1dnl Alpha mpn_addlsh1_n/mpn_sublsh1_n -- rp[] = up[] +- (vp[] << 1). 2 3dnl Copyright 2003, 2013 Free Software Foundation, Inc. 4 5dnl This file is part of the GNU MP Library. 6dnl 7dnl The GNU MP Library is free software; you can redistribute it and/or modify 8dnl it under the terms of either: 9dnl 10dnl * the GNU Lesser General Public License as published by the Free 11dnl Software Foundation; either version 3 of the License, or (at your 12dnl option) any later version. 13dnl 14dnl or 15dnl 16dnl * the GNU General Public License as published by the Free Software 17dnl Foundation; either version 2 of the License, or (at your option) any 18dnl later version. 19dnl 20dnl or both in parallel, as here. 21dnl 22dnl The GNU MP Library is distributed in the hope that it will be useful, but 23dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY 24dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 25dnl for more details. 26dnl 27dnl You should have received copies of the GNU General Public License and the 28dnl GNU Lesser General Public License along with the GNU MP Library. If not, 29dnl see https://www.gnu.org/licenses/. 30 31include(`../config.m4') 32 33C cycles/limb 34C EV4: ? 35C EV5: 7 36C EV6: 4 37 38C TODO 39C * Tune to reach 3.75 c/l on ev6. 40 41define(`rp',`r16') 42define(`up',`r17') 43define(`vp',`r18') 44define(`n', `r19') 45 46define(`u0', `r8') 47define(`u1', `r1') 48define(`v0', `r4') 49define(`v1', `r5') 50 51define(`cy0', `r0') 52define(`cy1', `r20') 53define(`cy', `r22') 54define(`rr', `r24') 55define(`ps', `r25') 56define(`sl', `r28') 57 58ifdef(`OPERATION_addlsh1_n',` 59 define(ADDSUB, addq) 60 define(CARRY, `cmpult $1,$2,$3') 61 define(func, mpn_addlsh1_n) 62') 63ifdef(`OPERATION_sublsh1_n',` 64 define(ADDSUB, subq) 65 define(CARRY, `cmpult $2,$1,$3') 66 define(func, mpn_sublsh1_n) 67') 68 69MULFUNC_PROLOGUE(mpn_addlsh1_n mpn_sublsh1_n) 70 71ASM_START() 72PROLOGUE(func) 73 and n, 2, cy0 74 blbs n, L(bx1) 75L(bx0): ldq v1, 0(vp) 76 ldq u1, 0(up) 77 lda r2, 0(r31) 78 bne cy0, L(b10) 79 80L(b00): lda vp, 48(vp) 81 lda up, -16(up) 82 lda rp, -8(rp) 83 lda cy0, 0(r31) 84 br r31, L(lo0) 85 86L(b10): lda vp, 32(vp) 87 lda rp, 8(rp) 88 lda cy0, 0(r31) 89 br r31, L(lo2) 90 91L(bx1): ldq v0, 0(vp) 92 ldq u0, 0(up) 93 lda r3, 0(r31) 94 beq cy0, L(b01) 95 96L(b11): lda vp, 40(vp) 97 lda up, -24(up) 98 lda rp, 16(rp) 99 lda cy1, 0(r31) 100 br r31, L(lo3) 101 102L(b01): lda n, -4(n) 103 lda cy1, 0(r31) 104 ble n, L(end) 105 lda vp, 24(vp) 106 lda up, -8(up) 107 108 ALIGN(16) 109L(top): addq v0, v0, r6 110 ldq v1, -16(vp) 111 addq r6, r3, sl C combined vlimb 112 ldq u1, 16(up) 113 ADDSUB u0, sl, ps C ulimb + (vlimb << 1) 114 cmplt v0, r31, r2 C high v bits 115 ADDSUB ps, cy1, rr C consume carry from previous operation 116 CARRY( ps, u0, cy0) C carry out #2 117 stq rr, 0(rp) 118 CARRY( rr, ps, cy) C carry out #3 119 lda vp, 32(vp) C bookkeeping 120 addq cy, cy0, cy0 C final carry out 121L(lo0): addq v1, v1, r7 122 ldq v0, -40(vp) 123 addq r7, r2, sl 124 ldq u0, 24(up) 125 ADDSUB u1, sl, ps 126 cmplt v1, r31, r3 127 ADDSUB ps, cy0, rr 128 CARRY( ps, u1, cy1) 129 stq rr, 8(rp) 130 CARRY( rr, ps, cy) 131 lda rp, 32(rp) C bookkeeping 132 addq cy, cy1, cy1 133L(lo3): addq v0, v0, r6 134 ldq v1, -32(vp) 135 addq r6, r3, sl 136 ldq u1, 32(up) 137 ADDSUB u0, sl, ps 138 cmplt v0, r31, r2 139 ADDSUB ps, cy1, rr 140 CARRY( ps, u0, cy0) 141 stq rr, -16(rp) 142 CARRY( rr, ps, cy) 143 lda up, 32(up) C bookkeeping 144 addq cy, cy0, cy0 145L(lo2): addq v1, v1, r7 146 ldq v0, -24(vp) 147 addq r7, r2, sl 148 ldq u0, 8(up) 149 ADDSUB u1, sl, ps 150 cmplt v1, r31, r3 151 ADDSUB ps, cy0, rr 152 CARRY( ps, u1, cy1) 153 stq rr, -8(rp) 154 CARRY( rr, ps, cy) 155 lda n, -4(n) C bookkeeping 156 addq cy, cy1, cy1 157 bgt n, L(top) 158 159L(end): addq v0, v0, r6 160 addq r6, r3, sl 161 ADDSUB u0, sl, ps 162 cmplt v0, r31, r2 163 ADDSUB ps, cy1, rr 164 CARRY( ps, u0, cy0) 165 stq rr, 0(rp) 166 CARRY( rr, ps, cy) 167 addq cy, cy0, cy0 168 addq cy0, r2, r0 169 170 ret r31,(r26),1 171EPILOGUE() 172ASM_END() 173