Menu
▾
▴
ggnfs-commits
|
From: <Ba...@us...> - 2009-04-18 03:58:56
|
Revision: 353
http://ggnfs.svn.sourceforge.net/ggnfs/?rev=353&view=rev
Author: Batalov
Date: 2009-04-18 03:58:53 +0000 (Sat, 18 Apr 2009)
Log Message:
-----------
My patches to lasieve_64
Modified Paths:
--------------
trunk/src/experimental/lasieve4_64/INSTALL
trunk/src/experimental/lasieve4_64/athlon64/Makefile
trunk/src/experimental/lasieve4_64/athlon64/slinie1A.asm
trunk/src/experimental/lasieve4_64/athlon64/slinie2A.asm
trunk/src/experimental/lasieve4_64/athlon64/slinie3A.asm
trunk/src/experimental/lasieve4_64/athlon64/tdslinie1.asm
trunk/src/experimental/lasieve4_64/athlon64/tdslinie2.asm
trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c
Modified: trunk/src/experimental/lasieve4_64/INSTALL
===================================================================
--- trunk/src/experimental/lasieve4_64/INSTALL 2009-04-18 03:42:10 UTC (rev 352)
+++ trunk/src/experimental/lasieve4_64/INSTALL 2009-04-18 03:58:53 UTC (rev 353)
@@ -2,6 +2,8 @@
follow these steps to build the binaries on a 64-bit machine.
These should build and run on Athlon64, or Core2 cpus.
+Prerequisites: gmp.h libgmp
+
cd athlon64
make liblasieve.a
make liblasieveI11.a
@@ -12,5 +14,6 @@
make liblasieveI16.a
cp *.a ..
cd ..
+ln -s athlon64 asm
make
Modified: trunk/src/experimental/lasieve4_64/athlon64/Makefile
===================================================================
--- trunk/src/experimental/lasieve4_64/athlon64/Makefile 2009-04-18 03:42:10 UTC (rev 352)
+++ trunk/src/experimental/lasieve4_64/athlon64/Makefile 2009-04-18 03:58:53 UTC (rev 353)
@@ -7,7 +7,7 @@
# Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
# 02111-1307, USA.
-CFLAGS= -O3 -march=amdfam10 -mtune=amdfam10
+CFLAGS= -O3 -march=k8 -mtune=k8 -I.
#CFLAGS= -g
CC=gcc
Modified: trunk/src/experimental/lasieve4_64/athlon64/slinie1A.asm
===================================================================
--- trunk/src/experimental/lasieve4_64/athlon64/slinie1A.asm 2009-04-18 03:42:10 UTC (rev 352)
+++ trunk/src/experimental/lasieve4_64/athlon64/slinie1A.asm 2009-04-18 03:58:53 UTC (rev 353)
@@ -75,10 +75,6 @@
popq %rbx
ret
ifelse(j_per_strip,1,`divert`'',`divert(-1)')dnl
-dnl In this case, it is possible to keep the sieve interval in a register
-dnl since the register keeping the root is not needed.
-define(sieve_interval_arg,sieve_interval)dnl
-define(`sieve_interval',root)dnl
function_head(slinie1)
cmpq aux_ptr,aux_ptr_ub
pushq %rbx
@@ -86,11 +82,15 @@
jbe slinie1_ende
subq $8,aux_ptr_ub
slinie1_fbi_loop:
+ movzwq proot_src,auxreg
movzwq prime_src,prime
- movzwq root_src,sieve_ptr
+ movzwq root_src,root
+ subq prime,auxreg
movb log_src,sieve_log
+ movq auxreg,proot
movq sieve_interval,sieve_ptr_ub
- leaq (sieve_interval,sieve_ptr),sieve_ptr
+ movq root,sieve_ptr
+ leaq (sieve_ptr_ub,sieve_ptr),sieve_ptr
addq $n_i,sieve_ptr_ub
subq prime,sieve_ptr_ub
movb (sieve_ptr),sv0
Modified: trunk/src/experimental/lasieve4_64/athlon64/slinie2A.asm
===================================================================
--- trunk/src/experimental/lasieve4_64/athlon64/slinie2A.asm 2009-04-18 03:42:10 UTC (rev 352)
+++ trunk/src/experimental/lasieve4_64/athlon64/slinie2A.asm 2009-04-18 03:58:53 UTC (rev 353)
@@ -78,10 +78,6 @@
popq %rbx
ret
ifelse(j_per_strip,1,`divert`'',`divert(-1)')dnl
-dnl In this case, it is possible to keep the sieve interval in a register
-dnl since the register keeping the root is not needed.
-define(sieve_interval_arg,sieve_interval)dnl
-define(`sieve_interval',root)dnl
function_head(slinie2)
cmpq aux_ptr,aux_ptr_ub
pushq %rbx
@@ -89,11 +85,15 @@
jbe slinie2_ende
subq $8,aux_ptr_ub
slinie2_fbi_loop:
+ movzwq proot_src,auxreg
movzwq prime_src,prime
- movzwq root_src,sieve_ptr
+ movzwq root_src,root
+ subq prime,auxreg
movb log_src,sieve_log
+ movq auxreg,proot
movq sieve_interval,sieve_ptr_ub
- leaq (sieve_interval,sieve_ptr),sieve_ptr
+ movq root,sieve_ptr
+ leaq (sieve_ptr_ub,sieve_ptr),sieve_ptr
addq $n_i,sieve_ptr_ub
movb (sieve_ptr,prime),sv0
subq prime,sieve_ptr_ub
Modified: trunk/src/experimental/lasieve4_64/athlon64/slinie3A.asm
===================================================================
--- trunk/src/experimental/lasieve4_64/athlon64/slinie3A.asm 2009-04-18 03:42:10 UTC (rev 352)
+++ trunk/src/experimental/lasieve4_64/athlon64/slinie3A.asm 2009-04-18 03:58:53 UTC (rev 353)
@@ -85,10 +85,6 @@
popq %rbx
ret
ifelse(j_per_strip,1,`divert`'',`divert(-1)')dnl
-dnl In this case, it is possible to keep the sieve interval in a register
-dnl since the register keeping the root is not needed.
-define(sieve_interval_arg,sieve_interval)dnl
-define(`sieve_interval',root)dnl
function_head(slinie3)
cmpq aux_ptr,aux_ptr_ub
pushq %rbx
@@ -96,11 +92,15 @@
jbe slinie3_ende
subq $8,aux_ptr_ub
slinie3_fbi_loop:
+ movzwq proot_src,auxreg
movzwq prime_src,prime
- movzwq root_src,sieve_ptr
+ movzwq root_src,root
+ subq prime,auxreg
movb log_src,sieve_log
+ movq auxreg,proot
movq sieve_interval,sieve_ptr_ub
- leaq (sieve_interval,sieve_ptr),sieve_ptr
+ movq root,sieve_ptr
+ leaq (sieve_ptr_ub,sieve_ptr),sieve_ptr
addq $n_i,sieve_ptr_ub
movb (sieve_ptr,prime),sv0
subq prime,sieve_ptr_ub
Modified: trunk/src/experimental/lasieve4_64/athlon64/tdslinie1.asm
===================================================================
--- trunk/src/experimental/lasieve4_64/athlon64/tdslinie1.asm 2009-04-18 03:42:10 UTC (rev 352)
+++ trunk/src/experimental/lasieve4_64/athlon64/tdslinie1.asm 2009-04-18 03:58:53 UTC (rev 353)
@@ -111,4 +111,4 @@
movq sieve_ptr,(tds_buffer,auxreg,8)
tdsline1s_next_j`'i:
')
- jmp tdslinie1_nextfbi
\ No newline at end of file
+ jmp tdslinie1_nextfbi
Modified: trunk/src/experimental/lasieve4_64/athlon64/tdslinie2.asm
===================================================================
--- trunk/src/experimental/lasieve4_64/athlon64/tdslinie2.asm 2009-04-18 03:42:10 UTC (rev 352)
+++ trunk/src/experimental/lasieve4_64/athlon64/tdslinie2.asm 2009-04-18 03:58:53 UTC (rev 353)
@@ -127,4 +127,4 @@
movq sieve_ptr,(auxreg)
tdslinie2s_next_j`'i:
')
- jmp tdslinie2_next_fbi
\ No newline at end of file
+ jmp tdslinie2_next_fbi
Modified: trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c
===================================================================
--- trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c 2009-04-18 03:42:10 UTC (rev 352)
+++ trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c 2009-04-18 03:58:53 UTC (rev 353)
@@ -1,6 +1,20 @@
-/*3:*/
-// #line 36 "gnfs-lasieve4e.w"
+/* gnfs-lasieve4e.c
+ By Jens Franke.
+ 6/13/04: Hacked up for use in GGNFS by Chris Monico.
+
+ This program is free software; you can redistribute it and/or
+ modify it under the terms of 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.
+
+ You should have received a copy of the GNU General Public License along
+ with this program; see the file COPYING. If not, write to the Free
+ Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
+ 02111-1307, USA.
+*/
+// #include "lasieve.h"
+
#include <stdio.h>
#include <sys/types.h>
#include <math.h>
@@ -9,6 +23,7 @@
#include <limits.h>
#include <string.h>
#include <time.h>
+#include <ctype.h>
#ifdef LINUX
#include <endian.h>
#endif
@@ -24,9 +39,6 @@
#define TDS_PRIMALITY_TEST TDS_IMMEDIATELY
#endif
-/*:3*//*4:*/
-// #line 61 "gnfs-lasieve4e.w"
-
#include "if.h"
#include "primgen32.h"
#include "asm/32bit.h"
@@ -37,17 +49,11 @@
#include "gmp-aux.h"
#include "lasieve-prepn.h"
-/*:4*//*5:*/
-// #line 75 "gnfs-lasieve4e.w"
-
#define TDS_IMMEDIATELY 0
#define TDS_BIGSS 1
#define TDS_ODDNESS_CLASS 2
#define TDS_SPECIAL_Q 3
-/*:5*//*6:*/
-// #line 82 "gnfs-lasieve4e.w"
-
#define GCD_SIEVE_BOUND 10
#include "asm/siever-config.c"
@@ -61,23 +67,15 @@
u32_t nss= 0,nzss[3]= {0,0,0};
#endif
-static float
-FB_bound[2],sieve_report_multiplier[2];
+static float FB_bound[2],sieve_report_multiplier[2];
static u16_t sieve_min[2],max_primebits[2],max_factorbits[2];
static u32_t*(FB[2]),*(proots[2]),FBsize[2];
-
-/*44:*/
-// #line 1610 "gnfs-lasieve4e.w"
-
static double*(tpoly_f[2]);
#define CANDIDATE_SEARCH_STEPS 128
static unsigned char**(sieve_report_bounds[2]);
static i32_t n_srb_i,n_srb_j;
-/*:44*/
-// #line 102 "gnfs-lasieve4e.w"
-
static u32_t b,first_spq,first_spq1,first_root,last_spq,sieve_count;
static mpz_t m,N,aux1,aux2,aux3,sr_a,sr_b;
@@ -89,7 +87,8 @@
i32_t poldeg[2],poldeg_max;
-u32_t keep_factorbase;
+u32_t keep_factorbase;
+u32_t g_resume;
#define MAX_LPFACTORS 3
static mpz_t rational_rest,algebraic_rest;
mpz_t factors[MAX_LPFACTORS];
@@ -103,49 +102,34 @@
static i64_t tdsi_clock[2]= {0,0},tds1_clock[2]= {0,0},tds2_clock[2]= {0,0};
static i64_t tds3_clock[2]= {0,0},tds4_clock[2]= {0,0};
-char*basename;
-char*input_line= NULL;
+char *base_name;
+char *input_line= NULL;
size_t input_line_alloc= 0;
-/*:6*//*7:*/
-// #line 133 "gnfs-lasieve4e.w"
static u32_t ncand;
static u16_t*cand;
static unsigned char*fss_sv;
+u32_t process_no;
+char *sysload_cmd;
+double sieveStartTime;
-/*:7*//*8:*/
-// #line 139 "gnfs-lasieve4e.w"
-
static int tdcand_cmp(const void*x,const void*y)
{
-return(int)(*((u16_t*)x))-(int)(*((u16_t*)y));
+ return(int)(*((u16_t*)x))-(int)(*((u16_t*)y));
}
-/*:8*//*9:*/
-// #line 155 "gnfs-lasieve4e.w"
-
typedef struct xFBstruct{
-u32_t p,pp,q,qq,r,l;
+ u32_t p,pp,q,qq,r,l;
}*xFBptr;
static volatile xFBptr xFB[2];
static volatile u32_t xFBs[2];
-/*:9*//*10:*/
-// #line 181 "gnfs-lasieve4e.w"
-
static void xFBtranslate(u16_t*rop,xFBptr op);
static int xFBcmp(const void*,const void*);
-/*:10*//*12:*/
-// #line 196 "gnfs-lasieve4e.w"
+static u32_t add_primepowers2xaFB(size_t*aFB_alloc_ptr,u32_t pp_bound,u32_t side,u32_t p,u32_t r);
-static u32_t add_primepowers2xaFB(size_t*aFB_alloc_ptr,
-u32_t pp_bound,u32_t side,u32_t p,u32_t r);
-
-/*:12*//*13:*/
-// #line 214 "gnfs-lasieve4e.w"
-
i32_t a0,a1,b0,b1;
#if 0
u32_t I_bits;
@@ -154,66 +138,36 @@
u32_t root_no;
float sigma;
-/*:13*//*14:*/
-// #line 234 "gnfs-lasieve4e.w"
-
static u32_t oddness_type;
static u32_t n_i,n_j,i_bits,j_bits;
-/*:14*//*15:*/
-// #line 239 "gnfs-lasieve4e.w"
-
-/*18:*/
-// #line 461 "gnfs-lasieve4e.w"
-
u32_t spq_i,spq_j,spq_x;
-/*:18*//*28:*/
-// #line 995 "gnfs-lasieve4e.w"
-
u32_t fbi1[2];
-/*:28*//*29:*/
-// #line 1000 "gnfs-lasieve4e.w"
-
u32_t fbis[2];
-/*:29*//*31:*/
-// #line 1101 "gnfs-lasieve4e.w"
-
static u32_t j_per_strip,jps_bits,jps_mask,n_strips;
static struct schedule_struct{
-u16_t***schedule;
-u32_t*fbi_bounds;
-u32_t n_pieces;
-unsigned char*schedlogs;
-u16_t n_strips,current_strip;
-size_t alloc,alloc1;
-u32_t*ri;
+ u16_t***schedule;
+ u32_t*fbi_bounds;
+ u32_t n_pieces;
+ unsigned char*schedlogs;
+ u16_t n_strips,current_strip;
+ size_t alloc,alloc1;
+ u32_t*ri;
}*(schedules[2]);
u32_t n_schedules[2];
-/*:31*//*32:*/
-// #line 1115 "gnfs-lasieve4e.w"
-
static u32_t*(LPri[2]);
#define RI_SIZE 2
-/*:32*//*33:*/
-// #line 1120 "gnfs-lasieve4e.w"
-
static u32_t*(current_ij[2]);
-/*:33*//*34:*/
-// #line 1127 "gnfs-lasieve4e.w"
-
static size_t sched_alloc[2];
#define SE_SIZE 2
#define SCHEDFBI_MAXSTEP 0x10000
-/*:34*//*38:*/
-// #line 1338 "gnfs-lasieve4e.w"
-
#define USE_MEDSCHED
#ifdef USE_MEDSCHED
static u16_t**(med_sched[2]);
@@ -223,9 +177,6 @@
static u16_t n_medsched_pieces[2];
#endif
-/*:38*//*40:*/
-// #line 1392 "gnfs-lasieve4e.w"
-
static unsigned char*sieve_interval= NULL,*(FB_logs[2]);
static unsigned char*tiny_sieve_buffer;
#define TINY_SIEVEBUFFER_SIZE 420
@@ -233,43 +184,22 @@
static double sieve_multiplier[2],FB_maxlog[2];
static u32_t j_offset;
-/*:40*//*48:*/
-// #line 1682 "gnfs-lasieve4e.w"
-
void do_scheduling(struct schedule_struct*,u32_t,u32_t,u32_t);
-/*:48*//*51:*/
-// #line 1729 "gnfs-lasieve4e.w"
-
static u16_t*(smallsieve_aux[2]),*(smallsieve_auxbound[2][5]);
static u16_t*(smallsieve_tinybound[2]);
-/*:51*//*52:*/
-// #line 1737 "gnfs-lasieve4e.w"
-
static u16_t*(smallsieve_aux1[2]),*(smallsieve_aux1_ub_odd[2]);
static u16_t*(smallsieve_aux1_ub[2]),*(smallsieve_tinybound1[2]);
-/*:52*//*53:*/
-// #line 1744 "gnfs-lasieve4e.w"
-
static u16_t*(smallsieve_aux2[2]),*(smallsieve_aux2_ub[2]);
-/*:53*//*54:*/
-// #line 1761 "gnfs-lasieve4e.w"
-
static u16_t*(smallpsieve_aux[2]),*(smallpsieve_aux_ub_pow1[2]);
static u16_t*(smallpsieve_aux_ub_odd[2]),*(smallpsieve_aux_ub[2]);
static unsigned char*horizontal_sievesums;
-/*:54*//*55:*/
-// #line 1768 "gnfs-lasieve4e.w"
-
static u16_t*(x2FB[2]),x2FBs[2];
-/*:55*//*56:*/
-// #line 1776 "gnfs-lasieve4e.w"
-
static u16_t*tinysieve_curpos;
#ifndef MMX_TD
static u16_t**(smalltdsieve_aux[2]);
@@ -278,46 +208,25 @@
#endif
#endif
-/*:56*//*57:*/
-// #line 1788 "gnfs-lasieve4e.w"
-
#ifdef GCD_SIEVE_BOUND
static u32_t np_gcd_sieve;
static unsigned char*gcd_sieve_buffer;
static void gcd_sieve(void);
#endif
-/*:57*//*95:*/
-// #line 2777 "gnfs-lasieve4e.w"
-
u16_t**schedbuf;
-/*:95*//*103:*/
-// #line 2872 "gnfs-lasieve4e.w"
-
static void store_candidate(u16_t,u16_t,unsigned char);
-/*:103*//*110:*/
-// #line 3050 "gnfs-lasieve4e.w"
-
void trial_divide(void);
-/*:110*//*124:*/
-// #line 3486 "gnfs-lasieve4e.w"
-
#ifndef SCHED_TDS_BUFSIZE
#define SCHED_TDS_BUFSIZE 1024
#endif
u16_t*(sched_tds_buffer[SCHED_TDS_BUFSIZE]);
-/*:124*//*140:*/
-// #line 3896 "gnfs-lasieve4e.w"
-
u32_t*mpz_trialdiv(mpz_t N,u32_t*pbuf,u32_t ncp,char*errmsg);
-/*:140*//*142:*/
-// #line 3949 "gnfs-lasieve4e.w"
-
static void output_tdsurvivor(u32_t*,u32_t*,u32_t*,u32_t*,mpz_t,mpz_t);
static void store_tdsurvivor(u32_t*,u32_t*,u32_t*,u32_t*,mpz_t,mpz_t);
static int primality_tests(void);
@@ -330,9 +239,6 @@
#define MAX_TDS_INCREMENT 1024
#define TDS_FBP_ALLOC_INCREMENT 8192
-/*:142*//*150:*/
-// #line 4229 "gnfs-lasieve4e.w"
-
#if 0
#define OFMT_CWI
#endif
@@ -340,23 +246,11 @@
static char u32_t2cwi(u32_t);
#endif
-/*:150*//*153:*/
-// #line 4263 "gnfs-lasieve4e.w"
-
void dumpsieve(u32_t j_offset,u32_t side);
-/*:153*/
-// #line 240 "gnfs-lasieve4e.w"
-
-/*114:*/
-// #line 3135 "gnfs-lasieve4e.w"
-
u32_t*(td_buf[2]),**td_buf1;
size_t td_buf_alloc[2]= {1024,1024};
-/*:114*//*119:*/
-// #line 3306 "gnfs-lasieve4e.w"
-
static unsigned char tds_coll[UCHAR_MAX];
u32_t**tds_fbi= NULL;
u32_t**tds_fbi_curpos= NULL;
@@ -365,26 +259,20 @@
static size_t tds_fbi_alloc= TDFBI_ALLOC;
#endif
-/*:119*//*138:*/
-// #line 3871 "gnfs-lasieve4e.w"
-
static mpz_t td_rests[L1_SIZE];
static mpz_t large_factors[2],*(large_primes[2]);
static mpz_t FBb_sq[2];
-/*:138*/
-// #line 241 "gnfs-lasieve4e.w"
-
void Usage()
{
-complain("Usage");
+ complain("Usage");
}
static u32_t n_prereports= 0,n_reports= 0,n_rep1= 0,n_rep2= 0;
static u32_t n_tdsurvivors[2]= {0,0};
-static FILE*ofile;
-static char*ofile_name;
+static FILE*g_ofile;
+static char*g_ofile_name;
#ifdef STC_DEBUG
FILE*debugfile;
@@ -413,8 +301,8 @@
static void
terminate_sieving(int signo)
{
-exitval= USER_INTERRUPT_EXITVAL;
-longjmp(termination_jb,USER_INTERRUPT);
+ exitval= USER_INTERRUPT_EXITVAL;
+ longjmp(termination_jb,USER_INTERRUPT);
}
static clock_t last_clock;
@@ -423,2135 +311,1995 @@
extern u64_t MMX_TdNloop;
#endif
-main(int argc,char**argv)
+/*******************************************************/
+double sTime()
+ /*******************************************************/
+#if 0 && !defined (_MSC_VER) && !defined (__MINGW32__) && !defined (MINGW32)
{
-u16_t zip_output,force_aFBcalc;
-u16_t catch_signals;
-u32_t all_spq_done;
-u32_t n_spq,n_spq_discard;
-double tStart, tNow, lastReport;
-char*sysload_cmd;
-u32_t process_no;
-#ifdef STC_DEBUG
-debugfile= fopen("rtdsdebug","w");
+ static struct timeval this_tv;
+ static struct timezone dumbTZ;
+ double t;
+
+ gettimeofday(&this_tv, &dumbTZ);
+ t = this_tv.tv_sec + 0.000001*this_tv.tv_usec;
+ return t;
+}
+#else
+{
+ return clock() / (double)CLOCKS_PER_SEC;
+}
#endif
-/*21:*/
-// #line 562 "gnfs-lasieve4e.w"
+/**************************************************/
+void logTotalTime()
+ /**************************************************/
{
-i32_t option;
-FILE*input_data;
-u32_t i;
+ double t=sTime()-sieveStartTime;
+ FILE *fp=fopen("ggnfs.log", "a");
+
+ fprintf(fp, "\tLatSieveTime: %ld\n", (long)t);
+ fclose(fp);
+}
-sieve_count= U32_MAX;
-ofile_name= NULL;
-zip_output= 0;
-special_q_side= NO_SIDE;
-sigma= 0;
-keep_factorbase= 0;
-basename= NULL;
-first_spq= 0;
-sieve_count= 1;
-force_aFBcalc= 0;
-sysload_cmd= NULL;
-process_no= 0;
-catch_signals= 0;
+/**************************************************/
+int parse_q_from_line(char *buf) {
+ /**************************************************/
+ char *p, *tmp, *next_field;
+ u32_t q, q0, i, side;
+ static int first=0;
+
+ for(p=tmp=buf; *p && isspace(*p); p++);
+ if(!*p) return 0; /* empty line, skip */
+
+ side = (special_q_side == RATIONAL_SIDE) ? 0 : 1;
+ for(i=0; *p; p++) {
+ if(*p == ':') {
+ if(i++ == side) tmp = p; /* we will only scan this section for a q0 */
+ } else if(!(*p == '-' || *p == ',' || isspace(*p) || isxdigit(*p))) {
+ if(first++ == 0) printf(" Warning! some corrupt lines in the original file\n");
+ return -1;
+ }
+ }
+ if(i!=2) {
+ printf(" Warning: an incomplete line in the original file; if just a few, it's ok, they will be skipped\n");
+ return -1; /* must have two ':' some ',' and hexdigits */
+ }
+
+ q0 = first_spq;
+ do {
+ q = strtoul(tmp + 1, &next_field, 16);
+ if(q >= first_spq && q < first_spq+sieve_count)
+ q0 = q;
+ tmp = next_field;
+ } while(tmp[0] == ',' && isxdigit(tmp[1]));
+
+ /* I've seen cases when q0 is not the last reported in the comma-separated list */
+ /* However, the closer it is to the end of the line the more likely it was the true q0 */
+ /* In 99% cases it is the last value, but we don't want to depend on that */
+
+ if(q0 > first_spq && q0 < first_spq+sieve_count) {
+ sieve_count -= (q0 - first_spq);
+ first_spq = q0;
+ }
+ return 1;
+}
-first_psp_side= 2;
-first_mpqs_side= 0;
-J_bits= U32_MAX;
-
+/**************************************************/
+int main(int argc, char **argv)
+ /**************************************************/
+{
+ u16_t zip_output,force_aFBcalc;
+ u16_t catch_signals;
+ u32_t all_spq_done;
+ u32_t n_spq, n_spq_discard;
+ double tStart, tNow, lastReport;
+
+#if defined (_MSC_VER) && defined (_DEBUG)
+ int tmpDbgFlag;
+ tmpDbgFlag = _CrtSetDbgFlag(_CRTDBG_REPORT_FLAG);
+ /*
+ tmpDbgFlag |= _CRTDBG_CHECK_ALWAYS_DF;
+ tmpDbgFlag |= _CRTDBG_CHECK_CRT_DF;
+ tmpDbgFlag |= _CRTDBG_DELAY_FREE_MEM_DF;
+ */
+ tmpDbgFlag |= _CRTDBG_LEAK_CHECK_DF;
+ _CrtSetDbgFlag(tmpDbgFlag);
+#endif
+
+ n_spq = 0;
+ n_spq_discard = 0;
+
+ sieveStartTime = sTime();
+
+#ifdef STC_DEBUG
+ debugfile= fopen("rtdsdebug","w");
+#endif
+ {
+ i32_t option;
+ FILE*input_data;
+ u32_t i;
+
+ sieve_count= U32_MAX;
+ g_ofile_name= NULL;
+ zip_output= 0;
+ special_q_side= NO_SIDE;
+ sigma= 0;
+ keep_factorbase= 0;
+ g_resume= 0;
+ base_name= NULL;
+ first_spq= 0;
+ sieve_count= 1;
+ force_aFBcalc= 0;
+ sysload_cmd= NULL;
+ process_no= 0;
+ catch_signals= 0;
+
+ first_psp_side= 2;
+ first_mpqs_side= 0;
+ J_bits= U32_MAX;
+
#define NumRead(x) if(sscanf(optarg,"%u",&x)!=1) Usage()
#define NumRead16(x) if(sscanf(optarg,"%hu",&x)!=1) Usage()
-
-append_output= 0;
-while((option= getopt(argc,argv,"FJ:L:M:N:P:S:ab:c:f:i:kn:o:q:rt:vz"))!=-1){
-switch(option){
-case'F':
-force_aFBcalc= 1;
-break;
-case'J':
-NumRead(J_bits);
-break;
-case'L':
-sysload_cmd= optarg;
-break;
-case'M':
-NumRead16(first_mpqs_side);
-break;
-case'P':
-NumRead16(first_psp_side);
-break;
-case'S':
-if(sscanf(optarg,"%f",&sigma)!=1){
-errprintf("Cannot read floating point number %s\n",optarg);
-Usage();
-}
-break;
-case'a':
-if(special_q_side!=NO_SIDE){
-errprintf("Ignoring -a\n");
-break;
-}
-special_q_side= ALGEBRAIC_SIDE;
-break;
-case'b':
-if(basename!=NULL)errprintf("Ignoring -b %s\n",basename);
-else basename= optarg;
-break;
-case'c':
-NumRead(sieve_count);
-break;
-case'f':
-if(sscanf(optarg,"%u:%u:%u",&first_spq,&first_spq1,&first_root)!=3){
-if(sscanf(optarg,"%u",&first_spq)==1){
-first_spq1= first_spq;
-first_root= 0;
-}else Usage();
-}else append_output= 1;
-break;
-case'i':
-if(sscanf(optarg,"%hu",&cmdline_first_sieve_side)!=1)
-complain("-i %s ???\n",optarg);
-break;
-case'k':
-keep_factorbase= 1;
-break;
-case'n':
-catch_signals= 1;
-case'N':
-NumRead(process_no);
-break;
-case'o':
-ofile_name= optarg;
-break;
-case'q':
-NumRead16(special_q_side);
-break;
-case'r':
-if(special_q_side!=NO_SIDE){
-errprintf("Ignoring -r\n");
-break;
-}
-special_q_side= RATIONAL_SIDE;
-break;
-case't':
-if(sscanf(optarg,"%hu",&cmdline_first_td_side)!=1)
-complain("-t %s ???\n",optarg);
-break;
-case'v':
-verbose++;
-break;
-case'z':
-zip_output= 1;
-break;
-}
-}
-if(J_bits==U32_MAX)J_bits= I_bits-1;
-if(first_psp_side==2)first_psp_side= first_mpqs_side;
+
+ append_output= 0;
+ while((option= getopt(argc,argv,"FJ:L:M:N:P:RS:ab:c:f:i:kn:o:q:rt:vz"))!=-1) {
+ switch(option) {
+ case 'R':
+ g_resume = 1; break;
+ case'F':
+ force_aFBcalc= 1;
+ break;
+ case'J':
+ NumRead(J_bits);
+ break;
+ case'L':
+ sysload_cmd= optarg;
+ break;
+ case'M':
+ NumRead16(first_mpqs_side);
+ break;
+ case'P':
+ NumRead16(first_psp_side);
+ break;
+ case'S':
+ if(sscanf(optarg,"%f",&sigma)!=1) {
+ errprintf("Cannot read floating point number %s\n",optarg);
+ Usage();
+ }
+ break;
+ case'a':
+ if(special_q_side!=NO_SIDE) {
+ errprintf("Ignoring -a\n");
+ break;
+ }
+ special_q_side= ALGEBRAIC_SIDE;
+ break;
+ case'b':
+ if(base_name!=NULL)errprintf("Ignoring -b %s\n",base_name);
+ else base_name= optarg;
+ break;
+ case'c':
+ NumRead(sieve_count);
+ break;
+ case'f':
+ if(sscanf(optarg,"%u:%u:%u",&first_spq,&first_spq1,&first_root)!=3) {
+ if(sscanf(optarg,"%u",&first_spq) == 1) {
+ first_spq1= first_spq;
+ first_root= 0;
+ } else Usage();
+ } else append_output= 1;
+ break;
+ case'i':
+ if(sscanf(optarg,"%hu",&cmdline_first_sieve_side)!=1)
+ complain("-i %s ???\n",optarg);
+ break;
+ case'k':
+ keep_factorbase= 1;
+ break;
+ case'n':
+ catch_signals= 1;
+ case'N':
+ NumRead(process_no);
+ break;
+ case'o':
+ g_ofile_name= optarg;
+ break;
+ case'q':
+ NumRead16(special_q_side);
+ break;
+ case'r':
+ if(special_q_side!=NO_SIDE) {
+ errprintf("Ignoring -r\n");
+ break;
+ }
+ special_q_side= RATIONAL_SIDE;
+ break;
+ case't':
+ if(sscanf(optarg,"%hu",&cmdline_first_td_side)!=1)
+ complain("-t %s ???\n",optarg);
+ break;
+ case'v':
+ verbose++;
+ break;
+ case'z':
+ zip_output= 1;
+ break;
+ }
+ }
+
+#define LINE_BUF_SIZE 300
+
+ if (g_resume != 0) {
+ char buf[LINE_BUF_SIZE];
+ int ret;
+
+ if (zip_output != 0)
+ complain("Cannot resume gzipped file. gunzip, and retry without -z\n");
+ if (g_ofile_name == NULL)
+ complain("Cannot resume without the file name\n");
+ if (strcmp(g_ofile_name, "-") == 0)
+ complain("Cannot resume using stdout\n");
+ if ((g_ofile = fopen(g_ofile_name, "ab+")) == NULL)
+ complain("Cannot open %s for append: %m\n", g_ofile_name);
+ while(fgets(buf, LINE_BUF_SIZE, g_ofile)) {
+ ret = parse_q_from_line(buf);
+ }
+ if(ret < 0) fprintf(g_ofile, "\n"); /* encapsulating the last incomplete line */
+ printf(" Resuming with -f %d -c %d\n", first_spq, sieve_count);
+ first_spq1 = first_spq;
+ }
+
+ if(J_bits == U32_MAX)
+ J_bits = I_bits-1;
+ if(first_psp_side == 2)
+ first_psp_side = first_mpqs_side;
#ifndef I_bits
#error Must #define I_bits
#endif
-
-if(optind<argc&&basename==NULL){
-basename= argv[optind];
-optind++;
-}
-if(optind<argc)fprintf(stderr,"Ignoring %u trailing command line args\n",
-argc-optind);
-if(basename==NULL)basename= "gnfs";
-if((input_data= fopen(basename,"r"))==NULL){
-complain("Cannot open %s for input of nfs polynomials: %m\n",basename);
-}
-mpz_init(N);
-mpz_init(m);
-mpz_init(aux1);
-mpz_init(aux2);
-mpz_init(aux3);
-mpz_init(sr_a);
-mpz_init(sr_b);
-mpz_ull_init();
-mpz_init(rational_rest);
-mpz_init(algebraic_rest);
-//input_poly(N,poly,poldeg,poly+1,poldeg+1,m,input_data);
+
+ if(optind<argc&&base_name == NULL) {
+ base_name= argv[optind];
+ optind++;
+ }
+ if(optind<argc)fprintf(stderr,"Ignoring %u trailing command line args\n",
+ argc-optind);
+ if(base_name == NULL)base_name= "gnfs";
+ if((input_data= fopen(base_name,"r")) == NULL) {
+ complain("Cannot open %s for input of nfs polynomials: %m\n",base_name);
+ }
+ mpz_init(N);
+ mpz_init(m);
+ mpz_init(aux1);
+ mpz_init(aux2);
+ mpz_init(aux3);
+ mpz_init(sr_a);
+ mpz_init(sr_b);
+ mpz_ull_init();
+ mpz_init(rational_rest);
+ mpz_init(algebraic_rest);
+ //input_poly(N,poly,poldeg,poly+1,poldeg+1,m,input_data);
#if 0
-if(poldeg[1]> 1){
-if(poldeg[0]==1){
-mpz_t*X;
-poldeg[0]= poldeg[1];
-poldeg[1]= 1;
-X= poly[0];
-poly[0]= poly[1];
-poly[1]= X;
-}else{
-complain("Degrees >1 on both sides not implemented\n");
-}
-}
+ if(poldeg[1]> 1) {
+ if(poldeg[0] == 1) {
+ mpz_t*X;
+ poldeg[0]= poldeg[1];
+ poldeg[1]= 1;
+ X= poly[0];
+ poly[0]= poly[1];
+ poly[1]= X;
+ } else {
+ complain("Degrees >1 on both sides not implemented\n");
+ }
+ }
#endif
-//skip_blanks_comments(&input_line,&input_line_alloc,input_data);
-close(input_data);
- { FILE *fp;
- char token[256], value[512], thisLine[1024];
-
-
- sieve_min[0] = sieve_min[1]=0;
-
- if (!(fp = fopen(basename, "rb"))) {
- printf("Error opening %s for read!\n", basename);
- return -1;
- }
- input_poly(N, poly, poldeg, poly + 1, poldeg + 1, m, fp);
- rewind(fp);
- while (!(feof(fp))) {
- thisLine[0] = 0;
- fgets(thisLine, 1023, fp);
- /* Special case: If there's a polynomial, handle it seperately: */
- if (strncmp(thisLine, "START_POLY", 10)==0) {
- while (!(feof(fp)) && strncmp(thisLine, "END_POLY", 8))
- fgets(thisLine, 1023, fp);
- } else if ((sscanf(thisLine, "%255s %511s", token, value)==2) &&
- (thisLine[0] != '#')) {
-
- token[sizeof(token)-1] = 0;
- if (strncmp(token, "skew:", 5)==0) {
- sigma = (float)atof(value);
- } else if (strncmp(token, "q0:", 3)==0) {
- first_spq = atol(value);
- first_spq1= first_spq;
- first_root= 0;
-
- } else if (strncmp(token, "qintsize:", 9)==0) {
- sieve_count = atol(value);
- } else if ((strncmp(token, "skip0:", 6)==0) ||
- (strncmp(token, "askip:", 6)==0)) {
- sieve_min[0] = atol(value);
- } else if ((strncmp(token, "skip1:", 6)==0) ||
- (strncmp(token, "rskip:", 6)==0)) {
- sieve_min[1] = atol(value);
- } else if ((strncmp(token, "lim0:", 5)==0) ||
- (strncmp(token, "alim:", 5)==0)) {
- FB_bound[0] = (float)atol(value);
- } else if ((strncmp(token, "lim1:", 5)==0)||
- (strncmp(token, "rlim:", 5)==0)) {
- FB_bound[1] = (float)atof(value);
- } else if ((strncmp(token, "lpb0:", 5)==0) ||
- (strncmp(token, "lpba:", 5)==0)) {
- max_primebits[0] = atoi(value);
- } else if ((strncmp(token, "lpb1:", 5)==0) ||
- (strncmp(token, "lpbr:", 5)==0)) {
- max_primebits[1] = atoi(value);
- } else if ((strncmp(token, "mfb0:", 5)==0) ||
- (strncmp(token, "mfba:", 5)==0)) {
- max_factorbits[0] = atoi(value);
- } else if ((strncmp(token, "mfb1:", 5)==0) ||
- (strncmp(token, "mfbr:", 5)==0)) {
- value[sizeof(value)-1] = 0;
- max_factorbits[1] = atoi(value);
- } else if ((strncmp(token, "lambda0:", 8)==0) ||
- (strncmp(token, "alambda:", 8)==0)) {
- sieve_report_multiplier[0] = (float)atof(value);
- } else if ((strncmp(token, "lambda1:", 8)==0) ||
- (strncmp(token, "rlambda:", 8)==0)) {
- sieve_report_multiplier[1] = (float)atof(value);
- }
+ //skip_blanks_comments(&input_line,&input_line_alloc,input_data);
+ close(input_data);
+ { FILE *fp;
+ char token[256], value[512], thisLine[1024];
+
+
+ sieve_min[0] = sieve_min[1]=0;
+
+ if (!(fp = fopen(base_name, "rb"))) {
+ printf("Error opening %s for read!\n", base_name);
+ return -1;
+ }
+ input_poly(N, poly, poldeg, poly + 1, poldeg + 1, m, fp);
+ rewind(fp);
+ while (!(feof(fp))) {
+ thisLine[0] = 0;
+ fgets(thisLine, 1023, fp);
+ /* Special case: If there's a polynomial, handle it seperately: */
+ if (strncmp(thisLine, "START_POLY", 10) == 0) {
+ while (!(feof(fp)) && strncmp(thisLine, "END_POLY", 8))
+ fgets(thisLine, 1023, fp);
+ } else if ((sscanf(thisLine, "%255s %511s", token, value) == 2) &&
+ (thisLine[0] != '#')) {
+
+ token[sizeof(token)-1] = 0;
+ if (strncmp(token, "skew:", 5) == 0) {
+ sigma = (float)atof(value);
+ } else if (strncmp(token, "q0:", 3) == 0) {
+ first_spq = atol(value);
+ first_spq1= first_spq;
+ first_root= 0;
+
+ } else if (strncmp(token, "qintsize:", 9) == 0) {
+ sieve_count = atol(value);
+ } else if ((strncmp(token, "skip0:", 6) == 0) ||
+ (strncmp(token, "askip:", 6) == 0)) {
+ sieve_min[0] = atol(value);
+ } else if ((strncmp(token, "skip1:", 6) == 0) ||
+ (strncmp(token, "rskip:", 6) == 0)) {
+ sieve_min[1] = atol(value);
+ } else if ((strncmp(token, "lim0:", 5) == 0) ||
+ (strncmp(token, "alim:", 5) == 0)) {
+ FB_bound[0] = (float)atol(value);
+ } else if ((strncmp(token, "lim1:", 5) == 0)||
+ (strncmp(token, "rlim:", 5) == 0)) {
+ FB_bound[1] = (float)atof(value);
+ } else if ((strncmp(token, "lpb0:", 5) == 0) ||
+ (strncmp(token, "lpba:", 5) == 0)) {
+ max_primebits[0] = atoi(value);
+ } else if ((strncmp(token, "lpb1:", 5) == 0) ||
+ (strncmp(token, "lpbr:", 5) == 0)) {
+ max_primebits[1] = atoi(value);
+ } else if ((strncmp(token, "mfb0:", 5) == 0) ||
+ (strncmp(token, "mfba:", 5) == 0)) {
+ max_factorbits[0] = atoi(value);
+ } else if ((strncmp(token, "mfb1:", 5) == 0) ||
+ (strncmp(token, "mfbr:", 5) == 0)) {
+ value[sizeof(value)-1] = 0;
+ max_factorbits[1] = atoi(value);
+ } else if ((strncmp(token, "lambda0:", 8) == 0) ||
+ (strncmp(token, "alambda:", 8) == 0)) {
+ sieve_report_multiplier[0] = (float)atof(value);
+ } else if ((strncmp(token, "lambda1:", 8) == 0) ||
+ (strncmp(token, "rlambda:", 8) == 0)) {
+ sieve_report_multiplier[1] = (float)atof(value);
+ }
#ifdef _NO
- else {
- printf("Warning: Ignoring input line:\n%s\n", thisLine);
- }
+ else {
+ printf("Warning: Ignoring input line:\n%s\n", thisLine);
+ }
#endif
- }
- }
- fclose(fp);
- }
- last_spq= first_spq+sieve_count;
-if(last_spq>=I32_MAX/2){
-complain("Cannot handle special q >= %d\n",I32_MAX/2);
-}
-for(i= 0;i<2;i++){
-if(FB_bound[i]<4||sieve_report_multiplier[i]<=0){
-complain("Please set all bounds to reasonable values!\n");
-}
+ }
+ }
+ fclose(fp);
+ }
+ last_spq= first_spq+sieve_count;
+ if(last_spq>=I32_MAX/2) {
+ complain("Cannot handle special q >= %d\n",I32_MAX/2);
+ }
+ for(i= 0;i<2;i++) {
+ if(FB_bound[i]<4||sieve_report_multiplier[i]<=0) {
+ complain("Please set all bounds to reasonable values!\n");
+ }
#if 1
-if(max_primebits[i]> 33){
-complain("Only large primes up to 33 bits are allowed.\n");
-}
+ if(max_primebits[i]> 33) {
+ complain("Only large primes up to 33 bits are allowed.\n");
+ }
#endif
-}
-if(sieve_count!=0){
-if(sigma==0)complain("Please set a skewness\n");
-if(special_q_side==NO_SIDE){
-errprintf("Please use -a or -r\n");
-Usage();
-}
-if(FB_bound[special_q_side]> first_spq){
-FB_bound[special_q_side]=(float) first_spq-1;
-printf("Warning: lowering FB_bound to %u.\n",first_spq-1);
-//complain("Special q lower bound %u below rFB bound %g\n",
-//first_spq,FB_bound[special_q_side]);
-}
-}
-fclose(input_data);
-if(poldeg[0]<poldeg[1])poldeg_max= poldeg[1];
-else poldeg_max= poldeg[0];
-
-
-
-i_shift= 1<<(I_bits-1);
-n_I= 1<<I_bits;
-n_i= n_I/2;
-i_bits= I_bits-1;
-n_J= 1<<J_bits;
-n_j= n_J/2;
-j_bits= J_bits-1;
-/*22:*/
-// #line 775 "gnfs-lasieve4e.w"
-
-{
-u32_t i,j;
-double x,y,z;
-
-x= sqrt(first_spq*sigma)*n_I;
-y= x/sigma;
-for(j= 0;j<2;j++){
-poly_f[j]= xmalloc((poldeg[j]+1)*sizeof(*poly_f[j]));
-
-for(i= 0,z= 1,poly_norm[j]= 0;
-i<=poldeg[j];i++){
-poly_f[j][i]= mpz_get_d(poly[j][i]);
-poly_norm[j]= poly_norm[j]*y+fabs(poly_f[j][i])*z;
-z*= x;
-}
-}
-}
-
-/*:22*/
-// #line 771 "gnfs-lasieve4e.w"
-
-}
-
-/*:21*/
-// #line 301 "gnfs-lasieve4e.w"
-
-siever_init();
-/*23:*/
-// #line 795 "gnfs-lasieve4e.w"
-
-if(sieve_count!=0){
-
-if(ofile_name==NULL){
-if(zip_output==0){
-asprintf(&ofile_name,"%s.lasieve-%u.%u-%u",basename,
-special_q_side,first_spq,last_spq);
-}else{
-asprintf(&ofile_name,
-append_output==0?
-"gzip --best --stdout > %s.lasieve-%u.%u-%u.gz":
-"gzip --best --stdout >> %s.lasieve-%u.%u-%u.gz",
-basename,special_q_side,first_spq,last_spq);
-}
-}else{
-if(strcmp(ofile_name,"-")==0){
-if(zip_output==0){
-ofile= stdout;
-ofile_name= "to stdout";
-goto done_opening_output;
-}else ofile_name= "gzip --best --stdout";
-}else{
-if(fnmatch("*.gz",ofile_name,0)==0){
-char*on1;
-
-zip_output= 1;
-on1= strdup(ofile_name);
-asprintf(&ofile_name,"gzip --best --stdout > %s",on1);
-free(on1);
-}else zip_output= 0;
-}
-}
-if(zip_output==0){
-if(append_output> 0){
-ofile= fopen(ofile_name,"a");
-}else ofile= fopen(ofile_name,"w");
-if(ofile==NULL)complain("Cannot open %s for output: %m\n",ofile_name);
-}else{
-if((ofile= popen(ofile_name,"w"))==NULL)
-complain("Cannot exec %s for output: %m\n",ofile_name);
-}
-done_opening_output:
-fprintf(ofile,"");
-}
-
-/*:23*/
-// #line 303 "gnfs-lasieve4e.w"
-
-/*24:*/
-// #line 845 "gnfs-lasieve4e.w"
-
-{
-size_t FBS_alloc= 4096;
-u32_t prime;
-pr32_struct ps;
-char*afbname;
-FILE*afbfile;
-u32_t side;
-
-initprime32(&ps);
-
-for(side= 0;side<2;side++){
-if(poldeg[side]==1){
-FB[side]= xmalloc(FBS_alloc*sizeof(u32_t));
-proots[side]= xmalloc(FBS_alloc*sizeof(u32_t));
-prime= firstprime32(&ps);
-for(prime= nextprime32(&ps),fbi1[side]= 0,FBsize[side]= 0;
-prime<FB_bound[side];prime= nextprime32(&ps)){
-u32_t x;
-x= mpz_fdiv_ui(poly[side][1],prime);
-if(x> 0){
-modulo32= prime;
-x= modmul32(modinv32(x),mpz_fdiv_ui(poly[side][0],prime));
-x= x> 0?prime-x:0;
-}else x= prime;
-if(prime<L1_SIZE)fbi1[side]= FBsize[side];
-if(prime<n_i)fbis[side]= FBsize[side];
-if(FBsize[side]==FBS_alloc){
-FBS_alloc*= 2;
-FB[side]= xrealloc(FB[side],FBS_alloc*sizeof(u32_t));
-proots[side]= xrealloc(proots[side],FBS_alloc*sizeof(u32_t));
-}
-proots[side][FBsize[side]]= x;
-FB[side][FBsize[side]++]= prime;
-}
-proots[side]= xrealloc(proots[side],FBsize[side]*sizeof(u32_t));
-FB[side]= xrealloc(FB[side],FBsize[side]*sizeof(u32_t));
-fbi1[side]++;
-fbis[side]++;
-if(fbi1[side]<fbis[side])fbi1[side]= fbis[side];
-}else{
-u32_t j,k,l;
-asprintf(&afbname,"%s.afb.%u",basename,side);
-if(force_aFBcalc> 0||(afbfile= fopen(afbname,"r"))==NULL){
-/*25:*/
-// #line 930 "gnfs-lasieve4e.w"
-
-u32_t*root_buffer;
-size_t aFB_alloc;
-
-root_buffer= xmalloc(poldeg[side]*sizeof(*root_buffer));
-aFB_alloc= 4096;
-FB[side]= xmalloc(aFB_alloc*sizeof(**FB));
-proots[side]= xmalloc(aFB_alloc*sizeof(**proots));
-for(prime= firstprime32(&ps),FBsize[side]= 0;
-prime<FB_bound[side];prime= nextprime32(&ps)){
-u32_t i,nr;
-
-nr= root_finder(root_buffer,poly[side],poldeg[side],prime);
-for(i= 0;i<nr;i++){
-if(aFB_alloc<=FBsize[side]){
-aFB_alloc*= 2;
-FB[side]= xrealloc(FB[side],aFB_alloc*sizeof(**FB));
-proots[side]= xrealloc(proots[side],aFB_alloc*sizeof(**proots));
-}
-FB[side][FBsize[side]]= prime;
-proots[side][FBsize[side]]= root_buffer[i];
-if(prime> 2)FBsize[side]++;
-}
-}
-FB[side]= xrealloc(FB[side],FBsize[side]*sizeof(**FB));
-proots[side]= xrealloc(proots[side],FBsize[side]*sizeof(**proots));
-free(root_buffer);
-
-/*:25*/
-// #line 889 "gnfs-lasieve4e.w"
-
-if(keep_factorbase> 0)/*27:*/
-// #line 975 "gnfs-lasieve4e.w"
-
-{
-if((afbfile= fopen(afbname,"w"))==NULL){
-complain("Cannot open %s for output of aFB: %m\n",afbname);
-}
-if(write_u32(afbfile,&(FBsize[side]),1)!=1){
-complain("Cannot write aFBsize to %s: %m\n",afbname);
-}
-if(write_u32(afbfile,FB[side],FBsize[side])!=FBsize[side]||
-write_u32(afbfile,proots[side],FBsize[side])!=FBsize[side]){
-complain("Cannot write aFB to %s: %m\n",afbname);
-}
-if(write_u32(afbfile,&xFBs[side],1)!=1){
-complain("Cannot write aFBsize to %s: %m\n",afbname);
-}
-fclose(afbfile);
-}
-
-/*:27*/
-// #line 890 "gnfs-lasieve4e.w"
-
-}else{
-/*26:*/
-// #line 959 "gnfs-lasieve4e.w"
-
-if(read_u32(afbfile,&(FBsize[side]),1)!=1){
-complain("Cannot read aFB size from %s: %m\n",afbname);
-}
-FB[side]= xmalloc(FBsize[side]*sizeof(u32_t));
-proots[side]= xmalloc(FBsize[side]*sizeof(u32_t));
-if(read_u32(afbfile,FB[side],FBsize[side])!=FBsize[side]||
-read_u32(afbfile,proots[side],FBsize[side])!=FBsize[side]){
-complain("Cannot read aFB from %s: %m\n",afbname);
-}
-if(read_u32(afbfile,&xFBs[side],1)!=1){
-complain("%s: Cannot read xFBsize\n",afbname);
-}
-fclose(afbfile);
-
-/*:26*/
-// #line 892 "gnfs-lasieve4e.w"
-
-}
-
-
-for(j= 0,k= 0,l= 0;j<FBsize[side];j++){
-if(FB[side][j]<L1_SIZE)k= j;
-if(FB[side][j]<n_i)l= j;
-if(FB[side][j]> L1_SIZE&&FB[side][j]> n_I)break;
-}
-if(FBsize[side]> 0){
-if(k<l)k= l;
-fbis[side]= l+1;
-fbi1[side]= k+1;
-}else{
-fbis[side]= 0;
-fbi1[side]= 0;
-}
-}
-}
-
-{
-u32_t i,srfbs,safbs;
-
-for(i= 0,srfbs= 0;i<xFBs[1];i++){
-if(xFB[1][i].p==xFB[1][i].pp)srfbs++;
-}
-for(i= 0,safbs= 0;i<xFBs[0];i++){
-if(xFB[0][i].p==xFB[0][i].pp)safbs++;
-}
-logbook(0,"FBsize %u+%u (deg %u), %u+%u (deg %u)\n",
-FBsize[0],safbs,poldeg[0],FBsize[1],srfbs,poldeg[1]);
-}
-free(afbname);
-
-/*45:*/
-// #line 1617 "gnfs-lasieve4e.w"
-
-{
-u32_t i;
-size_t si,sj;
-
-n_srb_i= 2*((n_i+2*CANDIDATE_SEARCH_STEPS-1)/(2*CANDIDATE_SEARCH_STEPS));
-n_srb_j= (n_J+2*CANDIDATE_SEARCH_STEPS-1)/(2*CANDIDATE_SEARCH_STEPS);
-sj= n_srb_j*sizeof(*(sieve_report_bounds[0]));
-si= n_srb_i*sizeof(**(sieve_report_bounds[0]));
-for(i= 0;i<2;i++){
-u32_t j;
-
-tpoly_f[i]= xmalloc((1+poldeg[i])*sizeof(**tpoly_f));
-sieve_report_bounds[i]= xmalloc(sj);
-for(j= 0;j<n_srb_j;j++)
-sieve_report_bounds[i][j]= xmalloc(si);
-}
-}
-
-/*:45*/
-// #line 926 "gnfs-lasieve4e.w"
-
-}
-
-/*:24*/
-// #line 304 "gnfs-lasieve4e.w"
-
-if(sieve_count==0)exit(0);
-/*30:*/
-// #line 1004 "gnfs-lasieve4e.w"
-
-{
-u32_t side,i;
-
-for(side= 0;side<2;side++){
-u32_t prime,nr;
-struct xFBstruct*s;
-u32_t*root_buffer;
-size_t xaFB_alloc= 0;
-FB_logs[side]= xmalloc(FBsize[side]);
-sieve_multiplier[side]= (UCHAR_MAX-50)/log(poly_norm[side]);
-
-root_buffer= xmalloc(poldeg[side]*sizeof(*root_buffer));
-prime= 2;
-nr= root_finder(root_buffer,poly[side],poldeg[side],prime);
-
-for(i= 0;i<nr;i++){
-adjust_bufsize((void**)&(xFB[side]),&xaFB_alloc,1+xFBs[side],
-16,sizeof(**xFB));
-s= xFB[side]+xFBs[side];
-s->p= prime;
-s->pp= prime;
-if(root_buffer[i]==prime){
-s->qq= prime;
-s->q= 1;
-s->r= 1;
-}else{
-s->qq= 1;
-s->q= prime;
-s->r= root_buffer[i];
-}
-xFBs[side]++;
-add_primepowers2xaFB(&xaFB_alloc,n_I,side,0,0);
-}
-free(root_buffer);
-for(i= 0;i<FBsize[side];i++){
-double l;
-u32_t l1;
-
-prime= FB[side][i];
-if(prime> n_I/prime)break;
-l= log(prime);
-l1= add_primepowers2xaFB(&xaFB_alloc,n_I,side,prime,proots[side][i]);
-FB_logs[side][i]= rint(l1*l*sieve_multiplier[side]);
-}
-while(i<FBsize[side]){
-double l;
-
-l= log(FB[side][i]);
-if(l> FB_maxlog[side])FB_maxlog[side]= l;
-FB_logs[side][i++]= rint(sieve_multiplier[side]*l);
-}
-FB_maxlog[side]*= sieve_multiplier[side];
-qsort(xFB[side],xFBs[side],sizeof(*(xFB[side])),xFBcmp);
-}
-}
-
-/*:30*/
-// #line 306 "gnfs-lasieve4e.w"
-
-/*35:*/
-// #line 1133 "gnfs-lasieve4e.w"
-
-sieve_interval= xvalloc(L1_SIZE);
-cand= xvalloc(L1_SIZE*sizeof(*cand));
-fss_sv= xvalloc(L1_SIZE);
-tiny_sieve_buffer= xmalloc(TINY_SIEVEBUFFER_SIZE);
-if(n_i> L1_SIZE)
-complain("Strip length %u exceeds L1 size %u\n",n_i,L1_SIZE);
-j_per_strip= L1_SIZE/n_i;
-jps_bits= L1_BITS-i_bits;
-jps_mask= j_per_strip-1;
-if(j_per_strip!=1<<jps_bits)
-Schlendrian("Expected %u j per strip, calculated %u\n",
-j_per_strip,1<<jps_bits);
-n_strips= n_j>>(L1_BITS-i_bits);
-rec_info_init(n_i,n_j);
-/*58:*/
-// #line 1796 "gnfs-lasieve4e.w"
-
-{
-u32_t s;
+ }
+ if(sieve_count!=0) {
+ if(sigma == 0)complain("Please set a skewness\n");
+ if(special_q_side == NO_SIDE) {
+ errprintf("Please use -a or -r\n");
+ Usage();
+ }
+ if(FB_bound[special_q_side]> first_spq) {
+ FB_bound[special_q_side]=(float) first_spq-1;
+ printf("Warning: lowering FB_bound to %u.\n",first_spq-1);
+ //complain("Special q lower bound %u below rFB bound %g\n",
+ //first_spq,FB_bound[special_q_side]);
+ }
+ }
+ fclose(input_data);
+ if(poldeg[0]<poldeg[1])poldeg_max= poldeg[1];
+ else poldeg_max= poldeg[0];
+
+
+
+ i_shift= 1<<(I_bits-1);
+ n_I= 1<<I_bits;
+ n_i= n_I/2;
+ i_bits= I_bits-1;
+ n_J= 1<<J_bits;
+ n_j= n_J/2;
+ j_bits= J_bits-1;
+ {
+ u32_t i,j;
+ double x,y,z;
+
+ x= sqrt(first_spq*sigma)*n_I;
+ y= x/sigma;
+ for(j= 0;j<2;j++) {
+ poly_f[j]= xmalloc((poldeg[j]+1)*sizeof(*poly_f[j]));
+
+ for(i= 0,z= 1,poly_norm[j]= 0;
+ i<=poldeg[j];i++) {
+ poly_f[j][i]= mpz_get_d(poly[j][i]);
+ poly_norm[j]= poly_norm[j]*y+fabs(poly_f[j][i])*z;
+ z*= x;
+ }
+ }
+ }
+
+ }
+
+ siever_init();
+ if(sieve_count!=0) {
+ if(g_ofile_name == NULL) {
+ if(zip_output == 0) {
+ asprintf(&g_ofile_name,"%s.lasieve-%u.%u-%u",base_name,
+ special_q_side,first_spq,last_spq);
+ } else {
+ asprintf(&g_ofile_name,
+ append_output == 0?
+ "gzip --best --stdout > %s.lasieve-%u.%u-%u.gz":
+ "gzip --best --stdout >> %s.lasieve-%u.%u-%u.gz",
+ base_name,special_q_side,first_spq,last_spq);
+ }
+ } else {
+ if(strcmp(g_ofile_name,"-") == 0) {
+ if(zip_output == 0) {
+ g_ofile= stdout;
+ g_ofile_name= "to stdout";
+ goto done_opening_output;
+ } else g_ofile_name= "gzip --best --stdout";
+ } else {
+ if(fnmatch("*.gz",g_ofile_name,0) == 0) {
+ char*on1;
+
+ zip_output= 1;
+ on1= strdup(g_ofile_name);
+ asprintf(&g_ofile_name,"gzip --best --stdout > %s",on1);
+ free(on1);
+ } else zip_output= 0;
+ }
+ }
+ if(zip_output == 0) {
+ if (g_resume != 0) {
+ goto done_opening_output;
+ }
+ if(append_output> 0) {
+ g_ofile= fopen(g_ofile_name,"a");
+ } else {
+ if ((g_ofile = fopen(g_ofile_name, "r")) != NULL)
+ complain(" Will not overwrite existing file %s for output; rename it, move it away, or use -R option (resume)\n", g_ofile_name);
+ g_ofile= fopen(g_ofile_name,"w");
+ }
+ if(g_ofile == NULL)complain("Cannot open %s for output: %m\n",g_ofile_name);
+ } else {
+ if ((g_ofile= popen(g_ofile_name,"w")) == NULL)
+ complain("Cannot exec %s for output: %m\n",g_ofile_name);
+ }
+ done_opening_output:
+ fprintf(g_ofile,"");
+ }
+
+ {
+ size_t FBS_alloc= 4096;
+ u32_t prime;
+ pr32_struct ps;
+ char*afbname;
+ FILE*afbfile;
+ u32_t side;
+
+ initprime32(&ps);
+
+ for(side= 0;side<2;side++) {
+ if(poldeg[side] == 1) {
+ FB[side]= xmalloc(FBS_alloc*sizeof(u32_t));
+ proots[side]= xmalloc(FBS_alloc*sizeof(u32_t));
+ prime= firstprime32(&ps);
+ for(prime= nextprime32(&ps),fbi1[side]= 0,FBsize[side]= 0;
+ prime<FB_bound[side];prime= nextprime32(&ps)) {
+ u32_t x;
+ x= mpz_fdiv_ui(poly[side][1],prime);
+ if(x> 0) {
+ modulo32= prime;
+ x= modmul32(modinv32(x),mpz_fdiv_ui(poly[side][0],prime));
+ x= x> 0?prime-x:0;
+ } else x= prime;
+ if(prime<L1_SIZE)fbi1[side]= FBsize[side];
+ if(prime<n_i)fbis[side]= FBsize[side];
+ if(FBsize[side] == FBS_alloc) {
+ FBS_alloc*= 2;
+ FB[side]= xrealloc(FB[side],FBS_alloc*sizeof(u32_t));
+ proots[side]= xrealloc(proots[side],FBS_alloc*sizeof(u32_t));
+ }
+ proots[side][FBsize[side]]= x;
+ FB[side][FBsize[side]++]= prime;
+ }
+ proots[side]= xrealloc(proots[side],FBsize[side]*sizeof(u32_t));
+ FB[side]= xrealloc(FB[side],FBsize[side]*sizeof(u32_t));
+ fbi1[side]++;
+ fbis[side]++;
+ if(fbi1[side]<fbis[side])fbi1[side]= fbis[side];
+ } else {
+ u32_t j,k,l;
+ asprintf(&afbname,"%s.afb.%u",base_name,side);
+ if(force_aFBcalc> 0||(afbfile= fopen(afbname,"r")) == NULL) {
+ u32_t*root_buffer;
+ size_t aFB_alloc;
+
+ root_buffer= xmalloc(poldeg[side]*sizeof(*root_buffer));
+ aFB_alloc= 4096;
+ FB[side]= xmalloc(aFB_alloc*sizeof(**FB));
+ proots[side]= xmalloc(aFB_alloc*sizeof(**proots));
+ for(prime= firstprime32(&ps),FBsize[side]= 0;
+ prime<FB_bound[side];prime= nextprime32(&ps)) {
+ u32_t i,nr;
+
+ nr= root_finder(root_buffer,poly[side],poldeg[side],prime);
+ for(i= 0;i<nr;i++) {
+ if(aFB_alloc<=FBsize[side]) {
+ aFB_alloc*= 2;
+ FB[side]= xrealloc(FB[side],aFB_alloc*sizeof(**FB));
+ proots[side]= xrealloc(proots[side],aFB_alloc*sizeof(**proots));
+ }
+ FB[side][FBsize[side]]= prime;
+ proots[side][FBsize[side]]= root_buffer[i];
+ if(prime> 2)FBsize[side]++;
+ }
+ }
+ FB[side]= xrealloc(FB[side],FBsize[side]*sizeof(**FB));
+ proots[side]= xrealloc(proots[side],FBsize[side]*sizeof(**proots));
+ free(root_buffer);
+
+ if(keep_factorbase> 0) {
+ if((afbfile= fopen(afbname,"w")) == NULL) {
+ complain("Cannot open %s for output of aFB: %m\n",afbname);
+ }
+ if(write_u32(afbfile,&(FBsize[side]),1)!=1) {
+ complain("Cannot write aFBsize to %s: %m\n",afbname);
+ }
+ if(write_u32(afbfile,FB[side],FBsize[side])!=FBsize[side]||
+ write_u32(afbfile,proots[side],FBsize[side])!=FBsize[side]) {
+ complain("Cannot write aFB to %s: %m\n",afbname);
+ }
+ if(write_u32(afbfile,&xFBs[side],1)!=1) {
+ complain("Cannot write aFBsize to %s: %m\n",afbname);
+ }
+ fclose(afbfile);
+ }
+
+ } else {
+ if(read_u32(afbfile,&(FBsize[side]),1)!=1) {
+ complain("Cannot read aFB size from %s: %m\n",afbname);
+ }
+ FB[side]= xmalloc(FBsize[side]*sizeof(u32_t));
+ proots[side]= xmalloc(FBsize[side]*sizeof(u32_t));
+ if(read_u32(afbfile,FB[side],FBsize[side])!=FBsize[side]||
+ read_u32(afbfile,proots[side],FBsize[side])!=FBsize[side]) {
+ complain("Cannot read aFB from %s: %m\n",afbname);
+ }
+ if(read_u32(afbfile,&xFBs[side],1)!=1) {
+ complain("%s: Cannot read xFBsize\n",afbname);
+ }
+ fclose(afbfile);
+
+ }
+
+
+ for(j= 0,k= 0,l= 0;j<FBsize[side];j++) {
+ if(FB[side][j]<L1_SIZE)k= j;
+ if(FB[side][j]<n_i)l= j;
+ if(FB[side][j]> L1_SIZE&&FB[side][j]> n_I)break;
+ }
+ if(FBsize[side]> 0) {
+ if(k<l)k= l;
+ fbis[side]= l+1;
+ fbi1[side]= k+1;
+ } else {
+ fbis[side]= 0;
+ fbi1[side]= 0;
+ }
+ }
+ }
+
+ {
+ u32_t i,srfbs,safbs;
+
+ for(i= 0,srfbs= 0;i<xFBs[1];i++) {
+ if(xFB[1][i].p == xFB[1][i].pp)srfbs++;
+ }
+ for(i= 0,safbs= 0;i<xFBs[0];i++) {
+ if(xFB[0][i].p == xFB[0][i].pp)safbs++;
+ }
+ logbook(0,"FBsize %u+%u (deg %u), %u+%u (deg %u)\n",
+ FBsize[0],safbs,poldeg[0],FBsize[1],srfbs,poldeg[1]);
+ }
+ free(afbname);
+
+ {
+ u32_t i;
+ size_t si,sj;
+
+ n_srb_i= 2*((n_i+2*CANDIDATE_SEARCH_STEPS-1)/(2*CANDIDATE_SEARCH_STEPS));
+ n_srb_j= (n_J+2*CANDIDATE_SEARCH_STEPS-1)/(2*CANDIDATE_SEARCH_STEPS);
+ sj= n_srb_j*sizeof(*(sieve_report_bounds[0]));
+ si= n_srb_i*sizeof(**(sieve_report_bounds[0]));
+ for(i= 0;i<2;i++) {
+ u32_t j;
+
+ tpoly_f[i]= xmalloc((1+poldeg[i])*sizeof(**tpoly_f));
+ sieve_report_bounds[i]= xmalloc(sj);
+ for(j= 0;j<n_srb_j;j++)
+ sieve_report_bounds[i][j]= xmalloc(si);
+ }
+ }
+
+ }
+
+ if(sieve_count == 0)exit(0);
+ {
+ u32_t side,i;
+
+ for(side= 0;side<2;side++) {
+ u32_t prime,nr;
+ struct xFBstruct*s;
+ u32_t*root_buffer;
+ size_t xaFB_alloc= 0;
+ FB_logs[side]= xmalloc(FBsize[side]);
+ sieve_multiplier[side]= (UCHAR_MAX-50)/log(poly_norm[side]);
+
+ root_buffer= xmalloc(poldeg[side]*sizeof(*root_buffer));
+ prime= 2;
+ nr= root_finder(root_buffer,poly[side],poldeg[side],prime);
+
+ for(i= 0;i<nr;i++) {
+ adjust_bufsize((void**)&(xFB[side]),&xaFB_alloc,1+xFBs[side],
+ 16,sizeof(**xFB));
+ s= xFB[side]+xFBs[side];
+ s->p= prime;
+ s->pp= prime;
+ if(root_buffer[i] == prime) {
+ s->qq= prime;
+ s->q= 1;
+ s->r= 1;
+ } else {
+ s->qq= 1;
+ s->q= prime;
+ s->r= root_buffer[i];
+ }
+ xFBs[side]++;
+ add_primepowers2xaFB(&xaFB_alloc,n_I,side,0,0);
+ }
+ free(root_buffer);
+ for(i= 0;i<FBsize[side];i++) {
+ double l;
+ u32_t l1;
+
+ prime= FB[side][i];
+ if(prime> n_I/prime)break;
+ l= log(prime);
+ l1= add_primepowers2xaFB(&xaFB_alloc,n_I,side,prime,proots[side][i]);
+ FB_logs[side][i]= rint(l1*l*sieve_multiplier[side]);
+ }
+ while(i<FBsize[side]) {
+ double l;
+
+ l= log(FB[side][i]);
+ if(l> FB_maxlog[side])FB_maxlog[side]= l;
+ FB_logs[side][i++]= rint(sieve_multiplier[side]*l);
+ }
+ FB_maxlog[side]*= sieve_multiplier[side];
+ qsort(xFB[side],xFBs[side],sizeof(*(xFB[side])),xFBcmp);
+ }
+ }
+
+ sieve_interval= xvalloc(L1_SIZE);
+ cand= xvalloc(L1_SIZE*sizeof(*cand));
+ fss_sv= xvalloc(L1_SIZE);
+ tiny_sieve_buffer= xmalloc(TINY_SIEVEBUFFER_SIZE);
+ if(n_i> L1_SIZE)
+ complain("Strip length %u exceeds L1 size %u\n",n_i,L1_SIZE);
+ j_per_strip= L1_SIZE/n_i;
+ jps_bits= L1_BITS-i_bits;
+ jps_mask= j_per_strip-1;
+ if(j_per_strip!=1<<jps_bits)
+ Schlendrian("Expected %u j per strip, calculated %u\n",
+ j_per_strip,1<<jps_bits);
+ n_strips= n_j>>(L1_BITS-i_bits);
+ rec_info_init(n_i,n_j);
+ {
+ u32_t s;
#define MAX_TINY_2POW 4
-
-if(poldeg[0]<poldeg[1])s= poldeg[1];
-else s= poldeg[0];
-tinysieve_curpos= xmalloc(TINY_SIEVE_MIN*s*sizeof(*tinysieve_curpos));
-horizontal_sievesums= xmalloc(j_per_strip*sizeof(*horizontal_sievesums));
-for(s= 0;s<2;s++){
-u32_t fbi;
-size_t maxent;
-
-smallsieve_aux[s]= xmalloc(4*fbis[s]*sizeof(*(smallsieve_aux[s])));
+
+ if(poldeg[0]<poldeg[1])s= poldeg[1];
+ else s= poldeg[0];
+ tinysieve_curpos= xmalloc(TINY_SIEVE_MIN*s*sizeof(*tinysieve_curpos));
+ horizontal_sievesums= xmalloc(j_per_strip*sizeof(*horizontal_sievesums));
+ for(s= 0;s<2;s++) {
+ u32_t fbi;
+ size_t maxent;
+
+ smallsieve_aux[s]= xmalloc(4*fbis[s]*sizeof(*(smallsieve_aux[s])));
#ifndef MMX_TD
#ifdef PREINVERT
-smalltd_pi[s]= xmalloc(fbis[s]*sizeof(*(smalltd_pi[s])));
+ smalltd_pi[s]= xmalloc(fbis[s]*sizeof(*(smalltd_pi[s])));
#endif
-smalltdsieve_aux[s]= xmalloc(j_per_strip*sizeof(*(smalltdsieve_aux[s])));
-for(fbi= 0;fbi<j_per_strip;fbi++)
-smalltdsieve_aux[s][fbi]=
-xmalloc(fbis[s]*sizeof(**(smalltdsieve_aux[s])));
+ smalltdsieve_aux[s]= xmalloc(j_per_strip*sizeof(*(smalltdsieve_aux[s])));
+ for(fbi= 0;fbi<j_per_strip;fbi++)
+ smalltdsieve_aux[s][fbi]=
+ xmalloc(fbis[s]*sizeof(**(smalltdsieve_aux[s])));
#else
-
-MMX_TdAllocate(j_per_strip,fbis[0],fbis[1]);
+
+ MMX_TdAllocate(j_per_strip,fbis[0],fbis[1]);
#endif
-smallsieve_aux1[s]= xmalloc(6*xFBs[s]*sizeof(*(smallsieve_aux1[s])));
-
-
-maxent= fbis[s];
-maxent+= xFBs[s];
-smallpsieve_aux[s]= xmalloc(3*maxent*sizeof(*(smallpsieve_aux[s])));
-maxent= 0;
-for(fbi= 0;fbi<xFBs[s];fbi++){
-if(xFB[s][fbi].p==2)
-maxent++;
-}
-smallsieve_aux2[s]= xmalloc(4*maxent*sizeof(*(smallsieve_aux2[s])));
-x2FB[s]= xmalloc(maxent*6*sizeof(*(x2FB[s])));
-}
-}
-
-/*:58*//*59:*/
-// #line 1839 "gnfs-lasieve4e.w"
-
+ smallsieve_aux1[s]= xmalloc(6*xFBs[s]*sizeof(*(smallsieve_aux1[s])));
+
+
+ maxent= fbis[s];
+ maxent+= xFBs[s];
+ smallpsieve_aux[s]= xmalloc(3*maxent*sizeof(*(smallpsieve_aux[s])));
+ maxent= 0;
+ for(fbi= 0;fbi<xFBs[s];fbi++) {
+ if(xFB[s][fbi].p == 2)
+ maxent++;
+ }
+ smallsieve_aux2[s]= xmalloc(4*maxent*sizeof(*(smallsieve_aux2[s])));
+ x2FB[s]= xmalloc(maxent*6*sizeof(*(x2FB[s])));
+ }
+ }
+
#ifdef GCD_SIEVE_BOUND
-{
-u32_t p,i;
-
-firstprime32(&special_q_ps);
-np_gcd_sieve= 0;
-for(p= nextprime32(&special_q_ps);p<GCD_SIEVE_BOUND;
-p= nextprime32(&special_q_ps))np_gcd_sieve++;
-gcd_sieve_buffer= xmalloc(2*np_gcd_sieve*sizeof(*gcd_sieve_buffer));
-
-firstprime32(&special_q_ps);
-i= 0;
-for(p= nextprime32(&special_q_ps);p<GCD_SIEVE_BOUND;
-p= nextprime32(&special_q_ps))gcd_sieve_buffer[2*i++]= p;
-}
+ {
+ u32_t p,i;
+
+ firstprime32(&special_q_ps);
+ np_gcd_sieve= 0;
+ for(p= nextprime32(&special_q_ps);p<GCD_SIEVE_BOUND;
+ p= nextprime32(&special_q_ps))np_gcd_sieve++;
+ gcd_sieve_buffer= xmalloc(2*np_gcd_sieve*sizeof(*gcd_sieve_buffer));
+
+ firstprime32(&special_q_ps);
+ i= 0;
+ for(p= nextprime32(&special_q_ps);p<GCD_SIEVE_BOUND;
+ p= nextprime32(&special_q_ps))gcd_sieve_buffer[2*i++]= p;
+ }
#endif
-
-/*:59*/
-// #line 1148 "gnfs-lasieve4e.w"
-
-{
-u32_t s;
-for(s= 0;s<2;s++){
-if(sieve_min[s]<TINY_SIEVE_MIN&&sieve_min[s]!=0){
-errprintf("Sieving with all primes on side %u since\n",s);
-errprintf("tiny sieve procedure is being used\n");
-sieve_min[s]= 0;
-}
-current_ij[s]= xmalloc(FBsize[s]*sizeof(*current_ij[s]));
-LPri[s]= xmalloc(FBsize[s]*sizeof(**LPri)*RI_SIZE);
-}
-}
-
-/*:35*//*36:*/
-// #line 1175 "gnfs-lasieve4e.w"
-
-{
-u32_t s;
-size_t total_alloc;
-u16_t*sched_buf;
-double pvl_max[2];
-
-total_alloc= 0;
-for(s= 0;s<2;s++){
-u32_t i,fbi_lb;
-
-if(sigma>=1)pvl_max[s]= poldeg[s]*log(last_spq*sqrt(sigma));
-else pvl_max[s]= poldeg[s]*log(last_spq/sqrt(sigma));
-pvl_max[s]+= log(poly_norm[s]);
-if(fbi1[s]>=FBsize[s]||i_bits+j_bits<=L1_BITS){
-n_schedules[s]= 0;
-continue;
-}
-for(i= 0;i<N_PRIMEBOUNDS;i++)
-if(FB_bound[s]<=schedule_primebounds[i]||
-i_bits+j_bits<=schedule_sizebits[i])
-break;
-n_schedules[s]= i+1;
-schedules[s]= xmalloc(n_schedules[s]*sizeof(**schedules));
-fbi_lb= fbi1[s];
-for(i= 0;i<n_schedules[s];i++){
-u32_t fbp_lb,fbp_ub;
-u32_t fbi,fbi_ub;
-u32_t sp_i;
-u32_t n,sl_i;
-u32_t ns;
-size_t allocate,all1;
-
-if(i==n_schedules[s]-1)fbp_ub= FB_bound[s];
-else fbp_ub= schedule_primebounds[i];
-if(i==0)fbp_lb= FB[s][fbi1[s]];
-else fbp_lb= schedule_primebounds[i-1];
-
-if(i_bits+j_bits<schedule_sizebits[i])ns= 1<<(i_bits+j_bits-L1_BITS);
-else ns= 1<<(schedule_sizebits[i]-L1_BITS);
-schedules[s][i].n_strips= ns;
-
-
-
+
+ {
+ u32_t s;
+ for(s= 0;s<2;s++) {
+ if(sieve_min[s]<TINY_SIEVE_MIN&&sieve_min[s]!=0) {
+ errprintf("Sieving with all primes on side %u since\n",s);
+ errprintf("tiny sieve procedure is being used\n");
+ sieve_min[s]= 0;
+ }
+ current_ij[s]= xmalloc(FBsize[s]*sizeof(*current_ij[s]));
+ LPri[s]= xmalloc(FBsize[s]*sizeof(**LPri)*RI_SIZE);
+ }
+ }
+
+ {
+ u32_t s;
+ size_t total_alloc;
+ u16_t*sched_buf;
+ double pvl_max[2];
+
+ total_alloc= 0;
+ for(s= 0;s<2;s++) {
+ u32_t i,fbi_lb;
+
+ if(sigma>=1)pvl_max[s]= poldeg[s]*log(last_spq*sqrt(sigma));
+ else pvl_max[s]= poldeg[s]*log(last_spq/sqrt(sigma));
+ pvl_max[s]+= log(poly_norm[s]);
+ if(fbi1[s]>=FBsize[s]||i_bits+j_bits<=L1_BITS) {
+ n_schedules[s]= 0;
+ continue;
+ }
+ for(i= 0;i<N_PRIMEBOUNDS;i++)
+ if(FB_bound[s]<=schedule_primebounds[i]||
+ i_bits+j_bits<=schedule_sizebits[i])
+ break;
+ n_schedules[s]= i+1;
+ schedules[s]= xmalloc(n_schedules[s]*sizeof(**schedules));
+ fbi_lb= fbi1[s];
+ for(i= 0;i<n_schedules[s];i++) {
+ u32_t fbp_lb,fbp_ub;
+ u32_t fbi,fbi_ub;
+ u32_t sp_i;
+ u32_t n,sl_i;
+ u32_t ns;
+ size_t allocate,all1;
+
+ if(i == n_schedules[s]-1)fbp_ub= FB_bound[s];
+ else fbp_ub= schedule_primebounds[i];
+ if(i == 0)fbp_lb= FB[s][fbi1[s]];
+ else fbp_lb= schedule_primebounds[i-1];
+
+ if(i_bits+j_bits<schedule_sizebits[i])ns= 1<<(i_bits+j_bits-L1_BITS);
+ else ns= 1<<(schedule_sizebits[i]-L1_BITS);
+ schedules[s][i].n_strips= ns;
+
+
+
#ifndef SCHED_TOL
#ifndef NO_SCHEDTOL
+/* these values are experimental; report SCHED_PATHOLOGY to http://mersenneforum.org/showthread.php?t=11430 */
+#define SCHED_PAD 48
+#if I_bits<15
#define SCHED_TOL 2
+#else
+#define SCHED_TOL 1.2
#endif
#endif
+#endif
#ifdef SCHED_TOL
-allocate= rint(SCHED_TOL*n_i*j_per_strip*log(log(fbp_ub)/log(fbp_lb)));
+ allocate = rint(SCHED_PAD + SCHED_TOL * n_i * j_per_strip * log(log(fbp_ub) / log(fbp_lb)));
#else
-allocate= rint(sched_tol[i]*n_i*j_per_strip*log(log(fbp_ub)/log(fbp_lb)));
+ allocate = rint(sched_tol[i]*n_i*j_per_strip*log(log(fbp_ub)/log(fbp_lb)));
#endif
-allocate*= SE_SIZE;
-
-
-
-all1= allocate+n_i*ceil(pvl_max[s]/log(fbp_lb))*SE_SIZE;
-schedules[s][i].alloc= allocate;
-schedules[s][i].alloc1= all1;
-
-for(n= 0,fbi= fbi_lb;fbi<FBsize[s];){
-u32_t fbi_ub1;
-fbi_ub1= fbi+SCHEDFBI_MAXSTEP;
-if(fbi_ub1>=FBsize[s])fbi_ub1= FBsize[s];
-else{
-if(FB[s][fbi_ub1]> fbp_ub){
-while(FB[s][fbi_ub1]> fbp_ub)
-fbi_ub1--;
-fbi_ub1++;
-}
-}
-if(FB_logs[s][fbi]==FB_logs[s][fbi_ub1-1]){
-n++;
-fbi= fbi_ub1;
-}else{
-u32_t l;
-n+= FB_logs[s][fbi_ub1-1]-FB_logs[s][fbi];
-fbi= fbi_ub1-1;
-l= FB_logs[s][fbi];
-while(FB_logs[s][fbi]==l)fbi--;
-fbi++;
-}
-if(fbi>=FBsize[s]||FB[s][fbi]> fbp_ub)break;
-}
-fbi_ub= fbi;
-schedules[s][i].n_pieces= n;
-n++;
-schedules[s][i].schedule= xmalloc(n*sizeof(*(schedules[s][i].schedule)));
-for(sl_i= 0;sl_i<n;sl_i++)
-schedules[s][i].schedule[sl_i]=
-xmalloc(ns*sizeof(**(schedules[s][i].schedule)));
-schedules[s][i].schedule[0][0]= (u16_t*)total_alloc;
-total_alloc+= all1;
-for(sp_i= 1;sp_i<ns;sp_i++){
-schedules[s][i].schedule[0][sp_i]= (u16_t*)total_alloc;
-total_alloc+= allocate;
-}
-schedules[s][i].fbi_bounds=
-xmalloc(n*sizeof(*(schedules[s][i].fbi_bounds)));
-schedules[s][i].schedlogs= xmalloc(n);
-for(n= 0,fbi= fbi_lb;fbi<fbi_ub;){
-u32_t fbi_ub1;
-fbi_ub1= fbi+SCHEDFBI_MAXSTEP;
-if(fbi_ub1> fbi_ub)fbi_ub1= fbi_ub;
-if(FB_logs[s][fbi]==FB_logs[s][fbi_ub1-1]){
-schedules[s][i].fbi_bounds[n++]= fbi;
-fbi= fbi_ub1;
-}else{
-u32_t l,lmax;
-
-lmax= FB_logs[s][fbi_ub1-1];
-for(l= FB_logs[s][fbi];l<lmax;l++){
-schedules[s][i].fbi_bounds[n++]= fbi;
-while(fbi<fbi_ub&&FB_logs[s][fbi]==l)fbi++;
-}
-}
-}
-if(n!=schedules[s][i].n_pieces)
-Schlendrian("Expected %u schedule pieces on side %u, have %u\n",
-schedules[s][i].n_pieces,s,n);
-schedules[s][i].fbi_bounds[n++]= fbi;
-for(n= 0;n<schedules[s][i].n_pieces;n++)
-schedules[s][i].schedlogs[n]= FB_logs[s][schedules[s][i].fbi_bounds[n]];
-schedules[s][i].ri=
-LPri[s]+(schedules[s][i].fbi_bounds[0]-fbis[s])*RI_SIZE;
-fbi_lb= fbi_ub;
-}
-}
-/*37:*/
-// #line 1323 "gnfs-lasieve4e.w"
-
-sched_buf= xmalloc((total_alloc+65536*SE_SIZE*j_per_strip)*
-sizeof(**((**schedules).schedule)));
-for(s= 0;s<2;s++){
-u32_t i;
-for(i= 0;i<n_schedules[s];i++){
-u32_t sp_i;
-
-for(sp_i= 0;sp_i<schedules[s][i].n_strips;sp_i++)
-schedules[s][i].schedule[0][sp_i]=
-sched_buf+(size_t)(schedules[s][i].schedule[0][sp_i]);
-}
-}
-
-/*:37*/
-// #line 1305 "gnfs-lasieve4e.w"
-
-/*39:*/
-// #line 1349 "gnfs-lasieve4e.w"
-
+ allocate*= SE_SIZE;
+
+ all1= allocate+n_i*ceil(pvl_max[s]/log(fbp_lb))*SE_SIZE;
+ schedules[s][i].alloc= allocate;
+ schedules[s][i].alloc1= all1;
+ //printf("### [%d][%d] alloc = %ld %ld %ld %ld\n",s,i,allocate,all1,n_i,j_per_strip);
+
+ for(n= 0,fbi= fbi_lb;fbi<FBsize[s];) {
+ u32_t fbi_ub1;
+ fbi_ub1= fbi+SCHEDFBI_MAXSTEP;
+ if(fbi_ub1>=FBsize[s])fbi_ub1= FBsize[s];
+ else{
+ if(FB[s][fbi_ub1]> fbp_ub) {
+ while(FB[s][fbi_ub1]> fbp_ub)
+ fbi_ub1--;
+ fbi_ub1++;
+ }
+ }
+ if(FB_logs[s][fbi] == FB_logs[s][fbi_ub1-1]) {
+ n++;
+ fbi= fbi_ub1;
+ } else {
+ u32_t l;
+ n+= FB_logs[s][fbi_ub1-1]-FB_logs[s][fbi];
+ fbi= fbi_ub1-1;
+ l= FB_logs[s][fbi];
+ while(FB_logs[s][fbi] == l)fbi--;
+ fbi++;
+ }
+ if(fbi>=FBsize[s]||FB[s][fbi]> fbp_ub)break;
+ }
+ fbi_ub= fbi;
+ schedules[s][i].n_pieces= n;
+ n++;
+ schedules[s][i].schedule...
[truncated message content] |
|
From: <Ba...@us...> - 2009-04-19 18:47:54
|
Revision: 354
http://ggnfs.svn.sourceforge.net/ggnfs/?rev=354&view=rev
Author: Batalov
Date: 2009-04-19 18:47:45 +0000 (Sun, 19 Apr 2009)
Log Message:
-----------
Two bug fixes
Modified Paths:
--------------
trunk/src/experimental/lasieve4_64/athlon64/siever-config.h
trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c
Modified: trunk/src/experimental/lasieve4_64/athlon64/siever-config.h
===================================================================
--- trunk/src/experimental/lasieve4_64/athlon64/siever-config.h 2009-04-18 03:58:53 UTC (rev 353)
+++ trunk/src/experimental/lasieve4_64/athlon64/siever-config.h 2009-04-19 18:47:45 UTC (rev 354)
@@ -27,6 +27,7 @@
#if 1
void MMX_TdAllocate(int,size_t,size_t);
u16_t*MMX_TdInit(int,u16_t*,u16_t*,u32_t*,int);
+u32_t*MMX_Td(u32_t *pbuf,int side,u16_t strip_i);
void MMX_TdUpdate(int,int);
#define MMX_TD
#define MMX_REGW 8
Modified: trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c
===================================================================
--- trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c 2009-04-18 03:58:53 UTC (rev 353)
+++ trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c 2009-04-19 18:47:45 UTC (rev 354)
@@ -593,7 +593,7 @@
}
#endif
//skip_blanks_comments(&input_line,&input_line_alloc,input_data);
- close(input_data);
+ fclose(input_data);
{ FILE *fp;
char token[256], value[512], thisLine[1024];
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
|
From: <Ba...@us...> - 2009-10-30 02:07:57
|
Revision: 370
http://ggnfs.svn.sourceforge.net/ggnfs/?rev=370&view=rev
Author: Batalov
Date: 2009-10-30 02:07:48 +0000 (Fri, 30 Oct 2009)
Log Message:
-----------
mpqs will not fail on a prime square
Modified Paths:
--------------
trunk/src/experimental/lasieve4_64/athlon64/siever-config.h
trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c
Modified: trunk/src/experimental/lasieve4_64/athlon64/siever-config.h
===================================================================
--- trunk/src/experimental/lasieve4_64/athlon64/siever-config.h 2009-10-29 05:23:37 UTC (rev 369)
+++ trunk/src/experimental/lasieve4_64/athlon64/siever-config.h 2009-10-30 02:07:48 UTC (rev 370)
@@ -20,7 +20,7 @@
void asm_getbc(u32_t,u32_t,u32_t,u32_t*,u32_t*,u32_t*,u32_t*);
#define ASM_SCHEDSIEVE
#define ASM_SCHEDTDSIEVE2
-u16_t**tdsieve_sched2buf(u16_t*,u16_t*,unsigned char*,u16_t**,u16_t**);
+u16_t**tdsieve_sched2buf(u16_t**,u16_t*,unsigned char*,u16_t**,u16_t**);
#define ASM_MPZ_TD
#define PREINVERT
Modified: trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c
===================================================================
--- trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c 2009-10-29 05:23:37 UTC (rev 369)
+++ trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c 2009-10-30 02:07:48 UTC (rev 370)
@@ -2708,12 +2708,12 @@
tNow = sTime();
if (tNow > lastReport + 5.0) {
lastReport = tNow;
- fprintf(stderr, "\rtotal yield: %u, q=%u (%1.5lf sec/rel)",
+ fprintf(stderr, "\rtotal yield: %u, q=%u (%1.5lf sec/rel) ",
(unsigned int)yield, (unsigned int)special_q, (tNow - tStart)/yield);
fflush(stderr);
}
}
- fprintf(stderr, "\rtotal yield: %u, q=%u (%1.5lf sec/rel)\n",
+ fprintf(stderr, "\rtotal yield: %u, q=%u (%1.5lf sec/rel) \n",
(unsigned int)yield, (unsigned int)special_q, (sTime() - tStart)/yield);
free(r);
}
@@ -3942,6 +3942,19 @@
mpz_neg(large_factors[s1],large_factors[s1]);
if((nf= mpqs_factor(large_factors[s1],max_primebits[s1],&mf))<0) {
+ /* did it fail on a square? */
+ mpz_sqrtrem(large_primes[s1][0],large_primes[s1][1],large_factors[s1]);
+ if(mpz_sgn(large_primes[s1][1]) == 0) { /* remainder == 0? */
+ mpz_set(large_primes[s1][1],large_primes[s1][0]);
+ nlp[s1]= 2;
+ if(verbose) {
+ fprintf(stderr," mpqs on a prime square ");
+ mpz_out_str(stderr,10,large_primes[s1][0]);
+ fprintf(stderr,"^2 ");
+ }
+ continue;
+ }
+ if(verbose) {
fprintf(stderr,"mpqs failed for ");
mpz_out_str(stderr,10,large_factors[s1]);
fprintf(stderr,"(a,b): ");
@@ -3949,6 +3962,7 @@
fprintf(stderr," ");
mpz_out_str(stderr,10,sr_b);
fprintf(stderr,"\n");
+ }
n_mpqsfail[s1]++;
break;
}
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
|
From: <Ba...@us...> - 2009-11-04 03:57:57
|
Revision: 376
http://ggnfs.svn.sourceforge.net/ggnfs/?rev=376&view=rev
Author: Batalov
Date: 2009-11-04 03:57:36 +0000 (Wed, 04 Nov 2009)
Log Message:
-----------
faster sieving (up to 20%)
Modified Paths:
--------------
trunk/src/experimental/lasieve4_64/athlon64/ls-defs.asm
trunk/src/experimental/lasieve4_64/athlon64/siever-config.h
trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c
trunk/src/experimental/lasieve4_64/mpqs.c
Modified: trunk/src/experimental/lasieve4_64/athlon64/ls-defs.asm
===================================================================
--- trunk/src/experimental/lasieve4_64/athlon64/ls-defs.asm 2009-11-01 21:30:54 UTC (rev 375)
+++ trunk/src/experimental/lasieve4_64/athlon64/ls-defs.asm 2009-11-04 03:57:36 UTC (rev 376)
@@ -1,4 +1,4 @@
-define(l1_bits,15)dnl
+define(l1_bits,16)dnl
define(n_i,eval(2**n_i_bits))dnl
define(n_i_mask,eval(n_i-1))dnl
define(l1_size,eval(2**l1_bits))dnl
Modified: trunk/src/experimental/lasieve4_64/athlon64/siever-config.h
===================================================================
--- trunk/src/experimental/lasieve4_64/athlon64/siever-config.h 2009-11-01 21:30:54 UTC (rev 375)
+++ trunk/src/experimental/lasieve4_64/athlon64/siever-config.h 2009-11-04 03:57:36 UTC (rev 376)
@@ -5,7 +5,7 @@
#define __SIEVER_CONFIG_H__
#include <stddef.h>
#include <limits.h>
-#define L1_BITS 15
+#define L1_BITS 16
#define ULONG_RI
typedef unsigned u32_t;
typedef int i32_t;
Modified: trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c
===================================================================
--- trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c 2009-11-01 21:30:54 UTC (rev 375)
+++ trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c 2009-11-04 03:57:36 UTC (rev 376)
@@ -147,7 +147,7 @@
u32_t fbis[2];
-#if I_bits<16
+#if I_bits<17
static u32_t j_per_strip,jps_bits;
#else
#define j_per_strip 1
@@ -1002,7 +1002,7 @@
tiny_sieve_buffer= xmalloc(TINY_SIEVEBUFFER_SIZE);
if(n_i> L1_SIZE)
complain("Strip length %u exceeds L1 size %u\n",n_i,L1_SIZE);
-#if I_bits<16
+#if I_bits<17
j_per_strip= L1_SIZE/n_i;
jps_bits= L1_BITS-i_bits;
#endif
@@ -2285,7 +2285,7 @@
p= x[0];
l= x[1];
d= x[2];
-#if I_bits<16
+#if I_bits<17
while(d<j_per_strip) {
horizontal_sievesums[d]+= l;
d+= p;
Modified: trunk/src/experimental/lasieve4_64/mpqs.c
===================================================================
--- trunk/src/experimental/lasieve4_64/mpqs.c 2009-11-01 21:30:54 UTC (rev 375)
+++ trunk/src/experimental/lasieve4_64/mpqs.c 2009-11-04 03:57:36 UTC (rev 376)
@@ -1227,10 +1227,13 @@
i1=sv[s1]; i2=sv[s2]; sv+=p;
if (i1||i2) {
if (i1) { nr=rels[i1][4]; rels[i1][nr+5]=mpqs_nFBk_1+i; rels[i1][4]++; }
- if (i2) { nr=rels[i2][4]; rels[i2][nr+5]=mpqs_nFBk_1+i; rels[i2][4]++; } }
+ if (i2) { nr=rels[i2][4]; rels[i2][nr+5]=mpqs_nFBk_1+i; rels[i2][4]++; }
+ }
i1=sv[s1]; i2=sv[s2]; sv+=p;
if (i1||i2) {
- if (i1) { nr=rels[i1][4]; rels[i1][nr+5]=mpqs_nFBk_1+i; rels[i1][4]++; } if (i2) { nr=rels[i2][4]; rels[i2][nr+5]=mpqs_nFBk_1+i; rels[i2][4]++; } }
+ if (i1) { nr=rels[i1][4]; rels[i1][nr+5]=mpqs_nFBk_1+i; rels[i1][4]++; }
+ if (i2) { nr=rels[i2][4]; rels[i2][nr+5]=mpqs_nFBk_1+i; rels[i2][4]++; }
+ }
}
svend+=p;
while (sv<svend) {
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
|
From: <Ba...@us...> - 2009-11-06 23:57:21
|
Revision: 382
http://ggnfs.svn.sourceforge.net/ggnfs/?rev=382&view=rev
Author: Batalov
Date: 2009-11-06 23:57:07 +0000 (Fri, 06 Nov 2009)
Log Message:
-----------
better L1_BITS definition
Modified Paths:
--------------
trunk/src/experimental/lasieve4_64/INSTALL
trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c
Modified: trunk/src/experimental/lasieve4_64/INSTALL
===================================================================
--- trunk/src/experimental/lasieve4_64/INSTALL 2009-11-06 23:52:37 UTC (rev 381)
+++ trunk/src/experimental/lasieve4_64/INSTALL 2009-11-06 23:57:07 UTC (rev 382)
@@ -2,6 +2,12 @@
follow these steps to build the binaries on a 64-bit machine.
These should build and run on Athlon64, or Core2 cpus.
+******************************
+NOTE: on Core2 replace in athlon64/ls-defs.asm
+- define(l1_bits,16)dnl
++ define(l1_bits,15)dnl
+****************************** (don't know how to automate this)
+
Prerequisites: gmp.h libgmp
cd athlon64
Modified: trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c
===================================================================
--- trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c 2009-11-06 23:52:37 UTC (rev 381)
+++ trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c 2009-11-06 23:57:07 UTC (rev 382)
@@ -147,7 +147,7 @@
u32_t fbis[2];
-#if I_bits<17
+#if I_bits<=L1_BITS
static u32_t j_per_strip,jps_bits;
#else
#define j_per_strip 1
@@ -1002,7 +1002,7 @@
tiny_sieve_buffer= xmalloc(TINY_SIEVEBUFFER_SIZE);
if(n_i> L1_SIZE)
complain("Strip length %u exceeds L1 size %u\n",n_i,L1_SIZE);
-#if I_bits<17
+#if I_bits<=L1_BITS
j_per_strip= L1_SIZE/n_i;
jps_bits= L1_BITS-i_bits;
#endif
@@ -2285,7 +2285,7 @@
p= x[0];
l= x[1];
d= x[2];
-#if I_bits<17
+#if I_bits<=L1_BITS
while(d<j_per_strip) {
horizontal_sievesums[d]+= l;
d+= p;
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
|
From: <Ba...@us...> - 2009-12-21 08:44:14
|
Revision: 383
http://ggnfs.svn.sourceforge.net/ggnfs/?rev=383&view=rev
Author: Batalov
Date: 2009-12-21 08:44:06 +0000 (Mon, 21 Dec 2009)
Log Message:
-----------
define L1_BITS 15; for k8+ CPUs read INSTALL
Modified Paths:
--------------
trunk/src/experimental/lasieve4_64/INSTALL
trunk/src/experimental/lasieve4_64/athlon64/ls-defs.asm
trunk/src/experimental/lasieve4_64/athlon64/siever-config.h
Modified: trunk/src/experimental/lasieve4_64/INSTALL
===================================================================
--- trunk/src/experimental/lasieve4_64/INSTALL 2009-11-06 23:57:07 UTC (rev 382)
+++ trunk/src/experimental/lasieve4_64/INSTALL 2009-12-21 08:44:06 UTC (rev 383)
@@ -3,11 +3,20 @@
These should build and run on Athlon64, or Core2 cpus.
******************************
-NOTE: on Core2 replace in athlon64/ls-defs.asm
-- define(l1_bits,16)dnl
-+ define(l1_bits,15)dnl
-****************************** (don't know how to automate this)
+NOTE for Phenom/K8 users: replace in athlon64/ls-defs.asm
+ define(l1_bits,15)dnl
+=>
+ define(l1_bits,16)dnl
+and in athlon64/siever-config.h
+#define L1_BITS 15
+=>
+#define L1_BITS 16
+
+This may add 5-10% speed to your sieving speed.
+But try it both ways and decide for yourself. Caveat emptor!
+******************************
+
Prerequisites: gmp.h libgmp
cd athlon64
Modified: trunk/src/experimental/lasieve4_64/athlon64/ls-defs.asm
===================================================================
--- trunk/src/experimental/lasieve4_64/athlon64/ls-defs.asm 2009-11-06 23:57:07 UTC (rev 382)
+++ trunk/src/experimental/lasieve4_64/athlon64/ls-defs.asm 2009-12-21 08:44:06 UTC (rev 383)
@@ -1,4 +1,4 @@
-define(l1_bits,16)dnl
+define(l1_bits,15)dnl
define(n_i,eval(2**n_i_bits))dnl
define(n_i_mask,eval(n_i-1))dnl
define(l1_size,eval(2**l1_bits))dnl
Modified: trunk/src/experimental/lasieve4_64/athlon64/siever-config.h
===================================================================
--- trunk/src/experimental/lasieve4_64/athlon64/siever-config.h 2009-11-06 23:57:07 UTC (rev 382)
+++ trunk/src/experimental/lasieve4_64/athlon64/siever-config.h 2009-12-21 08:44:06 UTC (rev 383)
@@ -6,15 +6,12 @@
#include <stddef.h>
#include <limits.h>
-#if defined(__amdfam10__) || defined(__k8__) || defined(__athlon__)
-#define L1_BITS 16
-#else
#define L1_BITS 15
+/* ===if you change L1_BITS here, change l1_bits define in ls-defs.asm ===*/
/* L1_BITS 15 for Intel Core2 (32Kb L1 data cache), AMD K6 */
/* L1_BITS 16 for Phenom/Opteron... AMD K7 and up */
/* L1_BITS 14 was here before. PentiumIII had 16Kb */
/* L1_BITS 13 for Pentium4 ? (only 8Kb L1 Dcache) */
-#endif
#define ULONG_RI
typedef unsigned u32_t;
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
|
From: <Ba...@us...> - 2012-02-12 22:23:50
|
Revision: 425
http://ggnfs.svn.sourceforge.net/ggnfs/?rev=425&view=rev
Author: Batalov
Date: 2012-02-12 22:23:43 +0000 (Sun, 12 Feb 2012)
Log Message:
-----------
minor
Modified Paths:
--------------
trunk/src/experimental/lasieve4_64/gmp-aux.h
trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c
Modified: trunk/src/experimental/lasieve4_64/gmp-aux.h
===================================================================
--- trunk/src/experimental/lasieve4_64/gmp-aux.h 2011-12-20 09:14:29 UTC (rev 424)
+++ trunk/src/experimental/lasieve4_64/gmp-aux.h 2012-02-12 22:23:43 UTC (rev 425)
@@ -3,7 +3,7 @@
void adjust_mpz_bufsize(mpz_t**x,size_t*alloc_ptr,size_t size,size_t increment);
int string2mpz(mpz_t rop,char*x,int base);
-#if __GNU_MP_VERSION < 3 || __GNU_MP_VERSION_MINOR == 0
+#if __GNU_MP_VERSION < 3 || GNU_MP_VERSION == 3 && GNU_MP_VERSION_MINOR == 0
#define NEED_MPZ_MUL_SI
void mpz_mul_si(mpz_t x,mpz_t y,long int z);
#endif
Modified: trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c
===================================================================
--- trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c 2011-12-20 09:14:29 UTC (rev 424)
+++ trunk/src/experimental/lasieve4_64/gnfs-lasieve4e.c 2012-02-12 22:23:43 UTC (rev 425)
@@ -3992,11 +3992,13 @@
if(mpz_sgn(large_primes[s1][1]) == 0) { /* remainder == 0? */
mpz_set(large_primes[s1][1],large_primes[s1][0]);
nlp[s1]= 2;
+#if 0 /* this is now tested well enough, no need for a message */
if(verbose > 1) {
fprintf(stderr," mpqs on a prime square ");
mpz_out_str(stderr,10,large_primes[s1][0]);
fprintf(stderr,"^2 ");
}
+#endif
continue;
}
if(verbose > 1) {
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
×
Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.
Thanks for helping keep SourceForge clean.
X