From 161de004453f214c6030bb9e9babd2d0048a7337 Mon Sep 17 00:00:00 2001 From: Torbjorn Granlund Date: Thu, 24 Nov 2011 12:19:09 +0100 Subject: Retune. --- mpn/powerpc32/p5/gmp-mparam.h | 137 +++++++++++------------ mpn/powerpc32/p6/gmp-mparam.h | 206 ++++++++++++++++++----------------- mpn/powerpc64/mode64/p3/gmp-mparam.h | 73 +++++++------ mpn/powerpc64/mode64/p6/gmp-mparam.h | 42 +++---- 4 files changed, 240 insertions(+), 218 deletions(-) diff --git a/mpn/powerpc32/p5/gmp-mparam.h b/mpn/powerpc32/p5/gmp-mparam.h index a8400ce65..ba210ecc4 100644 --- a/mpn/powerpc32/p5/gmp-mparam.h +++ b/mpn/powerpc32/p5/gmp-mparam.h @@ -30,114 +30,117 @@ along with the GNU MP Library. If not, see http://www.gnu.org/licenses/. */ #define MOD_1_UNNORM_THRESHOLD 0 /* always */ #define MOD_1N_TO_MOD_1_1_THRESHOLD 8 #define MOD_1U_TO_MOD_1_1_THRESHOLD 6 -#define MOD_1_1_TO_MOD_1_2_THRESHOLD 8 -#define MOD_1_2_TO_MOD_1_4_THRESHOLD 46 -#define PREINV_MOD_1_TO_MOD_1_THRESHOLD 15 +#define MOD_1_1_TO_MOD_1_2_THRESHOLD 9 +#define MOD_1_2_TO_MOD_1_4_THRESHOLD 50 +#define PREINV_MOD_1_TO_MOD_1_THRESHOLD 18 #define USE_PREINV_DIVREM_1 1 +#define DIV_QR_2_PI2_THRESHOLD MP_SIZE_T_MAX /* never */ #define DIVEXACT_1_THRESHOLD 0 /* always */ -#define BMOD_1_TO_MOD_1_THRESHOLD 62 +#define BMOD_1_TO_MOD_1_THRESHOLD 61 #define MUL_TOOM22_THRESHOLD 22 -#define MUL_TOOM33_THRESHOLD 78 +#define MUL_TOOM33_THRESHOLD 57 #define MUL_TOOM44_THRESHOLD 130 -#define MUL_TOOM6H_THRESHOLD 206 -#define MUL_TOOM8H_THRESHOLD 260 +#define MUL_TOOM6H_THRESHOLD 189 +#define MUL_TOOM8H_THRESHOLD 309 #define MUL_TOOM32_TO_TOOM43_THRESHOLD 89 #define MUL_TOOM32_TO_TOOM53_THRESHOLD 99 -#define MUL_TOOM42_TO_TOOM53_THRESHOLD 85 +#define MUL_TOOM42_TO_TOOM53_THRESHOLD 83 #define MUL_TOOM42_TO_TOOM63_THRESHOLD 88 -#define SQR_BASECASE_THRESHOLD 0 /* always */ -#define SQR_TOOM2_THRESHOLD 42 +#define SQR_BASECASE_THRESHOLD 6 +#define SQR_TOOM2_THRESHOLD 40 #define SQR_TOOM3_THRESHOLD 77 -#define SQR_TOOM4_THRESHOLD 169 -#define SQR_TOOM6_THRESHOLD 246 -#define SQR_TOOM8_THRESHOLD 381 +#define SQR_TOOM4_THRESHOLD 124 +#define SQR_TOOM6_THRESHOLD 140 +#define SQR_TOOM8_THRESHOLD 238 + +#define MULMID_TOOM42_THRESHOLD 40 #define MULMOD_BNM1_THRESHOLD 15 -#define SQRMOD_BNM1_THRESHOLD 18 +#define SQRMOD_BNM1_THRESHOLD 16 + +#define POWM_SEC_TABLE 4,29,252,840,2080 -#define MUL_FFT_MODF_THRESHOLD 380 /* k = 5 */ +#define MUL_FFT_MODF_THRESHOLD 412 /* k = 5 */ #define MUL_FFT_TABLE3 \ - { { 380, 5}, { 19, 6}, { 10, 5}, { 21, 6}, \ - { 13, 5}, { 27, 6}, { 21, 7}, { 11, 6}, \ - { 25, 7}, { 13, 6}, { 27, 7}, { 15, 6}, \ - { 31, 7}, { 21, 8}, { 11, 7}, { 27, 8}, \ - { 15, 7}, { 33, 8}, { 19, 7}, { 39, 8}, \ - { 23, 7}, { 47, 8}, { 27, 9}, { 15, 8}, \ - { 39, 9}, { 23, 8}, { 51,10}, { 15, 9}, \ - { 31, 8}, { 67, 9}, { 39, 8}, { 79, 9}, \ - { 47,10}, { 31, 9}, { 79,10}, { 47,11}, \ - { 31,10}, { 63, 9}, { 135,10}, { 79, 9}, \ - { 159,10}, { 95, 9}, { 191,11}, { 63,10}, \ - { 127, 9}, { 255,10}, { 143, 9}, { 287, 8}, \ - { 575,10}, { 159,11}, { 95, 9}, { 383,12}, \ - { 63,11}, { 127,10}, { 255, 9}, { 511,10}, \ - { 271, 9}, { 543,10}, { 287, 9}, { 575,11}, \ - { 159,10}, { 319, 9}, { 639,10}, { 335, 9}, \ - { 671,10}, { 351,11}, { 191,10}, { 383, 9}, \ - { 767,10}, { 415, 9}, { 831,11}, { 223,12}, \ - { 4096,13}, { 8192,14}, { 16384,15}, { 32768,16} } -#define MUL_FFT_TABLE3_SIZE 76 + { { 412, 5}, { 21, 6}, { 11, 5}, { 23, 6}, \ + { 12, 5}, { 25, 6}, { 21, 7}, { 11, 6}, \ + { 25, 7}, { 13, 6}, { 27, 7}, { 21, 8}, \ + { 11, 7}, { 27, 8}, { 15, 7}, { 33, 8}, \ + { 19, 7}, { 39, 8}, { 23, 7}, { 47, 8}, \ + { 27, 9}, { 15, 8}, { 39, 9}, { 23, 8}, \ + { 51,10}, { 15, 9}, { 31, 8}, { 67, 9}, \ + { 39, 8}, { 79, 9}, { 55,10}, { 31, 9}, \ + { 79,10}, { 47, 9}, { 95,11}, { 31,10}, \ + { 63, 9}, { 135,10}, { 79, 9}, { 159,10}, \ + { 95,11}, { 63,10}, { 127, 9}, { 255,10}, \ + { 143, 9}, { 287,10}, { 159,11}, { 95,10}, \ + { 191,12}, { 63,11}, { 127,10}, { 255, 9}, \ + { 511,10}, { 271, 9}, { 543,10}, { 287,11}, \ + { 159,10}, { 335, 9}, { 671,10}, { 351, 9}, \ + { 703,11}, { 191,10}, { 383, 9}, { 767,10}, \ + { 415, 9}, { 831,11}, { 223,12}, { 4096,13}, \ + { 8192,14}, { 16384,15}, { 32768,16} } +#define MUL_FFT_TABLE3_SIZE 71 #define MUL_FFT_THRESHOLD 4736 -#define SQR_FFT_MODF_THRESHOLD 316 /* k = 5 */ +#define SQR_FFT_MODF_THRESHOLD 340 /* k = 5 */ #define SQR_FFT_TABLE3 \ - { { 316, 5}, { 19, 6}, { 10, 5}, { 21, 6}, \ + { { 340, 5}, { 21, 6}, { 11, 5}, { 23, 6}, \ { 21, 7}, { 11, 6}, { 24, 7}, { 13, 6}, \ - { 27, 7}, { 15, 6}, { 31, 7}, { 21, 8}, \ - { 11, 7}, { 27, 8}, { 15, 7}, { 33, 8}, \ - { 19, 6}, { 77, 7}, { 39, 8}, { 23, 7}, \ - { 47, 8}, { 27, 9}, { 15, 8}, { 39, 9}, \ - { 23, 8}, { 47,10}, { 15, 7}, { 121, 9}, \ - { 31, 8}, { 67, 9}, { 39, 8}, { 79, 9}, \ - { 47,10}, { 31, 9}, { 79,10}, { 47,11}, \ - { 31,10}, { 63, 9}, { 127, 8}, { 255,10}, \ - { 79, 9}, { 159, 8}, { 319, 9}, { 175,10}, \ - { 95, 9}, { 191, 8}, { 383,11}, { 63,10}, \ + { 27, 7}, { 21, 8}, { 11, 7}, { 27, 8}, \ + { 15, 7}, { 33, 8}, { 19, 7}, { 39, 8}, \ + { 23, 7}, { 47, 8}, { 27, 9}, { 15, 8}, \ + { 39, 9}, { 23, 8}, { 47,10}, { 15, 9}, \ + { 31, 8}, { 67, 9}, { 47,10}, { 31, 9}, \ + { 71,10}, { 47,11}, { 31,10}, { 63, 9}, \ + { 127, 8}, { 255, 9}, { 135,10}, { 79, 9}, \ + { 159,10}, { 95, 9}, { 191,11}, { 63,10}, \ { 127, 9}, { 255, 8}, { 511, 9}, { 271,10}, \ { 143, 9}, { 287, 8}, { 575, 9}, { 303,10}, \ - { 159, 9}, { 319,10}, { 175,11}, { 95,10}, \ - { 191, 9}, { 383,10}, { 207,12}, { 63,11}, \ + { 159,11}, { 95,10}, { 191,12}, { 63,11}, \ { 127,10}, { 255, 9}, { 511,10}, { 271, 9}, \ { 543,10}, { 287, 9}, { 575,10}, { 303,11}, \ { 159,10}, { 319, 9}, { 639,10}, { 335, 9}, \ { 671,10}, { 351,11}, { 191,10}, { 383, 9}, \ { 767,10}, { 415,11}, { 223,10}, { 447,12}, \ { 4096,13}, { 8192,14}, { 16384,15}, { 32768,16} } -#define SQR_FFT_TABLE3_SIZE 88 +#define SQR_FFT_TABLE3_SIZE 76 #define SQR_FFT_THRESHOLD 3712 #define MULLO_BASECASE_THRESHOLD 2 #define MULLO_DC_THRESHOLD 68 #define MULLO_MUL_N_THRESHOLD 9236 -#define DC_DIV_QR_THRESHOLD 70 -#define DC_DIVAPPR_Q_THRESHOLD 238 +#define DC_DIV_QR_THRESHOLD 69 +#define DC_DIVAPPR_Q_THRESHOLD 220 #define DC_BDIV_QR_THRESHOLD 75 #define DC_BDIV_Q_THRESHOLD 188 #define INV_MULMOD_BNM1_THRESHOLD 54 -#define INV_NEWTON_THRESHOLD 250 -#define INV_APPR_THRESHOLD 246 +#define INV_NEWTON_THRESHOLD 230 +#define INV_APPR_THRESHOLD 230 -#define BINV_NEWTON_THRESHOLD 375 +#define BINV_NEWTON_THRESHOLD 278 #define REDC_1_TO_REDC_N_THRESHOLD 87 -#define MU_DIV_QR_THRESHOLD 1334 -#define MU_DIVAPPR_Q_THRESHOLD 1387 -#define MUPI_DIV_QR_THRESHOLD 114 -#define MU_BDIV_QR_THRESHOLD 1078 -#define MU_BDIV_Q_THRESHOLD 1334 +#define MU_DIV_QR_THRESHOLD 1210 +#define MU_DIVAPPR_Q_THRESHOLD 1308 +#define MUPI_DIV_QR_THRESHOLD 106 +#define MU_BDIV_QR_THRESHOLD 1017 +#define MU_BDIV_Q_THRESHOLD 1210 #define MATRIX22_STRASSEN_THRESHOLD 14 -#define HGCD_THRESHOLD 104 -#define GCD_DC_THRESHOLD 424 -#define GCDEXT_DC_THRESHOLD 321 +#define HGCD_THRESHOLD 110 +#define HGCD_APPR_THRESHOLD 138 +#define HGCD_REDUCE_THRESHOLD 2578 +#define GCD_DC_THRESHOLD 408 +#define GCDEXT_DC_THRESHOLD 298 #define JACOBI_BASE_METHOD 4 -#define GET_STR_DC_THRESHOLD 12 -#define GET_STR_PRECOMPUTE_THRESHOLD 23 -#define SET_STR_DC_THRESHOLD 454 -#define SET_STR_PRECOMPUTE_THRESHOLD 1074 +#define GET_STR_DC_THRESHOLD 13 +#define GET_STR_PRECOMPUTE_THRESHOLD 24 +#define SET_STR_DC_THRESHOLD 527 +#define SET_STR_PRECOMPUTE_THRESHOLD 1090 diff --git a/mpn/powerpc32/p6/gmp-mparam.h b/mpn/powerpc32/p6/gmp-mparam.h index 73951d0ae..529a66d19 100644 --- a/mpn/powerpc32/p6/gmp-mparam.h +++ b/mpn/powerpc32/p6/gmp-mparam.h @@ -29,115 +29,127 @@ along with the GNU MP Library. If not, see http://www.gnu.org/licenses/. */ #define MOD_1_NORM_THRESHOLD 3 #define MOD_1_UNNORM_THRESHOLD 0 /* always */ #define MOD_1N_TO_MOD_1_1_THRESHOLD 3 -#define MOD_1U_TO_MOD_1_1_THRESHOLD 8 -#define MOD_1_1_TO_MOD_1_2_THRESHOLD 0 /* never mpn_mod_1_1p */ -#define MOD_1_2_TO_MOD_1_4_THRESHOLD 15 -#define PREINV_MOD_1_TO_MOD_1_THRESHOLD 9 +#define MOD_1U_TO_MOD_1_1_THRESHOLD 3 +#define MOD_1_1_TO_MOD_1_2_THRESHOLD MP_SIZE_T_MAX +#define MOD_1_2_TO_MOD_1_4_THRESHOLD 0 /* never mpn_mod_1s_2p */ +#define PREINV_MOD_1_TO_MOD_1_THRESHOLD 8 #define USE_PREINV_DIVREM_1 1 +#define DIV_QR_2_PI2_THRESHOLD MP_SIZE_T_MAX /* never */ #define DIVEXACT_1_THRESHOLD 0 /* always */ #define BMOD_1_TO_MOD_1_THRESHOLD MP_SIZE_T_MAX /* never */ -#define MUL_TOOM22_THRESHOLD 34 -#define MUL_TOOM33_THRESHOLD 70 -#define MUL_TOOM44_THRESHOLD 187 -#define MUL_TOOM6H_THRESHOLD 286 -#define MUL_TOOM8H_THRESHOLD 321 +#define MUL_TOOM22_THRESHOLD 19 +#define MUL_TOOM33_THRESHOLD 55 +#define MUL_TOOM44_THRESHOLD 88 +#define MUL_TOOM6H_THRESHOLD 137 +#define MUL_TOOM8H_THRESHOLD 181 -#define MUL_TOOM32_TO_TOOM43_THRESHOLD 110 -#define MUL_TOOM32_TO_TOOM53_THRESHOLD 118 -#define MUL_TOOM42_TO_TOOM53_THRESHOLD 107 -#define MUL_TOOM42_TO_TOOM63_THRESHOLD 145 +#define MUL_TOOM32_TO_TOOM43_THRESHOLD 57 +#define MUL_TOOM32_TO_TOOM53_THRESHOLD 56 +#define MUL_TOOM42_TO_TOOM53_THRESHOLD 57 +#define MUL_TOOM42_TO_TOOM63_THRESHOLD 56 #define SQR_BASECASE_THRESHOLD 0 /* always */ -#define SQR_TOOM2_THRESHOLD 68 -#define SQR_TOOM3_THRESHOLD 113 -#define SQR_TOOM4_THRESHOLD 312 -#define SQR_TOOM6_THRESHOLD 330 -#define SQR_TOOM8_THRESHOLD 357 +#define SQR_TOOM2_THRESHOLD 30 +#define SQR_TOOM3_THRESHOLD 56 +#define SQR_TOOM4_THRESHOLD 130 +#define SQR_TOOM6_THRESHOLD 189 +#define SQR_TOOM8_THRESHOLD 296 -#define MULMOD_BNM1_THRESHOLD 19 -#define SQRMOD_BNM1_THRESHOLD 20 +#define MULMID_TOOM42_THRESHOLD 26 -#define MUL_FFT_MODF_THRESHOLD 304 /* k = 5 */ +#define MULMOD_BNM1_THRESHOLD 7 +#define SQRMOD_BNM1_THRESHOLD 12 + +#define POWM_SEC_TABLE 2,26,127,453,1068 + +#define MUL_FFT_MODF_THRESHOLD 212 /* k = 5 */ #define MUL_FFT_TABLE3 \ - { { 304, 5}, { 17, 6}, { 9, 5}, { 19, 6}, \ - { 10, 5}, { 21, 6}, { 17, 7}, { 9, 6}, \ - { 20, 7}, { 11, 6}, { 24, 7}, { 13, 8}, \ - { 7, 7}, { 21, 8}, { 11, 7}, { 27, 9}, \ - { 7, 8}, { 15, 7}, { 33, 8}, { 19, 7}, \ - { 41, 8}, { 23, 7}, { 47, 8}, { 27, 9}, \ + { { 212, 5}, { 13, 6}, { 7, 5}, { 15, 6}, \ + { 13, 7}, { 7, 6}, { 16, 7}, { 9, 6}, \ + { 19, 7}, { 13, 8}, { 7, 7}, { 19, 8}, \ + { 11, 7}, { 25, 9}, { 7, 8}, { 15, 7}, \ + { 31, 8}, { 19, 7}, { 39, 8}, { 23, 9}, \ { 15, 8}, { 39, 9}, { 23, 8}, { 47,10}, \ - { 15, 9}, { 31, 8}, { 67, 9}, { 39, 8}, \ - { 79, 9}, { 47, 8}, { 95,10}, { 31, 9}, \ - { 79,10}, { 47, 9}, { 95,11}, { 31,10}, \ - { 63, 9}, { 127, 8}, { 255, 9}, { 135,10}, \ - { 79, 9}, { 159, 8}, { 319,10}, { 95, 9}, \ - { 191, 8}, { 383,11}, { 63,10}, { 127, 9}, \ - { 255, 8}, { 511, 9}, { 271,10}, { 143, 9}, \ - { 287,10}, { 159, 9}, { 319,11}, { 95,10}, \ - { 191, 9}, { 383,12}, { 63,11}, { 127,10}, \ - { 255, 9}, { 511,10}, { 271, 9}, { 543,10}, \ - { 287,11}, { 159,10}, { 319, 9}, { 639,10}, \ - { 351,11}, { 191,10}, { 383, 9}, { 767,10}, \ - { 415,11}, { 223,10}, { 447,12}, { 4096,13}, \ - { 8192,14}, { 16384,15}, { 32768,16} } -#define MUL_FFT_TABLE3_SIZE 83 -#define MUL_FFT_THRESHOLD 4736 - -#define SQR_FFT_MODF_THRESHOLD 312 /* k = 5 */ -#define SQR_FFT_TABLE3 \ - { { 312, 5}, { 21, 6}, { 11, 5}, { 23, 6}, \ - { 21, 7}, { 11, 6}, { 24, 7}, { 13, 6}, \ - { 27, 7}, { 17, 6}, { 35, 7}, { 21, 8}, \ - { 11, 7}, { 27, 8}, { 15, 7}, { 33, 8}, \ - { 19, 7}, { 39, 8}, { 23, 7}, { 47, 8}, \ - { 27, 9}, { 15, 8}, { 39, 9}, { 23, 8}, \ - { 47,10}, { 15, 9}, { 31, 8}, { 67, 9}, \ - { 39, 8}, { 79, 9}, { 47,10}, { 31, 9}, \ + { 15, 9}, { 31, 8}, { 63, 9}, { 39, 8}, \ + { 79, 9}, { 47,10}, { 31, 9}, { 63, 8}, \ + { 127, 9}, { 71, 8}, { 143, 7}, { 287, 9}, \ { 79,10}, { 47,11}, { 31,10}, { 63, 9}, \ - { 127, 8}, { 255,10}, { 79, 9}, { 159, 8}, \ - { 319,10}, { 95, 9}, { 191,11}, { 63,10}, \ - { 127, 9}, { 255, 8}, { 511, 9}, { 271,10}, \ + { 127, 8}, { 255, 7}, { 511, 9}, { 143, 8}, \ + { 287,10}, { 79, 9}, { 159, 8}, { 319, 9}, \ + { 175, 8}, { 351,10}, { 95, 9}, { 191, 8}, \ + { 383, 9}, { 207,10}, { 111,11}, { 63,10}, \ + { 127, 9}, { 255, 8}, { 511,10}, { 143, 9}, \ + { 287, 8}, { 575,10}, { 159, 9}, { 319,10}, \ + { 175, 9}, { 351,11}, { 95,10}, { 191, 9}, \ + { 383,10}, { 207, 9}, { 415,12}, { 63,11}, \ + { 127,10}, { 255, 9}, { 511,10}, { 287, 9}, \ + { 575,11}, { 159,10}, { 351, 9}, { 703,11}, \ + { 191,10}, { 415, 9}, { 831,11}, { 223,10}, \ + { 447,12}, { 4096,13}, { 8192,14}, { 16384,15}, \ + { 32768,16} } +#define MUL_FFT_TABLE3_SIZE 89 +#define MUL_FFT_THRESHOLD 1728 + +#define SQR_FFT_MODF_THRESHOLD 184 /* k = 5 */ +#define SQR_FFT_TABLE3 \ + { { 184, 5}, { 6, 4}, { 13, 5}, { 13, 6}, \ + { 7, 5}, { 15, 6}, { 13, 7}, { 7, 6}, \ + { 16, 7}, { 9, 6}, { 19, 7}, { 11, 6}, \ + { 23, 7}, { 13, 8}, { 7, 7}, { 19, 8}, \ + { 11, 7}, { 23, 9}, { 7, 8}, { 23, 9}, \ + { 15, 8}, { 39, 9}, { 23,10}, { 15, 9}, \ + { 31, 8}, { 63, 9}, { 39, 8}, { 79, 9}, \ + { 47,10}, { 31, 9}, { 63, 8}, { 127, 7}, \ + { 255, 9}, { 71, 8}, { 143, 7}, { 287, 6}, \ + { 575, 9}, { 79,10}, { 47,11}, { 31,10}, \ + { 63, 9}, { 127, 8}, { 255, 9}, { 143, 8}, \ + { 287, 7}, { 575,10}, { 79, 9}, { 159, 8}, \ + { 319, 9}, { 175, 8}, { 351,10}, { 95, 9}, \ + { 191, 8}, { 383, 9}, { 207,10}, { 111, 9}, \ + { 223,11}, { 63,10}, { 127, 9}, { 255,10}, \ { 143, 9}, { 287, 8}, { 575,10}, { 159, 9}, \ - { 319,11}, { 95,10}, { 191, 9}, { 383,12}, \ - { 63,11}, { 127,10}, { 255, 9}, { 511,10}, \ - { 271, 9}, { 543,10}, { 287, 9}, { 575,11}, \ - { 159,10}, { 319, 9}, { 639,10}, { 351,11}, \ - { 191,10}, { 383, 9}, { 767,10}, { 415,11}, \ - { 223,10}, { 447,12}, { 4096,13}, { 8192,14}, \ - { 16384,15}, { 32768,16} } -#define SQR_FFT_TABLE3_SIZE 78 -#define SQR_FFT_THRESHOLD 2752 - -#define MULLO_BASECASE_THRESHOLD 0 /* always */ -#define MULLO_DC_THRESHOLD 151 -#define MULLO_MUL_N_THRESHOLD 1175 - -#define DC_DIV_QR_THRESHOLD 133 -#define DC_DIVAPPR_Q_THRESHOLD 442 -#define DC_BDIV_QR_THRESHOLD 130 -#define DC_BDIV_Q_THRESHOLD 324 - -#define INV_MULMOD_BNM1_THRESHOLD 116 -#define INV_NEWTON_THRESHOLD 507 -#define INV_APPR_THRESHOLD 454 - -#define BINV_NEWTON_THRESHOLD 507 -#define REDC_1_TO_REDC_N_THRESHOLD 118 - -#define MU_DIV_QR_THRESHOLD 1652 -#define MU_DIVAPPR_Q_THRESHOLD 1752 -#define MUPI_DIV_QR_THRESHOLD 225 -#define MU_BDIV_QR_THRESHOLD 762 -#define MU_BDIV_Q_THRESHOLD 1017 - -#define MATRIX22_STRASSEN_THRESHOLD 28 -#define HGCD_THRESHOLD 76 -#define GCD_DC_THRESHOLD 333 -#define GCDEXT_DC_THRESHOLD 245 + { 319,10}, { 175, 9}, { 351,11}, { 95,10}, \ + { 191, 9}, { 383,10}, { 207, 9}, { 415,10}, \ + { 223,12}, { 63,11}, { 127,10}, { 255, 9}, \ + { 511,10}, { 287, 9}, { 575,11}, { 159,10}, \ + { 351, 9}, { 703, 8}, { 1407,11}, { 191,10}, \ + { 415,11}, { 223,10}, { 447, 9}, { 895,12}, \ + { 4096,13}, { 8192,14}, { 16384,15}, { 32768,16} } +#define SQR_FFT_TABLE3_SIZE 92 +#define SQR_FFT_THRESHOLD 1600 + +#define MULLO_BASECASE_THRESHOLD 2 +#define MULLO_DC_THRESHOLD 57 +#define MULLO_MUL_N_THRESHOLD 3176 + +#define DC_DIV_QR_THRESHOLD 52 +#define DC_DIVAPPR_Q_THRESHOLD 187 +#define DC_BDIV_QR_THRESHOLD 64 +#define DC_BDIV_Q_THRESHOLD 146 + +#define INV_MULMOD_BNM1_THRESHOLD 68 +#define INV_NEWTON_THRESHOLD 182 +#define INV_APPR_THRESHOLD 182 + +#define BINV_NEWTON_THRESHOLD 186 +#define REDC_1_TO_REDC_N_THRESHOLD 60 + +#define MU_DIV_QR_THRESHOLD 924 +#define MU_DIVAPPR_Q_THRESHOLD 807 +#define MUPI_DIV_QR_THRESHOLD 73 +#define MU_BDIV_QR_THRESHOLD 667 +#define MU_BDIV_Q_THRESHOLD 823 + +#define MATRIX22_STRASSEN_THRESHOLD 8 +#define HGCD_THRESHOLD 61 +#define HGCD_APPR_THRESHOLD 50 +#define HGCD_REDUCE_THRESHOLD 974 +#define GCD_DC_THRESHOLD 195 +#define GCDEXT_DC_THRESHOLD 134 #define JACOBI_BASE_METHOD 4 -#define GET_STR_DC_THRESHOLD 10 -#define GET_STR_PRECOMPUTE_THRESHOLD 20 -#define SET_STR_DC_THRESHOLD 199 -#define SET_STR_PRECOMPUTE_THRESHOLD 478 +#define GET_STR_DC_THRESHOLD 9 +#define GET_STR_PRECOMPUTE_THRESHOLD 21 +#define SET_STR_DC_THRESHOLD 190 +#define SET_STR_PRECOMPUTE_THRESHOLD 411 diff --git a/mpn/powerpc64/mode64/p3/gmp-mparam.h b/mpn/powerpc64/mode64/p3/gmp-mparam.h index 221b0e1d8..cf1d8ca47 100644 --- a/mpn/powerpc64/mode64/p3/gmp-mparam.h +++ b/mpn/powerpc64/mode64/p3/gmp-mparam.h @@ -23,12 +23,13 @@ along with the GNU MP Library. If not, see http://www.gnu.org/licenses/. */ #define MOD_1_NORM_THRESHOLD 0 /* always */ #define MOD_1_UNNORM_THRESHOLD 0 /* always */ -#define MOD_1N_TO_MOD_1_1_THRESHOLD MP_SIZE_T_MAX /* never */ -#define MOD_1U_TO_MOD_1_1_THRESHOLD 5 -#define MOD_1_1_TO_MOD_1_2_THRESHOLD 16 +#define MOD_1N_TO_MOD_1_1_THRESHOLD 7 +#define MOD_1U_TO_MOD_1_1_THRESHOLD 3 +#define MOD_1_1_TO_MOD_1_2_THRESHOLD 14 #define MOD_1_2_TO_MOD_1_4_THRESHOLD 0 /* never mpn_mod_1s_2p */ -#define PREINV_MOD_1_TO_MOD_1_THRESHOLD MP_SIZE_T_MAX /* never */ +#define PREINV_MOD_1_TO_MOD_1_THRESHOLD 17 #define USE_PREINV_DIVREM_1 0 +#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 MP_SIZE_T_MAX /* never */ @@ -36,22 +37,26 @@ along with the GNU MP Library. If not, see http://www.gnu.org/licenses/. */ #define MUL_TOOM33_THRESHOLD 33 #define MUL_TOOM44_THRESHOLD 46 #define MUL_TOOM6H_THRESHOLD 77 -#define MUL_TOOM8H_THRESHOLD 115 +#define MUL_TOOM8H_THRESHOLD 139 #define MUL_TOOM32_TO_TOOM43_THRESHOLD 49 -#define MUL_TOOM32_TO_TOOM53_THRESHOLD 38 -#define MUL_TOOM42_TO_TOOM53_THRESHOLD 33 -#define MUL_TOOM42_TO_TOOM63_THRESHOLD 32 - -#define SQR_BASECASE_THRESHOLD 0 /* always */ -#define SQR_TOOM2_THRESHOLD 16 -#define SQR_TOOM3_THRESHOLD 49 -#define SQR_TOOM4_THRESHOLD 70 -#define SQR_TOOM6_THRESHOLD 93 +#define MUL_TOOM32_TO_TOOM53_THRESHOLD 48 +#define MUL_TOOM42_TO_TOOM53_THRESHOLD 49 +#define MUL_TOOM42_TO_TOOM63_THRESHOLD 49 + +#define SQR_BASECASE_THRESHOLD 0 /* always (native) */ +#define SQR_TOOM2_THRESHOLD 14 +#define SQR_TOOM3_THRESHOLD 45 +#define SQR_TOOM4_THRESHOLD 64 +#define SQR_TOOM6_THRESHOLD 85 #define SQR_TOOM8_THRESHOLD 139 +#define MULMID_TOOM42_THRESHOLD 22 + #define MULMOD_BNM1_THRESHOLD 8 -#define SQRMOD_BNM1_THRESHOLD 9 +#define SQRMOD_BNM1_THRESHOLD 10 + +#define POWM_SEC_TABLE 2,23,127,502,1421 #define MUL_FFT_MODF_THRESHOLD 220 /* k = 5 */ #define MUL_FFT_TABLE3 \ @@ -123,35 +128,37 @@ along with the GNU MP Library. If not, see http://www.gnu.org/licenses/. */ #define SQR_FFT_TABLE3_SIZE 118 #define SQR_FFT_THRESHOLD 1728 -#define MULLO_BASECASE_THRESHOLD 3 -#define MULLO_DC_THRESHOLD 28 -#define MULLO_MUL_N_THRESHOLD 4940 +#define MULLO_BASECASE_THRESHOLD 2 +#define MULLO_DC_THRESHOLD 27 +#define MULLO_MUL_N_THRESHOLD 2367 -#define DC_DIV_QR_THRESHOLD 27 -#define DC_DIVAPPR_Q_THRESHOLD 95 -#define DC_BDIV_QR_THRESHOLD 28 +#define DC_DIV_QR_THRESHOLD 26 +#define DC_DIVAPPR_Q_THRESHOLD 87 +#define DC_BDIV_QR_THRESHOLD 27 #define DC_BDIV_Q_THRESHOLD 62 -#define INV_MULMOD_BNM1_THRESHOLD 29 -#define INV_NEWTON_THRESHOLD 92 -#define INV_APPR_THRESHOLD 94 +#define INV_MULMOD_BNM1_THRESHOLD 34 +#define INV_NEWTON_THRESHOLD 91 +#define INV_APPR_THRESHOLD 91 #define BINV_NEWTON_THRESHOLD 115 -#define REDC_1_TO_REDC_N_THRESHOLD 30 +#define REDC_1_TO_REDC_N_THRESHOLD 31 #define MU_DIV_QR_THRESHOLD 551 #define MU_DIVAPPR_Q_THRESHOLD 551 -#define MUPI_DIV_QR_THRESHOLD 49 -#define MU_BDIV_QR_THRESHOLD 492 +#define MUPI_DIV_QR_THRESHOLD 50 +#define MU_BDIV_QR_THRESHOLD 474 #define MU_BDIV_Q_THRESHOLD 492 -#define MATRIX22_STRASSEN_THRESHOLD 9 -#define HGCD_THRESHOLD 55 -#define GCD_DC_THRESHOLD 150 -#define GCDEXT_DC_THRESHOLD 124 +#define MATRIX22_STRASSEN_THRESHOLD 8 +#define HGCD_THRESHOLD 53 +#define HGCD_APPR_THRESHOLD 55 +#define HGCD_REDUCE_THRESHOLD 688 +#define GCD_DC_THRESHOLD 148 +#define GCDEXT_DC_THRESHOLD 118 #define JACOBI_BASE_METHOD 1 -#define GET_STR_DC_THRESHOLD 17 +#define GET_STR_DC_THRESHOLD 16 #define GET_STR_PRECOMPUTE_THRESHOLD 27 -#define SET_STR_DC_THRESHOLD 354 +#define SET_STR_DC_THRESHOLD 375 #define SET_STR_PRECOMPUTE_THRESHOLD 812 diff --git a/mpn/powerpc64/mode64/p6/gmp-mparam.h b/mpn/powerpc64/mode64/p6/gmp-mparam.h index bf7f0fd0c..5392138f1 100644 --- a/mpn/powerpc64/mode64/p6/gmp-mparam.h +++ b/mpn/powerpc64/mode64/p6/gmp-mparam.h @@ -39,26 +39,26 @@ along with the GNU MP Library. If not, see http://www.gnu.org/licenses/. */ #define MUL_TOOM33_THRESHOLD 50 #define MUL_TOOM44_THRESHOLD 112 #define MUL_TOOM6H_THRESHOLD 274 -#define MUL_TOOM8H_THRESHOLD 430 +#define MUL_TOOM8H_THRESHOLD 339 #define MUL_TOOM32_TO_TOOM43_THRESHOLD 62 -#define MUL_TOOM32_TO_TOOM53_THRESHOLD 84 +#define MUL_TOOM32_TO_TOOM53_THRESHOLD 76 #define MUL_TOOM42_TO_TOOM53_THRESHOLD 73 -#define MUL_TOOM42_TO_TOOM63_THRESHOLD 66 +#define MUL_TOOM42_TO_TOOM63_THRESHOLD 78 #define SQR_BASECASE_THRESHOLD 0 /* always (native) */ #define SQR_TOOM2_THRESHOLD 24 #define SQR_TOOM3_THRESHOLD 49 #define SQR_TOOM4_THRESHOLD 136 -#define SQR_TOOM6_THRESHOLD 274 -#define SQR_TOOM8_THRESHOLD 410 +#define SQR_TOOM6_THRESHOLD 226 +#define SQR_TOOM8_THRESHOLD 393 #define MULMID_TOOM42_THRESHOLD 36 #define MULMOD_BNM1_THRESHOLD 14 #define SQRMOD_BNM1_THRESHOLD 14 -#define POWM_SEC_TABLE 4,19,228,713,919 +#define POWM_SEC_TABLE 4,23,213,840,2618 #define MUL_FFT_MODF_THRESHOLD 340 /* k = 5 */ #define MUL_FFT_TABLE3 \ @@ -112,35 +112,35 @@ along with the GNU MP Library. If not, see http://www.gnu.org/licenses/. */ #define MULLO_BASECASE_THRESHOLD 5 #define MULLO_DC_THRESHOLD 61 -#define MULLO_MUL_N_THRESHOLD 5558 +#define MULLO_MUL_N_THRESHOLD 3271 -#define DC_DIV_QR_THRESHOLD 29 -#define DC_DIVAPPR_Q_THRESHOLD 112 +#define DC_DIV_QR_THRESHOLD 59 +#define DC_DIVAPPR_Q_THRESHOLD 200 #define DC_BDIV_QR_THRESHOLD 70 #define DC_BDIV_Q_THRESHOLD 168 #define INV_MULMOD_BNM1_THRESHOLD 61 -#define INV_NEWTON_THRESHOLD 93 -#define INV_APPR_THRESHOLD 91 +#define INV_NEWTON_THRESHOLD 166 +#define INV_APPR_THRESHOLD 166 #define BINV_NEWTON_THRESHOLD 222 #define REDC_1_TO_REDC_N_THRESHOLD 63 -#define MU_DIV_QR_THRESHOLD 807 -#define MU_DIVAPPR_Q_THRESHOLD 807 -#define MUPI_DIV_QR_THRESHOLD 27 -#define MU_BDIV_QR_THRESHOLD 872 +#define MU_DIV_QR_THRESHOLD 998 +#define MU_DIVAPPR_Q_THRESHOLD 979 +#define MUPI_DIV_QR_THRESHOLD 59 +#define MU_BDIV_QR_THRESHOLD 889 #define MU_BDIV_Q_THRESHOLD 1078 #define MATRIX22_STRASSEN_THRESHOLD 13 -#define HGCD_THRESHOLD 94 -#define HGCD_APPR_THRESHOLD 55 -#define HGCD_REDUCE_THRESHOLD 2121 -#define GCD_DC_THRESHOLD 253 -#define GCDEXT_DC_THRESHOLD 217 +#define HGCD_THRESHOLD 109 +#define HGCD_APPR_THRESHOLD 108 +#define HGCD_REDUCE_THRESHOLD 1052 +#define GCD_DC_THRESHOLD 501 +#define GCDEXT_DC_THRESHOLD 249 #define JACOBI_BASE_METHOD 4 #define GET_STR_DC_THRESHOLD 16 #define GET_STR_PRECOMPUTE_THRESHOLD 29 #define SET_STR_DC_THRESHOLD 532 -#define SET_STR_PRECOMPUTE_THRESHOLD 1561 +#define SET_STR_PRECOMPUTE_THRESHOLD 1639 -- cgit v1.2.1