github.com/aergoio/aergo@v1.3.1/libtool/src/gmp-6.1.2/mpn/x86_64/README (about) 1 Copyright 2003, 2004, 2006, 2008 Free Software Foundation, Inc. 2 3 This file is part of the GNU MP Library. 4 5 The GNU MP Library is free software; you can redistribute it and/or modify 6 it under the terms of either: 7 8 * the GNU Lesser General Public License as published by the Free 9 Software Foundation; either version 3 of the License, or (at your 10 option) any later version. 11 12 or 13 14 * the GNU General Public License as published by the Free Software 15 Foundation; either version 2 of the License, or (at your option) any 16 later version. 17 18 or both in parallel, as here. 19 20 The GNU MP Library is distributed in the hope that it will be useful, but 21 WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY 22 or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 23 for more details. 24 25 You should have received copies of the GNU General Public License and the 26 GNU Lesser General Public License along with the GNU MP Library. If not, 27 see https://www.gnu.org/licenses/. 28 29 30 31 32 33 AMD64 MPN SUBROUTINES 34 35 36 This directory contains mpn functions for AMD64 chips. It is also useful 37 for 64-bit Pentiums, and "Core 2". 38 39 40 RELEVANT OPTIMIZATION ISSUES 41 42 The Opteron and Athlon64 can sustain up to 3 instructions per cycle, but in 43 practice that is only possible for integer instructions. But almost any 44 three integer instructions can issue simultaneously, including any 3 ALU 45 operations, including shifts. Up to two memory operations can issue each 46 cycle. 47 48 Scheduling typically requires that load-use instructions are split into 49 separate load and use instructions. That requires more decode resources, 50 and it is rarely a win. Opteron/Athlon64 have deep out-of-order core. 51 52 53 Optimizing for 64-bit Pentium4 is probably a waste of time, as the most 54 critical instructions are very poorly implemented here. Perhaps we could 55 save a cycle or two, but the most common loops now run at between 10 and 22 56 cycles, so a saved cycle isn't too exciting. 57 58 59 The new spin of the venerable P6 core, the "Core 2" is much better than the 60 Pentium4 for the GMP loops. Its integer pipeline is somewhat similar to to 61 the Opteron/Athlon64 pipeline, except that the GMP favourites ADC/SBB and 62 MUL are slower. Furthermore, an INC/DEC followed by ADC/SBB incur a 63 pipeline stall of around 10 cycles. The default mpn_add_n and mpn_sub_n 64 code suffers badly from the stall. The code in the core2 subdirectory uses 65 the almost forgotten instruction JRCXZ for loop control, and updates the 66 induction variable using LEA. 67 68 69 70 REFERENCES 71 72 "System V Application Binary Interface AMD64 Architecture Processor 73 Supplement", draft version 0.99, December 2007. 74 http://www.x86-64.org/documentation/abi.pdf