From 11da511c784eca003deb90c23570f0873954e0de Mon Sep 17 00:00:00 2001 From: Duncan Wilkie Date: Sat, 18 Nov 2023 06:11:09 -0600 Subject: Initial commit. --- gmp-6.3.0/mpn/pa64/aors_n.asm | 130 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 130 insertions(+) create mode 100644 gmp-6.3.0/mpn/pa64/aors_n.asm (limited to 'gmp-6.3.0/mpn/pa64/aors_n.asm') diff --git a/gmp-6.3.0/mpn/pa64/aors_n.asm b/gmp-6.3.0/mpn/pa64/aors_n.asm new file mode 100644 index 0000000..ab4536f --- /dev/null +++ b/gmp-6.3.0/mpn/pa64/aors_n.asm @@ -0,0 +1,130 @@ +dnl HP-PA 2.0 mpn_add_n, mpn_sub_n + +dnl Copyright 1997, 2000, 2002, 2003, 2009, 2010 Free Software Foundation, +dnl Inc. + +dnl This file is part of the GNU MP Library. +dnl +dnl The GNU MP Library is free software; you can redistribute it and/or modify +dnl it under the terms of either: +dnl +dnl * the GNU Lesser General Public License as published by the Free +dnl Software Foundation; either version 3 of the License, or (at your +dnl option) any later version. +dnl +dnl or +dnl +dnl * the GNU General Public License as published by the Free Software +dnl Foundation; either version 2 of the License, or (at your option) any +dnl later version. +dnl +dnl or both in parallel, as here. +dnl +dnl The GNU MP Library is distributed in the hope that it will be useful, but +dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY +dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License +dnl for more details. +dnl +dnl You should have received copies of the GNU General Public License and the +dnl GNU Lesser General Public License along with the GNU MP Library. If not, +dnl see https://www.gnu.org/licenses/. + + +dnl This runs at 2 cycles/limb on PA8000 and 1.6875 cycles/limb on PA8500. It +dnl should be possible to reach the cache bandwidth 1.5 cycles/limb at least +dnl with PA8500. The problem now is stalling of the first ADD,DC after LDO, +dnl where the processor gets confused about where carry comes from. + +include(`../config.m4') + +dnl INPUT PARAMETERS +define(`rp',`%r26') +define(`up',`%r25') +define(`vp',`%r24') +define(`n',`%r23') + +ifdef(`OPERATION_add_n', ` + define(ADCSBC, `add,dc') + define(INITCY, `addi -1,%r22,%r0') + define(func, mpn_add_n) + define(func_nc, mpn_add_nc)') +ifdef(`OPERATION_sub_n', ` + define(ADCSBC, `sub,db') + define(INITCY, `subi 0,%r22,%r0') + define(func, mpn_sub_n) + define(func_nc, mpn_sub_nc)') + +MULFUNC_PROLOGUE(mpn_add_n mpn_add_nc mpn_sub_n mpn_sub_nc) + +ifdef(`HAVE_ABI_2_0w', +` .level 2.0w +',` .level 2.0 +') +PROLOGUE(func_nc) +ifdef(`HAVE_ABI_2_0w', +` b L(com) + nop +',` b L(com) + ldw -52(%r30), %r22 +') +EPILOGUE() +PROLOGUE(func) + ldi 0, %r22 +LDEF(com) + sub %r0, n, %r21 + depw,z %r21, 30, 3, %r28 C r28 = 2 * (-n & 7) + depw,z %r21, 28, 3, %r21 C r21 = 8 * (-n & 7) + sub up, %r21, up C offset up + sub vp, %r21, vp C offset vp + sub rp, %r21, rp C offset rp + blr %r28, %r0 C branch into loop + INITCY + +LDEF(loop) + ldd 0(up), %r20 + ldd 0(vp), %r31 + ADCSBC %r20, %r31, %r20 + std %r20, 0(rp) +LDEF(7) ldd 8(up), %r21 + ldd 8(vp), %r19 + ADCSBC %r21, %r19, %r21 + std %r21, 8(rp) +LDEF(6) ldd 16(up), %r20 + ldd 16(vp), %r31 + ADCSBC %r20, %r31, %r20 + std %r20, 16(rp) +LDEF(5) ldd 24(up), %r21 + ldd 24(vp), %r19 + ADCSBC %r21, %r19, %r21 + std %r21, 24(rp) +LDEF(4) ldd 32(up), %r20 + ldd 32(vp), %r31 + ADCSBC %r20, %r31, %r20 + std %r20, 32(rp) +LDEF(3) ldd 40(up), %r21 + ldd 40(vp), %r19 + ADCSBC %r21, %r19, %r21 + std %r21, 40(rp) +LDEF(2) ldd 48(up), %r20 + ldd 48(vp), %r31 + ADCSBC %r20, %r31, %r20 + std %r20, 48(rp) +LDEF(1) ldd 56(up), %r21 + ldd 56(vp), %r19 + ADCSBC %r21, %r19, %r21 + ldo 64(up), up + std %r21, 56(rp) + ldo 64(vp), vp + addib,> -8, n, L(loop) + ldo 64(rp), rp + + add,dc %r0, %r0, %r29 +ifdef(`OPERATION_sub_n',` + subi 1, %r29, %r29 +') + bve (%r2) +ifdef(`HAVE_ABI_2_0w', +` copy %r29, %r28 +',` ldi 0, %r28 +') +EPILOGUE() -- cgit v1.2.3