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/powerpc32/addmul_1.asm | 159 +++++++++++++++++++++++++++++++++++ 1 file changed, 159 insertions(+) create mode 100644 gmp-6.3.0/mpn/powerpc32/addmul_1.asm (limited to 'gmp-6.3.0/mpn/powerpc32/addmul_1.asm') diff --git a/gmp-6.3.0/mpn/powerpc32/addmul_1.asm b/gmp-6.3.0/mpn/powerpc32/addmul_1.asm new file mode 100644 index 0000000..07486df --- /dev/null +++ b/gmp-6.3.0/mpn/powerpc32/addmul_1.asm @@ -0,0 +1,159 @@ +dnl PowerPC-32 mpn_addmul_1 -- Multiply a limb vector with a limb and add the +dnl result to a second limb vector. + +dnl Copyright 1995, 1997, 1998, 2000-2003, 2005 Free Software Foundation, 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/. + +include(`../config.m4') + +C cycles/limb +C 603e: ? +C 604e: 6.75 +C 75x (G3): 8.7-14.3 +C 7400,7410 (G4): 8.7-14.3 +C 744x,745x (G4+): 9.5 +C power4/ppc970: 6.25 +C power5: 6.25 + +C INPUT PARAMETERS +C rp r3 +C up r4 +C n r5 +C vl r6 + +C This is optimized for the PPC604. It has not been tuned for other +C PowerPC processors. +C +C Loop Analysis for the 604: +C 12 mem insn +C 8 serializing insn +C 8 int multiply +C 25 int reg write +C 9 int ops (8 of which serialize) +C +C The multiply insns need 16 cycles/4limb. +C The integer register writes will need 13 cycles/4limb. +C All-in-all, it should be possible to get to 4 or 5 cycles/limb on PPC604, +C but that will require some clever FPNOPS and BNOPS for exact +C issue control. + + +ASM_START() +PROLOGUE(mpn_addmul_1) + cmpwi cr0,r5,9 C more than 9 limbs? + bgt cr0,L(big) C branch if more than 9 limbs + + mtctr r5 + lwz r0,0(r4) + mullw r7,r0,r6 + mulhwu r10,r0,r6 + lwz r9,0(r3) + addc r8,r7,r9 + addi r3,r3,-4 + bdz L(end) +L(loop): + lwzu r0,4(r4) + stwu r8,4(r3) + mullw r8,r0,r6 + adde r7,r8,r10 + mulhwu r10,r0,r6 + lwz r9,4(r3) + addze r10,r10 + addc r8,r7,r9 + bdnz L(loop) +L(end): stw r8,4(r3) + addze r3,r10 + blr + +L(big): stwu r1,-16(r1) + addi r5,r5,-1 + stw r30,8(r1) + srwi r0,r5,2 + stw r31,12(r1) + mtctr r0 + + lwz r7,0(r4) + mullw r8,r7,r6 + mulhwu r0,r7,r6 + lwz r7,0(r3) + addc r8,r8,r7 + stw r8,0(r3) + +L(loopU): + lwz r7,4(r4) + lwz r12,8(r4) + lwz r30,12(r4) + lwzu r31,16(r4) + mullw r8,r7,r6 + mullw r9,r12,r6 + mullw r10,r30,r6 + mullw r11,r31,r6 + adde r8,r8,r0 C add cy_limb + mulhwu r0,r7,r6 + lwz r7,4(r3) + adde r9,r9,r0 + mulhwu r0,r12,r6 + lwz r12,8(r3) + adde r10,r10,r0 + mulhwu r0,r30,r6 + lwz r30,12(r3) + adde r11,r11,r0 + mulhwu r0,r31,r6 + lwz r31,16(r3) + addze r0,r0 C new cy_limb + addc r8,r8,r7 + stw r8,4(r3) + adde r9,r9,r12 + stw r9,8(r3) + adde r10,r10,r30 + stw r10,12(r3) + adde r11,r11,r31 + stwu r11,16(r3) + bdnz L(loopU) + + andi. r31,r5,3 + mtctr r31 + beq cr0,L(endx) + +L(loopE): + lwzu r7,4(r4) + mullw r8,r7,r6 + adde r8,r8,r0 C add cy_limb + mulhwu r0,r7,r6 + lwz r7,4(r3) + addze r0,r0 C new cy_limb + addc r8,r8,r7 + stwu r8,4(r3) + bdnz L(loopE) +L(endx): + addze r3,r0 + lwz r30,8(r1) + lwz r31,12(r1) + addi r1,r1,16 + blr +EPILOGUE(mpn_addmul_1) -- cgit v1.2.3