From a89a14ef5da44684a16b204e7a70460cc8c4922a Mon Sep 17 00:00:00 2001 From: Thomas Voss Date: Fri, 21 Jun 2024 23:36:36 +0200 Subject: Basic constant folding implementation --- vendor/gmp-6.3.0/mpn/x86_64/k10/gcd_11.asm | 37 ++++ vendor/gmp-6.3.0/mpn/x86_64/k10/gcd_22.asm | 142 +++++++++++++ vendor/gmp-6.3.0/mpn/x86_64/k10/gmp-mparam.h | 248 ++++++++++++++++++++++ vendor/gmp-6.3.0/mpn/x86_64/k10/hamdist.asm | 109 ++++++++++ vendor/gmp-6.3.0/mpn/x86_64/k10/lshift.asm | 37 ++++ vendor/gmp-6.3.0/mpn/x86_64/k10/lshiftc.asm | 37 ++++ vendor/gmp-6.3.0/mpn/x86_64/k10/popcount.asm | 138 ++++++++++++ vendor/gmp-6.3.0/mpn/x86_64/k10/rshift.asm | 37 ++++ vendor/gmp-6.3.0/mpn/x86_64/k10/sec_tabselect.asm | 37 ++++ 9 files changed, 822 insertions(+) create mode 100644 vendor/gmp-6.3.0/mpn/x86_64/k10/gcd_11.asm create mode 100644 vendor/gmp-6.3.0/mpn/x86_64/k10/gcd_22.asm create mode 100644 vendor/gmp-6.3.0/mpn/x86_64/k10/gmp-mparam.h create mode 100644 vendor/gmp-6.3.0/mpn/x86_64/k10/hamdist.asm create mode 100644 vendor/gmp-6.3.0/mpn/x86_64/k10/lshift.asm create mode 100644 vendor/gmp-6.3.0/mpn/x86_64/k10/lshiftc.asm create mode 100644 vendor/gmp-6.3.0/mpn/x86_64/k10/popcount.asm create mode 100644 vendor/gmp-6.3.0/mpn/x86_64/k10/rshift.asm create mode 100644 vendor/gmp-6.3.0/mpn/x86_64/k10/sec_tabselect.asm (limited to 'vendor/gmp-6.3.0/mpn/x86_64/k10') diff --git a/vendor/gmp-6.3.0/mpn/x86_64/k10/gcd_11.asm b/vendor/gmp-6.3.0/mpn/x86_64/k10/gcd_11.asm new file mode 100644 index 0000000..4723093 --- /dev/null +++ b/vendor/gmp-6.3.0/mpn/x86_64/k10/gcd_11.asm @@ -0,0 +1,37 @@ +dnl AMD64 mpn_gcd_11. + +dnl Copyright 2012 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') + +ABI_SUPPORT(DOS64) +ABI_SUPPORT(STD64) + +MULFUNC_PROLOGUE(mpn_gcd_11) +include_mpn(`x86_64/core2/gcd_11.asm') diff --git a/vendor/gmp-6.3.0/mpn/x86_64/k10/gcd_22.asm b/vendor/gmp-6.3.0/mpn/x86_64/k10/gcd_22.asm new file mode 100644 index 0000000..f58b4cc --- /dev/null +++ b/vendor/gmp-6.3.0/mpn/x86_64/k10/gcd_22.asm @@ -0,0 +1,142 @@ +dnl AMD64 mpn_gcd_22. Assumes useful bsf, useless shrd, no tzcnt, no shlx. + +dnl Copyright 2019 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/bit +C AMD K8,K9 ? +C AMD K10 7.4 +C AMD bd1 9.9 +C AMD bd2 ? +C AMD bd3 ? +C AMD bd4 ? +C AMD bt1 ? +C AMD bt2 ? +C AMD zn1 ? +C AMD zn2 ? +C Intel P4 ? +C Intel CNR ? +C Intel PNR ? +C Intel NHM 9.2 +C Intel WSM 9.0 +C Intel SBR ? +C Intel IBR ? +C Intel HWL ? +C Intel BWL ? +C Intel SKL ? +C Intel atom ? +C Intel SLM ? +C Intel GLM ? +C Intel GLM+ ? +C VIA nano ? + + +define(`u1', `%rdi') +define(`u0', `%rsi') +define(`v1', `%rdx') +define(`v0_param', `%rcx') + +define(`v0', `%rax') +define(`cnt', `%rcx') + +define(`s0', `%r8') +define(`s1', `%r9') +define(`t0', `%r10') +define(`t1', `%r11') + +dnl ABI_SUPPORT(DOS64) C returns mp_double_limb_t in memory +ABI_SUPPORT(STD64) + +ASM_START() + TEXT + ALIGN(64) +PROLOGUE(mpn_gcd_22) + FUNC_ENTRY(4) + mov v0_param, v0 + + ALIGN(16) +L(top): mov v0, t0 + sub u0, t0 + jz L(lowz) C jump when low limb result = 0 + mov v1, t1 + sbb u1, t1 + + mov u0, s0 + mov u1, s1 + + bsf t0, cnt + + sub v0, u0 + sbb v1, u1 + +L(bck): cmovc t0, u0 C u = |u - v| + cmovnc u1, t1 C u = |u - v| + cmovc s0, v0 C v = min(u,v) + cmovc s1, v1 C v = min(u,v) + + shr R8(cnt), u0 + mov t1, u1 + shr R8(cnt), u1 + neg cnt + shl R8(cnt), t1 + or t1, u0 + + test u1, u1 + jnz L(top) + test v1, v1 + jnz L(top) + +L(gcd_11): + mov v0, %rdi +C mov u0, %rsi + TCALL( mpn_gcd_11) + +L(lowz):C We come here when v0 - u0 = 0 + C 1. If v1 - u1 = 0, then gcd is u = v. + C 2. Else compute gcd_21({v1,v0}, |u1-v1|) + mov v1, t0 + sub u1, t0 + je L(end) + + xor t1, t1 + mov u0, s0 + mov u1, s1 + bsf t0, cnt + mov u1, u0 + xor u1, u1 + sub v1, u0 + jmp L(bck) + +L(end): C mov v0, %rax + C mov v1, %rdx + FUNC_EXIT() + ret +EPILOGUE() diff --git a/vendor/gmp-6.3.0/mpn/x86_64/k10/gmp-mparam.h b/vendor/gmp-6.3.0/mpn/x86_64/k10/gmp-mparam.h new file mode 100644 index 0000000..349bace --- /dev/null +++ b/vendor/gmp-6.3.0/mpn/x86_64/k10/gmp-mparam.h @@ -0,0 +1,248 @@ +/* AMD K10 gmp-mparam.h -- Compiler/machine parameter header file. + +Copyright 2019 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 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. + +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 General Public License +for more details. + +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/. */ + +#define GMP_LIMB_BITS 64 +#define GMP_LIMB_BYTES 8 + +#if 0 +#undef mpn_sublsh_n +#define mpn_sublsh_n(rp,up,vp,n,c) \ + (((rp) == (up)) ? mpn_submul_1 (rp, vp, n, CNST_LIMB(1) << (c)) \ + : MPN(mpn_sublsh_n)(rp,up,vp,n,c)) +#endif + +/* 3200-3600 MHz K10 Thuban */ +/* FFT tuning limit = 427,161,280 */ +/* Generated by tuneup.c, 2019-10-22, gcc 8.3 */ + +#define MOD_1_NORM_THRESHOLD 0 /* always */ +#define MOD_1_UNNORM_THRESHOLD 0 /* always */ +#define MOD_1N_TO_MOD_1_1_THRESHOLD 5 +#define MOD_1U_TO_MOD_1_1_THRESHOLD 3 +#define MOD_1_1_TO_MOD_1_2_THRESHOLD 17 +#define MOD_1_2_TO_MOD_1_4_THRESHOLD 28 +#define PREINV_MOD_1_TO_MOD_1_THRESHOLD 11 +#define USE_PREINV_DIVREM_1 1 /* native */ +#define DIV_QR_1_NORM_THRESHOLD 1 +#define DIV_QR_1_UNNORM_THRESHOLD MP_SIZE_T_MAX /* never */ +#define DIV_QR_2_PI2_THRESHOLD MP_SIZE_T_MAX /* never */ +#define DIVEXACT_1_THRESHOLD 0 /* always (native) */ +#define BMOD_1_TO_MOD_1_THRESHOLD 15 + +#define DIV_1_VS_MUL_1_PERCENT 324 + +#define MUL_TOOM22_THRESHOLD 27 +#define MUL_TOOM33_THRESHOLD 81 +#define MUL_TOOM44_THRESHOLD 232 +#define MUL_TOOM6H_THRESHOLD 363 +#define MUL_TOOM8H_THRESHOLD 478 + +#define MUL_TOOM32_TO_TOOM43_THRESHOLD 97 +#define MUL_TOOM32_TO_TOOM53_THRESHOLD 155 +#define MUL_TOOM42_TO_TOOM53_THRESHOLD 145 +#define MUL_TOOM42_TO_TOOM63_THRESHOLD 160 +#define MUL_TOOM43_TO_TOOM54_THRESHOLD 142 + +#define SQR_BASECASE_THRESHOLD 0 /* always (native) */ +#define SQR_TOOM2_THRESHOLD 30 +#define SQR_TOOM3_THRESHOLD 117 +#define SQR_TOOM4_THRESHOLD 280 +#define SQR_TOOM6_THRESHOLD 446 +#define SQR_TOOM8_THRESHOLD 547 + +#define MULMID_TOOM42_THRESHOLD 34 + +#define MULMOD_BNM1_THRESHOLD 15 +#define SQRMOD_BNM1_THRESHOLD 17 + +#define MUL_FFT_MODF_THRESHOLD 530 /* k = 5 */ +#define MUL_FFT_TABLE3 \ + { { 530, 5}, { 24, 6}, { 13, 5}, { 27, 6}, \ + { 27, 7}, { 14, 6}, { 29, 7}, { 15, 6}, \ + { 31, 7}, { 29, 8}, { 15, 7}, { 32, 8}, \ + { 17, 7}, { 36, 8}, { 19, 7}, { 39, 8}, \ + { 21, 7}, { 43, 8}, { 23, 7}, { 47, 8}, \ + { 25, 7}, { 51, 8}, { 29, 9}, { 15, 8}, \ + { 35, 9}, { 19, 8}, { 43, 9}, { 23, 8}, \ + { 51, 9}, { 27, 8}, { 55,10}, { 15, 9}, \ + { 31, 8}, { 63, 9}, { 35, 8}, { 71, 9}, \ + { 39, 8}, { 81, 9}, { 43,10}, { 23, 9}, \ + { 55,11}, { 15,10}, { 31, 9}, { 71,10}, \ + { 39, 9}, { 87,10}, { 47, 9}, { 99,10}, \ + { 55,11}, { 31,10}, { 87,11}, { 47,10}, \ + { 111,12}, { 31,11}, { 63,10}, { 143,11}, \ + { 79,10}, { 167,11}, { 95,10}, { 191,11}, \ + { 111,12}, { 63,11}, { 143,10}, { 287,11}, \ + { 159,12}, { 95,11}, { 207,13}, { 63,12}, \ + { 127,11}, { 255,10}, { 511,11}, { 271,10}, \ + { 543,11}, { 287,12}, { 159,11}, { 319,10}, \ + { 639,11}, { 335,10}, { 671,11}, { 351,10}, \ + { 703,12}, { 191,11}, { 383,10}, { 767,11}, \ + { 415,12}, { 223,11}, { 447,13}, { 127,12}, \ + { 255,11}, { 543,12}, { 287,11}, { 575,10}, \ + { 1151,11}, { 607,12}, { 319,11}, { 671,12}, \ + { 351,11}, { 703,13}, { 191,12}, { 383,11}, \ + { 767,12}, { 415,11}, { 831,12}, { 447,14}, \ + { 127,13}, { 255,12}, { 543,11}, { 1087,12}, \ + { 575,11}, { 1151,12}, { 607,13}, { 319,12}, \ + { 703,11}, { 1407,12}, { 735,13}, { 383,12}, \ + { 831,13}, { 447,12}, { 959,14}, { 255,13}, \ + { 511,12}, { 1087,13}, { 575,12}, { 1215,13}, \ + { 639,12}, { 1343,13}, { 703,12}, { 1471,14}, \ + { 383,13}, { 767,12}, { 1535,13}, { 831,12}, \ + { 1663,13}, { 959,14}, { 511,13}, { 1087,12}, \ + { 2175,13}, { 1215,14}, { 639,13}, { 1471,14}, \ + { 767,13}, { 1663,14}, { 895,13}, { 1855,15}, \ + { 511,14}, { 1023,13}, { 2175,14}, { 1151,13}, \ + { 2431,14}, { 1279,13}, { 2559,14}, { 1407,15}, \ + { 767,14}, { 1791,16}, { 511,15}, { 1023,14}, \ + { 2431,15}, { 1279,14}, { 2943,15}, { 1535,14}, \ + { 3199,15}, { 1791,14}, { 3583,16}, { 1023,15}, \ + { 2047,14}, { 4223,15}, { 2303,14}, { 4863,15}, \ + { 2559,14}, { 5247,15}, { 2815,16}, { 1535,15}, \ + { 3071,14}, { 6271,15}, { 3327,14}, { 6911,15}, \ + { 3583,17}, { 1023,16}, { 2047,15}, { 4351,14}, \ + { 8959,15}, { 4863,16}, { 2559,15}, { 5887,14}, \ + { 11775,16}, { 3071,15}, { 6911,16}, { 3583,15}, \ + { 7167,17}, { 2047,16}, { 4095,15}, { 8959,16}, \ + { 4607,15}, { 9983,16}, { 5631,15}, { 11775,17}, \ + { 3071,16}, { 6143,15}, { 12543,16}, { 65536,17}, \ + { 131072,18}, { 262144,19}, { 524288,20}, {1048576,21}, \ + {2097152,22}, {4194304,23}, {8388608,24} } +#define MUL_FFT_TABLE3_SIZE 207 +#define MUL_FFT_THRESHOLD 7552 + +#define SQR_FFT_MODF_THRESHOLD 476 /* k = 5 */ +#define SQR_FFT_TABLE3 \ + { { 476, 5}, { 21, 6}, { 11, 5}, { 23, 6}, \ + { 12, 5}, { 25, 6}, { 29, 7}, { 28, 8}, \ + { 15, 7}, { 32, 8}, { 17, 7}, { 35, 8}, \ + { 19, 7}, { 39, 8}, { 21, 7}, { 43, 8}, \ + { 23, 7}, { 47, 8}, { 29, 9}, { 15, 8}, \ + { 35, 9}, { 19, 8}, { 43, 9}, { 23, 8}, \ + { 49, 9}, { 27, 8}, { 55,10}, { 15, 9}, \ + { 31, 8}, { 63, 9}, { 43,10}, { 23, 9}, \ + { 55,11}, { 15,10}, { 31, 9}, { 67,10}, \ + { 39, 9}, { 83,10}, { 47, 9}, { 95,10}, \ + { 55,11}, { 31,10}, { 79,11}, { 47,10}, \ + { 103,12}, { 31,11}, { 63,10}, { 135,11}, \ + { 79,10}, { 167,11}, { 111,12}, { 63,11}, \ + { 127,10}, { 255,11}, { 143,10}, { 287, 9}, \ + { 575,11}, { 159,10}, { 319,12}, { 95,11}, \ + { 191,10}, { 383, 9}, { 767,10}, { 399,13}, \ + { 63,12}, { 127,11}, { 255,10}, { 511,11}, \ + { 271,10}, { 543,11}, { 287,10}, { 575,12}, \ + { 159,11}, { 319,10}, { 639,11}, { 335,10}, \ + { 671,11}, { 351,10}, { 703,11}, { 367,12}, \ + { 191,11}, { 383,10}, { 767,11}, { 415,10}, \ + { 831,12}, { 223,11}, { 447,13}, { 127,12}, \ + { 255,11}, { 511,10}, { 1023,11}, { 543,12}, \ + { 287,11}, { 575,10}, { 1151,11}, { 607,12}, \ + { 319,11}, { 639,10}, { 1279,11}, { 671,12}, \ + { 351,11}, { 703,10}, { 1407,13}, { 191,12}, \ + { 383,11}, { 799,12}, { 415,11}, { 831,12}, \ + { 447,14}, { 127,13}, { 255,12}, { 511,11}, \ + { 1023,12}, { 543,11}, { 1087,12}, { 575,11}, \ + { 1151,12}, { 607,13}, { 319,12}, { 639,11}, \ + { 1279,12}, { 671,11}, { 1343,12}, { 703,11}, \ + { 1407,12}, { 735,13}, { 383,12}, { 767,11}, \ + { 1535,12}, { 831,11}, { 1663,13}, { 447,12}, \ + { 959,14}, { 255,13}, { 511,12}, { 1087,13}, \ + { 575,12}, { 1215,13}, { 639,12}, { 1343,13}, \ + { 703,12}, { 1407,14}, { 383,13}, { 767,12}, \ + { 1535,13}, { 831,12}, { 1727,13}, { 895,12}, \ + { 1791,13}, { 959,15}, { 255,14}, { 511,13}, \ + { 1087,12}, { 2175,13}, { 1215,14}, { 639,13}, \ + { 1471,14}, { 767,13}, { 1727,14}, { 895,13}, \ + { 1791,15}, { 511,14}, { 1023,13}, { 2175,14}, \ + { 1151,13}, { 2303,14}, { 1279,13}, { 2559,14}, \ + { 1407,15}, { 767,14}, { 1791,16}, { 511,15}, \ + { 1023,14}, { 2303,15}, { 1279,14}, { 2815,15}, \ + { 1535,14}, { 3199,15}, { 1791,16}, { 1023,15}, \ + { 2047,14}, { 4223,15}, { 2303,14}, { 4863,15}, \ + { 2559,14}, { 5247,15}, { 2815,16}, { 1535,15}, \ + { 3071,14}, { 6271,15}, { 3327,14}, { 6911,17}, \ + { 1023,16}, { 2047,15}, { 4351,14}, { 8959,15}, \ + { 4863,16}, { 2559,15}, { 5887,14}, { 11775,16}, \ + { 3071,15}, { 6911,16}, { 3583,15}, { 7679,17}, \ + { 2047,16}, { 4095,15}, { 8959,16}, { 4607,15}, \ + { 9983,16}, { 5119,15}, { 10495,16}, { 5631,15}, \ + { 11775,17}, { 3071,16}, { 6143,15}, { 12287,16}, \ + { 65536,17}, { 131072,18}, { 262144,19}, { 524288,20}, \ + {1048576,21}, {2097152,22}, {4194304,23}, {8388608,24} } +#define SQR_FFT_TABLE3_SIZE 224 +#define SQR_FFT_THRESHOLD 5568 + +#define MULLO_BASECASE_THRESHOLD 0 /* always */ +#define MULLO_DC_THRESHOLD 61 +#define MULLO_MUL_N_THRESHOLD 14281 +#define SQRLO_BASECASE_THRESHOLD 9 +#define SQRLO_DC_THRESHOLD 0 /* never mpn_sqrlo_basecase */ +#define SQRLO_SQR_THRESHOLD 10950 + +#define DC_DIV_QR_THRESHOLD 54 +#define DC_DIVAPPR_Q_THRESHOLD 238 +#define DC_BDIV_QR_THRESHOLD 54 +#define DC_BDIV_Q_THRESHOLD 42 + +#define INV_MULMOD_BNM1_THRESHOLD 54 +#define INV_NEWTON_THRESHOLD 252 +#define INV_APPR_THRESHOLD 230 + +#define BINV_NEWTON_THRESHOLD 327 +#define REDC_1_TO_REDC_2_THRESHOLD 25 +#define REDC_2_TO_REDC_N_THRESHOLD 67 + +#define MU_DIV_QR_THRESHOLD 1620 +#define MU_DIVAPPR_Q_THRESHOLD 1620 +#define MUPI_DIV_QR_THRESHOLD 104 +#define MU_BDIV_QR_THRESHOLD 1528 +#define MU_BDIV_Q_THRESHOLD 1652 + +#define POWM_SEC_TABLE 1,22,321,473,2144 + +#define GET_STR_DC_THRESHOLD 15 +#define GET_STR_PRECOMPUTE_THRESHOLD 24 +#define SET_STR_DC_THRESHOLD 248 +#define SET_STR_PRECOMPUTE_THRESHOLD 1304 + +#define FAC_DSC_THRESHOLD 470 +#define FAC_ODD_THRESHOLD 25 + +#define MATRIX22_STRASSEN_THRESHOLD 17 +#define HGCD2_DIV1_METHOD 5 /* 8.38% faster than 4 */ +#define HGCD_THRESHOLD 115 +#define HGCD_APPR_THRESHOLD 146 +#define HGCD_REDUCE_THRESHOLD 3524 +#define GCD_DC_THRESHOLD 535 +#define GCDEXT_DC_THRESHOLD 460 +#define JACOBI_BASE_METHOD 1 /* 0.90% faster than 4 */ + +/* Tuneup completed successfully, took 448763 seconds */ diff --git a/vendor/gmp-6.3.0/mpn/x86_64/k10/hamdist.asm b/vendor/gmp-6.3.0/mpn/x86_64/k10/hamdist.asm new file mode 100644 index 0000000..f70494a --- /dev/null +++ b/vendor/gmp-6.3.0/mpn/x86_64/k10/hamdist.asm @@ -0,0 +1,109 @@ +dnl AMD64 mpn_hamdist -- hamming distance. + +dnl Copyright 2008, 2010-2012, 2017 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 AMD K8,K9 - +C AMD K10 2.0 = +C AMD bd1 ~4.4 = +C AMD bd2 ~4.4 = +C AMD bd3 +C AMD bd4 +C AMD bobcat 7.55 = +C AMD jaguar 2.52 - +C Intel P4 - +C Intel core2 - +C Intel NHM 2.03 + +C Intel SBR 2.01 + +C Intel IBR 1.96 + +C Intel HWL 1.64 = +C Intel BWL 1.56 - +C Intel SKL 1.52 = +C Intel atom +C Intel SLM 3.0 - +C VIA nano + +define(`ap', `%rdi') +define(`bp', `%rsi') +define(`n', `%rdx') + +ABI_SUPPORT(DOS64) +ABI_SUPPORT(STD64) + +ASM_START() + TEXT + ALIGN(32) +PROLOGUE(mpn_hamdist) + FUNC_ENTRY(3) + mov (ap), %r8 + xor (bp), %r8 + + lea (ap,n,8), ap C point at A operand end + lea (bp,n,8), bp C point at B operand end + neg n + + test $1, R8(n) + jz L(2) + +L(1): .byte 0xf3,0x49,0x0f,0xb8,0xc0 C popcnt %r8, %rax + xor R32(%r10), R32(%r10) + inc n + js L(top) + FUNC_EXIT() + ret + + ALIGN(16) +L(2): mov 8(ap,n,8), %r9 + .byte 0xf3,0x49,0x0f,0xb8,0xc0 C popcnt %r8, %rax + xor 8(bp,n,8), %r9 + .byte 0xf3,0x4d,0x0f,0xb8,0xd1 C popcnt %r9, %r10 + add $2, n + js L(top) + lea (%r10, %rax), %rax + FUNC_EXIT() + ret + + ALIGN(16) +L(top): mov (ap,n,8), %r8 + lea (%r10, %rax), %rax + mov 8(ap,n,8), %r9 + xor (bp,n,8), %r8 + xor 8(bp,n,8), %r9 + .byte 0xf3,0x49,0x0f,0xb8,0xc8 C popcnt %r8, %rcx + lea (%rcx, %rax), %rax + .byte 0xf3,0x4d,0x0f,0xb8,0xd1 C popcnt %r9, %r10 + add $2, n + js L(top) + + lea (%r10, %rax), %rax + FUNC_EXIT() + ret +EPILOGUE() diff --git a/vendor/gmp-6.3.0/mpn/x86_64/k10/lshift.asm b/vendor/gmp-6.3.0/mpn/x86_64/k10/lshift.asm new file mode 100644 index 0000000..cadf9b9 --- /dev/null +++ b/vendor/gmp-6.3.0/mpn/x86_64/k10/lshift.asm @@ -0,0 +1,37 @@ +dnl X86-64 mpn_lshift optimised for AMD K10. + +dnl Copyright 2012 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') + +ABI_SUPPORT(DOS64) +ABI_SUPPORT(STD64) + +MULFUNC_PROLOGUE(mpn_lshift) +include_mpn(`x86_64/fastsse/lshift-movdqu2.asm') diff --git a/vendor/gmp-6.3.0/mpn/x86_64/k10/lshiftc.asm b/vendor/gmp-6.3.0/mpn/x86_64/k10/lshiftc.asm new file mode 100644 index 0000000..48a92e5 --- /dev/null +++ b/vendor/gmp-6.3.0/mpn/x86_64/k10/lshiftc.asm @@ -0,0 +1,37 @@ +dnl X86-64 mpn_lshiftc optimised for AMD K10. + +dnl Copyright 2012 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') + +ABI_SUPPORT(DOS64) +ABI_SUPPORT(STD64) + +MULFUNC_PROLOGUE(mpn_lshiftc) +include_mpn(`x86_64/fastsse/lshiftc-movdqu2.asm') diff --git a/vendor/gmp-6.3.0/mpn/x86_64/k10/popcount.asm b/vendor/gmp-6.3.0/mpn/x86_64/k10/popcount.asm new file mode 100644 index 0000000..3814aea --- /dev/null +++ b/vendor/gmp-6.3.0/mpn/x86_64/k10/popcount.asm @@ -0,0 +1,138 @@ +dnl AMD64 mpn_popcount -- population count. + +dnl Copyright 2008, 2010-2012 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 AMD K8,K9 n/a +C AMD K10 1.125 +C Intel P4 n/a +C Intel core2 n/a +C Intel corei 1.25 +C Intel atom n/a +C VIA nano n/a + +C * The zero-offset of popcount is misassembled to the offset-less form, which +C is one byte shorter and therefore will mess up the switching code. +C * The outdated gas used in FreeBSD and NetBSD cannot handle the POPCNT insn, +C which is the main reason for our usage of '.byte'. + +C TODO +C * Improve switching code, the current code sucks. + +define(`up', `%rdi') +define(`n', `%rsi') + +ABI_SUPPORT(DOS64) +ABI_SUPPORT(STD64) + +ASM_START() + TEXT + ALIGN(32) +PROLOGUE(mpn_popcount) + FUNC_ENTRY(2) + +ifelse(1,1,` + lea (up,n,8), up + +C mov R32(n), R32(%rcx) +C neg R32(%rcx) + imul $-1, R32(n), R32(%rcx) + and $8-1, R32(%rcx) + + neg n + + mov R32(%rcx), R32(%rax) + neg %rax + lea (up,%rax,8),up + + xor R32(%rax), R32(%rax) + + lea (%rcx,%rcx,4), %rcx + + lea L(top)(%rip), %rdx + lea (%rdx,%rcx,2), %rdx + jmp *%rdx +',` + lea (up,n,8), up + + mov R32(n), R32(%rcx) + neg R32(%rcx) + and $8-1, R32(%rcx) + + neg n + + mov R32(%rcx), R32(%rax) + shl $3, R32(%rax) + sub %rax, up + + xor R32(%rax), R32(%rax) + +C add R32(%rcx), R32(%rcx) C 2x +C lea (%rcx,%rcx,4), %rcx C 10x + imul $10, R32(%rcx) + + lea L(top)(%rip), %rdx + add %rcx, %rdx + jmp *%rdx +') + + ALIGN(32) +L(top): +C 0 = n mod 8 + .byte 0xf3,0x4c,0x0f,0xb8,0x44,0xf7,0x00 C popcnt 0(up,n,8), %r8 + add %r8, %rax +C 7 = n mod 8 + .byte 0xf3,0x4c,0x0f,0xb8,0x4c,0xf7,0x08 C popcnt 8(up,n,8), %r9 + add %r9, %rax +C 6 = n mod 8 + .byte 0xf3,0x4c,0x0f,0xb8,0x44,0xf7,0x10 C popcnt 16(up,n,8), %r8 + add %r8, %rax +C 5 = n mod 8 + .byte 0xf3,0x4c,0x0f,0xb8,0x4c,0xf7,0x18 C popcnt 24(up,n,8), %r9 + add %r9, %rax +C 4 = n mod 8 + .byte 0xf3,0x4c,0x0f,0xb8,0x44,0xf7,0x20 C popcnt 32(up,n,8), %r8 + add %r8, %rax +C 3 = n mod 8 + .byte 0xf3,0x4c,0x0f,0xb8,0x4c,0xf7,0x28 C popcnt 40(up,n,8), %r9 + add %r9, %rax +C 2 = n mod 8 + .byte 0xf3,0x4c,0x0f,0xb8,0x44,0xf7,0x30 C popcnt 48(up,n,8), %r8 + add %r8, %rax +C 1 = n mod 8 + .byte 0xf3,0x4c,0x0f,0xb8,0x4c,0xf7,0x38 C popcnt 56(up,n,8), %r9 + add %r9, %rax + + add $8, n + js L(top) + FUNC_EXIT() + ret +EPILOGUE() diff --git a/vendor/gmp-6.3.0/mpn/x86_64/k10/rshift.asm b/vendor/gmp-6.3.0/mpn/x86_64/k10/rshift.asm new file mode 100644 index 0000000..249051a --- /dev/null +++ b/vendor/gmp-6.3.0/mpn/x86_64/k10/rshift.asm @@ -0,0 +1,37 @@ +dnl X86-64 mpn_rshift optimised for AMD K10. + +dnl Copyright 2012 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') + +ABI_SUPPORT(DOS64) +ABI_SUPPORT(STD64) + +MULFUNC_PROLOGUE(mpn_rshift) +include_mpn(`x86_64/fastsse/rshift-movdqu2.asm') diff --git a/vendor/gmp-6.3.0/mpn/x86_64/k10/sec_tabselect.asm b/vendor/gmp-6.3.0/mpn/x86_64/k10/sec_tabselect.asm new file mode 100644 index 0000000..e436034 --- /dev/null +++ b/vendor/gmp-6.3.0/mpn/x86_64/k10/sec_tabselect.asm @@ -0,0 +1,37 @@ +dnl X86-64 mpn_sec_tabselect. + +dnl Copyright 2012, 2013 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') + +ABI_SUPPORT(DOS64) +ABI_SUPPORT(STD64) + +MULFUNC_PROLOGUE(mpn_sec_tabselect) +include_mpn(`x86_64/fastsse/sec_tabselect.asm') -- cgit v1.2.3