head 1.1; branch 1.1.1; access; symbols netbsd-11-0-RC5:1.1.1.2 netbsd-11-0-RC4:1.1.1.2 netbsd-11-0-RC3:1.1.1.2 netbsd-11-0-RC2:1.1.1.2 netbsd-11-0-RC1:1.1.1.2 perseant-exfatfs-base-20250801:1.1.1.2 netbsd-11:1.1.1.2.0.18 netbsd-11-base:1.1.1.2 netbsd-10-1-RELEASE:1.1.1.2 perseant-exfatfs-base-20240630:1.1.1.2 perseant-exfatfs:1.1.1.2.0.16 perseant-exfatfs-base:1.1.1.2 netbsd-8-3-RELEASE:1.1.1.1 netbsd-9-4-RELEASE:1.1.1.2 netbsd-10-0-RELEASE:1.1.1.2 netbsd-10-0-RC6:1.1.1.2 netbsd-10-0-RC5:1.1.1.2 netbsd-10-0-RC4:1.1.1.2 netbsd-10-0-RC3:1.1.1.2 netbsd-10-0-RC2:1.1.1.2 netbsd-10-0-RC1:1.1.1.2 netbsd-10:1.1.1.2.0.14 netbsd-10-base:1.1.1.2 netbsd-9-3-RELEASE:1.1.1.2 gmp-6-2-1:1.1.1.2 cjep_sun2x-base1:1.1.1.2 cjep_sun2x:1.1.1.2.0.12 cjep_sun2x-base:1.1.1.2 cjep_staticlib_x-base1:1.1.1.2 netbsd-9-2-RELEASE:1.1.1.2 cjep_staticlib_x:1.1.1.2.0.10 cjep_staticlib_x-base:1.1.1.2 netbsd-9-1-RELEASE:1.1.1.2 gmp-6-2-0:1.1.1.2 phil-wifi-20200421:1.1.1.2 phil-wifi-20200411:1.1.1.2 is-mlppp:1.1.1.2.0.8 is-mlppp-base:1.1.1.2 phil-wifi-20200406:1.1.1.2 netbsd-8-2-RELEASE:1.1.1.1 netbsd-9-0-RELEASE:1.1.1.2 netbsd-9-0-RC2:1.1.1.2 netbsd-9-0-RC1:1.1.1.2 phil-wifi-20191119:1.1.1.2 netbsd-9:1.1.1.2.0.6 netbsd-9-base:1.1.1.2 phil-wifi-20190609:1.1.1.2 netbsd-8-1-RELEASE:1.1.1.1 netbsd-8-1-RC1:1.1.1.1 pgoyette-compat-merge-20190127:1.1.1.2 pgoyette-compat-20190127:1.1.1.2 pgoyette-compat-20190118:1.1.1.2 pgoyette-compat-1226:1.1.1.2 pgoyette-compat-1126:1.1.1.2 pgoyette-compat-1020:1.1.1.2 pgoyette-compat-0930:1.1.1.2 pgoyette-compat-0906:1.1.1.2 netbsd-7-2-RELEASE:1.1.1.1 pgoyette-compat-0728:1.1.1.2 netbsd-8-0-RELEASE:1.1.1.1 phil-wifi:1.1.1.2.0.4 phil-wifi-base:1.1.1.2 pgoyette-compat-0625:1.1.1.2 netbsd-8-0-RC2:1.1.1.1 pgoyette-compat-0521:1.1.1.2 pgoyette-compat-0502:1.1.1.2 pgoyette-compat-0422:1.1.1.2 netbsd-8-0-RC1:1.1.1.1 pgoyette-compat-0415:1.1.1.2 pgoyette-compat-0407:1.1.1.2 pgoyette-compat-0330:1.1.1.2 pgoyette-compat-0322:1.1.1.2 pgoyette-compat-0315:1.1.1.2 netbsd-7-1-2-RELEASE:1.1.1.1 pgoyette-compat:1.1.1.2.0.2 pgoyette-compat-base:1.1.1.2 netbsd-7-1-1-RELEASE:1.1.1.1 matt-nb8-mediatek:1.1.1.1.0.38 matt-nb8-mediatek-base:1.1.1.1 gmp-6-1-2:1.1.1.2 perseant-stdc-iso10646:1.1.1.1.0.36 perseant-stdc-iso10646-base:1.1.1.1 netbsd-8:1.1.1.1.0.34 netbsd-8-base:1.1.1.1 prg-localcount2-base3:1.1.1.1 prg-localcount2-base2:1.1.1.1 prg-localcount2-base1:1.1.1.1 prg-localcount2:1.1.1.1.0.32 prg-localcount2-base:1.1.1.1 pgoyette-localcount-20170426:1.1.1.1 bouyer-socketcan-base1:1.1.1.1 pgoyette-localcount-20170320:1.1.1.1 netbsd-7-1:1.1.1.1.0.30 netbsd-7-1-RELEASE:1.1.1.1 netbsd-7-1-RC2:1.1.1.1 netbsd-7-nhusb-base-20170116:1.1.1.1 bouyer-socketcan:1.1.1.1.0.28 bouyer-socketcan-base:1.1.1.1 pgoyette-localcount-20170107:1.1.1.1 netbsd-7-1-RC1:1.1.1.1 pgoyette-localcount-20161104:1.1.1.1 netbsd-7-0-2-RELEASE:1.1.1.1 localcount-20160914:1.1.1.1 netbsd-7-nhusb:1.1.1.1.0.26 netbsd-7-nhusb-base:1.1.1.1 pgoyette-localcount-20160806:1.1.1.1 pgoyette-localcount-20160726:1.1.1.1 pgoyette-localcount:1.1.1.1.0.24 pgoyette-localcount-base:1.1.1.1 netbsd-7-0-1-RELEASE:1.1.1.1 netbsd-7-0:1.1.1.1.0.22 netbsd-7-0-RELEASE:1.1.1.1 netbsd-7-0-RC3:1.1.1.1 netbsd-7-0-RC2:1.1.1.1 netbsd-7-0-RC1:1.1.1.1 netbsd-6-0-6-RELEASE:1.1.1.1 netbsd-6-1-5-RELEASE:1.1.1.1 netbsd-7:1.1.1.1.0.20 netbsd-7-base:1.1.1.1 yamt-pagecache-base9:1.1.1.1 yamt-pagecache-tag8:1.1.1.1 netbsd-6-1-4-RELEASE:1.1.1.1 netbsd-6-0-5-RELEASE:1.1.1.1 tls-earlyentropy:1.1.1.1.0.18 tls-earlyentropy-base:1.1.1.1 riastradh-xf86-video-intel-2-7-1-pre-2-21-15:1.1.1.1 riastradh-drm2-base3:1.1.1.1 netbsd-6-1-3-RELEASE:1.1.1.1 netbsd-6-0-4-RELEASE:1.1.1.1 gmp-5-1-3:1.1.1.1 netbsd-6-1-2-RELEASE:1.1.1.1 netbsd-6-0-3-RELEASE:1.1.1.1 netbsd-6-1-1-RELEASE:1.1.1.1 riastradh-drm2-base2:1.1.1.1 riastradh-drm2-base1:1.1.1.1 riastradh-drm2:1.1.1.1.0.12 riastradh-drm2-base:1.1.1.1 netbsd-6-1:1.1.1.1.0.16 netbsd-6-0-2-RELEASE:1.1.1.1 netbsd-6-1-RELEASE:1.1.1.1 netbsd-6-1-RC4:1.1.1.1 netbsd-6-1-RC3:1.1.1.1 agc-symver:1.1.1.1.0.14 agc-symver-base:1.1.1.1 netbsd-6-1-RC2:1.1.1.1 netbsd-6-1-RC1:1.1.1.1 yamt-pagecache-base8:1.1.1.1 netbsd-6-0-1-RELEASE:1.1.1.1 yamt-pagecache-base7:1.1.1.1 matt-nb6-plus-nbase:1.1.1.1 yamt-pagecache-base6:1.1.1.1 netbsd-6-0:1.1.1.1.0.10 netbsd-6-0-RELEASE:1.1.1.1 netbsd-6-0-RC2:1.1.1.1 tls-maxphys:1.1.1.1.0.8 tls-maxphys-base:1.1.1.1 matt-nb6-plus:1.1.1.1.0.6 matt-nb6-plus-base:1.1.1.1 netbsd-6-0-RC1:1.1.1.1 yamt-pagecache-base5:1.1.1.1 yamt-pagecache-base4:1.1.1.1 netbsd-6:1.1.1.1.0.4 netbsd-6-base:1.1.1.1 yamt-pagecache-base3:1.1.1.1 yamt-pagecache-base2:1.1.1.1 yamt-pagecache:1.1.1.1.0.2 yamt-pagecache-base:1.1.1.1 gmp-5-0-2:1.1.1.1 gmp:1.1.1; locks; strict; comment @# @; 1.1 date 2011.06.20.05.54.43; author mrg; state Exp; branches 1.1.1.1; next ; 1.1.1.1 date 2011.06.20.05.54.43; author mrg; state Exp; branches; next 1.1.1.2; 1.1.1.2 date 2017.08.22.09.40.49; author mrg; state Exp; branches; next ; commitid W5kmAIk8hwVpSb4A; desc @@ 1.1 log @Initial revision @ text @Copyright 2000, 2001 Free Software Foundation, Inc. This file is part of the GNU MP Library. The GNU MP Library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 3 of the License, or (at your option) any later version. The GNU MP Library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details. You should have received a copy of the GNU Lesser General Public License along with the GNU MP Library. If not, see http://www.gnu.org/licenses/. AMD K7 MPN SUBROUTINES This directory contains code optimized for the AMD Athlon CPU. The mmx subdirectory has routines using MMX instructions. All Athlons have MMX, the separate directory is just so that configure can omit it if the assembler doesn't support MMX. STATUS Times for the loops, with all code and data in L1 cache. cycles/limb mpn_add/sub_n 1.6 mpn_copyi 0.75 or 1.0 \ varying with data alignment mpn_copyd 0.75 or 1.0 / mpn_divrem_1 17.0 integer part, 15.0 fractional part mpn_mod_1 17.0 mpn_divexact_by3 8.0 mpn_l/rshift 1.2 mpn_mul_1 3.4 mpn_addmul/submul_1 3.9 mpn_mul_basecase 4.42 cycles/crossproduct (approx) mpn_sqr_basecase 2.3 cycles/crossproduct (approx) or 4.55 cycles/triangleproduct (approx) Prefetching of sources hasn't yet been tried. NOTES cmov, MMX, 3DNow and some extensions to MMX and 3DNow are available. Write-allocate L1 data cache means prefetching of destinations is unnecessary. Floating point multiplications can be done in parallel with integer multiplications, but there doesn't seem to be any way to make use of this. Unsigned "mul"s can be issued every 3 cycles. This suggests 3 is a limit on the speed of the multiplication routines. The documentation shows mul executing in IEU0 (or maybe in IEU0 and IEU1 together), so it might be that, to get near 3 cycles code has to be arranged so that nothing else is issued to IEU0. A busy IEU0 could explain why some code takes 4 cycles and other apparently equivalent code takes 5. OPTIMIZATIONS Unrolled loops are used to reduce looping overhead. The unrolling is configurable up to 32 limbs/loop for most routines and up to 64 for some. The K7 has 64k L1 code cache so quite big unrolling is allowable. Computed jumps into the unrolling are used to handle sizes not a multiple of the unrolling. An attractive feature of this is that times increase smoothly with operand size, but it may be that some routines should just have simple loops to finish up, especially when PIC adds between 2 and 16 cycles to get %eip. Position independent code is implemented using a call to get %eip for the computed jumps and a ret is always done, rather than an addl $4,%esp or a popl, so the CPU return address branch prediction stack stays synchronised with the actual stack in memory. Branch prediction, in absence of any history, will guess forward jumps are not taken and backward jumps are taken. Where possible it's arranged that the less likely or less important case is under a taken forward jump. CODING Instructions in general code have been shown grouped if they can execute together, which means up to three direct-path instructions which have no successive dependencies. K7 always decodes three and has out-of-order execution, but the groupings show what slots might be available and what dependency chains exist. When there's vector-path instructions an effort is made to get triplets of direct-path instructions in between them, even if there's dependencies, since this maximizes decoding throughput and might save a cycle or two if decoding is the limiting factor. INSTRUCTIONS adcl direct divl 39 cycles back-to-back lodsl,etc vector loop 1 cycle vector (decl/jnz opens up one decode slot) movd reg vector movd mem direct mull issue every 3 cycles, latency 4 cycles low word, 6 cycles high word popl vector (use movl for more than one pop) pushl direct, will pair with a load shrdl %cl vector, 3 cycles, seems to be 3 decode too xorl r,r false read dependency recognised REFERENCES "AMD Athlon Processor X86 Code Optimization Guide", AMD publication number 22007, revision K, February 2002. Available on-line, http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/22007.pdf "3DNow Technology Manual", AMD publication number 21928G/0-March 2000. This describes the femms and prefetch instructions. Available on-line, http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/21928.pdf "AMD Extensions to the 3DNow and MMX Instruction Sets Manual", AMD publication number 22466, revision D, March 2000. This describes instructions added in the Athlon processor, such as pswapd and the extra prefetch forms. Available on-line, http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/22466.pdf "3DNow Instruction Porting Guide", AMD publication number 22621, revision B, August 1999. This has some notes on general Athlon optimizations as well as 3DNow. Available on-line, http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/22621.pdf ---------------- Local variables: mode: text fill-column: 76 End: @ 1.1.1.1 log @initial import of GMP 5.0.2. GNU MP is a library for arbitrary precision arithmetic, operating on signed integers, rational numbers, and floating point numbers. It has a rich set of functions, and the functions have a regular interface. GMP is necessary for GCC >= 4.2. @ text @@ 1.1.1.2 log @initial import of GMP 6.1.2. main changes from 5.1.3 below. notes: - support for thumb-less ARM chips was in our port of 5.1.3, but a similar method has been provided upstream now - someone should look at the AVX failure reports, and fix them Changes between GMP version 6.1.0 and 6.1.1 FEATURES * Work around faulty cpuid on some recent Intel chips (this allows GMP to run on Skylake Pentiums). * Support thumb-less ARM chips. Changes between GMP version 6.0.* and 6.1.0 BUGS FIXED * The public function mpn_com is now correctly declared in gmp.h. * Healed possible failures of mpn_sec_sqr for non-cryptographic sizes for some obsolete CPUs. * Various problems related to precision for mpf have been fixed. * Fixed ABI incompatible stack alignment in calls from assembly code. * Fixed PIC bug in popcount affecting Intel processors using the 32-bit ABI. SPEEDUPS * Speedup for Intel Broadwell and Skylake through assembly code making use of new ADX instructions. * Square root is now faster when the remainder is not needed. Also the speed to compute the k-th root improved, for small sizes. FEATURES * New C++ functions gcd and lcm for mpz_class. * New public mpn functions mpn_divexact_1, mpn_zero_p, and mpn_cnd_swap. * New public mpq_cmp_z function, to efficiently compare rationals with integers. * Support for more 32-bit arm processors. * Support for AVX-less modern x86 CPUs. (Such support might be missing either because the CPU vendor chose to disable AVX, or because the running kernel lacks AVX context switch support.) * Support for NetBSD under Xen; we switch off AVX unconditionally under NetBSD since a bug in NetBSD makes AVX fail under Xen. MISC * Tuned values for FFT multiplications are provided for larger number on many platforms. Changes between GMP version 5.1.* and 6.0.0 BUGS FIXED * The function mpz_invert now considers any number invertible in Z/1Z. * The mpn multiply code now handles operands of more than 2^31 limbs correctly. (Note however that the mpz code is limited to 2^32 bits on 32-bit hosts and 2^37 bits on 64-bit hosts.) SPEEDUPS * Plain division of large operands is faster and more monotonous in operand size. * Major speedup for ARM, in particular ARM Cortex-A15, thanks to improved assembly. * Speedup for Intel Sandy Bridge, Ivy Bridge, Haswell, thanks to rewritten and vastly expanded assembly support. Speedup also for the older Core 2 and Nehalem. * Faster mixed arithmetic between mpq_class and double. FEATURES * Support for new Intel and AMD CPUs. * New public functions mpn_sec_mul and mpn_sec_sqr, implementing side-channel silent multiplication and squaring. * New public functions mpn_sec_div_qr and mpn_sec_div_r, implementing side-channel silent division. * New public functions mpn_cnd_add_n and mpn_cnd_sub_n. Side-channel silent conditional addition and subtraction. * New public function mpn_sec_powm, implementing side-channel silent modexp. * New public function mpn_sec_invert, implementing side-channel silent modular inversion. * Better support for applications which use the mpz_t type, but nevertheless need to call some of the lower-level mpn functions. See the documentation for mpz_limbs_read and related functions. @ text @d6 3 a8 13 it under the terms of either: * the GNU Lesser General Public License as published by the Free Software Foundation; either version 3 of the License, or (at your option) any later version. or * the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. or both in parallel, as here. d12 2 a13 2 or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. d15 2 a16 3 You should have received copies of the GNU General Public License and the GNU Lesser General Public License along with the GNU MP Library. If not, see https://www.gnu.org/licenses/. @