1dnl ARM mpn_mul_1 -- Multiply a limb vector with a limb and store the result 2dnl in a second limb vector. 3dnl Contributed by Robert Harley. 4 5dnl Copyright 1998, 2000, 2001, 2003, 2012 Free Software Foundation, Inc. 6 7dnl This file is part of the GNU MP Library. 8 9dnl The GNU MP Library is free software; you can redistribute it and/or modify 10dnl it under the terms of the GNU Lesser General Public License as published 11dnl by the Free Software Foundation; either version 3 of the License, or (at 12dnl your option) any later version. 13 14dnl The GNU MP Library is distributed in the hope that it will be useful, but 15dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY 16dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public 17dnl License for more details. 18 19dnl You should have received a copy of the GNU Lesser General Public License 20dnl along with the GNU MP Library. If not, see http://www.gnu.org/licenses/. 21 22include(`../config.m4') 23 24C cycles/limb 25C StrongARM 6-8 26C XScale ? 27C Cortex-A8 ? 28C Cortex-A9 4.75 29C Cortex-A15 ? 30 31C We should rewrite this along the lines of addmul_1.asm. That should save a 32C cycle on StrongARM, and several cycles on XScale. 33 34define(`rp',`r0') 35define(`up',`r1') 36define(`n',`r2') 37define(`vl',`r3') 38 39 40ASM_START() 41PROLOGUE(mpn_mul_1) 42 stmfd sp!, { r8, r9, lr } 43 ands r12, n, #1 44 beq L(skip1) 45 ldr lr, [up], #4 46 umull r9, r12, lr, vl 47 str r9, [rp], #4 48L(skip1): 49 tst n, #2 50 beq L(skip2) 51 mov r8, r12 52 ldmia up!, { r12, lr } 53 mov r9, #0 54 umlal r8, r9, r12, vl 55 mov r12, #0 56 umlal r9, r12, lr, vl 57 stmia rp!, { r8, r9 } 58L(skip2): 59 bics n, n, #3 60 beq L(rtn) 61 stmfd sp!, { r6, r7 } 62 63L(top): mov r6, r12 64 ldmia up!, { r8, r9, r12, lr } 65 ldr r7, [rp, #12] C cache allocate 66 mov r7, #0 67 umlal r6, r7, r8, vl 68 mov r8, #0 69 umlal r7, r8, r9, vl 70 mov r9, #0 71 umlal r8, r9, r12, vl 72 mov r12, #0 73 umlal r9, r12, lr, vl 74 subs n, n, #4 75 stmia rp!, { r6, r7, r8, r9 } 76 bne L(top) 77 78 ldmfd sp!, { r6, r7 } 79 80L(rtn): mov r0, r12 81 ldmfd sp!, { r8, r9, pc } 82EPILOGUE() 83