YAFU full changelog

· new sse2 code: faster small prime sieving in siqs
· new sse4.1 code: even faster small prime sieving in siqs
· new sse4.1 code: faster large prime bucket sieving in siqs
· makefile additions to include sse4.1 code in the fat binary on compatible hardware
· runtime flag to utilize sse4.1 code on compatible hardware
· enabled multipliers for fermat factorization
· fixed bug in qs filtering
· fixed a bug in .job file filling - handle no line break on last line
· fixed quot;too many refactorizationsquot; bug
· added a function to factor all single precision integers within a specified range
· frontend calculator now uses GMP
· Updated quot;gnfs.hquot; to use GMP
· automatic processing of several SNFS forms: N = a*b^n +/- c, for b < 100, c < 2^30, N < 1024 bits N = b^n +/- 1, for b > 100, N < 1024 bits N = a^n +/- b^n, for gcd(a,b) = 1, a,b

· made quot;found polyquot; messages much less verbose
· using /r instead of printing backspaces now in ecm.c and SIQS.c
· ggnfs jobs launched by yafu will now print out individual .last_spq files per thread, although they are still not used for anything
· get rid of blk_rel_count experiment code in siqs
· add the beginnings of CUDA squfof support - although it is far from working and probably not even beneficial at this point. currently protected by HAVE_CUDA definition
· more work on tinySIQS, but still not fully operational.
· added more fclose's
· fixed bugs that caused crashes when inputs numbers approached or exceeded 1024 characters in batchfiles.
· updates to text output of factor() to prevent window scrolling
· (re)support builds without NFS=1
· got rid of cat.exe warning messages in windows that don't have unxutils
· slight cleanup of nfs state machine
· improved min_rels calculation
· added ability to parse user supplied job files and supply missing parameters

· added fclose's and fixed gethostname alloc problem
· removed the quot;found time recordquot; messages from poly select
· remove the printing of rels during .dat parsing (available with verbose mode -v -v)
· changed siqs cutoff back to 115 bits

· fixed restart-with-factor bug by logging factors removed during restart with add_to_factor_list
· fiddled with q-range table again, and changed multi-threaded nfs sieving so that q-ranges are split over the threads, instead of each thread getting its own range.
· split blocksize dependent code in siqs into separate functions; runtime decisions are made based on cpu architecture as to which function to use. This eliminates the need for separate 32k/64k executables.
· much improved fermat factorization routine: 50x+ faster and accepts user supplied multipliers
· watch for ggnfs siever crash error code
· added ETA estimate to ecm for larger B1 values
· added ETA estimate to the filtering stage of NFS, while sieving
· cosmetic changes to factor() messages
· resume nfs in factor() if input matches .job file -ns bug fix and small changes to nfs state machine

· fixed counting of decimal digits for logging purposes
· massive overhaul of factor(): pretesting to (customizable) digit level instead of timed ecm pretesting, more ecm pretesting levels, better usage of ecm levels, printing of computed t levels. See docfile.txt for more info.
· removed -qs_ecm_ratio and -gnfs_ecm_ratio, added -pretest_ratio and -xover options - see docfile for more info
· added some logging info of new factor() state machine
· added -work option to make factor() aware of prior pretesting work (specified as a t-level)
· handle aborts and errors within nfs the same as with other factoring routines (print_factors and exit), to avoid infinite loops in the new factor() state machine
· a little bit more verbose when resuming nfs (with vflag > 0)
· skip looking for last special-q if argument -ns is given + prevent quot;time=xquot; lines from being written to .job files

· Includes fixes for the most recently reported bugs.
· Also removed zlib support, so data files will be as before.

· fixed bug in primes() routine
· smallmpqs now prints its factors to the logfile like it should
· disabled 8x med prime trial division due to windows x64 bug
· ported rho.c entirely to gmp removed my homegrown monty code from the project
· main now returns 0 instead of 1 (thanks yoyo!)

· fixed bug in primes() routine
· smallmpqs now prints its factors to the logfile like it should
· disabled 8x med prime trial division due to windows x64 bug
· ported rho.c entirely to gmp removed my homegrown monty code from the project
· main now returns 0 instead of 1 (thanks yoyo!)

· fixed a bug in LEGCD that called spDivide when v == 1, causing a crash
· tweaked poly_a generation to allow siqs to work on much smaller inputs
· modified siqs to use smallmpqs (instead of mpqs) below a threshold, and lowered the threshold to take advantage of new parameters for small siqs jobs.
· fixed a bug in nextprime that caused some small primes to be identifed as not prime
· smallmpqs called standalone now returns factors and residue
· smallmpqs now uses gmp throughout (for a large speedup)
· parameter adjustments to smallmpqs resulting in a small speedup
· fixed a bug in the SoE to allocate memory better
· double large prime cutoffs changed to uint64s, simplifying the code and providing a small speedup for larger jobs
· better polynomial root update and bucket sieving assembler code, providing a small speedup
· fixed a couple more bugs in prime counting and printing (
· added some extra info to the -v printout at the end of DLP siqs factorizations