, USA. ;! MA 02111-1307, USA.
; INPUT PARAMETERS ;! INPUT PARAMETERS
; res_ptr gr26 ;! res_ptr gr26
; s_ptr gr25 ;! s_ptr gr25
; size gr24 ;! size gr24
; cnt gr23 ;! cnt gr23
.code .text
.export __mpn_rshift .export __mpn_rshift
__mpn_rshift __mpn_rshift:
.proc .proc
.callinfo frame=64,no_calls .callinfo frame=64,no_calls
.entry .entry
@ -36,12 +36,12 @@ __mpn_rshift
ldws,ma 4(0,%r25),%r22 ldws,ma 4(0,%r25),%r22
mtsar %r23 mtsar %r23
addib,= -1,%r24,L$0004 addib,= -1,%r24,L$0004
vshd %r22,%r0,%r28 ; compute carry out limb vshd %r22,%r0,%r28 ;! compute carry out limb
ldws,ma 4(0,%r25),%r29 ldws,ma 4(0,%r25),%r29
addib,= -1,%r24,L$0002 addib,= -1,%r24,L$0002
vshd %r29,%r22,%r20 vshd %r29,%r22,%r20
L$loop ldws,ma 4(0,%r25),%r22 L$loop: ldws,ma 4(0,%r25),%r22
stws,ma %r20,4(0,%r26) stws,ma %r20,4(0,%r26)
addib,= -1,%r24,L$0003 addib,= -1,%r24,L$0003
vshd %r22,%r29,%r20 vshd %r22,%r29,%r20
@ -50,12 +50,12 @@ L$loop ldws,ma 4(0,%r25),%r22
addib,<> -1,%r24,L$loop addib,<> -1,%r24,L$loop
vshd %r29,%r22,%r20 vshd %r29,%r22,%r20
L$0002 stws,ma %r20,4(0,%r26) L$0002: stws,ma %r20,4(0,%r26)
vshd %r0,%r29,%r20 vshd %r0,%r29,%r20
bv 0(%r2) bv 0(%r2)
stw %r20,0(0,%r26) stw %r20,0(0,%r26)
L$0003 stws,ma %r20,4(0,%r26) L$0003: stws,ma %r20,4(0,%r26)
L$0004 vshd %r0,%r22,%r20 L$0004: vshd %r0,%r22,%r20
bv 0(%r2) bv 0(%r2)
stw %r20,0(0,%r26) stw %r20,0(0,%r26)

68
sysdeps/hppa/sub_n.s
View File

@ -1,56 +1,56 @@
; HP-PA __mpn_sub_n -- Subtract two limb vectors of the same length > 0 and ;! HP-PA __mpn_sub_n -- Subtract two limb vectors of the same length > 0 and
; store difference in a third limb vector. ;! store difference in a third limb vector.
; Copyright (C) 1992, 1994 Free Software Foundation, Inc. ;! Copyright (C) 1992, 1994 Free Software Foundation, Inc.
; This file is part of the GNU MP Library. ;! This file is part of the GNU MP Library.
; The GNU MP Library is free software; you can redistribute it and/or modify ;! The GNU MP 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 ;! 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 ;! the Free Software Foundation; either version 2 of the License, or (at your
; option) any later version. ;! option) any later version.
; The GNU MP Library is distributed in the hope that it will be useful, but ;! The GNU MP Library is distributed in the hope that it will be useful, but
; WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY ;! WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
; or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public ;! or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
; License for more details. ;! License for more details.
; You should have received a copy of the GNU Library General Public License ;! You should have received a copy of the GNU Library General Public License
; along with the GNU MP Library; see the file COPYING.LIB. If not, write to ;! along with the GNU MP Library; see the file COPYING.LIB. If not, write to
; the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, ;! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
; MA 02111-1307, USA. ;! MA 02111-1307, USA.
; INPUT PARAMETERS ;! INPUT PARAMETERS
; res_ptr gr26 ;! res_ptr gr26
; s1_ptr gr25 ;! s1_ptr gr25
; s2_ptr gr24 ;! s2_ptr gr24
; size gr23 ;! size gr23
; One might want to unroll this as for other processors, but it turns ;! One might want to unroll this as for other processors, but it turns
; out that the data cache contention after a store makes such ;! out that the data cache contention after a store makes such
; unrolling useless. We can't come under 5 cycles/limb anyway. ;! unrolling useless. We can't come under 5 cycles/limb anyway.
.code .text
.export __mpn_sub_n .export __mpn_sub_n
__mpn_sub_n __mpn_sub_n:
.proc .proc
.callinfo frame=0,no_calls .callinfo frame=0,no_calls
.entry .entry
ldws,ma 4(0,%r25),%r20 ldws,ma 4(0,%r25),%r21
ldws,ma 4(0,%r24),%r19 ldws,ma 4(0,%r24),%r20
addib,= -1,%r23,L$end ; check for (SIZE == 1) addib,= -1,%r23,L$end ;! check for (SIZE == 1)
sub %r20,%r19,%r28 ; subtract first limbs ignoring cy sub %r21,%r20,%r28 ;! subtract first limbs ignoring cy
L$loop ldws,ma 4(0,%r25),%r20 L$loop: ldws,ma 4(0,%r25),%r21
ldws,ma 4(0,%r24),%r19 ldws,ma 4(0,%r24),%r20
stws,ma %r28,4(0,%r26) stws,ma %r28,4(0,%r26)
addib,<> -1,%r23,L$loop addib,<> -1,%r23,L$loop
subb %r20,%r19,%r28 subb %r21,%r20,%r28
L$end stws %r28,0(0,%r26) L$end: stws %r28,0(0,%r26)
addc %r0,%r0,%r28 addc %r0,%r0,%r28
bv 0(%r2) bv 0(%r2)
subi 1,%r28,%r28 subi 1,%r28,%r28

82
sysdeps/hppa/udiv_qrnnd.s
View File

@ -1,45 +1,45 @@
; HP-PA __udiv_qrnnd division support, used from longlong.h. ;! HP-PA __udiv_qrnnd division support, used from longlong.h.
; This version runs fast on pre-PA7000 CPUs. ;! This version runs fast on pre-PA7000 CPUs.
; Copyright (C) 1993, 1994 Free Software Foundation, Inc. ;! Copyright (C) 1993, 1994 Free Software Foundation, Inc.
; This file is part of the GNU MP Library. ;! This file is part of the GNU MP Library.
; The GNU MP Library is free software; you can redistribute it and/or modify ;! The GNU MP 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 ;! 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 ;! the Free Software Foundation; either version 2 of the License, or (at your
; option) any later version. ;! option) any later version.
; The GNU MP Library is distributed in the hope that it will be useful, but ;! The GNU MP Library is distributed in the hope that it will be useful, but
; WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY ;! WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
; or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public ;! or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
; License for more details. ;! License for more details.
; You should have received a copy of the GNU Library General Public License ;! You should have received a copy of the GNU Library General Public License
; along with the GNU MP Library; see the file COPYING.LIB. If not, write to ;! along with the GNU MP Library; see the file COPYING.LIB. If not, write to
; the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, ;! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
; MA 02111-1307, USA. ;! MA 02111-1307, USA.
; INPUT PARAMETERS ;! INPUT PARAMETERS
; rem_ptr gr26 ;! rem_ptr gr26
; n1 gr25 ;! n1 gr25
; n0 gr24 ;! n0 gr24
; d gr23 ;! d gr23
; The code size is a bit excessive. We could merge the last two ds;addc ;! The code size is a bit excessive. We could merge the last two ds;addc
; sequences by simply moving the "bb,< Odd" instruction down. The only ;! sequences by simply moving the "bb,< Odd" instruction down. The only
; trouble is the FFFFFFFF code that would need some hacking. ;! trouble is the FFFFFFFF code that would need some hacking.
.code .text
.export __udiv_qrnnd .export __udiv_qrnnd
__udiv_qrnnd __udiv_qrnnd:
.proc .proc
.callinfo frame=0,no_calls .callinfo frame=0,no_calls
.entry .entry
comb,< %r23,0,L$largedivisor comb,< %r23,0,L$largedivisor
sub %r0,%r23,%r1 ; clear cy as side-effect sub %r0,%r23,%r1 ;! clear cy as side-effect
ds %r0,%r1,%r0 ds %r0,%r1,%r0
addc %r24,%r24,%r24 addc %r24,%r24,%r24
ds %r25,%r23,%r25 ds %r25,%r23,%r25
@ -111,12 +111,12 @@ __udiv_qrnnd
bv 0(%r2) bv 0(%r2)
addc %r28,%r28,%r28 addc %r28,%r28,%r28
L$largedivisor L$largedivisor:
extru %r24,31,1,%r19 ; r19 = n0 & 1 extru %r24,31,1,%r20 ;! r20 = n0 & 1
bb,< %r23,31,L$odd bb,< %r23,31,L$odd
extru %r23,30,31,%r22 ; r22 = d >> 1 extru %r23,30,31,%r22 ;! r22 = d >> 1
shd %r25,%r24,1,%r24 ; r24 = new n0 shd %r25,%r24,1,%r24 ;! r24 = new n0
extru %r25,30,31,%r25 ; r25 = new n1 extru %r25,30,31,%r25 ;! r25 = new n1
sub %r0,%r22,%r21 sub %r0,%r22,%r21
ds %r0,%r21,%r0 ds %r0,%r21,%r0
addc %r24,%r24,%r24 addc %r24,%r24,%r24
@ -185,14 +185,14 @@ L$largedivisor
ds %r25,%r22,%r25 ds %r25,%r22,%r25
comclr,>= %r25,%r0,%r0 comclr,>= %r25,%r0,%r0
addl %r25,%r22,%r25 addl %r25,%r22,%r25
sh1addl %r25,%r19,%r25 sh1addl %r25,%r20,%r25
stws %r25,0(0,%r26) stws %r25,0(0,%r26)
bv 0(%r2) bv 0(%r2)
addc %r24,%r24,%r28 addc %r24,%r24,%r28
L$odd addib,sv,n 1,%r22,L$FF.. ; r22 = (d / 2 + 1) L$odd: addib,sv,n 1,%r22,L$FF.. ;! r22 = (d / 2 + 1)
shd %r25,%r24,1,%r24 ; r24 = new n0 shd %r25,%r24,1,%r24 ;! r24 = new n0
extru %r25,30,31,%r25 ; r25 = new n1 extru %r25,30,31,%r25 ;! r25 = new n1
sub %r0,%r22,%r21 sub %r0,%r22,%r21
ds %r0,%r21,%r0 ds %r0,%r21,%r0
addc %r24,%r24,%r24 addc %r24,%r24,%r24
@ -262,8 +262,8 @@ L$odd addib,sv,n 1,%r22,L$FF.. ; r22 = (d / 2 + 1)
addc %r24,%r24,%r28 addc %r24,%r24,%r28
comclr,>= %r25,%r0,%r0 comclr,>= %r25,%r0,%r0
addl %r25,%r22,%r25 addl %r25,%r22,%r25
sh1addl %r25,%r19,%r25 sh1addl %r25,%r20,%r25
; We have computed (n1,,n0) / (d + 1), q' = r28, r' = r25 ;! We have computed (n1,,n0) / (d + 1), q' = r28, r' = r25
add,nuv %r28,%r25,%r25 add,nuv %r28,%r25,%r25
addl %r25,%r1,%r25 addl %r25,%r1,%r25
addc %r0,%r28,%r28 addc %r0,%r28,%r28
@ -273,9 +273,9 @@ L$odd addib,sv,n 1,%r22,L$FF.. ; r22 = (d / 2 + 1)
bv 0(%r2) bv 0(%r2)
addc %r0,%r28,%r28 addc %r0,%r28,%r28
; This is just a special case of the code above. ;! This is just a special case of the code above.
; We come here when d == 0xFFFFFFFF ;! We come here when d == 0xFFFFFFFF
L$FF.. add,uv %r25,%r24,%r24 L$FF..: add,uv %r25,%r24,%r24
sub,<< %r24,%r23,%r0 sub,<< %r24,%r23,%r0
ldo 1(%r24),%r24 ldo 1(%r24),%r24
stws %r24,0(0,%r26) stws %r24,0(0,%r26)