diff options
Diffstat (limited to 'sysdeps/sparc/sparc64/addmul_1.S')
| -rw-r--r-- | sysdeps/sparc/sparc64/addmul_1.S | 36 |
1 files changed, 16 insertions, 20 deletions
diff --git a/sysdeps/sparc/sparc64/addmul_1.S b/sysdeps/sparc/sparc64/addmul_1.S index 6782db77df..db8f53656e 100644 --- a/sysdeps/sparc/sparc64/addmul_1.S +++ b/sysdeps/sparc/sparc64/addmul_1.S @@ -1,7 +1,7 @@ /* SPARC v9 __mpn_addmul_1 -- Multiply a limb vector with a single limb and add the product to a second limb vector. - Copyright (C) 1996 Free Software Foundation, Inc. + Copyright (C) 1996, 1999 Free Software Foundation, Inc. This file is part of the GNU MP Library. @@ -31,29 +31,26 @@ ENTRY(__mpn_addmul_1) - !#PROLOGUE# 0 save %sp,-192,%sp - !#PROLOGUE# 1 sub %g0,%i2,%o7 - sllx %o7,3,%g5 - sub %i1,%g5,%o3 - sub %i0,%g5,%o4 mov 0,%o0 ! zero cy_limb - + sllx %o7,3,%o7 + sethi %hi(0x80000000),%o2 srl %i3,0,%o1 ! extract low 32 bits of s2_limb + sub %i1,%o7,%o3 srlx %i3,32,%i3 ! extract high 32 bits of s2_limb - mov 1,%o2 - sllx %o2,32,%o2 ! o2 = 0x100000000 + sub %i0,%o7,%o4 + add %o2,%o2,%o2 ! o2 = 0x100000000 ! hi ! ! mid-1 ! ! mid-2 ! ! lo ! 1: - sllx %o7,3,%g1 - ldx [%o3+%g1],%g5 + ldx [%o3+%o7],%g5 srl %g5,0,%i0 ! zero hi bits + ldx [%o4+%o7],%l1 srlx %g5,32,%g5 mulx %o1,%i0,%i4 ! lo product mulx %i3,%i0,%i1 ! mid-1 product @@ -64,25 +61,24 @@ ENTRY(__mpn_addmul_1) addcc %i1,%l2,%i1 ! add mid products mov 0,%l0 ! we need the carry from that add... movcs %xcc,%o2,%l0 ! ...compute it and... + sllx %i1,32,%i0 ! align low bits of mid product add %i5,%l0,%i5 ! ...add to bit 32 of the hi product - sllx %i1,32,%i0 ! align low bits of mid product srl %i4,0,%g5 ! zero high 32 bits of lo product add %i0,%g5,%i0 ! combine into low 64 bits of result srlx %i1,32,%i1 ! extract high bits of mid product... + addcc %i0,%o0,%i0 ! add cy_limb to low 64 bits of result add %i5,%i1,%i1 ! ...and add them to the high result - addcc %i0,%o0,%i0 ! add cy_limb to low 64 bits of result mov 0,%g5 movcs %xcc,1,%g5 - add %o7,1,%o7 - ldx [%o4+%g1],%l1 addcc %l1,%i0,%i0 - movcs %xcc,1,%g5 - stx %i0,[%o4+%g1] - brnz %o7,1b + stx %i0,[%o4+%o7] + add %g5,1,%l1 + movcs %xcc,%l1,%g5 + addcc %o7,8,%o7 + bne,pt %xcc,1b add %i1,%g5,%o0 ! compute new cy_limb - mov %o0,%i0 jmpl %i7+8, %g0 - restore + restore %o0,%g0,%o0 END(__mpn_addmul_1) |