--- /dev/null
+! SPARC __mpn_add_n -- Add two limb vectors of the same length > 0 and store
+! sum in a third limb vector.
+
+! Copyright (C) 1995, 1996 Free Software Foundation, Inc.
+
+! This file is part of the GNU MP Library.
+
+! 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
+! the Free Software Foundation; either version 2 of the License, or (at your
+! option) any later version.
+
+! 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
+! 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 the GNU MP Library; see the file COPYING.LIB. If not, write to
+! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
+! MA 02111-1307, USA.
+
+
+! INPUT PARAMETERS
+#define res_ptr %o0
+#define s1_ptr %o1
+#define s2_ptr %o2
+#define size %o3
+
+#include "sysdep.h"
+
+ .text
+ .align 4
+ .global C_SYMBOL_NAME(__mpn_add_n)
+C_SYMBOL_NAME(__mpn_add_n):
+ xor s2_ptr,res_ptr,%g1
+ andcc %g1,4,%g0
+ bne L1 ! branch if alignment differs
+ nop
+! ** V1a **
+L0: andcc res_ptr,4,%g0 ! res_ptr unaligned? Side effect: cy=0
+ be L_v1 ! if no, branch
+ nop
+/* Add least significant limb separately to align res_ptr and s2_ptr */
+ ld [s1_ptr],%g4
+ add s1_ptr,4,s1_ptr
+ ld [s2_ptr],%g2
+ add s2_ptr,4,s2_ptr
+ add size,-1,size
+ addcc %g4,%g2,%o4
+ st %o4,[res_ptr]
+ add res_ptr,4,res_ptr
+L_v1: addx %g0,%g0,%o4 ! save cy in register
+ cmp size,2 ! if size < 2 ...
+ bl Lend2 ! ... branch to tail code
+ subcc %g0,%o4,%g0 ! restore cy
+
+ ld [s1_ptr+0],%g4
+ addcc size,-10,size
+ ld [s1_ptr+4],%g1
+ ldd [s2_ptr+0],%g2
+ blt Lfin1
+ subcc %g0,%o4,%g0 ! restore cy
+/* Add blocks of 8 limbs until less than 8 limbs remain */
+Loop1: addxcc %g4,%g2,%o4
+ ld [s1_ptr+8],%g4
+ addxcc %g1,%g3,%o5
+ ld [s1_ptr+12],%g1
+ ldd [s2_ptr+8],%g2
+ std %o4,[res_ptr+0]
+ addxcc %g4,%g2,%o4
+ ld [s1_ptr+16],%g4
+ addxcc %g1,%g3,%o5
+ ld [s1_ptr+20],%g1
+ ldd [s2_ptr+16],%g2
+ std %o4,[res_ptr+8]
+ addxcc %g4,%g2,%o4
+ ld [s1_ptr+24],%g4
+ addxcc %g1,%g3,%o5
+ ld [s1_ptr+28],%g1
+ ldd [s2_ptr+24],%g2
+ std %o4,[res_ptr+16]
+ addxcc %g4,%g2,%o4
+ ld [s1_ptr+32],%g4
+ addxcc %g1,%g3,%o5
+ ld [s1_ptr+36],%g1
+ ldd [s2_ptr+32],%g2
+ std %o4,[res_ptr+24]
+ addx %g0,%g0,%o4 ! save cy in register
+ addcc size,-8,size
+ add s1_ptr,32,s1_ptr
+ add s2_ptr,32,s2_ptr
+ add res_ptr,32,res_ptr
+ bge Loop1
+ subcc %g0,%o4,%g0 ! restore cy
+
+Lfin1: addcc size,8-2,size
+ blt Lend1
+ subcc %g0,%o4,%g0 ! restore cy
+/* Add blocks of 2 limbs until less than 2 limbs remain */
+Loope1: addxcc %g4,%g2,%o4
+ ld [s1_ptr+8],%g4
+ addxcc %g1,%g3,%o5
+ ld [s1_ptr+12],%g1
+ ldd [s2_ptr+8],%g2
+ std %o4,[res_ptr+0]
+ addx %g0,%g0,%o4 ! save cy in register
+ addcc size,-2,size
+ add s1_ptr,8,s1_ptr
+ add s2_ptr,8,s2_ptr
+ add res_ptr,8,res_ptr
+ bge Loope1
+ subcc %g0,%o4,%g0 ! restore cy
+Lend1: addxcc %g4,%g2,%o4
+ addxcc %g1,%g3,%o5
+ std %o4,[res_ptr+0]
+ addx %g0,%g0,%o4 ! save cy in register
+
+ andcc size,1,%g0
+ be Lret1
+ subcc %g0,%o4,%g0 ! restore cy
+/* Add last limb */
+ ld [s1_ptr+8],%g4
+ ld [s2_ptr+8],%g2
+ addxcc %g4,%g2,%o4
+ st %o4,[res_ptr+8]
+
+Lret1: retl
+ addx %g0,%g0,%o0 ! return carry-out from most sign. limb
+
+L1: xor s1_ptr,res_ptr,%g1
+ andcc %g1,4,%g0
+ bne L2
+ nop
+! ** V1b **
+ mov s2_ptr,%g1
+ mov s1_ptr,s2_ptr
+ b L0
+ mov %g1,s1_ptr
+
+! ** V2 **
+/* If we come here, the alignment of s1_ptr and res_ptr as well as the
+ alignment of s2_ptr and res_ptr differ. Since there are only two ways
+ things can be aligned (that we care about) we now know that the alignment
+ of s1_ptr and s2_ptr are the same. */
+
+L2: cmp size,1
+ be Ljone
+ nop
+ andcc s1_ptr,4,%g0 ! s1_ptr unaligned? Side effect: cy=0
+ be L_v2 ! if no, branch
+ nop
+/* Add least significant limb separately to align s1_ptr and s2_ptr */
+ ld [s1_ptr],%g4
+ add s1_ptr,4,s1_ptr
+ ld [s2_ptr],%g2
+ add s2_ptr,4,s2_ptr
+ add size,-1,size
+ addcc %g4,%g2,%o4
+ st %o4,[res_ptr]
+ add res_ptr,4,res_ptr
+
+L_v2: addx %g0,%g0,%o4 ! save cy in register
+ addcc size,-8,size
+ blt Lfin2
+ subcc %g0,%o4,%g0 ! restore cy
+/* Add blocks of 8 limbs until less than 8 limbs remain */
+Loop2: ldd [s1_ptr+0],%g2
+ ldd [s2_ptr+0],%o4
+ addxcc %g2,%o4,%g2
+ st %g2,[res_ptr+0]
+ addxcc %g3,%o5,%g3
+ st %g3,[res_ptr+4]
+ ldd [s1_ptr+8],%g2
+ ldd [s2_ptr+8],%o4
+ addxcc %g2,%o4,%g2
+ st %g2,[res_ptr+8]
+ addxcc %g3,%o5,%g3
+ st %g3,[res_ptr+12]
+ ldd [s1_ptr+16],%g2
+ ldd [s2_ptr+16],%o4
+ addxcc %g2,%o4,%g2
+ st %g2,[res_ptr+16]
+ addxcc %g3,%o5,%g3
+ st %g3,[res_ptr+20]
+ ldd [s1_ptr+24],%g2
+ ldd [s2_ptr+24],%o4
+ addxcc %g2,%o4,%g2
+ st %g2,[res_ptr+24]
+ addxcc %g3,%o5,%g3
+ st %g3,[res_ptr+28]
+ addx %g0,%g0,%o4 ! save cy in register
+ addcc size,-8,size
+ add s1_ptr,32,s1_ptr
+ add s2_ptr,32,s2_ptr
+ add res_ptr,32,res_ptr
+ bge Loop2
+ subcc %g0,%o4,%g0 ! restore cy
+
+Lfin2: addcc size,8-2,size
+ blt Lend2
+ subcc %g0,%o4,%g0 ! restore cy
+Loope2: ldd [s1_ptr+0],%g2
+ ldd [s2_ptr+0],%o4
+ addxcc %g2,%o4,%g2
+ st %g2,[res_ptr+0]
+ addxcc %g3,%o5,%g3
+ st %g3,[res_ptr+4]
+ addx %g0,%g0,%o4 ! save cy in register
+ addcc size,-2,size
+ add s1_ptr,8,s1_ptr
+ add s2_ptr,8,s2_ptr
+ add res_ptr,8,res_ptr
+ bge Loope2
+ subcc %g0,%o4,%g0 ! restore cy
+Lend2: andcc size,1,%g0
+ be Lret2
+ subcc %g0,%o4,%g0 ! restore cy
+/* Add last limb */
+Ljone: ld [s1_ptr],%g4
+ ld [s2_ptr],%g2
+ addxcc %g4,%g2,%o4
+ st %o4,[res_ptr]
+
+Lret2: retl
+ addx %g0,%g0,%o0 ! return carry-out from most sign. limb
--- /dev/null
+! SPARC __mpn_addmul_1 -- Multiply a limb vector with a limb and add
+! the result to a second limb vector.
+
+! Copyright (C) 1992, 1993, 1994 Free Software Foundation, Inc.
+
+! This file is part of the GNU MP Library.
+
+! 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
+! the Free Software Foundation; either version 2 of the License, or (at your
+! option) any later version.
+
+! 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
+! 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 the GNU MP Library; see the file COPYING.LIB. If not, write to
+! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
+! MA 02111-1307, USA.
+
+
+! INPUT PARAMETERS
+! res_ptr o0
+! s1_ptr o1
+! size o2
+! s2_limb o3
+
+#include "sysdep.h"
+
+.text
+ .align 4
+ .global C_SYMBOL_NAME(__mpn_addmul_1)
+C_SYMBOL_NAME(__mpn_addmul_1):
+ ! Make S1_PTR and RES_PTR point at the end of their blocks
+ ! and put (- 4 x SIZE) in index/loop counter.
+ sll %o2,2,%o2
+ add %o0,%o2,%o4 ! RES_PTR in o4 since o0 is retval
+ add %o1,%o2,%o1
+ sub %g0,%o2,%o2
+
+ cmp %o3,0xfff
+ bgu Large
+ nop
+
+ ld [%o1+%o2],%o5
+ mov 0,%o0
+ b L0
+ add %o4,-4,%o4
+Loop0:
+ addcc %o5,%g1,%g1
+ ld [%o1+%o2],%o5
+ addx %o0,%g0,%o0
+ st %g1,[%o4+%o2]
+L0: wr %g0,%o3,%y
+ sra %o5,31,%g2
+ and %o3,%g2,%g2
+ andcc %g1,0,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,0,%g1
+ sra %g1,20,%g4
+ sll %g1,12,%g1
+ rd %y,%g3
+ srl %g3,20,%g3
+ or %g1,%g3,%g1
+
+ addcc %g1,%o0,%g1
+ addx %g2,%g4,%o0 ! add sign-compensation and cy to hi limb
+ addcc %o2,4,%o2 ! loop counter
+ bne Loop0
+ ld [%o4+%o2],%o5
+
+ addcc %o5,%g1,%g1
+ addx %o0,%g0,%o0
+ retl
+ st %g1,[%o4+%o2]
+
+
+Large: ld [%o1+%o2],%o5
+ mov 0,%o0
+ sra %o3,31,%g4 ! g4 = mask of ones iff S2_LIMB < 0
+ b L1
+ add %o4,-4,%o4
+Loop:
+ addcc %o5,%g3,%g3
+ ld [%o1+%o2],%o5
+ addx %o0,%g0,%o0
+ st %g3,[%o4+%o2]
+L1: wr %g0,%o5,%y
+ and %o5,%g4,%g2
+ andcc %g0,%g0,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%g0,%g1
+ rd %y,%g3
+ addcc %g3,%o0,%g3
+ addx %g2,%g1,%o0
+ addcc %o2,4,%o2
+ bne Loop
+ ld [%o4+%o2],%o5
+
+ addcc %o5,%g3,%g3
+ addx %o0,%g0,%o0
+ retl
+ st %g3,[%o4+%o2]
--- /dev/null
+! sparc __mpn_lshift --
+
+! Copyright (C) 1995, 1996 Free Software Foundation, Inc.
+
+! This file is part of the GNU MP Library.
+
+! 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
+! the Free Software Foundation; either version 2 of the License, or (at your
+! option) any later version.
+
+! 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
+! 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 the GNU MP Library; see the file COPYING.LIB. If not, write to
+! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
+! MA 02111-1307, USA.
+
+
+! INPUT PARAMETERS
+! res_ptr %o0
+! src_ptr %o1
+! size %o2
+! cnt %o3
+
+#include "sysdep.h"
+
+ .text
+ .align 4
+ .global C_SYMBOL_NAME(__mpn_lshift)
+C_SYMBOL_NAME(__mpn_lshift):
+ sll %o2,2,%g1
+ add %o1,%g1,%o1 ! make %o1 point at end of src
+ ld [%o1-4],%g2 ! load first limb
+ sub %g0,%o3,%o5 ! negate shift count
+ add %o0,%g1,%o0 ! make %o0 point at end of res
+ add %o2,-1,%o2
+ andcc %o2,4-1,%g4 ! number of limbs in first loop
+ srl %g2,%o5,%g1 ! compute function result
+ be L0 ! if multiple of 4 limbs, skip first loop
+ st %g1,[%sp+80]
+
+ sub %o2,%g4,%o2 ! adjust count for main loop
+
+Loop0: ld [%o1-8],%g3
+ add %o0,-4,%o0
+ add %o1,-4,%o1
+ addcc %g4,-1,%g4
+ sll %g2,%o3,%o4
+ srl %g3,%o5,%g1
+ mov %g3,%g2
+ or %o4,%g1,%o4
+ bne Loop0
+ st %o4,[%o0+0]
+
+L0: tst %o2
+ be Lend
+ nop
+
+Loop: ld [%o1-8],%g3
+ add %o0,-16,%o0
+ addcc %o2,-4,%o2
+ sll %g2,%o3,%o4
+ srl %g3,%o5,%g1
+
+ ld [%o1-12],%g2
+ sll %g3,%o3,%g4
+ or %o4,%g1,%o4
+ st %o4,[%o0+12]
+ srl %g2,%o5,%g1
+
+ ld [%o1-16],%g3
+ sll %g2,%o3,%o4
+ or %g4,%g1,%g4
+ st %g4,[%o0+8]
+ srl %g3,%o5,%g1
+
+ ld [%o1-20],%g2
+ sll %g3,%o3,%g4
+ or %o4,%g1,%o4
+ st %o4,[%o0+4]
+ srl %g2,%o5,%g1
+
+ add %o1,-16,%o1
+ or %g4,%g1,%g4
+ bne Loop
+ st %g4,[%o0+0]
+
+Lend: sll %g2,%o3,%g2
+ st %g2,[%o0-4]
+ retl
+ ld [%sp+80],%o0
--- /dev/null
+! SPARC __mpn_mul_1 -- Multiply a limb vector with a limb and store
+! the result in a second limb vector.
+
+! Copyright (C) 1992, 1993, 1994 Free Software Foundation, Inc.
+
+! This file is part of the GNU MP Library.
+
+! 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
+! the Free Software Foundation; either version 2 of the License, or (at your
+! option) any later version.
+
+! 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
+! 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 the GNU MP Library; see the file COPYING.LIB. If not, write to
+! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
+! MA 02111-1307, USA.
+
+
+! INPUT PARAMETERS
+! res_ptr o0
+! s1_ptr o1
+! size o2
+! s2_limb o3
+
+! ADD CODE FOR SMALL MULTIPLIERS!
+!1: ld
+! st
+!
+!2: ld ,a
+! addxcc a,a,x
+! st x,
+!
+!3_unrolled:
+! ld ,a
+! addxcc a,a,x1 ! 2a + cy
+! addx %g0,%g0,x2
+! addcc a,x1,x ! 3a + c
+! st x,
+!
+! ld ,a
+! addxcc a,a,y1
+! addx %g0,%g0,y2
+! addcc a,y1,x
+! st x,
+!
+!4_unrolled:
+! ld ,a
+! srl a,2,x1 ! 4a
+! addxcc y2,x1,x
+! sll a,30,x2
+! st x,
+!
+! ld ,a
+! srl a,2,y1
+! addxcc x2,y1,y
+! sll a,30,y2
+! st x,
+!
+!5_unrolled:
+! ld ,a
+! srl a,2,x1 ! 4a
+! addxcc a,x1,x ! 5a + c
+! sll a,30,x2
+! addx %g0,x2,x2
+! st x,
+!
+! ld ,a
+! srl a,2,y1
+! addxcc a,y1,x
+! sll a,30,y2
+! addx %g0,y2,y2
+! st x,
+!
+!8_unrolled:
+! ld ,a
+! srl a,3,x1 ! 8a
+! addxcc y2,x1,x
+! sll a,29,x2
+! st x,
+!
+! ld ,a
+! srl a,3,y1
+! addxcc x2,y1,y
+! sll a,29,y2
+! st x,
+
+#include "sysdep.h"
+
+.text
+ .align 4
+ .global C_SYMBOL_NAME(__mpn_mul_1)
+C_SYMBOL_NAME(__mpn_mul_1):
+ ! Make S1_PTR and RES_PTR point at the end of their blocks
+ ! and put (- 4 x SIZE) in index/loop counter.
+ sll %o2,2,%o2
+ add %o0,%o2,%o4 ! RES_PTR in o4 since o0 is retval
+ add %o1,%o2,%o1
+ sub %g0,%o2,%o2
+
+ cmp %o3,0xfff
+ bgu Large
+ nop
+
+ ld [%o1+%o2],%o5
+ mov 0,%o0
+ b L0
+ add %o4,-4,%o4
+Loop0:
+ st %g1,[%o4+%o2]
+L0: wr %g0,%o3,%y
+ sra %o5,31,%g2
+ and %o3,%g2,%g2
+ andcc %g1,0,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,0,%g1
+ sra %g1,20,%g4
+ sll %g1,12,%g1
+ rd %y,%g3
+ srl %g3,20,%g3
+ or %g1,%g3,%g1
+
+ addcc %g1,%o0,%g1
+ addx %g2,%g4,%o0 ! add sign-compensation and cy to hi limb
+ addcc %o2,4,%o2 ! loop counter
+ bne,a Loop0
+ ld [%o1+%o2],%o5
+
+ retl
+ st %g1,[%o4+%o2]
+
+
+Large: ld [%o1+%o2],%o5
+ mov 0,%o0
+ sra %o3,31,%g4 ! g4 = mask of ones iff S2_LIMB < 0
+ b L1
+ add %o4,-4,%o4
+Loop:
+ st %g3,[%o4+%o2]
+L1: wr %g0,%o5,%y
+ and %o5,%g4,%g2 ! g2 = S1_LIMB iff S2_LIMB < 0, else 0
+ andcc %g0,%g0,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%g0,%g1
+ rd %y,%g3
+ addcc %g3,%o0,%g3
+ addx %g2,%g1,%o0 ! add sign-compensation and cy to hi limb
+ addcc %o2,4,%o2 ! loop counter
+ bne,a Loop
+ ld [%o1+%o2],%o5
+
+ retl
+ st %g3,[%o4+%o2]
--- /dev/null
+! sparc __mpn_rshift --
+
+! Copyright (C) 1995, 1996 Free Software Foundation, Inc.
+
+! This file is part of the GNU MP Library.
+
+! 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
+! the Free Software Foundation; either version 2 of the License, or (at your
+! option) any later version.
+
+! 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
+! 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 the GNU MP Library; see the file COPYING.LIB. If not, write to
+! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
+! MA 02111-1307, USA.
+
+
+! INPUT PARAMETERS
+! res_ptr %o0
+! src_ptr %o1
+! size %o2
+! cnt %o3
+
+#include "sysdep.h"
+
+ .text
+ .align 4
+ .global C_SYMBOL_NAME(__mpn_rshift)
+C_SYMBOL_NAME(__mpn_rshift):
+ ld [%o1],%g2 ! load first limb
+ sub %g0,%o3,%o5 ! negate shift count
+ add %o2,-1,%o2
+ andcc %o2,4-1,%g4 ! number of limbs in first loop
+ sll %g2,%o5,%g1 ! compute function result
+ be L0 ! if multiple of 4 limbs, skip first loop
+ st %g1,[%sp+80]
+
+ sub %o2,%g4,%o2 ! adjust count for main loop
+
+Loop0: ld [%o1+4],%g3
+ add %o0,4,%o0
+ add %o1,4,%o1
+ addcc %g4,-1,%g4
+ srl %g2,%o3,%o4
+ sll %g3,%o5,%g1
+ mov %g3,%g2
+ or %o4,%g1,%o4
+ bne Loop0
+ st %o4,[%o0-4]
+
+L0: tst %o2
+ be Lend
+ nop
+
+Loop: ld [%o1+4],%g3
+ add %o0,16,%o0
+ addcc %o2,-4,%o2
+ srl %g2,%o3,%o4
+ sll %g3,%o5,%g1
+
+ ld [%o1+8],%g2
+ srl %g3,%o3,%g4
+ or %o4,%g1,%o4
+ st %o4,[%o0-16]
+ sll %g2,%o5,%g1
+
+ ld [%o1+12],%g3
+ srl %g2,%o3,%o4
+ or %g4,%g1,%g4
+ st %g4,[%o0-12]
+ sll %g3,%o5,%g1
+
+ ld [%o1+16],%g2
+ srl %g3,%o3,%g4
+ or %o4,%g1,%o4
+ st %o4,[%o0-8]
+ sll %g2,%o5,%g1
+
+ add %o1,16,%o1
+ or %g4,%g1,%g4
+ bne Loop
+ st %g4,[%o0-4]
+
+Lend: srl %g2,%o3,%g2
+ st %g2,[%o0-0]
+ retl
+ ld [%sp+80],%o0
--- /dev/null
+! SPARC __mpn_sub_n -- Subtract two limb vectors of the same length > 0 and
+! store difference in a third limb vector.
+
+! Copyright (C) 1995, 1996 Free Software Foundation, Inc.
+
+! This file is part of the GNU MP Library.
+
+! 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
+! the Free Software Foundation; either version 2 of the License, or (at your
+! option) any later version.
+
+! 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
+! 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 the GNU MP Library; see the file COPYING.LIB. If not, write to
+! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
+! MA 02111-1307, USA.
+
+
+! INPUT PARAMETERS
+#define res_ptr %o0
+#define s1_ptr %o1
+#define s2_ptr %o2
+#define size %o3
+
+#include "sysdep.h"
+
+ .text
+ .align 4
+ .global C_SYMBOL_NAME(__mpn_sub_n)
+C_SYMBOL_NAME(__mpn_sub_n):
+ xor s2_ptr,res_ptr,%g1
+ andcc %g1,4,%g0
+ bne L1 ! branch if alignment differs
+ nop
+! ** V1a **
+ andcc res_ptr,4,%g0 ! res_ptr unaligned? Side effect: cy=0
+ be L_v1 ! if no, branch
+ nop
+/* Add least significant limb separately to align res_ptr and s2_ptr */
+ ld [s1_ptr],%g4
+ add s1_ptr,4,s1_ptr
+ ld [s2_ptr],%g2
+ add s2_ptr,4,s2_ptr
+ add size,-1,size
+ subcc %g4,%g2,%o4
+ st %o4,[res_ptr]
+ add res_ptr,4,res_ptr
+L_v1: addx %g0,%g0,%o4 ! save cy in register
+ cmp size,2 ! if size < 2 ...
+ bl Lend2 ! ... branch to tail code
+ subcc %g0,%o4,%g0 ! restore cy
+
+ ld [s1_ptr+0],%g4
+ addcc size,-10,size
+ ld [s1_ptr+4],%g1
+ ldd [s2_ptr+0],%g2
+ blt Lfin1
+ subcc %g0,%o4,%g0 ! restore cy
+/* Add blocks of 8 limbs until less than 8 limbs remain */
+Loop1: subxcc %g4,%g2,%o4
+ ld [s1_ptr+8],%g4
+ subxcc %g1,%g3,%o5
+ ld [s1_ptr+12],%g1
+ ldd [s2_ptr+8],%g2
+ std %o4,[res_ptr+0]
+ subxcc %g4,%g2,%o4
+ ld [s1_ptr+16],%g4
+ subxcc %g1,%g3,%o5
+ ld [s1_ptr+20],%g1
+ ldd [s2_ptr+16],%g2
+ std %o4,[res_ptr+8]
+ subxcc %g4,%g2,%o4
+ ld [s1_ptr+24],%g4
+ subxcc %g1,%g3,%o5
+ ld [s1_ptr+28],%g1
+ ldd [s2_ptr+24],%g2
+ std %o4,[res_ptr+16]
+ subxcc %g4,%g2,%o4
+ ld [s1_ptr+32],%g4
+ subxcc %g1,%g3,%o5
+ ld [s1_ptr+36],%g1
+ ldd [s2_ptr+32],%g2
+ std %o4,[res_ptr+24]
+ addx %g0,%g0,%o4 ! save cy in register
+ addcc size,-8,size
+ add s1_ptr,32,s1_ptr
+ add s2_ptr,32,s2_ptr
+ add res_ptr,32,res_ptr
+ bge Loop1
+ subcc %g0,%o4,%g0 ! restore cy
+
+Lfin1: addcc size,8-2,size
+ blt Lend1
+ subcc %g0,%o4,%g0 ! restore cy
+/* Add blocks of 2 limbs until less than 2 limbs remain */
+Loope1: subxcc %g4,%g2,%o4
+ ld [s1_ptr+8],%g4
+ subxcc %g1,%g3,%o5
+ ld [s1_ptr+12],%g1
+ ldd [s2_ptr+8],%g2
+ std %o4,[res_ptr+0]
+ addx %g0,%g0,%o4 ! save cy in register
+ addcc size,-2,size
+ add s1_ptr,8,s1_ptr
+ add s2_ptr,8,s2_ptr
+ add res_ptr,8,res_ptr
+ bge Loope1
+ subcc %g0,%o4,%g0 ! restore cy
+Lend1: subxcc %g4,%g2,%o4
+ subxcc %g1,%g3,%o5
+ std %o4,[res_ptr+0]
+ addx %g0,%g0,%o4 ! save cy in register
+
+ andcc size,1,%g0
+ be Lret1
+ subcc %g0,%o4,%g0 ! restore cy
+/* Add last limb */
+ ld [s1_ptr+8],%g4
+ ld [s2_ptr+8],%g2
+ subxcc %g4,%g2,%o4
+ st %o4,[res_ptr+8]
+
+Lret1: retl
+ addx %g0,%g0,%o0 ! return carry-out from most sign. limb
+
+L1: xor s1_ptr,res_ptr,%g1
+ andcc %g1,4,%g0
+ bne L2
+ nop
+! ** V1b **
+ andcc res_ptr,4,%g0 ! res_ptr unaligned? Side effect: cy=0
+ be L_v1b ! if no, branch
+ nop
+/* Add least significant limb separately to align res_ptr and s1_ptr */
+ ld [s2_ptr],%g4
+ add s2_ptr,4,s2_ptr
+ ld [s1_ptr],%g2
+ add s1_ptr,4,s1_ptr
+ add size,-1,size
+ subcc %g2,%g4,%o4
+ st %o4,[res_ptr]
+ add res_ptr,4,res_ptr
+L_v1b: addx %g0,%g0,%o4 ! save cy in register
+ cmp size,2 ! if size < 2 ...
+ bl Lend2 ! ... branch to tail code
+ subcc %g0,%o4,%g0 ! restore cy
+
+ ld [s2_ptr+0],%g4
+ addcc size,-10,size
+ ld [s2_ptr+4],%g1
+ ldd [s1_ptr+0],%g2
+ blt Lfin1b
+ subcc %g0,%o4,%g0 ! restore cy
+/* Add blocks of 8 limbs until less than 8 limbs remain */
+Loop1b: subxcc %g2,%g4,%o4
+ ld [s2_ptr+8],%g4
+ subxcc %g3,%g1,%o5
+ ld [s2_ptr+12],%g1
+ ldd [s1_ptr+8],%g2
+ std %o4,[res_ptr+0]
+ subxcc %g2,%g4,%o4
+ ld [s2_ptr+16],%g4
+ subxcc %g3,%g1,%o5
+ ld [s2_ptr+20],%g1
+ ldd [s1_ptr+16],%g2
+ std %o4,[res_ptr+8]
+ subxcc %g2,%g4,%o4
+ ld [s2_ptr+24],%g4
+ subxcc %g3,%g1,%o5
+ ld [s2_ptr+28],%g1
+ ldd [s1_ptr+24],%g2
+ std %o4,[res_ptr+16]
+ subxcc %g2,%g4,%o4
+ ld [s2_ptr+32],%g4
+ subxcc %g3,%g1,%o5
+ ld [s2_ptr+36],%g1
+ ldd [s1_ptr+32],%g2
+ std %o4,[res_ptr+24]
+ addx %g0,%g0,%o4 ! save cy in register
+ addcc size,-8,size
+ add s1_ptr,32,s1_ptr
+ add s2_ptr,32,s2_ptr
+ add res_ptr,32,res_ptr
+ bge Loop1b
+ subcc %g0,%o4,%g0 ! restore cy
+
+Lfin1b: addcc size,8-2,size
+ blt Lend1b
+ subcc %g0,%o4,%g0 ! restore cy
+/* Add blocks of 2 limbs until less than 2 limbs remain */
+Loope1b:subxcc %g2,%g4,%o4
+ ld [s2_ptr+8],%g4
+ subxcc %g3,%g1,%o5
+ ld [s2_ptr+12],%g1
+ ldd [s1_ptr+8],%g2
+ std %o4,[res_ptr+0]
+ addx %g0,%g0,%o4 ! save cy in register
+ addcc size,-2,size
+ add s1_ptr,8,s1_ptr
+ add s2_ptr,8,s2_ptr
+ add res_ptr,8,res_ptr
+ bge Loope1b
+ subcc %g0,%o4,%g0 ! restore cy
+Lend1b: subxcc %g2,%g4,%o4
+ subxcc %g3,%g1,%o5
+ std %o4,[res_ptr+0]
+ addx %g0,%g0,%o4 ! save cy in register
+
+ andcc size,1,%g0
+ be Lret1b
+ subcc %g0,%o4,%g0 ! restore cy
+/* Add last limb */
+ ld [s2_ptr+8],%g4
+ ld [s1_ptr+8],%g2
+ subxcc %g2,%g4,%o4
+ st %o4,[res_ptr+8]
+
+Lret1b: retl
+ addx %g0,%g0,%o0 ! return carry-out from most sign. limb
+
+! ** V2 **
+/* If we come here, the alignment of s1_ptr and res_ptr as well as the
+ alignment of s2_ptr and res_ptr differ. Since there are only two ways
+ things can be aligned (that we care about) we now know that the alignment
+ of s1_ptr and s2_ptr are the same. */
+
+L2: cmp size,1
+ be Ljone
+ nop
+ andcc s1_ptr,4,%g0 ! s1_ptr unaligned? Side effect: cy=0
+ be L_v2 ! if no, branch
+ nop
+/* Add least significant limb separately to align s1_ptr and s2_ptr */
+ ld [s1_ptr],%g4
+ add s1_ptr,4,s1_ptr
+ ld [s2_ptr],%g2
+ add s2_ptr,4,s2_ptr
+ add size,-1,size
+ subcc %g4,%g2,%o4
+ st %o4,[res_ptr]
+ add res_ptr,4,res_ptr
+
+L_v2: addx %g0,%g0,%o4 ! save cy in register
+ addcc size,-8,size
+ blt Lfin2
+ subcc %g0,%o4,%g0 ! restore cy
+/* Add blocks of 8 limbs until less than 8 limbs remain */
+Loop2: ldd [s1_ptr+0],%g2
+ ldd [s2_ptr+0],%o4
+ subxcc %g2,%o4,%g2
+ st %g2,[res_ptr+0]
+ subxcc %g3,%o5,%g3
+ st %g3,[res_ptr+4]
+ ldd [s1_ptr+8],%g2
+ ldd [s2_ptr+8],%o4
+ subxcc %g2,%o4,%g2
+ st %g2,[res_ptr+8]
+ subxcc %g3,%o5,%g3
+ st %g3,[res_ptr+12]
+ ldd [s1_ptr+16],%g2
+ ldd [s2_ptr+16],%o4
+ subxcc %g2,%o4,%g2
+ st %g2,[res_ptr+16]
+ subxcc %g3,%o5,%g3
+ st %g3,[res_ptr+20]
+ ldd [s1_ptr+24],%g2
+ ldd [s2_ptr+24],%o4
+ subxcc %g2,%o4,%g2
+ st %g2,[res_ptr+24]
+ subxcc %g3,%o5,%g3
+ st %g3,[res_ptr+28]
+ addx %g0,%g0,%o4 ! save cy in register
+ addcc size,-8,size
+ add s1_ptr,32,s1_ptr
+ add s2_ptr,32,s2_ptr
+ add res_ptr,32,res_ptr
+ bge Loop2
+ subcc %g0,%o4,%g0 ! restore cy
+
+Lfin2: addcc size,8-2,size
+ blt Lend2
+ subcc %g0,%o4,%g0 ! restore cy
+Loope2: ldd [s1_ptr+0],%g2
+ ldd [s2_ptr+0],%o4
+ subxcc %g2,%o4,%g2
+ st %g2,[res_ptr+0]
+ subxcc %g3,%o5,%g3
+ st %g3,[res_ptr+4]
+ addx %g0,%g0,%o4 ! save cy in register
+ addcc size,-2,size
+ add s1_ptr,8,s1_ptr
+ add s2_ptr,8,s2_ptr
+ add res_ptr,8,res_ptr
+ bge Loope2
+ subcc %g0,%o4,%g0 ! restore cy
+Lend2: andcc size,1,%g0
+ be Lret2
+ subcc %g0,%o4,%g0 ! restore cy
+/* Add last limb */
+Ljone: ld [s1_ptr],%g4
+ ld [s2_ptr],%g2
+ subxcc %g4,%g2,%o4
+ st %o4,[res_ptr]
+
+Lret2: retl
+ addx %g0,%g0,%o0 ! return carry-out from most sign. limb
--- /dev/null
+! SPARC __mpn_submul_1 -- Multiply a limb vector with a limb and subtract
+! the result from a second limb vector.
+
+! Copyright (C) 1992, 1993, 1994 Free Software Foundation, Inc.
+
+! This file is part of the GNU MP Library.
+
+! 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
+! the Free Software Foundation; either version 2 of the License, or (at your
+! option) any later version.
+
+! 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
+! 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 the GNU MP Library; see the file COPYING.LIB. If not, write to
+! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
+! MA 02111-1307, USA.
+
+
+! INPUT PARAMETERS
+! res_ptr o0
+! s1_ptr o1
+! size o2
+! s2_limb o3
+
+#include "sysdep.h"
+
+.text
+ .align 4
+ .global C_SYMBOL_NAME(__mpn_submul_1)
+C_SYMBOL_NAME(__mpn_submul_1):
+ ! Make S1_PTR and RES_PTR point at the end of their blocks
+ ! and put (- 4 x SIZE) in index/loop counter.
+ sll %o2,2,%o2
+ add %o0,%o2,%o4 ! RES_PTR in o4 since o0 is retval
+ add %o1,%o2,%o1
+ sub %g0,%o2,%o2
+
+ cmp %o3,0xfff
+ bgu Large
+ nop
+
+ ld [%o1+%o2],%o5
+ mov 0,%o0
+ b L0
+ add %o4,-4,%o4
+Loop0:
+ subcc %o5,%g1,%g1
+ ld [%o1+%o2],%o5
+ addx %o0,%g0,%o0
+ st %g1,[%o4+%o2]
+L0: wr %g0,%o3,%y
+ sra %o5,31,%g2
+ and %o3,%g2,%g2
+ andcc %g1,0,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,%o5,%g1
+ mulscc %g1,0,%g1
+ sra %g1,20,%g4
+ sll %g1,12,%g1
+ rd %y,%g3
+ srl %g3,20,%g3
+ or %g1,%g3,%g1
+
+ addcc %g1,%o0,%g1
+ addx %g2,%g4,%o0 ! add sign-compensation and cy to hi limb
+ addcc %o2,4,%o2 ! loop counter
+ bne Loop0
+ ld [%o4+%o2],%o5
+
+ subcc %o5,%g1,%g1
+ addx %o0,%g0,%o0
+ retl
+ st %g1,[%o4+%o2]
+
+
+Large: ld [%o1+%o2],%o5
+ mov 0,%o0
+ sra %o3,31,%g4 ! g4 = mask of ones iff S2_LIMB < 0
+ b L1
+ add %o4,-4,%o4
+Loop:
+ subcc %o5,%g3,%g3
+ ld [%o1+%o2],%o5
+ addx %o0,%g0,%o0
+ st %g3,[%o4+%o2]
+L1: wr %g0,%o5,%y
+ and %o5,%g4,%g2
+ andcc %g0,%g0,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%o3,%g1
+ mulscc %g1,%g0,%g1
+ rd %y,%g3
+ addcc %g3,%o0,%g3
+ addx %g2,%g1,%o0
+ addcc %o2,4,%o2
+ bne Loop
+ ld [%o4+%o2],%o5
+
+ subcc %o5,%g3,%g3
+ addx %o0,%g0,%o0
+ retl
+ st %g3,[%o4+%o2]
--- /dev/null
+! SPARC v7 __udiv_qrnnd division support, used from longlong.h.
+! This is for v7 CPUs with a floating-point unit.
+
+! Copyright (C) 1993, 1994, 1996 Free Software Foundation, Inc.
+
+! This file is part of the GNU MP Library.
+
+! 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
+! the Free Software Foundation; either version 2 of the License, or (at your
+! option) any later version.
+
+! 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
+! 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 the GNU MP Library; see the file COPYING.LIB. If not, write to
+! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
+! MA 02111-1307, USA.
+
+
+! INPUT PARAMETERS
+! rem_ptr i0
+! n1 i1
+! n0 i2
+! d i3
+
+#include "sysdep.h"
+#undef ret /* Kludge for glibc */
+
+ .text
+ .align 8
+LC0: .double 0r4294967296
+LC1: .double 0r2147483648
+
+ .align 4
+ .global C_SYMBOL_NAME(__udiv_qrnnd)
+C_SYMBOL_NAME(__udiv_qrnnd):
+ !#PROLOGUE# 0
+ save %sp,-104,%sp
+ !#PROLOGUE# 1
+ st %i1,[%fp-8]
+ ld [%fp-8],%f10
+ sethi %hi(LC0),%o7
+ fitod %f10,%f4
+ ldd [%o7+%lo(LC0)],%f8
+ cmp %i1,0
+ bge L248
+ mov %i0,%i5
+ faddd %f4,%f8,%f4
+L248:
+ st %i2,[%fp-8]
+ ld [%fp-8],%f10
+ fmuld %f4,%f8,%f6
+ cmp %i2,0
+ bge L249
+ fitod %f10,%f2
+ faddd %f2,%f8,%f2
+L249:
+ st %i3,[%fp-8]
+ faddd %f6,%f2,%f2
+ ld [%fp-8],%f10
+ cmp %i3,0
+ bge L250
+ fitod %f10,%f4
+ faddd %f4,%f8,%f4
+L250:
+ fdivd %f2,%f4,%f2
+ sethi %hi(LC1),%o7
+ ldd [%o7+%lo(LC1)],%f4
+ fcmped %f2,%f4
+ nop
+ fbge,a L251
+ fsubd %f2,%f4,%f2
+ fdtoi %f2,%f2
+ st %f2,[%fp-8]
+ b L252
+ ld [%fp-8],%i4
+L251:
+ fdtoi %f2,%f2
+ st %f2,[%fp-8]
+ ld [%fp-8],%i4
+ sethi %hi(-2147483648),%g2
+ xor %i4,%g2,%i4
+L252:
+ wr %g0,%i4,%y
+ sra %i3,31,%g2
+ and %i4,%g2,%g2
+ andcc %g0,0,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,%i3,%g1
+ mulscc %g1,0,%g1
+ add %g1,%g2,%i0
+ rd %y,%g3
+ subcc %i2,%g3,%o7
+ subxcc %i1,%i0,%g0
+ be L253
+ cmp %o7,%i3
+
+ add %i4,-1,%i0
+ add %o7,%i3,%o7
+ st %o7,[%i5]
+ ret
+ restore
+L253:
+ blu L246
+ mov %i4,%i0
+ add %i4,1,%i0
+ sub %o7,%i3,%o7
+L246:
+ st %o7,[%i5]
+ ret
+ restore
--- /dev/null
+! SPARC v7 __udiv_qrnnd division support, used from longlong.h.
+! This is for v7 CPUs without a floating-point unit.
+
+! Copyright (C) 1993, 1994, 1996 Free Software Foundation, Inc.
+
+! This file is part of the GNU MP Library.
+
+! 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
+! the Free Software Foundation; either version 2 of the License, or (at your
+! option) any later version.
+
+! 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
+! 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 the GNU MP Library; see the file COPYING.LIB. If not, write to
+! the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
+! MA 02111-1307, USA.
+
+
+! INPUT PARAMETERS
+! rem_ptr o0
+! n1 o1
+! n0 o2
+! d o3
+
+#include "sysdep.h"
+
+ .text
+ .align 4
+ .global C_SYMBOL_NAME(__udiv_qrnnd)
+C_SYMBOL_NAME(__udiv_qrnnd):
+ tst %o3
+ bneg Largedivisor
+ mov 8,%g1
+
+ b Lp1
+ addxcc %o2,%o2,%o2
+
+Lplop: bcc Ln1
+ addxcc %o2,%o2,%o2
+Lp1: addx %o1,%o1,%o1
+ subcc %o1,%o3,%o4
+ bcc Ln2
+ addxcc %o2,%o2,%o2
+Lp2: addx %o1,%o1,%o1
+ subcc %o1,%o3,%o4
+ bcc Ln3
+ addxcc %o2,%o2,%o2
+Lp3: addx %o1,%o1,%o1
+ subcc %o1,%o3,%o4
+ bcc Ln4
+ addxcc %o2,%o2,%o2
+Lp4: addx %o1,%o1,%o1
+ addcc %g1,-1,%g1
+ bne Lplop
+ subcc %o1,%o3,%o4
+ bcc Ln5
+ addxcc %o2,%o2,%o2
+Lp5: st %o1,[%o0]
+ retl
+ xnor %g0,%o2,%o0
+
+Lnlop: bcc Lp1
+ addxcc %o2,%o2,%o2
+Ln1: addx %o4,%o4,%o4
+ subcc %o4,%o3,%o1
+ bcc Lp2
+ addxcc %o2,%o2,%o2
+Ln2: addx %o4,%o4,%o4
+ subcc %o4,%o3,%o1
+ bcc Lp3
+ addxcc %o2,%o2,%o2
+Ln3: addx %o4,%o4,%o4
+ subcc %o4,%o3,%o1
+ bcc Lp4
+ addxcc %o2,%o2,%o2
+Ln4: addx %o4,%o4,%o4
+ addcc %g1,-1,%g1
+ bne Lnlop
+ subcc %o4,%o3,%o1
+ bcc Lp5
+ addxcc %o2,%o2,%o2
+Ln5: st %o4,[%o0]
+ retl
+ xnor %g0,%o2,%o0
+
+Largedivisor:
+ and %o2,1,%o5 ! %o5 = n0 & 1
+
+ srl %o2,1,%o2
+ sll %o1,31,%g2
+ or %g2,%o2,%o2 ! %o2 = lo(n1n0 >> 1)
+ srl %o1,1,%o1 ! %o1 = hi(n1n0 >> 1)
+
+ and %o3,1,%g2
+ srl %o3,1,%g3 ! %g3 = floor(d / 2)
+ add %g3,%g2,%g3 ! %g3 = ceil(d / 2)
+
+ b LLp1
+ addxcc %o2,%o2,%o2
+
+LLplop: bcc LLn1
+ addxcc %o2,%o2,%o2
+LLp1: addx %o1,%o1,%o1
+ subcc %o1,%g3,%o4
+ bcc LLn2
+ addxcc %o2,%o2,%o2
+LLp2: addx %o1,%o1,%o1
+ subcc %o1,%g3,%o4
+ bcc LLn3
+ addxcc %o2,%o2,%o2
+LLp3: addx %o1,%o1,%o1
+ subcc %o1,%g3,%o4
+ bcc LLn4
+ addxcc %o2,%o2,%o2
+LLp4: addx %o1,%o1,%o1
+ addcc %g1,-1,%g1
+ bne LLplop
+ subcc %o1,%g3,%o4
+ bcc LLn5
+ addxcc %o2,%o2,%o2
+LLp5: add %o1,%o1,%o1 ! << 1
+ tst %g2
+ bne Oddp
+ add %o5,%o1,%o1
+ st %o1,[%o0]
+ retl
+ xnor %g0,%o2,%o0
+
+LLnlop: bcc LLp1
+ addxcc %o2,%o2,%o2
+LLn1: addx %o4,%o4,%o4
+ subcc %o4,%g3,%o1
+ bcc LLp2
+ addxcc %o2,%o2,%o2
+LLn2: addx %o4,%o4,%o4
+ subcc %o4,%g3,%o1
+ bcc LLp3
+ addxcc %o2,%o2,%o2
+LLn3: addx %o4,%o4,%o4
+ subcc %o4,%g3,%o1
+ bcc LLp4
+ addxcc %o2,%o2,%o2
+LLn4: addx %o4,%o4,%o4
+ addcc %g1,-1,%g1
+ bne LLnlop
+ subcc %o4,%g3,%o1
+ bcc LLp5
+ addxcc %o2,%o2,%o2
+LLn5: add %o4,%o4,%o4 ! << 1
+ tst %g2
+ bne Oddn
+ add %o5,%o4,%o4
+ st %o4,[%o0]
+ retl
+ xnor %g0,%o2,%o0
+
+Oddp: xnor %g0,%o2,%o2
+ ! q' in %o2. r' in %o1
+ addcc %o1,%o2,%o1
+ bcc LLp6
+ addx %o2,0,%o2
+ sub %o1,%o3,%o1
+LLp6: subcc %o1,%o3,%g0
+ bcs LLp7
+ subx %o2,-1,%o2
+ sub %o1,%o3,%o1
+LLp7: st %o1,[%o0]
+ retl
+ mov %o2,%o0
+
+Oddn: xnor %g0,%o2,%o2
+ ! q' in %o2. r' in %o4
+ addcc %o4,%o2,%o4
+ bcc LLn6
+ addx %o2,0,%o2
+ sub %o4,%o3,%o4
+LLn6: subcc %o4,%o3,%g0
+ bcs LLn7
+ subx %o2,-1,%o2
+ sub %o4,%o3,%o4
+LLn7: st %o4,[%o0]
+ retl
+ mov %o2,%o0
projects / ghc-hetmet.git / commitdiff