From: simonm Date: Fri, 5 Jun 1998 14:44:12 +0000 (+0000) Subject: [project @ 1998-06-05 14:44:10 by simonm] X-Git-Tag: Approx_2487_patches~594 X-Git-Url: http://git.megacz.com/?a=commitdiff_plain;h=00c8f6943bb7b5c6d70aa7164ce4c0a92dd118e4;p=ghc-hetmet.git [project @ 1998-06-05 14:44:10 by simonm] Initial revision --- diff --git a/ghc/rts/gmp/mpn/sparc64/add_n.s b/ghc/rts/gmp/mpn/sparc64/add_n.s new file mode 100644 index 0000000..01d1f49 --- /dev/null +++ b/ghc/rts/gmp/mpn/sparc64/add_n.s @@ -0,0 +1,58 @@ +! SPARC v9 __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 +! res_ptr %o0 +! s1_ptr %o1 +! s2_ptr %o2 +! size %o3 + +.section ".text" + .align 4 + .global __mpn_add_n + .type __mpn_add_n,#function + .proc 04 +__mpn_add_n: + sub %g0,%o3,%g3 + sllx %o3,3,%g1 + add %o1,%g1,%o1 ! make s1_ptr point at end + add %o2,%g1,%o2 ! make s2_ptr point at end + add %o0,%g1,%o0 ! make res_ptr point at end + mov 0,%o4 ! clear carry variable + sllx %g3,3,%o5 ! compute initial address index + +.Loop: ldx [%o2+%o5],%g1 ! load s2 limb + add %g3,1,%g3 ! increment loop count + ldx [%o1+%o5],%g2 ! load s1 limb + addcc %g1,%o4,%g1 ! add s2 limb and carry variable + movcc %xcc,0,%o4 ! if carry-out, o4 was 1; clear it + addcc %g1,%g2,%g1 ! add s1 limb to sum + stx %g1,[%o0+%o5] ! store result + add %o5,8,%o5 ! increment address index + brnz,pt %g3,.Loop + movcs %xcc,1,%o4 ! if s1 add gave carry, record it + + retl + mov %o4,%o0 +.LLfe1: + .size __mpn_add_n,.LLfe1-__mpn_add_n diff --git a/ghc/rts/gmp/mpn/sparc64/addmul_1.s b/ghc/rts/gmp/mpn/sparc64/addmul_1.s new file mode 100644 index 0000000..8d86390 --- /dev/null +++ b/ghc/rts/gmp/mpn/sparc64/addmul_1.s @@ -0,0 +1,89 @@ +! 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. + +! 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 + +.section ".text" + .align 4 + .global __mpn_addmul_1 + .type __mpn_addmul_1,#function + .proc 016 +__mpn_addmul_1: + !#PROLOGUE# 0 + save %sp,-160,%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 + + srl %i3,0,%o1 ! extract low 32 bits of s2_limb + srlx %i3,32,%i3 ! extract high 32 bits of s2_limb + mov 1,%o2 + sllx %o2,32,%o2 ! o2 = 0x100000000 + + ! hi ! + ! mid-1 ! + ! mid-2 ! + ! lo ! +.Loop: + sllx %o7,3,%g1 + ldx [%o3+%g1],%g5 + srl %g5,0,%i0 ! zero hi bits + srlx %g5,32,%g5 + mulx %o1,%i0,%i4 ! lo product + mulx %i3,%i0,%i1 ! mid-1 product + mulx %o1,%g5,%l2 ! mid-2 product + mulx %i3,%g5,%i5 ! hi product + srlx %i4,32,%i0 ! extract high 32 bits of lo product... + add %i1,%i0,%i1 ! ...and add it to the mid-1 product + addcc %i1,%l2,%i1 ! add mid products + mov 0,%l0 ! we need the carry from that add... + movcs %xcc,%o2,%l0 ! ...compute it and... + 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... + 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,.Loop + add %i1,%g5,%o0 ! compute new cy_limb + + mov %o0,%i0 + ret + restore +.LLfe1: + .size __mpn_addmul_1,.LLfe1-__mpn_addmul_1 diff --git a/ghc/rts/gmp/mpn/sparc64/lshift.s b/ghc/rts/gmp/mpn/sparc64/lshift.s new file mode 100644 index 0000000..ad1f667 --- /dev/null +++ b/ghc/rts/gmp/mpn/sparc64/lshift.s @@ -0,0 +1,96 @@ +! SPARC v9 __mpn_lshift -- + +! Copyright (C) 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 + +.section ".text" + .align 4 + .global __mpn_lshift + .type __mpn_lshift,#function + .proc 04 +__mpn_lshift: + sllx %o2,3,%g1 + add %o1,%g1,%o1 ! make %o1 point at end of src + ldx [%o1-8],%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 + and %o2,4-1,%g4 ! number of limbs in first loop + srlx %g2,%o5,%g1 ! compute function result + brz,pn %g4,.L0 ! if multiple of 4 limbs, skip first loop + stx %g1,[%sp+80] + + sub %o2,%g4,%o2 ! adjust count for main loop + +.Loop0: ldx [%o1-16],%g3 + add %o0,-8,%o0 + add %o1,-8,%o1 + add %g4,-1,%g4 + sllx %g2,%o3,%o4 + srlx %g3,%o5,%g1 + mov %g3,%g2 + or %o4,%g1,%o4 + brnz,pt %g4,.Loop0 + stx %o4,[%o0+0] + +.L0: brz,pn %o2,.Lend + nop + +.Loop: ldx [%o1-16],%g3 + add %o0,-32,%o0 + add %o2,-4,%o2 + sllx %g2,%o3,%o4 + srlx %g3,%o5,%g1 + + ldx [%o1-24],%g2 + sllx %g3,%o3,%g4 + or %o4,%g1,%o4 + stx %o4,[%o0+24] + srlx %g2,%o5,%g1 + + ldx [%o1-32],%g3 + sllx %g2,%o3,%o4 + or %g4,%g1,%g4 + stx %g4,[%o0+16] + srlx %g3,%o5,%g1 + + ldx [%o1-40],%g2 + sllx %g3,%o3,%g4 + or %o4,%g1,%o4 + stx %o4,[%o0+8] + srlx %g2,%o5,%g1 + + add %o1,-32,%o1 + or %g4,%g1,%g4 + brnz,pt %o2,.Loop + stx %g4,[%o0+0] + +.Lend: sllx %g2,%o3,%g2 + stx %g2,[%o0-8] + retl + ldx [%sp+80],%o0 +.LLfe1: + .size __mpn_lshift,.LLfe1-__mpn_lshift diff --git a/ghc/rts/gmp/mpn/sparc64/mul_1.s b/ghc/rts/gmp/mpn/sparc64/mul_1.s new file mode 100644 index 0000000..91d6eb0 --- /dev/null +++ b/ghc/rts/gmp/mpn/sparc64/mul_1.s @@ -0,0 +1,86 @@ +! SPARC v9 __mpn_mul_1 -- Multiply a limb vector with a single limb and +! store the product in a second 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 +! res_ptr o0 +! s1_ptr o1 +! size o2 +! s2_limb o3 + +.section ".text" + .align 4 + .global __mpn_mul_1 + .type __mpn_mul_1,#function + .proc 016 +__mpn_mul_1: + !#PROLOGUE# 0 + save %sp,-160,%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 + + srl %i3,0,%o1 ! extract low 32 bits of s2_limb + srlx %i3,32,%i3 ! extract high 32 bits of s2_limb + mov 1,%o2 + sllx %o2,32,%o2 ! o2 = 0x100000000 + + ! hi ! + ! mid-1 ! + ! mid-2 ! + ! lo ! +.Loop: + sllx %o7,3,%g1 + ldx [%o3+%g1],%g5 + srl %g5,0,%i0 ! zero hi bits + srlx %g5,32,%g5 + mulx %o1,%i0,%i4 ! lo product + mulx %i3,%i0,%i1 ! mid-1 product + mulx %o1,%g5,%l2 ! mid-2 product + mulx %i3,%g5,%i5 ! hi product + srlx %i4,32,%i0 ! extract high 32 bits of lo product... + add %i1,%i0,%i1 ! ...and add it to the mid-1 product + addcc %i1,%l2,%i1 ! add mid products + mov 0,%l0 ! we need the carry from that add... + movcs %xcc,%o2,%l0 ! ...compute it and... + 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... + 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 + stx %i0,[%o4+%g1] + brnz %o7,.Loop + add %i1,%g5,%o0 ! compute new cy_limb + + mov %o0,%i0 + ret + restore +.LLfe1: + .size __mpn_mul_1,.LLfe1-__mpn_mul_1 diff --git a/ghc/rts/gmp/mpn/sparc64/rshift.s b/ghc/rts/gmp/mpn/sparc64/rshift.s new file mode 100644 index 0000000..ff6a380 --- /dev/null +++ b/ghc/rts/gmp/mpn/sparc64/rshift.s @@ -0,0 +1,93 @@ +! SPARC v9 __mpn_rshift -- + +! Copyright (C) 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 + +.section ".text" + .align 4 + .global __mpn_rshift + .type __mpn_rshift,#function + .proc 04 +__mpn_rshift: + ldx [%o1],%g2 ! load first limb + sub %g0,%o3,%o5 ! negate shift count + add %o2,-1,%o2 + and %o2,4-1,%g4 ! number of limbs in first loop + sllx %g2,%o5,%g1 ! compute function result + brz,pn %g4,.L0 ! if multiple of 4 limbs, skip first loop + stx %g1,[%sp+80] + + sub %o2,%g4,%o2 ! adjust count for main loop + +.Loop0: ldx [%o1+8],%g3 + add %o0,8,%o0 + add %o1,8,%o1 + add %g4,-1,%g4 + srlx %g2,%o3,%o4 + sllx %g3,%o5,%g1 + mov %g3,%g2 + or %o4,%g1,%o4 + brnz,pt %g4,.Loop0 + stx %o4,[%o0-8] + +.L0: brz,pn %o2,.Lend + nop + +.Loop: ldx [%o1+8],%g3 + add %o0,32,%o0 + add %o2,-4,%o2 + srlx %g2,%o3,%o4 + sllx %g3,%o5,%g1 + + ldx [%o1+16],%g2 + srlx %g3,%o3,%g4 + or %o4,%g1,%o4 + stx %o4,[%o0-32] + sllx %g2,%o5,%g1 + + ldx [%o1+24],%g3 + srlx %g2,%o3,%o4 + or %g4,%g1,%g4 + stx %g4,[%o0-24] + sllx %g3,%o5,%g1 + + ldx [%o1+32],%g2 + srlx %g3,%o3,%g4 + or %o4,%g1,%o4 + stx %o4,[%o0-16] + sllx %g2,%o5,%g1 + + add %o1,32,%o1 + or %g4,%g1,%g4 + brnz %o2,.Loop + stx %g4,[%o0-8] + +.Lend: srlx %g2,%o3,%g2 + stx %g2,[%o0-0] + retl + ldx [%sp+80],%o0 +.LLfe1: + .size __mpn_rshift,.LLfe1-__mpn_rshift diff --git a/ghc/rts/gmp/mpn/sparc64/sub_n.s b/ghc/rts/gmp/mpn/sparc64/sub_n.s new file mode 100644 index 0000000..d4842b8 --- /dev/null +++ b/ghc/rts/gmp/mpn/sparc64/sub_n.s @@ -0,0 +1,58 @@ +! SPARC v9 __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 +! res_ptr %o0 +! s1_ptr %o1 +! s2_ptr %o2 +! size %o3 + +.section ".text" + .align 4 + .global __mpn_sub_n + .type __mpn_sub_n,#function + .proc 04 +__mpn_sub_n: + sub %g0,%o3,%g3 + sllx %o3,3,%g1 + add %o1,%g1,%o1 ! make s1_ptr point at end + add %o2,%g1,%o2 ! make s2_ptr point at end + add %o0,%g1,%o0 ! make res_ptr point at end + mov 0,%o4 ! clear carry variable + sllx %g3,3,%o5 ! compute initial address index + +.Loop: ldx [%o2+%o5],%g1 ! load s2 limb + add %g3,1,%g3 ! increment loop count + ldx [%o1+%o5],%g2 ! load s1 limb + addcc %g1,%o4,%g1 ! add s2 limb and carry variable + movcc %xcc,0,%o4 ! if carry-out, o4 was 1; clear it + subcc %g1,%g2,%g1 ! subtract s1 limb from sum + stx %g1,[%o0+%o5] ! store result + add %o5,8,%o5 ! increment address index + brnz,pt %g3,.Loop + movcs %xcc,1,%o4 ! if s1 subtract gave carry, record it + + retl + mov %o4,%o0 +.LLfe1: + .size __mpn_sub_n,.LLfe1-__mpn_sub_n diff --git a/ghc/rts/gmp/mpn/sparc64/submul_1.s b/ghc/rts/gmp/mpn/sparc64/submul_1.s new file mode 100644 index 0000000..e796243 --- /dev/null +++ b/ghc/rts/gmp/mpn/sparc64/submul_1.s @@ -0,0 +1,89 @@ +! SPARC v9 __mpn_submul_1 -- Multiply a limb vector with a single limb and +! subtract the product from a second limb vector. + +! Copyright (C) 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 +! s1_ptr o1 +! size o2 +! s2_limb o3 + +.section ".text" + .align 4 + .global __mpn_submul_1 + .type __mpn_submul_1,#function + .proc 016 +__mpn_submul_1: + !#PROLOGUE# 0 + save %sp,-160,%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 + + srl %i3,0,%o1 ! extract low 32 bits of s2_limb + srlx %i3,32,%i3 ! extract high 32 bits of s2_limb + mov 1,%o2 + sllx %o2,32,%o2 ! o2 = 0x100000000 + + ! hi ! + ! mid-1 ! + ! mid-2 ! + ! lo ! +.Loop: + sllx %o7,3,%g1 + ldx [%o3+%g1],%g5 + srl %g5,0,%i0 ! zero hi bits + srlx %g5,32,%g5 + mulx %o1,%i0,%i4 ! lo product + mulx %i3,%i0,%i1 ! mid-1 product + mulx %o1,%g5,%l2 ! mid-2 product + mulx %i3,%g5,%i5 ! hi product + srlx %i4,32,%i0 ! extract high 32 bits of lo product... + add %i1,%i0,%i1 ! ...and add it to the mid-1 product + addcc %i1,%l2,%i1 ! add mid products + mov 0,%l0 ! we need the carry from that add... + movcs %xcc,%o2,%l0 ! ...compute it and... + 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... + 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 + subcc %l1,%i0,%i0 + movcs %xcc,1,%g5 + stx %i0,[%o4+%g1] + brnz %o7,.Loop + add %i1,%g5,%o0 ! compute new cy_limb + + mov %o0,%i0 + ret + restore +.LLfe1: + .size __mpn_submul_1,.LLfe1-__mpn_submul_1