Debian Codegraph @

ChangeLog

raw
1 ------------------------------------------------------------------------
2 r529 | rgbatduke | 2011-04-01 13:49:31 -0400 (Fri, 01 Apr 2011) | 117 lines
3
4 OK, this is a fairly enormously major brutal checkin. Both dieharder
5 and libdieharder are ALMOST -Wall -pedantic clean. To get it there I
6 had to learn several things, such as how to get gcc to ignore "unused
7 variables" that are conveniently in a shared include file but aren't
8 really used in all the modules that share it, the fact that the various
9 flavors of C have varying "maximum string size guaranteed to be
10 supported" limits (none of which are really relevant to gcc, but it
11 complains about them anyway), and more. And of course I had to delete
12 all the cruftish lines of e.g. unused loop variables. I'm not quite
13 done with cleanup -- I may have gone overboard in a place or two and may
14 need to put some things back or address things that might affect
15 function -- but I want to get this all checked in.
16
17 There are two build errors left -- one is in dieharder/rdieharder.c (and
18 hence is yours, Dirk) and the other is in the skein code (and hence is
19 yours, David). David, I also need you to check a fix I made to the
20 rng_threefish code -- I finally took the time to figure out the dread
21 "dereferencing type-punned pointer breaks strict aliasing rules"
22 warning. I replaced the offending line:
23
24 *((unsigned long int *) state->block) = s;
25
26 with
27
28 unsigned long int *blockptr;
29 ...
30 blockptr = (unsigned long int*) &state->block;
31 *blockptr = s;
32
33 That is, I read what you were trying to do as "Set the contents of
34 state->block, cast to an unsigned long int pointer, equal to unsigned
35 long int pointer s" which might work but gcc -Wall hated it even before
36 -pedantic and (from what I've read) can have undesired side effects. So
37 I introduced an actual unsigned long int pointer, put the address of
38 state->block in it, and the set its contents equal to s. It didn't seem
39 to break threefish -- I tested the first few returns before and after
40 the fix with -S 1 and they were the same -- and I'm using threefish
41 right now in a validate.sh run to make sure that I didn't egregiously
42 break dieharder with all of the changes.
43
44 Changes you should be aware of:
45
46 * To avoid most of the "too long string" errors I went to -std=c95,
47 which permits strings a page in size (4095 bytes). That accommodates
48 the auto-documentation strings in the test headers. There may be
49 another way of doing this -- in fact there are probably two or three --
50 but to alter the dh headers at this point would (marginally) break the
51 API so it will need to wait for v4, I think. Apparently gcc is about to
52 be dressed up with an __attribute__ that will probably enable extra
53 large or unlimited data strings without complaints which is sensible
54 enough since it works on them anyway AFAICT.
55
56 * c95 turned off uint translation . I went through a huge block of
57 code turning uint into the two words unsigned int before getting
58 irritated enough to look at the headers where I discovered that yeah,
59 you can turn on the uint -> unsigned int macro with a suitable define.
60 So I did.
61
62 * c95 turned of BSD math macros in math.h, including M_PI. That
63 seemed really silly, so I turned them back on with a suitable define. I
64 didn't turn on the long forms (they only really make sense for long
65 doubles) but we can do that if we ever need PI to 24 places or whatever
66 it was.
67
68 * -Wall -pedantic really hates any sort of data that is included in a
69 source file where it isn't used. If we were all perfect programmers, I
70 suppose that we would create enough include files and control where they
71 were included precisely enough that no source file even included an
72 include file with a variable it didn't actually use. Alas, I'm not a
73 perfect programmer and lots of the data structures used only inside
74 certain tests or by certain generators are shared via libdieharder.h
75 with program modules that don't use them. Adding
76 __attribute__((unused)) after the definition but before the = sign
77 basically tells the compiler "Yes, I know, I planned it that way, now
78 shut the fuck up" and passes them through -pedantic without complaint.
79
80 I suppose that the virtue of the check is that it helps prevent
81 namespace collision, but of course the compiler checks for that anyway
82 and general local vs global rules seem like they would handle any
83 accidents that crop up the right way. If I feel really, really
84 energetic someday I may go and segregate out the data and either add it
85 to the sources directly (in a lot of cases that's a good place for it
86 anyway) or put it in a separate include file per module. OTOH things
87 like the dh headers are shared because I DO access content from them in
88 lots of places and want to be able to get to it from anyplace, so there
89 will always be some ((unused)) attribute variables in the program.
90 Printing out the test description string for any given test, or looking
91 up the default tsamples or psamples, for example, is something any sort
92 of application that uses the libdieharder library might want to do.
93
94 * As per current GBT recommendations and Dirk's suggestions, all of
95 the auto-whatever stuff in autogen.sh is now basically a single
96 autoreconf call. In fact, it looks like they made autoreconf just
97 because getting all of the things just right after a major GBT update
98 is, in fact, the pain in the ass that it has been to me from the
99 beginning, so this is rather a relief. I did leave the configure call
100 in the bottom, so running autogen.sh should still take one from a clean
101 checkout to make-ready, or of course you can enter autoreconf by hand
102 and run configure by hand as per usual. Hopefully this will all make
103 Debianheads happy...;-)
104
105 Things that I have NOT done yet -- this checkin is basically six hours
106 of work from 9 to 3 am plus another couple of hours today, so I'm
107 working as fast as I can as it is -- include debugging the endian
108 problem in the threefish (or was it AES?) code on e.g. a sparc or
109 powerpc set to the other endianness and dealing with a few real bug
110 reports that have come in from users already. I wanted to get the code
111 clean first as who knows, maybe doing so will help solve the problem?
112
113 SO, if you guys could each fix the two remaining problems (or tell me to
114 play through in spite of the fact that I'm not sure what is being
115 accomplished and what would break what) then I'll try to move on to the
116 next step.
117
118 rgb
119
120
121 ------------------------------------------------------------------------
122 r523 | rgbatduke | 2011-03-10 11:09:12 -0500 (Thu, 10 Mar 2011) | 4 lines
123
124 A last minute oops. I wanted to mark operm5 as good, and mark all of
125 the monkeys suspect (as they can pretty easily be run to failure for
126 good generators still).
127
128 ------------------------------------------------------------------------
129 r510 | rgbatduke | 2011-01-07 16:19:40 -0500 (Fri, 07 Jan 2011) | 31 lines
130
131 This is a WORKING snap and bump to 3.29.6beta. I actually fixed several
132 things that I broke before in the rng selection process. New features:
133
134 rng_kiss -- a damn fine rng. Faster than mt, better than mt except for
135 period.
136
137 rng_XOR -- Select this rng, and a list of others, e.g.
138
139 ./dieharder -g 207 -g 208 -g 14 -g 6 -g 205 -a
140
141 dieharder will then return the output of 208 (kiss), 14 (mt19937_1999),
142 6 (gfsr4) and 205 (aes) all xor'd together. Period infinite, no LESS
143 random than the MOST random of the generators alone. The price you pay
144 is sure, 2, 3, 4 times slower. But this is now the official gold
145 standard dieharder testing generator, as finding something randomer will
146 be difficult and of longer period impossible (what is the least common
147 multiple of 19937, 121, and all the rest? 2 to that power, like that).
148
149 I'm working on superkiss, a vectorized version of kiss with an insanely
150 long period, but the double precision part is broken and I don't see
151 why. The integer part works. I'll figure it out maybe tonight, and
152 have a few other Marsaglia generators to add.
153
154 Then I'll return (finally) to tests, with the gold standard generator
155 well and truly in place. At least three dieharder (diehard) tests are
156 broken, and I'd like to fix at least ONE of them before I get bogged
157 down teaching again.
158
159 rgb
160
161
162 ------------------------------------------------------------------------
163 r508 | rgbatduke | 2010-02-19 13:13:56 -0500 (Fri, 19 Feb 2010) | 9 lines
164
165 Oops, forgot to update FIRST. This should get me back in sync. You
166 guys should ignore this; I'm rearranging my whole source tree on my
167 laptop(s), dieharder included, and am just trying to make sure that the
168 rebuilt one is clean.
169
170 I also haven't completely forgotten the last post/request for interface
171 room -- I've just been insanely busy and haven't had time to even think
172 about it for the last few weeks. But I will get back there, I promise.
173
174 ------------------------------------------------------------------------
175 r498 | rgbatduke | 2009-10-28 01:48:25 -0400 (Wed, 28 Oct 2009) | 28 lines
176
177 Wow, a lot of stuff. This checkin contains a working -Y 2 option for
178 "test to destruction" where ttd is by default a return pvalue of
179 0.000001 or less OR getting to 10000 samples alive (both parameters
180 can be set on the command line with -X tolerance and -Z cutoff). I
181 actually did it two ways, and will keep the second (better) one and
182 shortly remove the cruft in std_test.c. In addition I had to update the
183 help, I updated the output routines in output_rnds so I could dump a
184 list of formatted floats (to test another rng tester that alas was so
185 broken it couldn't read any format I tried anyway), I fixed and updated
186 the man page, I got rid of the old overlap variable (no longer
187 desireable or necessary, although I have a bit of cruft left behind to
188 clean up still).
189
190 As a result of the initial ttd test, I am certain that there is a
191 problem with diehard_dna, one that causes it to fail aes at 0.000001 in
192 1500-4000 samples. This is odd, since this test has an "exactly"
193 computed mean and sigma target. I may try threefish in a second to see
194 if it fails too, in the same general order. Haven't done the auto-xor
195 generator thing yet.
196
197 I still have to implement -Y 1 (resolve ambiguity mode) where it will
198 force a test to fail or come back up with more samples, but it should be
199 straightforward. However, it is almost 2 am and I teach way too early,
200 so it is off to bed.
201
202 rgb
203
204
205 ------------------------------------------------------------------------
206 r497 | rgbatduke | 2009-10-22 09:45:52 -0400 (Thu, 22 Oct 2009) | 20 lines
207
208 This is a not-quite-yet-broken checkin of -Xtreme mode changes. Three new
209 control variables are in place. They are parsed (untested). They are
210 used in std_test to allocate much larger pvalue vectors (Xcutoff in
211 size) at test initialization time. I'm JUST READY to hack into the main
212 std_test execution loop with case switches or other conditionals and
213 implement at least resolve ambiguity and ttd modes. But as usual, I
214 have to go in and teach. At the moment, though, DH still builds and
215 runs -a correctly, so it seems like a good idea to check in a still
216 working snap in case I break everything and want to start over from
217 here.
218
219 Oh, I also am cleaning up a bit and made the multiply_p variable (-m
220 option) a double, so you CAN enter -m 0.1 and run only 10 psamples for a
221 fast version of -a(ll). At this point a lot of debugging is just
222 ensuring that all the tests run, and it is a PITA to wait 30+ minutes
223 for a -a(ll) run to get through. So you now CAN test fewer than the
224 default number of psamples in an -all run, even though most people won't
225 use the option in actual testing. The usual usage, -m 10 or -m 100,
226 works fine still.
227
228 ------------------------------------------------------------------------
229 r495 | rgbatduke | 2009-10-20 14:25:32 -0400 (Tue, 20 Oct 2009) | 34 lines
230
231 I'm checking in a lot of changes down below. -m is implemented and
232 documented. -k is implemented and documented. The man page is fixed
233 (post good-kstest and aes/threefish). The endian bug went away when I
234 refreshed the include files, making me wonder if it wasn't some sort of
235 strange GBT stuff and not a real problem -- I left in the endian code in
236 configure.ac but don't use it. I re-fixed diehard_runs.c -- it was
237 broken post patching but now seems good. I filed some documentation and
238 bug reports. I fixed a number of pernicious warnings about needing
239 casts (one still remains in threefish, but it is David's and I don't
240 know how to fix it). I worked on dieharder.html.in pretty substantially
241 to get it to match all of the above.
242
243 Next, -x (and maybe -X).
244
245 (p.s. -- Welcome David.)
246
247 (p.p.s. -- I'm still testing -- sigh, forever -- but it looks like all
248 non-deprecated tests are working OK in this snap, and that the -m
249 feature works nicely. I documented timings for the k options, and
250 basically it comes down the kstest being too slow to do large numbers of
251 samples without switching over to the asymptotic form of the test at
252 some point. I mean, going from three minute runs to over three hours
253 and still counting when I quit for a factor of 10 difference in the
254 number of samples, really serious nonlinear gains in the amount of
255 work/time required, and this was still -k 1 with Marsaglia's more modest
256 speedup, not even the "exact" mode with no speedup at all.
257
258 This will quite possibly require some further hacking of the boundaries
259 for a crossover that is "practical" and not too inaccurate as we gain
260 experience with our own patience, especially as we implement a -x like
261 option that just keeps crankin' on the number of samples to hit a
262 prespecified tolerance for failure.)
263
264
265 ------------------------------------------------------------------------
266 r494 | rgbatduke | 2009-10-19 09:46:39 -0400 (Mon, 19 Oct 2009) | 3 lines
267
268 This is all of the Bauer patches. Some are tested, but the testing
269 continues.
270
271 ------------------------------------------------------------------------
272 r493 | rgbatduke | 2009-10-18 10:43:52 -0400 (Sun, 18 Oct 2009) | 108 lines
273
274 This is checking in what will be 3.29.4beta.
275
276 Primary fixes so far:
277
278 Several changes to configure.ac to eliminate all reference to libaes
279 and to set macros ENDIAN_BIG or ENDIAN_LITTLE to 1 in the configure
280 stage of the build. I plan to insert a very simple prequel in Brian
281 Goodman's brg_endian.h header file that handles endian issues cleanly
282 and skips most of the stuff below for little endian. I do need to
283 ensure that it builds on i386 as well (when I'm done) as I have a report
284 that 3beta doesn't build clean on that architecture due to problems in
285 this header file.
286
287 A fix due to Glenn Emelko, [email protected], where I correctly
288 bumped filecount to type off_t in libdieharder.h but failed to redefine
289 rtot and rptr accordingly in the rng_file_input.c struct and code. He
290 was running 18 GB raw files and obviously this overflowed uint variables
291 with bad results. Oops, and thanks Glenn.
292
293 I am trying to get sts_serial.h to run at 24 bits by default, not 16
294 (I think that this will still take a not unreasonable amount of time).
295 The problem is that sts_serial doesn't use bits.c calls to parse out the
296 next 24 bits, it just grabs 16 bit, then the next 16 bits, out of 32 bit
297 uints. This is fast but not scalable. I have to go in and edit the
298 code to use bits calls to get the next 24 bits, no foolin, or better yet
299 use -n ntuple to set the maximum number of bits teste (that's my real
300 goal, with 24 being the default).
301
302 Plans: David Bauer sent me a fairly extensive patch against 3.29.3beta
303 that fixes some memory leaks and/or speed issues in bits (?) as well as
304 fixing some parts of the diehard OP code -- probably fixing Marsaglia's
305 bugs and not mine, but hard to say. There are bugs in there and I've
306 already squashed several so it wouldn't surprise me if there are more
307 (even more of my own:-). I'm going to TRY to implement most of his
308 fixes if they work well and seem to fix something that makes sense to
309 me, although I'd feel better (per fix) if I could find a test case that
310 illustrates the failure. I may have to ask him how he found the bugs so
311 I can document them in svn somewhere, later. Memory leaks of course are
312 relatively easy and again, I could easily have created some -- getting
313 rid of them is definitely called for. David is also looking at the
314 rgb_bitdist tests (which SHOULD be as sensitive as the OP tests if
315 cranked up to the correct degree) -- there may be some fixes there
316 coming.
317
318 Finally, I have a few operational changes in mind -- primarily adding or
319 altering the new interface in a couple of small ways. Kuiper will go
320 away as an option (but not the code -- I'll leave a macro in place that
321 can switch it back on in case there is ever any point in reconsidering
322 the test, if for example I or David or somebody else figures out an
323 exact CDF for it so it becomes as accurate and perhaps faster than KS,
324 or it is needed for a specific rng test in the future. -k flags will be
325 used to control how hard ks works (and hence how fast vs accurate
326 dieharder is) with a default of pretty fast, pretty accurate and
327 alternatives of slow but to-convergence-exact and really fast but only
328 accurate enough for the short version of the -a(ll) run.
329
330 I'd also like to introduce two new run modes controlled by flags. One
331 of them, -m(ultiplier), will allow the user to enter a scale factor to
332 be applied to the default 100 -psamples used in -a(ll) runs (otherwise
333 ignored). So if one want to run all the tests but with 1000 psamples
334 per test (or 10x whatever the per-test default is) one runs -a -m 10.
335 This should make it MUCH easier to test to destruction, increase test
336 resolution, etc.
337
338 Second, I want to introduce a flag that runs a test "to failure" --
339 something I've planned to do for a long time. David has already hacked
340 in his own version of doing this, and I used to do something very
341 similar in my numerical simulations. The idea of running in -x(treme)
342 mode (or whatever I name the flag) would be to start with e.g. 100
343 pvalues and then add 100 pvalues at a time to the test run until the
344 final pvalue fails a fairly stringent (user selectable) cutoff.
345
346 -d 1 -x 0.000001
347
348 would add psamples to the birthdays test until the final pvalue is under
349 0.000001.
350
351 -d 1 -X 0.000001
352
353 would do the same thing, but it would run the test to this degree of
354 failure psamples = 100 times with different rng seeds (if appropriate)
355 and return something like max, min, mean number of psamples required to
356 cause a test failure. SOMETHING like this is going to be needed,
357 because I think it is entirely plausible that some tests have
358 "poisonous" seeds that have just the right prime modulus to introduce
359 correlations in their stream, but that NEARLY ALWAYS are started with
360 seeds that yield good streams.
361
362 I'd like to have these last two options work for -a -m runs as well, so
363
364 -a -m 10 -X 0.000001
365
366 runs all tests until they fail low or high at one part in a million,
367 1000 times for different seeds per run, returning the average number of
368 psamples required to reach failure. I'd even like to be able to plot
369 the distribution of this number so one can pick out e.g. bimodal
370 distributions (bad seeds!) etc.
371
372 At some point being able to do everything that dieharder will want/need
373 to do is going to require a GUI -- something that can generate scatter
374 plots, candlesticks, real non-ascii histograms, line graphs, 2d/3d
375 surfaces. But that's still a ways in the future. -X is going to be
376 pretty tricky as well, as dieharder isn't equipped to return anything
377 but a final cumulative "pvalue" in [0,1] for a test. But it is probably
378 better to do it now in the beta phase where this doesn't really damage
379 any other future dependent interfaces (e.g R).
380
381
382 ------------------------------------------------------------------------
383 r492 | rgbatduke | 2009-10-12 18:53:13 -0400 (Mon, 12 Oct 2009) | 8 lines
384
385 THis is mostly to check in the dieharder NSF proposal from last year as
386 it has a roadmap for future dieharder development, and I'm thinking hard
387 about adding a few of the many missing generators now that kstest is
388 reliable. I'm still working on kstest, mind you, but it is mostly on
389 the details, not on the basic code.
390
391 rgb
392
393 ------------------------------------------------------------------------
394 r491 | rgbatduke | 2009-10-12 14:55:55 -0400 (Mon, 12 Oct 2009) | 6 lines
395
396 This actually works PRECISELY for all count ranges. It is still in
397 testing -- I've got a bit of work to do to be ready to release this
398 globally (including letting David test it and see if he agrees) but it
399 should COMPLETELY FIX dieharders final kstest (and I'll give it one
400 last opportunity to fix diehard_sums():-).
401
402 ------------------------------------------------------------------------
403 r490 | rgbatduke | 2009-10-12 12:39:51 -0400 (Mon, 12 Oct 2009) | 10 lines
404
405 Checking in some key papers (and some stuff getting rid of broken
406 diehard_sums altogether for now -- leaving in the test but strongly
407 deprecating it in dieharder). The papers SHOULD permit us to compute
408 the exact CDF for the one-sided KS test against a uniform distribution
409 for small N and thereby make the KS test reliable for all sample sizes.
410
411 In particular ks_CDF_N.pdf looks like it will do the trick.
412
413 rgb
414
415 ------------------------------------------------------------------------
416 r489 | rgbatduke | 2009-10-11 11:05:24 -0400 (Sun, 11 Oct 2009) | 32 lines
417
418 FINALLY I got threefish to work. brg_endian.h was broken as shit; it
419 starts off by pulling something from crypt.h that is obviously broken on
420 modern linux boxes (at least my Fedora 11 x86_64 box). The remaining
421 code looks quite general and seems to work, although I have to admit I
422 absolutely hate crap code like this -- it smacks of aimk, imake, and
423 other crap tools that detect platform type using some sort of transient
424 trace from one tool or another that breaks three years later (or rather,
425 requires yet another conditional). I'll leave this in for now in case
426 somebody wants to port to sparc or some bigendian platform, but since we
427 are using threefish only to make random numbers and don't care to ever
428 decrypt the stream of 0's or whatever it is applied to, I honestly doubt
429 that it matters. Getting endianness wrong sounds like it is at worst an
430 extra byte shuffle.
431
432 Either way, this will be 3.29.2beta and I'll put it up on the dieharder
433 website in a few minutes (after a full -a run of threefish passes). I
434 may add a comment to brg_endian.h indicating my hack, lest people be
435 tempted to use it as if it weren't modified.
436
437 Grrr. I'm REALLY tempted to just strip it to the two line definition
438 that is all that matters in skein_port.h and screw the whole "automagic"
439 thing. Robust code is robust code, and there are bound to be
440 intrinsically portable ways to handle endianness IF it is really
441 necessary in the first place.
442
443 At least this finally liberates me to move on and work on kstest and
444 kuiper again. That's been on hold for a few days, but I'm feeling like
445 we're getting close to having one or the other work "perfectly" (if I
446 can find and add the missing O(1/N) correction terms from the
447 literature).
448
449
450 ------------------------------------------------------------------------
451 r488 | rgbatduke | 2009-10-08 12:50:29 -0400 (Thu, 08 Oct 2009) | 25 lines
452
453 This is most of the threefish stuff required, but I'm still having
454 trouble with the big/smallendian conversions apparently needed by skein
455 in threefish. One function that is supposed to be defined automagically
456 is coming out UNdefined in the linker, which is "bad". I may have to
457 ask David Bauer how he got this to compile. Note that I've added both
458 bauer and emelko's current round of bug reports and remarks to the Bugs
459 directory below. David in particular has been really looking hard at
460 kstests, and with good justification. The kstest is apparently very
461 poorly defined even in stats texts and the literature. It is apparently
462 more broken in R than it is in dieharder, and it is still a bit broken
463 in dieharder.
464
465 As is so often the case in dieharder problems, pushing the test suite to
466 new limits exposes weaknesses in code that has long been taken for
467 granted because it has never been used for a rigorous analysis of this
468 sort. But it NEEDS a precise ks or kuiper test, not just a sorta-useful
469 approximate one, or one cannot rely on its statements of weakness or
470 failure.
471
472 Anyway, this checkin is still broken but is within one #define or so of
473 working, I think, once I figure out how to do it without violating the
474 code in the brg_endian.h include that is supposed to automagically
475 select the right Skein function that is currently undefined.
476
477
478 ------------------------------------------------------------------------
479 r487 | rgbatduke | 2009-10-07 12:19:44 -0400 (Wed, 07 Oct 2009) | 16 lines
480
481 David Bauer contribued two cryptographic grade GSL wrapped rngs (one of
482 which I had been working on myself, but his has no dependencies and it
483 works already). rng_aes appears to work, very respectably. It has
484 minimal controls (compared to aespipe) but aespipe is still there if
485 people want to play with it directly. It isn't too shabby speedwise,
486 actually, for what should be a world-class rng. I'm going to see if he
487 (David) cares if I contribute it back to the GSL -- it needs a few
488 generators like this in its collection. Although as it is GPL the
489 answer is obviously not, I think.
490
491 In a second I'm going to insert rng_3fish as a second one. These are
492 enormously useful for testing dieharder itself, and as GPL sources will
493 be useful just being part of the dieharder unless/until they make the
494 GSL.
495
496
497 ------------------------------------------------------------------------
498 r486 | rgbatduke | 2009-10-06 14:26:39 -0400 (Tue, 06 Oct 2009) | 3 lines
499
500 Little fixes, ignore. Added -d 204 to -all properly, fixed its
501 autodocumentation a bit.
502
503 ------------------------------------------------------------------------
504 r485 | rgbatduke | 2009-10-06 14:17:43 -0400 (Tue, 06 Oct 2009) | 35 lines
505
506 This records a validation script to use with aespipe to produce a
507 "standard run" of dieharder in -v3. aespipe with the fixed, trivial
508 256-bit key in aeskey below, is used to encrypt /dev/zero into a stdin
509 stream and fed to dieharder -a. The encrypted stream should be as close
510 to "truly random" as we can currently manage with simple, reasonably
511 fast tools. The interesting thing is that this stream actually PASSES
512 ALL OF THE TESTS in dieharder, even the "known bad" tests such as
513 diehard_operm5. This makes it very, very useful for comparison
514 purposes. For example, for the first time ever, I feel like I can now
515 say that mt19937 actually FAILS dieharder (or has weaknesses that are
516 explicitly exposed by dieharder) when it consistently has tests (even
517 very specific tests for very specific length ntuples) on which it is
518 weak or fails or exhibits high bias in its output pvalues.
519
520 To be fair, passing all of the tests isn't necessarily a good thing,
521 since there are over 100 of them including ntuples. One expects 1/100
522 or 1/200 or thereabouts to be weak for a PERFECT RNG on most runs.
523 Eyeballing the distribution of final P in the aespipe run reveals that
524 dieharder still produces a weak high bias in the final distribution of
525 pvalues, but this is very much in line for the bias revealed by
526 rgb_kstest_test and is therefore very likely an artifact of using -p 100
527 as the default for most of the tests in -a(ll).
528
529 I'm going to run the validation line:
530
531 cat /dev/zero | aespipe -P aeskey | ./dieharder -g 200 -a
532
533 with -p 1000 just for grins (which will take the rest of the day, I
534 expect) and see if it doesn't push the final distribution right back
535 where it belongs, with less visible bias towards the 0.9-1.0 range and
536 away from 0.0-0.1 on the bottom.
537
538 Still, a perfect PASS for a nearly perfect generator. How cool is that?
539
540
541 ------------------------------------------------------------------------
542 r484 | rgbatduke | 2009-10-06 12:45:36 -0400 (Tue, 06 Oct 2009) | 40 lines
543
544 This is a set of changes that:
545
546 a) Fix (for the time being) a problem with ltmain.sh, badly. I
547 suspect that I'll need to add a libtoolize command to the autogen.sh
548 script in order to prevent drift from local libtools in the long run, or
549 give in and make it a link to /usr/share/libtool/config/ltmain.sh and
550 pray that this is portable.
551
552 b) Changes the default ks test in dieharder from broken Kuiper or
553 broken KS to fixed KS. This is an CRITICAL fix and needs to backport to
554 2.28 as with it dieharder will FINALLY give much more nearly correct
555 pvalues for the relatively small number of pvalue samples in the kstest
556 at the end of each test. With the old code one needed two or three
557 orders of magnitude more samples -- at LEAST -p 10000 -- in order for
558 the final pvalue to be not VISIBLY high biased when applied to perfect
559 uniform deviates. With the fix -p 100 works "OK" although -p 1000 would
560 be better and will probably be the default -a(ll) option in 3.x.
561
562 The actual fixes are a single line in dieharder/set_globals.c (change
563 the comment name but not the default number of the ks_test global), a
564 single line in libdieharder/kstest.c, and switching the order in
565 libdieharder/std_test.c so that ks_test == 0 runs kstest, not
566 kuiper_kstest. Fixing the documentation is probably not worth it in
567 2.28.
568
569 I would suggest still holding out on actually making the fixes for a
570 bit, as I'm actively playing with things and testing out the new code
571 (in a moment with aespipe as I still haven't finished rng_aes). The
572 changes are preserved and saved as 3.29.1beta, though, with the addition
573 of a very useful and useable rgb_kstest_test routine that can be used to
574 further debug and/or improve the kstest used to generate final test
575 pronouncements of pass/fail/weak etc. And we still need to decide if it
576 is time to move on to v3, as a lot of people are using it and it seems
577 to be stable and usable and has lots of bug fixes and feature
578 enhancements (including much better future scalability as I add tests
579 and generators).
580
581 rgb
582
583
584 ------------------------------------------------------------------------
585 r483 | rgbatduke | 2009-10-06 09:08:59 -0400 (Tue, 06 Oct 2009) | 7 lines
586
587 The new rgb_kstest_test in this version actually works, but it looks
588 like we have some sort of libtool derived bug in the build. I'm
589 checking in clean so I can rerun libtoolize, which will hopefully get me
590 a new ltmain.sh, which will hopefully build a libtools script that
591 contains the correct ECHO/echo lines and perhaps deals with MODE
592 correctly.
593
594 ------------------------------------------------------------------------
595 r482 | rgbatduke | 2009-10-04 09:44:12 -0400 (Sun, 04 Oct 2009) | 15 lines
596
597 Ignore today's checkins, Dirk. I'm adding a new test (rgb_kstest_test)
598 to test the kstest routines (as well as to MAYBE function as a new test
599 in the suite, but I doubt that it will be sensitive enough to be any
600 use). Basically, I plan to fill a vector with tsamples uniform
601 deviates, run a kstest on them (which tests for uniformity and generates
602 a pvalue that should itself be a uniform deviate) to fill in the usual
603 vector of pvalues and run the final kstest on that. A kstest SHOULD
604 recursively take uniform deviates to a uniform deviate, for a large
605 enough set of uniform deviates, and I want to find out a) if this is
606 true; and b) if it is true just what a "large enough set" is. This test
607 should help me find out both, and if a) is incorrect, to perhaps "fix"
608 the kstest as this is their THEORETICAL behavior and failure to
609 accomplish this indicates a bug in the code or a real problem in the
610 theory...
611
612 ------------------------------------------------------------------------
613 r481 | rgbatduke | 2009-10-02 16:32:30 -0400 (Fri, 02 Oct 2009) | 3 lines
614
615 This checks in what might be a VERY IMPORTANT fix to kstest, due to
616 David Bauer. Needs more testing, though, with a world class crypt.
617
618 ------------------------------------------------------------------------
619 r480 | rgbatduke | 2009-03-17 08:27:23 -0400 (Tue, 17 Mar 2009) | 2 lines
620
621 Checking in so I can leave.
622
623 ------------------------------------------------------------------------
624 r479 | rgbatduke | 2009-03-17 00:28:25 -0400 (Tue, 17 Mar 2009) | 3 lines
625
626 This is broken as far as the aes generator is concerned, AND I'll
627 probably need to put libaes into the dieharder packaging.
628
629 ------------------------------------------------------------------------
630 r478 | rgbatduke | 2009-01-29 10:57:43 -0500 (Thu, 29 Jan 2009) | 8 lines
631
632 Checking in a LOT of changes and additions associated with v3 -- I've
633 been holding them so as not to screw up the RDH side of things before
634 everything stabilizes. A lot of the stuff below is documentation
635 intended to guide future development and additions. Some of it is fixes
636 (data and otherwise) in diehard tests. Some of it fixes the way
637 dieharder (the binary, not library) initializes (and adds local tests)
638 and runs all tests.
639
640 ------------------------------------------------------------------------
641 r477 | rgbatduke | 2008-10-08 15:11:30 -0400 (Wed, 08 Oct 2008) | 3 lines
642
643 Sending in a minor change to START fixing up parsecl.c to be more
644 robust.
645
646 ------------------------------------------------------------------------
647 r476 | rgbatduke | 2008-09-29 22:22:38 -0400 (Mon, 29 Sep 2008) | 26 lines
648
649 This checkin should make Mattias "perfectly happy". It enables:
650
651 rgb@lilith|B:1140>./dieharder -a -D default -D -1 -D prefix -D
652 no_whitespace -D show_num -s 1
653 0|rng_name|num|rands/second|
654 1|mt19937|13|1.17e+08|
655 0|test_name|num|ntup|tsamples|psamples|p-value|Assessment|Seed
656 2|diehard_birthdays|0|0|100|100|0.16302070|PASSED|3542794731
657 2|diehard_operm5|1|5|1000000|100|0.04115096|PASSED|2304163927
658 2|diehard_rank_32x32|2|0|40000|100|0.92631752|PASSED|2245496723
659 2|diehard_rank_6x8|3|0|100000|100|0.86585575|PASSED|3223183182
660 2|diehard_bitstream|4|0|2097152|100|0.60520232|PASSED|2615297461
661 2|diehard_opso|5|0|2097152|100|0.05852624|PASSED|1542897414
662
663 which is, AFAICT, exactly what he wants. Oh, he wants the full test
664 name as an output field option instead of the short name, but he might
665 have to wait on that...
666
667 This also checks in a couple of minor bugfixes reported by Mattias and
668 Marc Abel. Marc has another feature request I haven't looked at yet.
669 Both of them are using dieharder quite heavily in the beta version, so
670 I'm hoping that it is shaking out. I'm also hoping this round of
671 changes didn't break anything.
672
673 Not quite ready for a release, but perhaps getting closer.
674
675 ------------------------------------------------------------------------
676 r475 | rgbatduke | 2008-09-22 19:10:01 -0400 (Mon, 22 Sep 2008) | 2 lines
677
678 Added small section to man page on output control.
679
680 ------------------------------------------------------------------------
681 r474 | rgbatduke | 2008-09-22 07:24:15 -0400 (Mon, 22 Sep 2008) | 12 lines
682
683 Small changes to add dieharder-config.in to the Makefile.am and to
684 get the rpm to autobuild with a split between @VERSION@ and
685 @DIEHARDER_LT_VERSION@ -- I basically twinned the latter into
686 @DIEHARDER_LIB_VERSION@. A slight pain, but it means the library can
687 have a different version (no beta) compared to the program (with beta).
688
689 The successful RPM build means that everything is in place, although
690 there is still cruft in include and probably libdieharder and there may
691 be NON-cruft that isn't in the repo. But I gotta go and won't find it
692 now.
693
694
695 ------------------------------------------------------------------------
696 r473 | rgbatduke | 2008-09-22 03:29:33 -0400 (Mon, 22 Sep 2008) | 24 lines
697
698 OK, this is PARTIALLY decrufted -- I doubt that it is finished yet, and
699 I haven't even started to tackly the proper decrufting of the library.
700 I've cleaned up the dieharder man page, checked all the autodocumenting
701 features of dieharder, and run a bunch of tests. I've preemptively
702 fixed around three or four bugs, and finished implementing a couple of
703 features that were missing on the previous checkin, e.g. the ability to
704 use any of:
705
706 dieharder -d diehard_sums -g 6
707 dieharder -d 14 -g gfsr4
708 ...
709
710 (all tests AND rngs selectable by name OR number. It is 3:20 am, and I
711 have to get up by 6:40. It is therefore bedtime.
712
713 If I haven't forgotten to checkin any files, it should build and run
714 pretty well. Probably not perfectly, but pretty well. Matthias should
715 be happy -- if he uses -c ',' and -D prefix, he'll get close to exactly
716 what he wants. Everything seems to be working as far as I can tell with
717 limited testing.
718
719 Might be a day or three before I can really tackle this again. G'night.
720
721
722 ------------------------------------------------------------------------
723 r472 | rgbatduke | 2008-09-22 00:13:56 -0400 (Mon, 22 Sep 2008) | 92 lines
724
725 OK, this is a checkin of dieharder 3.28.0beta. It is NOT fully
726 decrufted, but seems to mostly nearly hopefully all work. That probably
727 means that there are only a dozen or two bugs. There are also a few
728 API features I haven't implemented in the UI yet -- specifically the
729 reporting of errors (like a rewind of a file in mid-test). So this
730 ISN'T really a beta -- more like an alpha.
731
732 Dirk, please do not start converting this over into Rdh yet. I'm
733 checking it in for two reasons -- one is that I have to remove a whole
734 pile of files to decruft and svn won't let me until I check in. Another
735 is that I NEED to checkin -- it makes me nervous to have this large a
736 delta not checked in. There are probably a couple of critical sources
737 I've forgotten to add entirely and I won't know until I check in and
738 check out and build a fresh clean copy.
739
740 MOST of what I've done, from Rdh's point of view, should be invisible
741 after Rdh is (fairly minimally) hacked one last time. Basically Rdh
742 should use its own version of set_globals.c (or patch mine, or ifdef
743 mine). Note that there are a lot fewer variables, and this list may
744 shrink a bit more.
745
746 ntuple's meaning hasn't changed, and you already handle that.
747
748 Seed and strategy work together -- the latter is a new variable and
749 SHOULDN'T affect Rdh, but just in case, here's what it does.
750
751 The default strategy for dieharder is to reseed once when a rng is
752 chosen. In the default output view, the seed is written to the rng
753 information part of the header. That way if one wishes to reproduce
754 a test result, one can enter the seed with -S seed.
755
756 However, this is actually a PROBLEM if one runs multiple tests from this
757 one seed. If one runs tests out of order, the results will be
758 different. This is true for me running all the tests in order via
759 dieharder -a if I should ever change test order, and is true for Rdh if
760 one runs first one test, then another in different orders, from the same
761 single specification of rng and/or Seed.
762
763 Also, there may be situations where one wants to run a single test
764 multiple times, each time from a newly selected seed to (in essence)
765 determine if some seeds are "bad" for a given rng. dieharder doesn't
766 yet support that, but I think that in R it would be pretty easy.
767
768 Setting strategy to anything nonzero (say, 1) causes dieharder to reseed
769 the random number generator at the beginning of any test. If -S seed
770 was NOT specified, it just generates a new random seed, so that one
771 could run e.g. diehard_birthdays 100 times in a loop and each one would
772 reseed anew with a new random seed. If -S seed IS specified, it uses
773 the specified seed. If a file is being used for input (not stdin) it
774 forces a rewind at the beginning of each test, which is actually not a
775 bad thing to do as it conserves rands. (I hope, I haven't yet tested
776 this latter feature much yet but it should work.:-).
777
778 SO, Rdh will probably just leave strategy = 0 alone and either set the
779 global value of Seed for one-time initialization with a fixed seed or
780 not, accepting a one-time random seed. But you CAN support strategy if
781 you ever think you need to.
782
783 (From my point of view its primary purpose is to make the creation of a
784 validation run trivial -- if I run
785
786 dieharder -a -S 1 -s 1
787
788 I generate a validation table. If I run
789
790 dieharder -a -S 1 -s 1 -D test_name -D pvalues
791
792 I generate a very sparse validation table (basically, just test name and
793 pvalue). You can throw a -c ' ' in there if you want white space
794 separation or -c ',' if you prefer comma separated values, etc. You
795 have nearly complete control over dieharder's output at this point, see:
796
797 dieharder -F
798
799 for a listing of output control flags. dieharder -l and dieharder -h
800 and dieharder -g -1 all work as before, but I've completely flattened
801 and rationalized test-space so it works just like rng-space (and I mean
802 JUST like it -- very similar setup.
803
804 I completely changed (seriously streamlined and cleaned up) the test
805 call procedure, so a SINGLE run_test() routine does pretty much all the
806 work, a SINGLE output() routine does all of THAT work, and so that all
807 the dieharder CLI-specific stuff is done in parsecl(), and then only if
808 you enter specific commands, or in dieharder.c (main()). I tried to
809 label things that are CLI specific there as well.
810
811 If you want to grab a copy of this and build it and play with it, feel
812 free. As soon as this checkin is complete, though, I'm going to start
813 decrufting and checking to be sure I have all the required modules
814 actually in the repo.
815
816
817 ------------------------------------------------------------------------
818 r453 | rgb | 2008-09-10 07:16:17 -0400 (Wed, 10 Sep 2008) | 14 lines
819
820 Dearie me. This checkin actually works, although I still haven't
821 implemented the output.c patch needed by Dirk or fixed the missing .h
822 file in include/dieharder. Still, I >>HAVE<< turned all the tests into
823 type int's (still no returns) and stripped dieharder.h and split off a
824 globals.h file that I don't think I'm going to need, actually, although
825 it was useful while stripping dieharder.h as a reservoir of codelets
826 that I needed to put back.
827
828 Anyway, it is entirely possible that Dirk will read these words as I
829 ALSO have dieharder.googlecode.com set up with him on it, and while this
830 checkin is still local (about to be svnsync'd up, not directly checked
831 in) VERY SOON NOW I may try checkout out from the google repo, which
832 will of course check BACK into the google repo thereafter.
833
834 ------------------------------------------------------------------------
835 r452 | rgb | 2008-09-09 18:38:14 -0400 (Tue, 09 Sep 2008) | 4 lines
836
837 This checks in a code fragment that reseeds the rng at the beginning of
838 each run_whatever segment. This fragment "guarantees" that every test
839 run uses a fixed Seed if it is set.
840
841 ------------------------------------------------------------------------
842 r451 | rgb | 2008-09-08 01:18:25 -0400 (Mon, 08 Sep 2008) | 2 lines
843
844 This now works. 2.28.1 indeed I dub thee.
845
846 ------------------------------------------------------------------------
847 r450 | rgb | 2008-09-07 23:53:50 -0400 (Sun, 07 Sep 2008) | 3 lines
848
849 This seems to fix the output of sts_serial so it is consistent. I do
850 have a few small bugs to clean up to get a "perfect" display.
851
852 ------------------------------------------------------------------------
853 r449 | rgb | 2008-09-07 13:58:56 -0400 (Sun, 07 Sep 2008) | 2 lines
854
855 So this is 2.28.1, for the moment and sake of argument.
856
857 ------------------------------------------------------------------------
858 r448 | rgb | 2008-09-07 13:58:00 -0400 (Sun, 07 Sep 2008) | 2 lines
859
860 This is ready to get some sort of rev boost.
861
862 ------------------------------------------------------------------------
863 r447 | rgb | 2008-09-07 10:00:25 -0400 (Sun, 07 Sep 2008) | 6 lines
864
865 This adds a "new" test -- the rgb_lagged_sums test, which is the
866 user test but wrapped up to run on a whole sequence of values, the
867 way I need to make sts_serial run very shortly. It is sufficient
868 to CLEARLY SHOW that mt19937 is actually a weak generator -- it
869 is "too uniform".
870
871 ------------------------------------------------------------------------
872 r444 | rgb | 2008-09-06 19:50:52 -0400 (Sat, 06 Sep 2008) | 2 lines
873
874 This SHOULD be everything. Table output mode should now work for -a.
875
876 ------------------------------------------------------------------------
877 r443 | rgb | 2008-09-06 18:43:39 -0400 (Sat, 06 Sep 2008) | 3 lines
878
879 This fixes four more tests. I can now run a LOT of the way through -a
880 before reverting to the old style output.
881
882 ------------------------------------------------------------------------
883 r442 | rgb | 2008-09-06 14:37:53 -0400 (Sat, 06 Sep 2008) | 8 lines
884
885 This is coming along nicely. I have pretty much everything set up for
886 table vs report output and am streamlining the run_whatever routines to
887 the point where only a tiny bit of test-specific initiation
888 differentiates them. If I could move it into the test itself, I could
889 pretty much completely simplify the dieharder CLI code to a single
890 generic test shell call plus a SMALL set of specialized calls for e.g.
891 benchmarks or non-standard tests that don't return pvalues per se.
892
893 ------------------------------------------------------------------------
894 r441 | rgb | 2008-09-05 15:08:47 -0400 (Fri, 05 Sep 2008) | 7 lines
895
896 This actually now works to display tables. Time to hack hack hack
897 and make ALL the tests work with a table. Right now most of them will
898 ignore it. I really should combine report and table in "results" and
899 put a SINGLE CALL to results in run_whatever.c. Results must then
900 "do it all". Also, I think I'm going to run the benchmarker on ALL
901 non-filesystem tests to get a timing, IF the table/timing flag is set.
902
903 ------------------------------------------------------------------------
904 r440 | rgb | 2008-09-05 08:39:13 -0400 (Fri, 05 Sep 2008) | 3 lines
905
906 This applies a small patch from Dirk that just cleans up a few issues in
907 dieharder.h.
908
909 ------------------------------------------------------------------------
910 r439 | rgb | 2008-09-05 07:20:58 -0400 (Fri, 05 Sep 2008) | 3 lines
911
912 OK, we'll try ONE LAST change -- commenting out the fclose -- before
913 sending this to Dirk.
914
915 ------------------------------------------------------------------------
916 r438 | rgb | 2008-09-05 07:16:01 -0400 (Fri, 05 Sep 2008) | 15 lines
917
918 This is a 2.27.14 checkin. I'll shoot it off to Dirk who is waiting on
919 it. I noted, however, that my NEW permutations test FAILS (or should
920 fail) mt19937! It produces too GOOD a spread of permutations,
921 consistently! This is fascinating information. It means that perhaps
922 operm5 is NOT broken; maybe getting permutations to work out
923 multinomially is the most difficult test one can put an rng to.
924
925 Something to look at later.
926
927 This checkin should fix dieharder and libdieharder so that one can run
928 multiple tests (including invocation of the same test or different rngs)
929 from a single dieharder call. This cannot happen in dieharder -- don't
930 worry. But it can and will happen in Rdh and in gdieharder, so it is an
931 important fix nonetheless.
932
933 ------------------------------------------------------------------------
934 r437 | rgb | 2008-09-04 17:07:26 -0400 (Thu, 04 Sep 2008) | 20 lines
935
936 This actually checks in so that we can pop the snapshot number in just a
937 minute -- a major bugfix relative to Rdh. The two problems addressed
938 herein are:
939
940 a) In order to be able to run many tests, one after another, on many
941 rngs, one after another, I have to be more careful than I have been
942 about allocating and freeing test resources on the one hand, rngs (which
943 must be freed) on the other hand, and resetting the static bit buffers
944 used in bits.c on the third hand.
945
946 b) I also needed to set up startup with a split into two sections, one
947 that runs only one time period and one that runs every time a new test
948 is created, executed, and destroyed (with everything reset at the end to
949 run a new test).
950
951 The good thing beyond Rdh in these changes is that several of them are
952 equally necessary in a GUI version e.g. gdieharder. So nothing being
953 done here is wasted...
954
955
956 ------------------------------------------------------------------------
957 r436 | rgb | 2008-08-19 12:44:28 -0400 (Tue, 19 Aug 2008) | 4 lines
958
959 This really, really should be IT! I dub thee 2.27.12 as I've got to get
960 on with things. This version works for BOTH x86_64 and i386. It should
961 build into debian packages. It should rpmbuild --rebuild.
962
963 ------------------------------------------------------------------------
964 r435 | rgb | 2008-08-19 12:34:56 -0400 (Tue, 19 Aug 2008) | 5 lines
965
966 This is a nearly final (for now) checkin for 2.27.12. It builds
967 decently. I probably need to put back config.sub as it is one of the
968 many things that should get automagically rebuilt -- if I have a
969 placeholder for it already present.
970
971 ------------------------------------------------------------------------
972 r434 | rgb | 2008-08-19 12:16:18 -0400 (Tue, 19 Aug 2008) | 4 lines
973
974 We'll give this a try -- this defines __auto_build_post at the top of
975 the specfile and SEEMS to prevent the check-buildroot crash in rpm
976 building without my particular .rpmmacros.
977
978 ------------------------------------------------------------------------
979 r433 | rgb | 2008-08-19 11:50:28 -0400 (Tue, 19 Aug 2008) | 2 lines
980
981 With any luck at all, this will be ready to fly.
982
983 ------------------------------------------------------------------------
984 r432 | rgb | 2008-08-19 11:42:31 -0400 (Tue, 19 Aug 2008) | 2 lines
985
986 We're working our way back to not losing everything we just did, dammit.
987
988 ------------------------------------------------------------------------
989 r431 | rgb | 2008-08-19 11:34:04 -0400 (Tue, 19 Aug 2008) | 2 lines
990
991 Try try again...
992
993 ------------------------------------------------------------------------
994 r430 | rgb | 2008-08-19 11:31:38 -0400 (Tue, 19 Aug 2008) | 3 lines
995
996 Sending this towards lucifer. We really need to make sure everything is
997 in subversion!
998
999 ------------------------------------------------------------------------
1000 r429 | rgb | 2008-08-19 07:40:00 -0400 (Tue, 19 Aug 2008) | 4 lines
1001
1002 This MAY fix things up so that they rebuild more cleanly. I'll try
1003 sending them on to Dirk directly in tarball form to see if they make him
1004 all happy.
1005
1006 ------------------------------------------------------------------------
1007 r428 | rgb | 2008-08-19 06:43:42 -0400 (Tue, 19 Aug 2008) | 4 lines
1008
1009 Well, perhaps we should send this in since we could be suffering from
1010 early hard disk crash syndrome. In fact, I should drag out my 8 GB USB
1011 flash (at least) and back up Src if nothing else.
1012
1013 ------------------------------------------------------------------------
1014 r427 | rgb | 2008-08-19 06:22:07 -0400 (Tue, 19 Aug 2008) | 2 lines
1015
1016 Moving math down to bugs, as it is really a set of patches.
1017
1018 ------------------------------------------------------------------------
1019 r426 | rgb | 2008-08-19 06:21:33 -0400 (Tue, 19 Aug 2008) | 14 lines
1020
1021 We're rearranging this so I can track bugs a bit better. I've really
1022 got only two outstanding ones -- getting aclocal to find libtool on a
1023 system I don't control or understand (problem possible solved by
1024 rerunning ./autogen.sh on that system) and getting the hard path of
1025 missing out of the toplevel Makefile and replacing it with a relative
1026 one so that it can be built without rerunning autogen.sh, maybe. Oh,
1027 and some sort of rpm build wierdness.
1028
1029 Also, Dirk wants me to decruft to where -Wall produces no unused
1030 variable complaints. We'll see about that -- MAYBE I'll do it,
1031 eventually, but unused variables are really pretty harmless and are
1032 sometimes useful and I usually ignore them until I have too much time on
1033 my hands.
1034
1035 ------------------------------------------------------------------------
1036 r425 | rgb | 2008-08-19 06:01:36 -0400 (Tue, 19 Aug 2008) | 4 lines
1037
1038 This removes RDieHarder (which Dirk is maintaining separately at this
1039 point). It fixes the newly improved man page (which was broken). It
1040 removes a redundant R directory.
1041
1042 ------------------------------------------------------------------------
1043 r424 | rgb | 2008-08-18 18:10:44 -0400 (Mon, 18 Aug 2008) | 3 lines
1044
1045 This fixes a few small errors in the 2.27.11 abstract; no changes to the
1046 actual program, though.
1047
1048 ------------------------------------------------------------------------
1049 r423 | rgb | 2008-08-18 16:15:49 -0400 (Mon, 18 Aug 2008) | 19 lines
1050
1051 This checks in the ChangeLog itself, as well as the last few
1052 minor changes in the tests run with the -a command and the
1053 test listing function in the CLI. I think I declare this to be:
1054
1055 2.27.11
1056
1057 and we'll start working on 12. I think this actually fixes
1058 all the outstanding bugs except for the problem in sums and
1059 the problem in operm5 in diehard (and of course there are tests
1060 half implemented as well).
1061
1062 I THOUGHT I'd fixed diehard_sums, actually, but it still most definitely
1063 fails on large -p. I'll have to check sums out explicitly to see why it
1064 doesn't seem to work at the moment. It's possible I dropped the fixes
1065 when recently changing systems.
1066
1067 Anyway, here it is. Bring on the bugs.
1068
1069
1070 ------------------------------------------------------------------------
1071 r422 | rgb | 2008-08-18 15:50:15 -0400 (Mon, 18 Aug 2008) | 21 lines
1072
1073 This might be ready for a production snapshot, or a testing snapshot,
1074 or something like that. At any rate, I have BOTH fixed the generator
1075 problem (where rngs have numbers that can change as new generators
1076 are added. I fixed this by significantly increasing the size of the
1077 types[] vector and working moderately hard to pack it with generators
1078 sparsely and sanely -- sparse so there is plenty of room, sane because
1079 the space is now blocked out by general category with room to grown.
1080
1081 This will NOT GUARANTEE that the GSL will not change its internal order,
1082 and dieharder will continue to follow the GSL order where they overlap
1083 as nothing else makes sense.
1084
1085 I >>also<< think I've fixed -r 5, rgb_permutations. This is a
1086 non-overlapping permutations test and runs to more or less arbitrary
1087 order (and seems to work quite nicely where it runs, although it gets
1088 quite slow when we get to very large permutations as one might expect).
1089
1090 I think it is time to bump the minor-minor number to 11 and post it, for
1091 the time being with -r 5 still marked out as experimental in the
1092 help listing.
1093
1094 ------------------------------------------------------------------------
1095 r421 | rgb | 2008-08-18 14:33:06 -0400 (Mon, 18 Aug 2008) | 5 lines
1096
1097 This fixes the only real bug, I think -- dieharder now dies with a
1098 complaint if it is invoked with an invalid -g entry, sort of. Hmmm,
1099 actually I might need to tweak this fix just a bit more to check
1100 on the complete range.
1101
1102 ------------------------------------------------------------------------
1103 r420 | rgb | 2008-08-18 14:29:17 -0400 (Mon, 18 Aug 2008) | 3 lines
1104
1105 This actually works and has the number spaces previously discussed
1106 all implemented and everything.
1107
1108 ------------------------------------------------------------------------
1109 r419 | rgb | 2008-08-17 20:15:08 -0400 (Sun, 17 Aug 2008) | 6 lines
1110
1111 This is a step in the process of figuring out how to "fix" the numbers
1112 for rngs to test. I think I'm going to have to move my own setup
1113 routine to libdieharder. But I'm still trying to figure out whether or
1114 not I can or should replace the gsl code itself, or run something in
1115 parallel.
1116
1117 ------------------------------------------------------------------------
1118 r417 | rgb | 2008-08-17 13:30:41 -0400 (Sun, 17 Aug 2008) | 3 lines
1119
1120 This rearranges the Bugs stuff. I need to create a "cleared" directory
1121 in Bugs so I can see what is done, as well.
1122
1123 ------------------------------------------------------------------------
1124 r416 | rgb | 2008-08-17 12:31:18 -0400 (Sun, 17 Aug 2008) | 5 lines
1125
1126 This updates the abstract, which is also important documentation and has
1127 been added to the %doc lines in the specfile. I should really create a
1128 make installdoc target in Makefile.am that can install a bunch of
1129 documents, but that is for another day.
1130
1131 ------------------------------------------------------------------------
1132 r415 | rgb | 2008-08-15 19:03:26 -0400 (Fri, 15 Aug 2008) | 4 lines
1133
1134 This is 2.27.10-1, and adds stdin support and fixes a few bugs in the
1135 specfile. I'm posting it for distribution, although it is still not
1136 "finished" -- this is very much development snapshot time.
1137
1138 ------------------------------------------------------------------------
1139 r414 | rgb | 2008-08-15 14:14:45 -0400 (Fri, 15 Aug 2008) | 4 lines
1140
1141 This adds the capability to input a binary stream into stdin. This
1142 patch is due to Paul Serice. He also included a (standalone?) rng to
1143 test with, which I'll mess with in a moment.
1144
1145 ------------------------------------------------------------------------
1146 r413 | rgb | 2008-08-03 08:40:35 -0400 (Sun, 03 Aug 2008) | 3 lines
1147
1148 Added a version string to help. Probably should add one elsewhere, e.g.
1149 at the top of rng listing and test listing.
1150
1151 ------------------------------------------------------------------------
1152 r412 | rgb | 2008-08-01 09:11:10 -0400 (Fri, 01 Aug 2008) | 7 lines
1153
1154 Had to back off the svn2cl in the rpm build -- it now has to be run BY
1155 HAND before just MY rpmbuild (dummy me) so that it won't fail for end
1156 users without svn2cl and an svn repo to match. It still builds
1157 perfectly, and I'll do a full rpm build cycle including this step to get
1158 this last comment into the changelog.
1159
1160
1161 ------------------------------------------------------------------------
1162 r411 | rgb | 2008-07-29 12:01:51 -0400 (Tue, 29 Jul 2008) | 46 lines
1163
1164 ...And it is. So henceforth ALL of my svn checkins will be duly noted
1165 in the ChangeLog, which will therefore get to be very long indeed, as I
1166 check in a lot.
1167
1168 Damn! I guess this means no more swearing, and I sure hope I haven't
1169 impugned anyone's character in the first 400 plus checkins. At least
1170 not too much. Hmmm, I wonder just what I HAVE said down there. I tend
1171 to treat SVN logs as sort of a free-association commentary about the
1172 project in addition to being a place to document changes. Some changes
1173 have "real" documentation in the form of the svn diffs, as well -- I
1174 haven't been too specific about just EXACTLY what I've changed every
1175 time.
1176
1177 Well, better a ChangeLog up to date and verbose than one months to years
1178 old, and I'll be hogtied and hung by the nose before I write a precisely
1179 formatted description of changes day to day as I work on this -- twice.
1180 So SVN it is (dating back to when it was really CVS, BTW, imported to
1181 SVN).
1182
1183 Next, I suppose we should provide at least >>a<< fix for the problem of
1184 shifting generator numbers, and on my own I want to clean up some
1185 issues with file read -- arrange it so we can read in very large files
1186 and eliminate/document file-based usage redundancies such as needing -g
1187 65 AND -f filename. This will be a bit tricky, as dieharder can read in
1188 raw (binary) as well as ascii in many formats, and using the -g flag to
1189 differentiate at least raw vs cooked has been an easy solution.
1190
1191 Regarding numbers, it has been suggested that we do two things:
1192
1193 a) Take action to ensure that the numbers don't change any more, even if
1194 the gsl or I or a user add new generators. This is actually a bit of a
1195 PITA (much as I see the motivation for doing so, as moving numbers break
1196 user scripts) because the way generators are added is to run through all
1197 the supported ones of each type and just add them in gsl or call order.
1198 However, there are several ways I can do it (some requiring more
1199 programming than others) and I'll see which one looks to be easiest and
1200 most robust.
1201
1202 b) Arrange it so we can call generators by name, not just by number.
1203 Names would presumably not change even if numbers were inserted, and
1204 many users will be interested in testing a single generator they are
1205 thinking of using by name.
1206
1207 rgb
1208
1209
1210 ------------------------------------------------------------------------
1211 r410 | rgb | 2008-07-29 11:40:30 -0400 (Tue, 29 Jul 2008) | 5 lines
1212
1213 OK, this seems to work, autobuilding ChangeLog from the svn log of my
1214 timestamp file, that basically picks up a message like this one every
1215 time I do an svn checkin. We'll see in a moment if this is reflected in
1216 ChangeLog on the next build.
1217
1218 ------------------------------------------------------------------------
1219 r409 | rgb | 2008-07-29 11:26:48 -0400 (Tue, 29 Jul 2008) | 2 lines
1220
1221 Just testing...
1222
1223 ------------------------------------------------------------------------
1224 r407 | rgb | 2008-07-22 09:06:14 -0400 (Tue, 22 Jul 2008) | 5 lines
1225
1226 This is, in fact, a major checkin. A serious won't-build-on-lib64 bug
1227 was identified and resolved, the resolution being the addition of
1228 --libdir=... to the ./configure line in the toplevel specfile.
1229 Truthfully, I should probably add this to autogen.sh.
1230
1231 ------------------------------------------------------------------------
1232 r406 | rgb | 2008-03-27 12:33:04 -0400 (Thu, 27 Mar 2008) | 8 lines
1233
1234 This is getting close to a release snapshot once again. We do need to
1235 set the number of tests correctly. I may have to give up on operm5 or
1236 rgb_operm for the time being (again) but at least I have
1237 rgb_permutations and rgb_bitdist, both of which test similar things
1238 along with sts_serial ditto. I do need to try to fix sums. I'd love to
1239 add a couple of the contributed tests. I do need to go through the
1240 opso, dna, etc. tests and "fix" them wrt overlap and so on.
1241
1242 ------------------------------------------------------------------------
1243 r405 | rgb | 2008-03-26 21:20:57 -0400 (Wed, 26 Mar 2008) | 4 lines
1244
1245 This APPEARS to be a functional permutations test, as expected/hoped.
1246 It uses NON-overlapping samples, although at a guess it would work
1247 for overlapping ones, just not as well.
1248
1249 ------------------------------------------------------------------------
1250 r404 | rgb | 2008-03-26 17:09:43 -0400 (Wed, 26 Mar 2008) | 4 lines
1251
1252 This is one, last, poor attempt to get SOME sort of functioning
1253 permutations test. I actually had this working "inside" rgb_operm()
1254 so I imagine I will have it working again here shortly.
1255
1256 ------------------------------------------------------------------------
1257 r403 | rgb | 2008-03-26 13:39:14 -0400 (Wed, 26 Mar 2008) | 6 lines
1258
1259 This appears to be "fixed" so that bitstream runs without complaint. I
1260 will need to check for memory leaks in the long run (heh heh). This
1261 once again takes me to where "everything works" but sums operm5, I
1262 think. I'll do another standard S 1 run in a second to make a new
1263 snapshot with a fixed/reference bitstring.
1264
1265 ------------------------------------------------------------------------
1266 r402 | rgb | 2008-03-24 15:39:49 -0400 (Mon, 24 Mar 2008) | 6 lines
1267
1268 This fixes up all the tests that seem to require chisq with explicit
1269 cutoffs (frequently one that will be ignored). Cutoff does matter
1270 in the rgb_bitdist test, though! I think all of these tests are at last
1271 valid through -p 100 and possibly -p 1000, except for sums, bitstream
1272 (which I'll fix shortly) and the perennial operm5.
1273
1274 ------------------------------------------------------------------------
1275 r401 | rgb | 2008-03-24 00:18:51 -0400 (Mon, 24 Mar 2008) | 4 lines
1276
1277 I am HOPING that this is all good, finally. It runs and gives very
1278 reasonable answers (once) at -p 1000. Now to see if this is
1279 semiconsistent. I've fixed a few zillion small bugs...
1280
1281 ------------------------------------------------------------------------
1282 r400 | rgb | 2008-03-22 18:33:47 -0400 (Sat, 22 Mar 2008) | 6 lines
1283
1284 This repairs one serious bug -- I was misinitializing itail -- and makes
1285 it a bit more precise as far as counting dofs.
1286
1287
1288
1289
1290 ------------------------------------------------------------------------
1291 r399 | rgb | 2008-03-22 15:54:02 -0400 (Sat, 22 Mar 2008) | 6 lines
1292
1293 This MAY at long last, actually fix rgb_bitdist by simultaneously fixing
1294 vtests in general. Our chisq now AUTOMATICALLY bundles the tail(s) or
1295 low spots in a chisq, independent of where they occur. It looks like I
1296 failed to set ndof correctly, so there will be another round of tests
1297 and a quick checkin in a moment.
1298
1299 ------------------------------------------------------------------------
1300 r398 | rgb | 2008-03-19 02:23:40 -0400 (Wed, 19 Mar 2008) | 13 lines
1301
1302 THIS IS A WORKING STS SERIAL as far as I can tell. It correctly reveals
1303 known weak generators. It seems to pass known "decent" generators. It
1304 is very sensitive to certain kinds of failure.
1305
1306 I now should redo rgb_bitdist to be the NON-overlapping version of this
1307 test, at the very least comparing the freq output (generated with the
1308 same code, I think, ported the other way) to a multinomial chisq on the
1309 expected values. If I can actually check the distribution of each
1310 number against an asymptotically normal curve, that would be great too.
1311 If I could IN THE SAME TEST check the asymptotic fraction -- something
1312 like the monkey tests do -- that would be just fabulous! Especially if
1313 I use a normal and make the comparison very good and not empirical.
1314
1315 ------------------------------------------------------------------------
1316 r397 | rgb | 2008-03-18 20:45:03 -0400 (Tue, 18 Mar 2008) | 4 lines
1317
1318 This is very, very close to working perfectly WITH LABELS for
1319 all the histograms generated by sts_serial. We'll save it just because
1320 it looks really useful.
1321
1322 ------------------------------------------------------------------------
1323 r396 | rgb | 2008-03-18 17:13:51 -0400 (Tue, 18 Mar 2008) | 9 lines
1324
1325 As far as I can tell, sts_serial now works perfectly. Time to clean up
1326 its act. I will start by decrufting some of the superfluous output from
1327 the library call, then I'm going to FIX the test struct so that each
1328 test has its OWN descriptor line that can be set along with its pvalue
1329 and that will be used by the histogram and/or single line output routine
1330 to identify the associated statistic. I have to make it a bit easier
1331 to make a single test call and get back the MATRIX of p-values for a set
1332 of statistics all generated with the single call.
1333
1334 ------------------------------------------------------------------------
1335 r395 | rgb | 2008-03-17 00:58:13 -0400 (Mon, 17 Mar 2008) | 4 lines
1336
1337 This, finally, appears to work perfectly. The last thing I need to do
1338 is to put all of the pvalues somewhere to be passed back to the caller,
1339 in order.
1340
1341 ------------------------------------------------------------------------
1342 r394 | rgb | 2008-03-16 21:44:35 -0400 (Sun, 16 Mar 2008) | 3 lines
1343
1344 This fixes a build problem on x86_64 -- must not have an explicit
1345 -L/usr/lib on x86_64 systems, empirically.
1346
1347 ------------------------------------------------------------------------
1348 r393 | rgb | 2008-03-16 20:55:43 -0400 (Sun, 16 Mar 2008) | 8 lines
1349
1350 This is much closer. Fixed a few small bugs, and one fairly large one
1351 from the transition to a loop. This is "probably" right at this point.
1352 The interesting question is whether and how we can generate statistics,
1353 pvalues, and results for the whole vector of results. In principle we
1354 can, but I don't know that I've done 14 in a row. OR, maybe I'll figure
1355 out that all I need to do is m = 16 (as it includes all its smaller
1356 cousins, supposedly).
1357
1358 ------------------------------------------------------------------------
1359 r392 | rgb | 2008-03-16 15:40:29 -0400 (Sun, 16 Mar 2008) | 3 lines
1360
1361 This isn't really done, but we're going to check it in since it sort of
1362 mostly works.
1363
1364 ------------------------------------------------------------------------
1365 r391 | rgb | 2008-03-12 17:47:53 -0400 (Wed, 12 Mar 2008) | 2 lines
1366
1367 Let's see if this revision works...
1368
1369 ------------------------------------------------------------------------
1370 r388 | rgb | 2008-03-12 13:34:59 -0400 (Wed, 12 Mar 2008) | 5 lines
1371
1372 Sending it ALL in to Duke and upstairs. There is a bug
1373 of sorts in our chisq -- we don't have a good way of
1374 specifying the number of degrees of freedom for some
1375 of the tests and we need to.
1376
1377 ------------------------------------------------------------------------
1378 r385 | rgb | 2008-03-12 07:09:20 -0400 (Wed, 12 Mar 2008) | 3 lines
1379
1380 This is about as fixed as I can make it without doing algebra. Algebra
1381 later, fix code up now.
1382
1383 ------------------------------------------------------------------------
1384 r384 | rgb | 2008-03-11 19:26:01 -0400 (Tue, 11 Mar 2008) | 7 lines
1385
1386 This APPEARS to check with John E. Davis's patches. diehard_dna is no
1387 problem -- the drop-in replacement there produces identical numbers,
1388 runs in half the time. rgb_bitdist gives DIFFERENT numbers but is MUCH
1389 faster. I still have several "small bugs" to check out including bugs
1390 in specific tests, and I probably will need to test the static routines
1391 as well.
1392
1393 ------------------------------------------------------------------------
1394 r383 | rgb | 2008-03-11 15:49:34 -0400 (Tue, 11 Mar 2008) | 5 lines
1395
1396 Preparing to do the next round(s) of patches. I suppose we'll start
1397 with the get_bit_whatever patches. First I need to generate a "before"
1398 for at least one test that uses them, e.g. birthdays. Then we'll
1399 compare it to after with the fixed seed of 1.
1400
1401 ------------------------------------------------------------------------
1402 r382 | rgb | 2008-03-11 00:53:27 -0400 (Tue, 11 Mar 2008) | 3 lines
1403
1404 This is beefing up the tarball so it just works with the usual
1405 preexisting configure.
1406
1407 ------------------------------------------------------------------------
1408 r381 | rgb | 2008-03-11 00:25:48 -0400 (Tue, 11 Mar 2008) | 5 lines
1409
1410 RIGHT THIS INSTANT we finally have a clean F8 build with the new
1411 Makefile.am's and a hacked specfile. Alas, to fix the rpath problem
1412 I had to resort to using chrpath, which sucks big time. Unless there
1413 is a libtool directive for fixing it, though, I'm SOL.
1414
1415 ------------------------------------------------------------------------
1416 r380 | rgb | 2008-03-10 09:41:26 -0400 (Mon, 10 Mar 2008) | 5 lines
1417
1418 I don't know -- I think that we're getting closer and closer. We will
1419 have the code completely and cleanly GBT based by the end of the
1420 morning, I think. Well, maybe not since I'm going in to see if there
1421 is anything entrepreneurial going on with Thom LaBean.
1422
1423 ------------------------------------------------------------------------
1424 r379 | rgb | 2008-03-10 08:08:18 -0400 (Mon, 10 Mar 2008) | 7 lines
1425
1426 OK, this is pretty much all the patches, I think. We should be very
1427 close to BSD-ready. I wonder if I can add the word "beta" to the
1428 "release" or the "8" in VERSION?
1429
1430 Oh, one other thing -- I very definitely need to fix the rpm build
1431 (specfile). Let's install into /tmp and see what we've got.
1432
1433 ------------------------------------------------------------------------
1434 r378 | rgb | 2008-03-10 07:54:46 -0400 (Mon, 10 Mar 2008) | 6 lines
1435
1436 Decrufted the makefiles, added a /dev/arandom rng for BSD, fixed tiny
1437 modulus bug in rng list display, added sensible error message extensions
1438 for people running across distros for the /dev/?random devices (I
1439 "could", of course, use fstat to determine if the devices are there and
1440 refuse to add them if they aren't, but that's for the future).
1441
1442 ------------------------------------------------------------------------
1443 r377 | rgb | 2008-03-09 22:07:15 -0400 (Sun, 09 Mar 2008) | 4 lines
1444
1445 Checking in a mildly broken set. FWIW, a full build actually works.
1446 OTOH, make rpm does not. Perhaps not too unsurprising. I need to do
1447 a full install into /tmp and see what comes up.
1448
1449 ------------------------------------------------------------------------
1450 r376 | rgb | 2008-03-09 15:35:38 -0400 (Sun, 09 Mar 2008) | 4 lines
1451
1452 This actually is getting close to being done "the right way". It still
1453 builds, and I think builds an RPM, but I need to work on the library
1454 side to clean it up as well.
1455
1456 ------------------------------------------------------------------------
1457 r375 | rgb | 2008-03-09 14:57:03 -0400 (Sun, 09 Mar 2008) | 2 lines
1458
1459 OK, this is really close to being what freebsd wants.
1460
1461 ------------------------------------------------------------------------
1462 r374 | rgb | 2008-03-09 13:55:12 -0400 (Sun, 09 Mar 2008) | 5 lines
1463
1464 OK, looks like the BSD patch for rngs_gnu_r.c works, and I'm guessing
1465 that it will work in general UNLESS it has to be the gnu error()
1466 program to interface with R, in which case the patch will have to be
1467 ifdef'd.
1468
1469 ------------------------------------------------------------------------
1470 r373 | rgb | 2008-03-09 13:38:06 -0400 (Sun, 09 Mar 2008) | 6 lines
1471
1472 OK, we are in the middle of a Major Patch Job. We're fixing various
1473 small things that are wrong with our GBT port, trying to get it so that
1474 it will automagically build on a wider set of systems. The stuff I've
1475 done so far still builds (I need to test the rpm build though) and
1476 "should" build on gentoo as well as Debian and Fedora/RH.
1477
1478 ------------------------------------------------------------------------
1479 r372 | rgb | 2008-02-12 19:01:04 -0500 (Tue, 12 Feb 2008) | 9 lines
1480
1481 This fixes the Crispin bug, -- a bad number in one column of s[i][j]
1482 relative to the original diehard. Sure hope it is the correct one; the
1483 later (c version) diehard matched what I had.
1484
1485 I really need to compute the r and s directly. But then, I think I
1486 understand the operm5 test well enough to do it "directly" without
1487 overlap as a non-overlapping permutations test. It might even be fun to
1488 make overlap just an option.
1489
1490 ------------------------------------------------------------------------
1491 r366 | rgb | 2008-01-30 07:47:48 -0500 (Wed, 30 Jan 2008) | 2 lines
1492
1493 This fixes the abstract AGAIN...
1494
1495 ------------------------------------------------------------------------
1496 r365 | rgb | 2008-01-02 11:23:16 -0500 (Wed, 02 Jan 2008) | 4 lines
1497
1498 This should be the absolute minimum needed to build clean from tarball
1499 using "just" ./configure followed by make. Need to fix the configure
1500 script to check for the gsl and anything else that might be needed.
1501
1502 ------------------------------------------------------------------------
1503 r362 | rgb | 2007-12-07 12:40:35 -0500 (Fri, 07 Dec 2007) | 2 lines
1504
1505 This checkin just updates the release number.
1506
1507 ------------------------------------------------------------------------
1508 r361 | rgb | 2007-12-06 14:39:53 -0500 (Thu, 06 Dec 2007) | 7 lines
1509
1510 This patches two small things -- adds fflush(stdout); so tees of the
1511 results don't make you wait. Removes a development line from
1512 diehard_operm5 that shouldn't be there in any KIND of real release.
1513
1514 Need to bump the snap revision and reinstall.
1515
1516
1517 ------------------------------------------------------------------------
1518 r360 | rgb | 2007-11-19 07:38:36 -0500 (Mon, 19 Nov 2007) | 29 lines
1519
1520 This is pretty much a fully functional version. I fixed the last bug,
1521 which turned out to be that I had the right idea but missed the:
1522
1523 I + P + P* + P^2 + P*^2 + ...
1524
1525 and was accidentally doing (when I empirically noted that including k=0
1526 in the sum was necessary) either:
1527
1528 I + P + P^2...
1529
1530 or
1531
1532 2I + P + P* + ...
1533
1534 Added a conditional to eliminate the second k = 0 instance in the core
1535 loop summing the fi*fj and got perfect correspondance through -n 3. The
1536 -n 3 matrix is different, but pretty obviously a permutation of the
1537 AKM matrix as to be expected since the permutation order is arbitrary
1538 and they probably didn't use lexical ordering. The eigenvalues,
1539 however, are dead on the same!
1540
1541 So I'm back up to the point where I should be able to form a test
1542 statistic somehow. I have to covariance matrix precisely in hand --
1543 something at this point about weak inverses (weak because it is singular
1544 and has no inverse?). A bit curious, actually. We'll see.
1545
1546 rgb
1547
1548
1549 ------------------------------------------------------------------------
1550 r359 | rgb | 2007-11-19 01:02:46 -0500 (Mon, 19 Nov 2007) | 13 lines
1551
1552 OK, this SEEMS to repair everything, except for some spurious
1553 normalization stuff that is at least consistent between the exact
1554 and experimental matrices. They now come out the same, and I suspect
1555 they are proportional to the nilpotent markov processes paper result.
1556 For k=2 they clearly are -- a factor of 2. For k=3 it is harder to see
1557 for sure. Time to go back to the paper with a calculator.
1558
1559 At least now the matrices are properly symmetric, as well. That means
1560 that their eigenvalues can be computed with my existing code with its
1561 symmetric matrix eigenvalue calls. It does leave me with the
1562 substantial problem of computing a test statistic automagically, but one
1563 small step at a time...
1564
1565 ------------------------------------------------------------------------
1566 r358 | rgb | 2007-11-14 07:01:57 -0500 (Wed, 14 Nov 2007) | 7 lines
1567
1568 This works, but my assumptions concerning the inverse of C are incorrect
1569 because (gasp) C isn't invertible. It has eigenvalues and eigenvectors.
1570 I can compute them, apparently -- that's what this snapshot does (I'm
1571 going to compare the results now to what's in my nilpotent processes
1572 paper). I still don't quite see how to make my results into a p-value,
1573 though, and may need help with this...
1574
1575 ------------------------------------------------------------------------
1576 r357 | rgb | 2007-11-13 11:53:48 -0500 (Tue, 13 Nov 2007) | 7 lines
1577
1578 It is SO key to preserve this in precisely this state. At this instant
1579 in time, cexact and cexpt are identically formed, one running over all
1580 permutations and the other from sampling, and they work perfectly EXECPT
1581 for the #!)#%@ sign error. I had the sign right, and don't quite see
1582 how I changed the code, so I'm about to look. In the meantime, though,
1583 we mustn't lose this working form...
1584
1585 ------------------------------------------------------------------------
1586 r356 | rgb | 2007-10-31 22:11:57 -0400 (Wed, 31 Oct 2007) | 3 lines
1587
1588 This is a broken checkin, but I'm on track for getting operm finished at
1589 last, if only I stick to it.
1590
1591 ------------------------------------------------------------------------
1592 r354 | rgb | 2007-10-24 15:19:05 -0400 (Wed, 24 Oct 2007) | 2 lines
1593
1594 Need to send to duke, dummy.s
1595
1596 ------------------------------------------------------------------------
1597 r353 | rgb | 2007-10-24 15:18:41 -0400 (Wed, 24 Oct 2007) | 2 lines
1598
1599 This is needed for development, I think.
1600
1601 ------------------------------------------------------------------------
1602 r352 | rgb | 2007-10-24 15:17:37 -0400 (Wed, 24 Oct 2007) | 2 lines
1603
1604 Trying again. Dunno how this got screwed up but it is.
1605
1606 ------------------------------------------------------------------------
1607 r351 | rgb | 2007-10-03 16:16:38 -0400 (Wed, 03 Oct 2007) | 3 lines
1608
1609 This is actually a pretty solid implementation of
1610 rgb_minimum_distance...
1611
1612 ------------------------------------------------------------------------
1613 r350 | rgb | 2007-09-29 17:43:55 -0400 (Sat, 29 Sep 2007) | 3 lines
1614
1615 works perfectly. fixed memory leak. about to improve 2dsphere with
1616 sort, then run -a.
1617
1618 ------------------------------------------------------------------------
1619 r349 | rgb | 2007-09-29 11:24:16 -0400 (Sat, 29 Sep 2007) | 2 lines
1620
1621 This is IT and works... rgb_minimum_distance is there.
1622
1623 ------------------------------------------------------------------------
1624 r347 | rgb | 2007-09-20 05:37:04 -0400 (Thu, 20 Sep 2007) | 3 lines
1625
1626 I need to fix this, but probably not today. In doc there is a "fix me"
1627 recipe for 2d spheres. I'm currently adding the Linux Magazine invoice.
1628
1629 ------------------------------------------------------------------------
1630 r346 | rgb | 2007-09-19 14:44:10 -0400 (Wed, 19 Sep 2007) | 3 lines
1631
1632 Checking in a fix for the BROKEN 2d spheres test. I should be able to
1633 fix 3dspheres as well from it.
1634
1635 ------------------------------------------------------------------------
1636 r345 | rgb | 2007-08-28 12:54:51 -0400 (Tue, 28 Aug 2007) | 3 lines
1637
1638 Bug reported by a user, sort of. The README still said you could just
1639 type "make" and you can't. Fixed.
1640
1641 ------------------------------------------------------------------------
1642 r344 | rgb | 2007-08-13 18:30:55 -0400 (Mon, 13 Aug 2007) | 2 lines
1643
1644 Will this go in? Time will tell...
1645
1646 ------------------------------------------------------------------------
1647 r343 | rgb | 2007-07-16 11:20:54 -0400 (Mon, 16 Jul 2007) | 45 lines
1648
1649 We are working HARD or confirming or denying dieharder_operm5 AND
1650 working out a more general rgb_operm. The convertoperm.pl script
1651 once AGAIN confirms that I'm unpacking and repacking (for C) the
1652 r and s matrices of diehard_operm correctly.
1653
1654 However, AFAICT, the diehard_operm5 JUST tallies up t[120], the count of
1655 the different permutation indices. However, all the permutations should
1656 occur with a SYMMETRIC probability -- I verified this directly by doing
1657 the actual integrals, but in retrospect it is pretty obvious from
1658 symmetry alone. The t-vector therefore has NOTHING TO DO with the
1659 covariance matrix no matter what you do with it, which may be why
1660 diehard_operm5() is a bad test. FWIW, I've added output statements to
1661 diehard_operm5() that dump the t-vector and verified empirically that
1662 yes, it produces a more or less uniform distribution of bin-probability
1663 1/120 -- one could in fact do a chisq test on this directly and very
1664 likely obtain an actual meaningful test, but that is not at all what is
1665 done.
1666
1667 Note well that the sampled distribution of t[] will not SIGNIFICANTLY
1668 vary if computed from overlapping samples! This is simple to see -- it
1669 would be just like computing five distinct samplings of non-overlapping
1670 samples, each of which would be expected to have exactly the same mean
1671 and variance. Because of the overlap there could be a tiny bit of
1672 correlation between the FLUCTUATIONS in the MEASURED mean values for any
1673 given bin, but the bin mean would have to be unchanged for each
1674 overlapping sampling, and for that matter the bin variance for each
1675 sampling would ALSO have to be the same.
1676
1677 The only effect of combining the overlapping five samplings (each with
1678 identical mean and variance!) would therefore be to AT MOST alter the
1679 sampled VARIANCE of the bin distribution, per bin, a tiny bit from what
1680 one might expect analytically from truly iid samples. One certainly
1681 isn't going to see this variation in the variance from two runs -- the
1682 expected statistical noise would overwhelm the signal as this is a
1683 second moment/cumulant effect, a susceptibility if you like, and hence
1684 VERY difficult to resolve.
1685
1686 What we in fact must do is (I'm pretty sure) compute c[][] analytically,
1687 (which I have done) and then simulate the SAME c[][] directly from the
1688 data. Then difference them, which should make the components of
1689 c_exp[][] - c_exact[][] presented as vector a zero mean chisq
1690 distributed object with N! - (N-1)! DoF or something like that -- 120-24
1691 = 96 for N = 5? Here I'll defer to experts, eventually.
1692
1693
1694 ------------------------------------------------------------------------
1695 r342 | rgb | 2007-06-16 01:11:57 -0400 (Sat, 16 Jun 2007) | 2 lines
1696
1697 Hopefully this has all the files needed for rgb_operm() checked in.
1698
1699 ------------------------------------------------------------------------
1700 r341 | rgb | 2007-06-16 01:09:13 -0400 (Sat, 16 Jun 2007) | 2 lines
1701
1702 This actually builds. Let's see if the new test shows up...
1703
1704 ------------------------------------------------------------------------
1705 r340 | rgb | 2007-06-15 20:12:58 -0400 (Fri, 15 Jun 2007) | 23 lines
1706
1707 Checking in several things. rgb_operm.c is a template for my eventual
1708 replacement for operm5. I have about concluded that operm5 is basically
1709 a joke -- it is conceptually incorrect. The correct statistic should be
1710 the overlapping permutations covariance matrix itself. This test just
1711 counts the occurrence of permutations -- the "overlap" part is
1712 completely irrelevant in this context. It returns an aggregate count
1713 vector whose mean is 1/120 (obviously, permutations of 5 objects = 5! =
1714 120). It makes NO DIFFERENCE AT ALL if this vector is evaluated from
1715 overlapping samples or from independent samples -- the mean will be the
1716 same. All that MIGHT change is the variance -- since the samples are
1717 drawn from an overlapping population, one would expect that their
1718 variance would be strictly less than one would expect from truly
1719 independent samples.
1720
1721 In a minute I'll verify this with the overlap flag and a couple of runs
1722 with mt, but I now understand why the test fails. It remains to figure
1723 out how to build a CORRECT overlapping permutations test that uses the
1724 COVARIANCE of overlapping samples as a statistic. I actually think that
1725 this will be pretty easy, but I do have to also understand why the
1726 definition in the nilpotent markov chains paper doesn't (I think) lead
1727 to the numbers that they publish there. Or rather, why my MUCH SIMPLER
1728 algorithm doesn't lead to the same numbers they get.
1729
1730 ------------------------------------------------------------------------
1731 r333 | rgb | 2007-06-14 14:54:23 -0400 (Thu, 14 Jun 2007) | 2 lines
1732
1733 This should have no operm in it.
1734
1735 ------------------------------------------------------------------------
1736 r332 | rgb | 2007-06-14 14:36:51 -0400 (Thu, 14 Jun 2007) | 2 lines
1737
1738 Probably a good time to update to lucifer, since it is actually up...
1739
1740 ------------------------------------------------------------------------
1741 r331 | rgb | 2007-06-08 10:14:04 -0400 (Fri, 08 Jun 2007) | 3 lines
1742
1743 OK, NOW let's actually try to get rng_uvag in, THEN we'll try to
1744 increase its speed...
1745
1746 ------------------------------------------------------------------------
1747 r330 | rgb | 2007-05-24 13:35:50 -0400 (Thu, 24 May 2007) | 4 lines
1748
1749 I'm gradually refining the bitstream variance -- 290 is from over
1750 260000 samples. I'm shooting for a s.d. of a couple of tenths, although
1751 honestly it is overkill.
1752
1753 ------------------------------------------------------------------------
1754 r329 | rgb | 2007-05-23 01:02:21 -0400 (Wed, 23 May 2007) | 12 lines
1755
1756 OK, this is a major enough checkin. I have at long last fixed
1757 bitstream.c. Marsaglia's published value of sigma is off by close to a
1758 factor of two. sigma is in fact 288.6. I have verified this by
1759 simulation with very excellent RNGs, e.g. MT or gfsr4 AND with HARDWARE
1760 generators (which were failing bitstream before).
1761
1762 Marsaglia couldn't resolve the error (which could just have been due to
1763 a typo at some point that was perpetuated) because his test didn't run
1764 enough times or plot the pvalues in a histogram, but if you plot the
1765 obtained means in a histogram and fit a normal to them, there is
1766 absolutely no doubt.
1767
1768 ------------------------------------------------------------------------
1769 r328 | rgb | 2007-05-23 00:35:33 -0400 (Wed, 23 May 2007) | 2 lines
1770
1771 Forgot to check this one in...
1772
1773 ------------------------------------------------------------------------
1774 r327 | rgb | 2007-05-23 00:31:03 -0400 (Wed, 23 May 2007) | 2 lines
1775
1776 OK, this is one of two ways to get things to metatron to run faster.
1777
1778 ------------------------------------------------------------------------
1779 r326 | rgb | 2007-05-22 21:12:07 -0400 (Tue, 22 May 2007) | 3 lines
1780
1781 This fixes several problems, and is working its way towards a new
1782 release.
1783
1784 ------------------------------------------------------------------------
1785 r325 | rgb | 2007-05-22 10:04:12 -0400 (Tue, 22 May 2007) | 2 lines
1786
1787 Silly checkin, as I dropped a file here that goes below.
1788
1789 ------------------------------------------------------------------------
1790 r324 | rgb | 2007-05-21 19:46:44 -0400 (Mon, 21 May 2007) | 5 lines
1791
1792 Well, we're making LOTS of little changes here. We're splitting off
1793 Exit() into its own file so it can be shred, we're moving
1794 db_gnu_r_rngs.c to gnu_r_rngs.c, we're adding prototypes and changing
1795 the namespace for RDH. All at once.
1796
1797 ------------------------------------------------------------------------
1798 r323 | rgb | 2007-05-21 18:14:19 -0400 (Mon, 21 May 2007) | 3 lines
1799
1800 Fixed small problem in makefile. Otherwise I'm not sure why these files
1801 are changing -- they really shouldn't.
1802
1803 ------------------------------------------------------------------------
1804 r322 | rgb | 2007-05-21 15:04:01 -0400 (Mon, 21 May 2007) | 3 lines
1805
1806 I do believe that this now is ready to rock and roll. AFAICT, it "just
1807 works".
1808
1809 ------------------------------------------------------------------------
1810 r321 | rgb | 2007-05-21 14:39:08 -0400 (Mon, 21 May 2007) | 6 lines
1811
1812 This is checking in dieharder WITH RDieHarder included and under version
1813 control. RDieHarder now depends on dieharder sources. I still haven't
1814 got the Makefile, with dependencies, built into the existing Makefile
1815 but I will. Basically, I plan three new targets: rdhprep, rdhlib (as
1816 root only) and rdhtgz (to make a "package" out of it).
1817
1818 ------------------------------------------------------------------------
1819 r320 | rgb | 2007-05-21 09:34:44 -0400 (Mon, 21 May 2007) | 10 lines
1820
1821 Well, we've been working hard here. Let me try to summarize. I've
1822 instrumented the code to make it "easy" to move it to RDieHarder. Dirk
1823 doesn't understand, quite, how I plan to make RDH a "part of" the
1824 dieharder source tree, but he will when I'm done as it will then be
1825 really easy to prep his cran/debian package. I think, I hope.
1826
1827 I think that I'm going to need to pursue dieharder this summer pretty
1828 aggressively in addition to working on axioms and other books. I should
1829 probably look hard at getting money to support it, as well.
1830
1831 ------------------------------------------------------------------------
1832 r319 | rgb | 2007-05-19 08:14:56 -0400 (Sat, 19 May 2007) | 3 lines
1833
1834 We've got to get this to lucifer and then try to unpack it on
1835 metatron...
1836
1837 ------------------------------------------------------------------------
1838 r318 | rgb | 2007-04-29 14:06:32 -0400 (Sun, 29 Apr 2007) | 2 lines
1839
1840 This too has to go to lucifer...
1841
1842 ------------------------------------------------------------------------
1843 r317 | rgb | 2007-03-12 09:50:39 -0400 (Mon, 12 Mar 2007) | 4 lines
1844
1845 This is the last set of changes associated with 2.24.3 -- which is
1846 hopefully getting really, really close to a release I can live with.
1847 I suppose I should fix the man pages, at least to some extent.
1848
1849 ------------------------------------------------------------------------
1850 r316 | rgb | 2007-03-05 10:35:05 -0500 (Mon, 05 Mar 2007) | 6 lines
1851
1852 This looks like it successfully splits off the manual so it is no longer
1853 autobuilt and is not in the rpms.
1854
1855 From here out the manual will be a SEPARATE project aimed square at lulu
1856 publication as a book I might even make money from.
1857
1858 ------------------------------------------------------------------------
1859 r315 | rgb | 2007-03-05 10:32:43 -0500 (Mon, 05 Mar 2007) | 2 lines
1860
1861 We'll work our way there by hook or by crook.
1862
1863 ------------------------------------------------------------------------
1864 r314 | rgb | 2007-03-05 10:32:03 -0500 (Mon, 05 Mar 2007) | 2 lines
1865
1866 Grrr. I just lost a whole bunch of work...
1867
1868 ------------------------------------------------------------------------
1869 r313 | rgb | 2007-03-01 20:00:11 -0500 (Thu, 01 Mar 2007) | 11 lines
1870
1871 This has now been fixed, a la Charles Karney's suggestion and Janusz's
1872 paper demonstrates, to use overlapping samples by default. Some tests
1873 it really DOES matter, if those tests are valid at all (operm5 being a
1874 classic example). I still doubt the bitstream-dna test series -- I
1875 think that there is NO difference between overlapping and
1876 non-overlapping samples there, or at most a tiny shrinking of the
1877 distribution variance in the end, as they don't measure "overlap", only
1878 a CLT-limited mean. Overlapping 20 bit samples should make a tiny shift
1879 in the sample standard deviation and have no other effect in these
1880 tests.
1881
1882 ------------------------------------------------------------------------
1883 r312 | rgb | 2007-03-01 19:38:44 -0500 (Thu, 01 Mar 2007) | 4 lines
1884
1885 This fixes the Bug reported by Brian J. Wong, making bl and bu of a
1886 function in bits.c static uint (preserved between calls) instead of
1887 dynamic. Probably my bad...
1888
1889 ------------------------------------------------------------------------
1890 r309 | rgb | 2007-02-28 10:39:58 -0500 (Wed, 28 Feb 2007) | 2 lines
1891
1892 Sending this into SVN on general principles.
1893
1894 ------------------------------------------------------------------------
1895 r308 | rgb | 2007-02-27 22:38:22 -0500 (Tue, 27 Feb 2007) | 10 lines
1896
1897 This actually seems to work. So we now have 66 random number generators
1898 to test. Tony Pasqualoni's Cellular Automaton rng isn't anywhere NEARLY
1899 as fast as he alleges compared to the GSL routines inside identical
1900 packaging -- in fact it is about par for the course -- but so far it
1901 seems to be 1 bit and 2 bit random, and we're working on a -a run that
1902 will expose it to the latest/greatest operm5. It remains to be seen
1903 whether or not K. Janusz's paper contains algorithms that will let me
1904 check and/or correct Marsaglia's operm5, or perhaps extend the operm5
1905 into a RANGE of tests like rgb_bitdist.
1906
1907 ------------------------------------------------------------------------
1908 r307 | rgb | 2007-02-27 22:19:36 -0500 (Tue, 27 Feb 2007) | 6 lines
1909
1910 This is a very first cut of Tony Pasqualoni's Cellular Automaton
1911 random number generator, ported into a gsl standard wrapper. This will
1912 let me test Tony's assertion that this is a good, very fast rng in
1913 exactly the same wrapper that one can study the other "good" (but
1914 slower) rng's.
1915
1916 ------------------------------------------------------------------------
1917 r306 | rgb | 2007-02-27 02:10:12 -0500 (Tue, 27 Feb 2007) | 2 lines
1918
1919 This is a really modest checkin, but we'll see if it can manage...
1920
1921 ------------------------------------------------------------------------
1922 r305 | rgb | 2007-02-27 02:04:24 -0500 (Tue, 27 Feb 2007) | 13 lines
1923
1924 TAG 2.24.2
1925
1926 I fixed the -q(uiet) flag so that it works, and decided to bundle all of
1927 the above bugfixes and feature shifts into a new minor minor bump. This
1928 is the "official" GBT build, if Dirk concurs when he tries it.
1929
1930 As far as I know, we're at least back where we were when we started
1931 implementing GBT and then some.
1932
1933 Now, if only I could figure out the trunk thing. Maybe I'll try doing a
1934 checkout and build without the trunk movement...
1935
1936
1937 ------------------------------------------------------------------------
1938 r304 | rgb | 2007-02-27 01:31:45 -0500 (Tue, 27 Feb 2007) | 5 lines
1939
1940 TAG 2.24.1-2
1941
1942 Fixed documentation and added a hard record of Janusz's bug in operm5.
1943
1944
1945 ------------------------------------------------------------------------
1946 r303 | rgb | 2007-02-27 01:06:25 -0500 (Tue, 27 Feb 2007) | 2 lines
1947
1948 Sending this to Duke, at least...
1949
1950 ------------------------------------------------------------------------
1951 r302 | rgb | 2007-02-27 01:06:09 -0500 (Tue, 27 Feb 2007) | 47 lines
1952
1953 This contains a change suggested by the following:
1954
1955 %<snip snip -----
1956 From: Janusz Kawczak <[email protected]>
1957 To: Robert G. Brown <[email protected]>
1958 Cc: Dirk Eddelbuettel <[email protected]>
1959 Subject: Re: Random R-Package
1960
1961 Hello, without going to much into peculiarities (for now, at least) of
1962 the rngs and numbers coming from Marsaglia and you (Robert), I would
1963 like to give the result from our paper that is directly applicable to
1964 the operm5 header you cited below:
1965
1966 "...(with # rank 99)..." - this is the point I questioned in our paper.
1967 I asked Marsaglia how he got this rank, but I never received the reply
1968 from him. I can only presume that he either guessed this number or
1969 estimated it (somehow). It remains a mystery to me; like a Marsaglia's
1970 voodoo step. The answer should be 96=5!-4!. It is not a big deal
1971 of a difference in this case, however, the theorem we proved allows one
1972 to consider cased of perm6, perm7 ... and higher with the exact number
1973 of degree of freedom calculations. Thus, exact and not guessed results.
1974
1975 It is my opinion that the operm-type test is very powerful for detecting
1976 local correlations (dependencies) and it should be used for such
1977 purpose. So, any algorithmic PRNG will suffer and most likely do not
1978 dwell well when exposed to this test. I have not had a chance to do
1979 serious experiments on the "natural-random" generators, e.g.
1980 alpha-particle counting, radio noise signals and many others. But, I
1981 suspect that the operm-type test should do well in this situation.
1982
1983 I cannot promise at this time much, but I am going to look at the issues
1984 you addressed in your e-mail. However, the questions you've raised are a
1985 bit different animals to deal with.
1986
1987 In summary, it would be useful, once and for all, to correct 99 degrees
1988 to 96 in the operm5 Marsaglia's header file. Maybe, one day, this will
1989 happen :-)
1990 %<snip snip -----
1991
1992 I've implemented this change -- 99 to 96 -- although I welcome any
1993 comments. The test still fails the big four -- gsfr4, taus2, ranlxd2,
1994 and mt19937 -- nearly all the time, overlapping or not. I still suspect
1995 either errors in r and s or (quite possibly) programming errors, but it
1996 ALMOST works and lacking a gold standard rng it is difficult to resolve
1997 the question numerically.
1998
1999
2000 ------------------------------------------------------------------------
2001 r301 | rgb | 2007-02-27 00:16:46 -0500 (Tue, 27 Feb 2007) | 4 lines
2002
2003 This is an important change -- I fixed an annoyance (if not worse) with
2004 the libdieharder setting of the major revision number. I may need to
2005 propagate the expr back to the toplevel Make.
2006
2007 ------------------------------------------------------------------------
2008 r300 | rgb | 2007-02-26 23:39:47 -0500 (Mon, 26 Feb 2007) | 7 lines
2009
2010 Dirk found that I was still screwing up the PREFIX thing by stubbornly
2011 refusing to use (lower case) prefix, libdir, and includedir in my make
2012 install target. I think I've stripped out my own versions of these
2013 completely so that they can a) be set by configure and run with "just"
2014 make install or b) overridden on the make line, e.g make prefix=/tmp/usr
2015 install (as apparently he likes to do to make debian packages).
2016
2017 ------------------------------------------------------------------------
2018 r299 | rgb | 2007-02-25 14:42:52 -0500 (Sun, 25 Feb 2007) | 3 lines
2019
2020 This is definitely experimental, trying to write a web installer
2021 modification that I cannot test until I get back the net.
2022
2023 ------------------------------------------------------------------------
2024 r298 | rgb | 2007-02-25 14:32:05 -0500 (Sun, 25 Feb 2007) | 4 lines
2025
2026 Ok, this is TAG=2.24.1, the first Gnu Build Tools release that builds
2027 rpms clean from (we hope) SVN. I do need to validate this with one
2028 last SVN checkout, but I'm pretty sure it is the case...
2029
2030 ------------------------------------------------------------------------
2031 r297 | rgb | 2007-02-25 13:35:17 -0500 (Sun, 25 Feb 2007) | 2 lines
2032
2033 Well, a time to check in. We're about to see if an rpm build works...
2034
2035 ------------------------------------------------------------------------
2036 r296 | rgb | 2007-02-25 13:15:29 -0500 (Sun, 25 Feb 2007) | 6 lines
2037
2038 This may be a really nice advance. The dieharder build now uses
2039 ../include and ../libdieharder as -I and -L respectively, and plain old
2040 "make" in both cases should work, from a clean checkout. I'm guessing
2041 that I can add a simple Makefile.am to include to do the actual install
2042 of the include files.
2043
2044 ------------------------------------------------------------------------
2045 r295 | rgb | 2007-02-25 12:47:10 -0500 (Sun, 25 Feb 2007) | 4 lines
2046
2047 This should be REALLY REALLY close. We'll checkin, do a full checkout,
2048 and try building. If/when we get there, we'll work strictly from build
2049 to build once again.
2050
2051 ------------------------------------------------------------------------
2052 r294 | rgb | 2007-02-25 12:27:25 -0500 (Sun, 25 Feb 2007) | 4 lines
2053
2054 OK, this is one whole cut closer, and ALMOST works, but we're definitely
2055 going to have to try this one more time... We also need at some point
2056 to stop svn control of Makefiles and maintain only Makefile.ams
2057
2058 ------------------------------------------------------------------------
2059 r293 | rgb | 2007-02-25 12:17:22 -0500 (Sun, 25 Feb 2007) | 3 lines
2060
2061 Adding what MAY be all the things needed to enable full automated Gnu
2062 style builds from the autogen.sh script.
2063
2064 ------------------------------------------------------------------------
2065 r292 | rgb | 2007-02-25 12:16:05 -0500 (Sun, 25 Feb 2007) | 4 lines
2066
2067 Finished a first cut at adding GBT support. Need to check in the .in
2068 files and configure.ac files one level down, and do a full checkout to
2069 see if it autobuilds from an svn checkout.
2070
2071 ------------------------------------------------------------------------
2072 r291 | rgb | 2007-02-19 09:23:15 -0500 (Mon, 19 Feb 2007) | 3 lines
2073
2074 This dumps buildroot. We don't need or want this in the tarball or rpm
2075 sources.
2076
2077 ------------------------------------------------------------------------
2078 r290 | rgb | 2007-02-19 03:47:31 -0500 (Mon, 19 Feb 2007) | 4 lines
2079
2080 We are working quite hard on getting the ChangeLog to be automagically
2081 registered. I'm guessing it should just go into %doc and we should
2082 pretty much forget the tag otherwise.
2083
2084 ------------------------------------------------------------------------
2085 r289 | rgb | 2007-02-19 03:05:19 -0500 (Mon, 19 Feb 2007) | 5 lines
2086
2087 This fixes a variety of problems with using shared libraries correctly,
2088 and moves the project much closer to where it could e.g. be included in
2089 FC extras. The library builds nicely now, for example. However, I do
2090 need to switch to using autoconf, however much I generally dislike it.
2091
2092 ------------------------------------------------------------------------
2093 r288 | rgb | 2007-02-19 01:32:29 -0500 (Mon, 19 Feb 2007) | 2 lines
2094
2095 OK, this fixes some serious uglyness. Let's see if the rpm builds...
2096
2097 ------------------------------------------------------------------------
2098 r287 | rgb | 2007-02-18 16:49:23 -0500 (Sun, 18 Feb 2007) | 3 lines
2099
2100 Updating the svn tree and sync'ing, perhaps for the last time. Will
2101 this now reside on code.google.com? We'll see.
2102
2103 ------------------------------------------------------------------------
2104 r286 | rgb | 2007-02-18 08:37:35 -0500 (Sun, 18 Feb 2007) | 3 lines
2105
2106 Let's call this TAG=2.5.24 -- bumping the first minor to announce that
2107 we've cleaned up several bugs and repackaged.
2108
2109 ------------------------------------------------------------------------
2110 r285 | rgb | 2007-02-15 13:55:48 -0500 (Thu, 15 Feb 2007) | 28 lines
2111
2112 This fixes a bug reported by Matthias Braeunig <[email protected]>, who
2113 was running the test on a file of data just one time per test using more
2114 or less this line:
2115
2116 for t in $(echo "-d"{1..19} "-r"{1..4} "-s"{1..2}); do ./dieharder -q -t9375 -p1 $t;done > results
2117
2118 He noted that kstest_kuiper returned 2.0 for input single pvalues of
2119 0.0, and that the test overall returned a pass even if p=1.0 exactly
2120 (which it also should not do for single pvalues).
2121
2122 I made the following changes. kstest_kuiper now returns a single input
2123 pvalue as its output pvalue -- clearly that's all one can do and the
2124 right thing to do. The test.c assessment that prints out the results
2125 also no longer calls the return a KStest result, and reports failure
2126 a bit differently, flagging the high-p failures as well as the low p
2127 failures, and adjusting the reported failure range accordingly.
2128
2129 To do this I had to use -static in the Makefile to work in the
2130 development tree. I'm guessing that I need to add LDFLAG = -dynamic to
2131 the build make line in the specfile and should LEAVE it -static in the
2132 actual tree so people don't get stymied if they download and build in
2133 the tarball directory ("people" including me) while rpm's will still
2134 autobuild dynamic correctly.
2135
2136 Finally, Matthias reported that the -q flag doesn't work. He's right,
2137 but I'm not sure I'm going to fix that. Rather I should probably just
2138 kill it and let people filter the output results by hand...
2139
2140 ------------------------------------------------------------------------
2141 r284 | rgb | 2007-02-05 23:42:45 -0500 (Mon, 05 Feb 2007) | 4 lines
2142
2143 I've just fired this up to the web. I deem it finished enough.
2144 Now to backport the library fix to wulfware, and call it a night, very
2145 much a night.
2146
2147 ------------------------------------------------------------------------
2148 r283 | rgb | 2007-02-05 22:06:02 -0500 (Mon, 05 Feb 2007) | 5 lines
2149
2150 OK, it installs and builds PERFECTLY, and THIS time it almost certainly
2151 is dynamically linking to libdieharder.so, as it linked without the .a
2152 library present at all. I'll now have to go retrofit this to wulfware,
2153 as it is not building correctly.
2154
2155 ------------------------------------------------------------------------
2156 r282 | rgb | 2007-02-05 21:39:00 -0500 (Mon, 05 Feb 2007) | 3 lines
2157
2158 This actually rand and built! Now to try to rebuild it WITHOUT the .a
2159 library installed...
2160
2161 ------------------------------------------------------------------------
2162 r281 | rgb | 2007-02-05 21:30:42 -0500 (Mon, 05 Feb 2007) | 3 lines
2163
2164 Just in case I drop something, I finally seem to have "fixed" the
2165 libdieharder makefile. On to fame and glory.
2166
2167 ------------------------------------------------------------------------
2168 r280 | rgb | 2007-02-05 15:06:05 -0500 (Mon, 05 Feb 2007) | 2 lines
2169
2170 OK, this is crawling on closer to ready to go...
2171
2172 ------------------------------------------------------------------------
2173 r279 | rgb | 2007-02-05 14:51:33 -0500 (Mon, 05 Feb 2007) | 2 lines
2174
2175 Make this go away, please.
2176
2177 ------------------------------------------------------------------------
2178 r277 | rgb | 2007-01-28 10:58:14 -0500 (Sun, 28 Jan 2007) | 6 lines
2179
2180 This has a brand new ultra-cool target, "installrepo" that makes the
2181 rpms and installs them in the repo, from when yum can install them.
2182
2183 BTW, I think I forgot the "requires" tag in the dieharder sources,
2184 partly because it seems that it isn't, in fact, required. Hmmm.
2185
2186 ------------------------------------------------------------------------
2187 r276 | rgb | 2007-01-27 19:11:19 -0500 (Sat, 27 Jan 2007) | 2 lines
2188
2189 This works!
2190
2191 ------------------------------------------------------------------------
2192 r275 | rgb | 2007-01-27 17:39:42 -0500 (Sat, 27 Jan 2007) | 2 lines
2193
2194 This is getting really close, worth checking in...
2195
2196 ------------------------------------------------------------------------
2197 r274 | rgb | 2007-01-27 13:27:14 -0500 (Sat, 27 Jan 2007) | 4 lines
2198
2199 This now builds a perfectly rebuildable tarball. We can think about
2200 just what else we'd like to add to that tarball in a moment, but first
2201 we need to FINALLY get the rpm to build, maybe.
2202
2203 ------------------------------------------------------------------------
2204 r273 | rgb | 2007-01-27 12:59:42 -0500 (Sat, 27 Jan 2007) | 2 lines
2205
2206 This updates the abstract.
2207
2208 ------------------------------------------------------------------------
2209 r272 | rgb | 2007-01-27 12:50:16 -0500 (Sat, 27 Jan 2007) | 3 lines
2210
2211 OK, this just maybe is working now with a target that rebuilds
2212 both specfiles and Makefiles when the toplevel Makefile is altered.
2213
2214 ------------------------------------------------------------------------
2215 r271 | rgb | 2007-01-27 11:57:23 -0500 (Sat, 27 Jan 2007) | 8 lines
2216
2217 This is just peachy. make and make install targets work for BOTH
2218 dieharder_src and libdieharder directories, which is pretty cool,
2219 really. The remaining problem will be how to force a rebuild
2220 of the library in such a way that it works when we're developing but
2221 doesn't barf when we're rpm-ifying.
2222
2223 At this point in time it is high time to try the rpm build.
2224
2225 ------------------------------------------------------------------------
2226 r270 | rgb | 2007-01-27 11:41:39 -0500 (Sat, 27 Jan 2007) | 3 lines
2227
2228 OK, this is making progress. Time to go back to libdieharder and get
2229 a build to work there...
2230
2231 ------------------------------------------------------------------------
2232 r269 | rgb | 2007-01-27 11:13:18 -0500 (Sat, 27 Jan 2007) | 3 lines
2233
2234 Just making sure this is all ready to run when I start to edit
2235 the Makefile in libdieharder.
2236
2237 ------------------------------------------------------------------------
2238 r268 | rgb | 2007-01-27 11:08:45 -0500 (Sat, 27 Jan 2007) | 11 lines
2239
2240 OK, this is a very painful move. We will, of course, mothball and
2241 preserve libdieharder's original svn tree, but now that we're figuring
2242 out how to do one specfile, many packages from a single toplevel
2243 source tree we no longer wish to maintain libdieharder in a separate
2244 subversion project. So we're checking it into this one. All the
2245 change history is preserved, but in pieces -- CVSROOT first,
2246 subversion's libdieharder second, and from now on, here in the one
2247 true dieharder tree and its subversion controlled project.
2248
2249 Next we have to get this so that a make install does the right thing.
2250
2251 ------------------------------------------------------------------------
2252 r267 | rgb | 2007-01-27 10:53:36 -0500 (Sat, 27 Jan 2007) | 7 lines
2253
2254 This is a first cut at making dieharder actually build, after
2255 libdieharder is built and installed. From now on BOTH will use ONLY
2256 the include files that are stored in ./include, which will actually
2257 simplify life tremendously. I may symlink them through to the
2258 source directory(s) and may even svn control the symlinks, if svn can
2259 manage them. CVS couldn't...
2260
2261 ------------------------------------------------------------------------
2262 r266 | rgb | 2007-01-27 10:31:10 -0500 (Sat, 27 Jan 2007) | 5 lines
2263
2264 OK, we are within a single step (removing or moving some include files)
2265 of being cleaned up and ready to proceed. I'll probably copy at least
2266 part of the sm Makefile to get the hang of looping a make through source
2267 directories in order to achieve the make install in the right sequence.
2268
2269 ------------------------------------------------------------------------
2270 r145 | rgb | 2005-03-14 08:57:05 -0500 (Mon, 14 Mar 2005) | 5 lines
2271
2272 This sends in parse.c. Y'know, I'll bet that this is what breaks
2273 Dan's access. Shit, looks like I'll have to kiss "make sync" goodbye...
2274 or give him an "account" locally with the same uid/gid -- nope that
2275 won't work either. Sigh.
2276
2277 ------------------------------------------------------------------------
2278 r144 | rgb | 2005-03-10 22:42:00 -0500 (Thu, 10 Mar 2005) | 2 lines
2279
2280 This goes home and into laptop...
2281
2282 ------------------------------------------------------------------------
2283 r143 | rgb | 2005-03-10 22:05:41 -0500 (Thu, 10 Mar 2005) | 2 lines
2284
2285 Sending this all home, so I can make it on metatron?
2286
2287 ------------------------------------------------------------------------
2288 r141 | rgb | 2004-11-30 09:14:46 -0500 (Tue, 30 Nov 2004) | 5 lines
2289
2290 I think we'll add this to the repository as the "reference run" against
2291 the default generator, dieharder -a > dieharder.all. If we update this
2292 at the end of every new addition, we'll be able to advance to toolset
2293 in a fairly systematic (but generally reliable) way.
2294
2295 ------------------------------------------------------------------------
2296 r140 | rgb | 2004-11-29 16:36:23 -0500 (Mon, 29 Nov 2004) | 2 lines
2297
2298 This looks like it ran nicely.
2299
2300 ------------------------------------------------------------------------
2301 r139 | rgb | 2004-11-29 14:30:09 -0500 (Mon, 29 Nov 2004) | 3 lines
2302
2303 This MIGHT just be a reference run followed by tag bump and checkin.
2304 Looks pretty nifty, right up to a first draft of histogram.
2305
2306 ------------------------------------------------------------------------
2307 r138 | rgb | 2004-11-29 14:03:49 -0500 (Mon, 29 Nov 2004) | 4 lines
2308
2309 Adding the published diehard F90 source code to the tree for
2310 porting convenience, although we will not use any of it verbatim,
2311 obviously, in a c port.
2312
2313 ------------------------------------------------------------------------
2314 r137 | rgb | 2004-11-29 12:18:04 -0500 (Mon, 29 Nov 2004) | 3 lines
2315
2316 I'm hoping that this is the needed binary rank program that analyzes
2317 binary (bitlevel) matrices for the diehard binary rank tests.
2318
2319 ------------------------------------------------------------------------
2320 r136 | rgb | 2004-11-24 01:14:41 -0500 (Wed, 24 Nov 2004) | 5 lines
2321
2322 This seems to "work", although it is consistently producing an overall
2323 p-value that is in the .9 range and hence "too high". I'm going to
2324 start up a full run of 100 x 40000 in a second to see if I get a
2325 "normal" pvalue.
2326
2327 ------------------------------------------------------------------------
2328 r135 | rgb | 2004-11-24 01:00:50 -0500 (Wed, 24 Nov 2004) | 2 lines
2329
2330 Well, not JUST that. I suppose that next we'll have to actually debug.
2331
2332 ------------------------------------------------------------------------
2333 r134 | rgb | 2004-11-24 00:58:30 -0500 (Wed, 24 Nov 2004) | 6 lines
2334
2335 This is ALMOST working. I'd say the binary rank part is working, hence
2336 the checkin. It is the rank_32x32 part that isn't, but I'll work on it.
2337
2338 I suspect something really simple, like needing to normalize y by
2339 tsamples...
2340
2341 ------------------------------------------------------------------------
2342 r133 | rgb | 2004-11-22 15:08:03 -0500 (Mon, 22 Nov 2004) | 4 lines
2343
2344 This is simply lovely, simply lovely, with a good start on adding
2345 Diehard's Binary Rank test. All I need is a suitably scaled matrix
2346 and a few zillion rands...
2347
2348 ------------------------------------------------------------------------
2349 r132 | rgb | 2004-11-22 14:23:11 -0500 (Mon, 22 Nov 2004) | 2 lines
2350
2351 Why didn't these make it to Duke?
2352
2353 ------------------------------------------------------------------------
2354 r131 | rgb | 2004-11-22 03:35:19 -0500 (Mon, 22 Nov 2004) | 6 lines
2355
2356 This is 0.5.8 stable, I hope. Time to go beddy-bye, hoping that this is
2357 now ready for real development and grantseeking work. I should be
2358 able to add a couple more diehard tests "easily" at this point, I think.
2359
2360 rgb
2361
2362 ------------------------------------------------------------------------
2363 r129 | rgb | 2004-11-22 01:47:07 -0500 (Mon, 22 Nov 2004) | 5 lines
2364
2365 This is VERY VERY close to being a fairly serious tagged checkin.
2366 We've resolved the problem of multiline strings in gcc, snagged a
2367 C tutorial for gcc, completely fixed the documentation -- this is
2368 all pretty awesomely ready for a brave new release.
2369
2370 ------------------------------------------------------------------------
2371 r128 | rgb | 2004-11-21 23:24:18 -0500 (Sun, 21 Nov 2004) | 11 lines
2372
2373 This is really close to working with all the changes in the command
2374 line options and associated global variables. In fact, it might BE
2375 working.
2376
2377 Two things that I really really need are a routine that can take a
2378 data object that is one big string and display it to the screen
2379 (something gcc refuses to do any more, which sucks big time) and a 20x
2380 histogram of p-values. Let's see if I can find them on the web before
2381 tackling them -- in particular the former seems like it must have been
2382 written by somebody...
2383
2384 ------------------------------------------------------------------------
2385 r127 | rgb | 2004-11-20 13:17:55 -0500 (Sat, 20 Nov 2004) | 6 lines
2386
2387 This is now really truly ready to go, EXCEPT that NOW I have to alter
2388 all sorts of command options according to the latest prescriptions in
2389 the checkin logs and abstract/web page. And get it "working perfectly"
2390 once again. I might even finish this this weekend, if I really hammer
2391 at it.
2392
2393 ------------------------------------------------------------------------
2394 r126 | rgb | 2004-11-20 10:55:07 -0500 (Sat, 20 Nov 2004) | 3 lines
2395
2396 This is the REALLY final checkin of rand_rate, I think, before we clone
2397 it into dieharder.
2398
2399 ------------------------------------------------------------------------
2400 r125 | rgb | 2004-11-20 10:50:28 -0500 (Sat, 20 Nov 2004) | 87 lines
2401
2402 This is the last checkin of rand_rate AS rand_rate. I'm going to change
2403 the name of this suite to dieharder. I'm also going to change the
2404 test numbering schema and option naming so that e.g.
2405
2406 -d diehard test number
2407 -r rgb test number
2408 -s sts test number
2409
2410 where -1 runs all tests of a given kind, 0 lists a description of all
2411 tests in the suite, -2 runs all tests of a given kind EXACTLY as they
2412 are run in the original code, if possible -- I'm not sure I'm going to
2413 test overlapping bit strings drawn from a single int just to bump the
2414 count of "random numbers" unless the test explicitly calls for it and it
2415 makes sense, as there is this thing about each sample being
2416 "independent" that worries me with overlapping draws.
2417
2418 -p number of p-values in final KS (and possibly other) test(s). This
2419 is the number of times each "test" is run with independent random number
2420 seeds (as the DEFAULT from now on). This defaults to 100, which is
2421 actually a lot and very reasonable but which can be increased if one is
2422 in doubt about whether the distribution of p returned by "the test" is
2423 in fact uniform.
2424
2425 -t number of test samples that go into making up a single test. This
2426 is NOT always a free parameter -- in many of the diehard tests
2427 especially, the number of e.g. points drawn from a 2d or 3d volume in
2428 the minimum distance tests is fixed, and varying it would vary the
2429 target distribution and test statistic. Although this is a bit
2430 unfortunate, since varying the number of test samples is an excellent
2431 way to make marginal failures in the distribution of p resolve into
2432 clear failures, we either live with it or derive general forms for the
2433 asymptotic target distributions as a function of the number of samples
2434 or do simulations and empirically deduce forms ditto (as Marsaglia and
2435 others appear to have done). For the moment we'll live with it.
2436
2437 -b bitstring width. Some tests are applied only to samples that are
2438 "bitstrings" (as opposed to e.g. lists of unsigned integers) of
2439 user-specifiable length. One reason to limit the tests in this way is
2440 to avoid numerical difficulties in e.g. evaluating P_binomial(k,p,n),
2441 where one can easily under or overflow and end up with garbage or a gsl
2442 fault. Another is to "free" some of the existing tests that have a very
2443 specific size of bitstring that they look at so that this can be varied
2444 when the target distribution can still be computed as a function of
2445 bitstring size. This will be overridden as necessary, like -t, for
2446 tests that really do require a fixed bitstring size to approach the
2447 known target distribution.
2448
2449 -n ntuple or window width. A number of tests look at bit ntuples. An
2450 ntuple is a set of n consecutive bits drawn from a bitstring (possibly
2451 of -b specified width) or vector of random uints (possibly of -t
2452 specified length). ntuples are always drawn relative to a bit offset
2453 specified from the right (least significant) with 0 being the rightmost
2454 bit, with cyclic boundary conditions on the entire bitstring >>or<<
2455 sample uint vector, so drawing an ntuple cannot fail for any offset
2456 within the number of significant bits returned by the generator (which
2457 MAY NOT BE 32, or even 31 -- some generators return as few as 16
2458 significant bits!).
2459
2460 For example, an 8-bit bitstring might be:
2461
2462 01100111
2463
2464 and all the 3-tuples drawn from it are given by
2465
2466 offset 3-tuple
2467 =====================
2468 0 111
2469 1 011
2470 2 001
2471 3 100
2472 4 110
2473 5 011
2474 6 101 <- Note wrap around!
2475 7 110 <- Ditto!
2476
2477 A general purpose
2478
2479 get_bit_ntuple(*bitstring,bitstring size,ntuple size,offset)
2480
2481 routine is provided that is used by many tests to get ntuples from a
2482 uint *bitstring of given >>uint vector length<< (not bitstring length).
2483
2484 Other test controls may be added as well, but these are what I'm going
2485 to document right now. Mostly I'm checking in all the placeholders
2486 required for the rest of the diehard tests so I can start to knock them
2487 off systematically. Sure hope I'm past major rewrites!
2488
2489 ------------------------------------------------------------------------
2490 r124 | rgb | 2004-11-20 03:07:19 -0500 (Sat, 20 Nov 2004) | 2 lines
2491
2492 Fixed silly spelling error (sigh).
2493
2494 ------------------------------------------------------------------------
2495 r123 | rgb | 2004-11-20 02:50:00 -0500 (Sat, 20 Nov 2004) | 3 lines
2496
2497 Well, that was quick. A nasty (but easy) little bug in diehard_runs
2498 squashed (size -> tsamples).
2499
2500 ------------------------------------------------------------------------
2501 r122 | rgb | 2004-11-20 02:47:34 -0500 (Sat, 20 Nov 2004) | 14 lines
2502
2503 This should/maybe be a serious v 0.5.3 checkin. We are about to try
2504 -v -1. If it works, it will cycle through all of the working tests
2505 (and all the tests are working). All the tests are now written in
2506 a way that they can use sample and kstest_kuiper() to do the validation
2507 of the p-values obtained from running a possibly size-variable test
2508 on bits or frequencies or runs or whatever.
2509
2510 If this test works, it is off to the website, I'm off to bed, and next
2511 we go back to moving in new diehard tests. With the magical sliding
2512 bitwindow (which really does seem to work pretty well:-) implementing
2513 at least the n-tuple diehard tests should now be pretty easy, and I
2514 can probably do a more rigorous job than GM did because I don't have
2515 to scrimp on rands.
2516
2517 ------------------------------------------------------------------------
2518 r120 | rgb | 2004-11-20 02:34:39 -0500 (Sat, 20 Nov 2004) | 2 lines
2519
2520 Try again (network down).
2521
2522 ------------------------------------------------------------------------
2523 r119 | rgb | 2004-11-20 02:34:07 -0500 (Sat, 20 Nov 2004) | 4 lines
2524
2525 This clears diehard_birthdays AND diehard_2dsphere. Only one diehard
2526 to go and we'll have EVERYTHING running with sample and the new test
2527 format (but of course rgb_persist, which doesn't count).
2528
2529 ------------------------------------------------------------------------
2530 r118 | rgb | 2004-11-19 22:11:12 -0500 (Fri, 19 Nov 2004) | 2 lines
2531
2532 Fixed up diehard_runs so it uses the new test format. Works charmlike.
2533
2534 ------------------------------------------------------------------------
2535 r117 | rgb | 2004-11-19 18:38:19 -0500 (Fri, 19 Nov 2004) | 5 lines
2536
2537 This is v 0.5.1, which is nicely fixed for BOTH sts_monobit AND
2538 sts_runs, both in the new format, with a consistent Xtest_eval()
2539 routine. This fixes lots of things -- both tests are very likely
2540 to be "good".
2541
2542 ------------------------------------------------------------------------
2543 r115 | rgb | 2004-11-19 18:34:38 -0500 (Fri, 19 Nov 2004) | 7 lines
2544
2545 OK, this looks like sts_runs is now "good" in the new format. However,
2546 it may have broken sts_monobit. The problem is, is there or is there
2547 not a sqrt(2.0) in the erfc relative to sigma, and if there is is
2548 there an EXTRA one in sts_runs vs sts_monobit. Need to clear this up
2549 and either always put it into Xtest or always put it into xtest.sigma,
2550 but not both.
2551
2552 ------------------------------------------------------------------------
2553 r114 | rgb | 2004-11-19 16:06:34 -0500 (Fri, 19 Nov 2004) | 2 lines
2554
2555 Sending this home, I hope.
2556
2557 ------------------------------------------------------------------------
2558 r113 | rgb | 2004-11-19 15:09:15 -0500 (Fri, 19 Nov 2004) | 8 lines
2559
2560 This is now squeaky clean for rgb_bitdist and sts_monobit, and we're
2561 working on sts_runs. Then a quick dash through the diehards and we'll
2562 be back to where we were but a bit cleaner.
2563
2564 I still MIGHT try to get cleaner still. I'm not at all convinced
2565 that I need the test structs, for example, although perhaps they
2566 allow some encapsulation that is useful.
2567
2568 ------------------------------------------------------------------------
2569 r112 | rgb | 2004-11-19 13:55:01 -0500 (Fri, 19 Nov 2004) | 3 lines
2570
2571 This is a checkin to Duke of the nifty neato cool new improved
2572 version. It may be time to change the name and everything.
2573
2574 ------------------------------------------------------------------------
2575 r111 | rgb | 2004-11-19 12:35:19 -0500 (Fri, 19 Nov 2004) | 4 lines
2576
2577 OK so there was one more teeny bug in rgb_bitdist() -- wrong order in
2578 the final output statement. Fixed. I also added a feature, reseeding
2579 the rng in sample on an -i flag.
2580
2581 ------------------------------------------------------------------------
2582 r110 | rgb | 2004-11-19 12:28:09 -0500 (Fri, 19 Nov 2004) | 4 lines
2583
2584 This is just ensuring that the tag for version 0.5.0 is noted. This
2585 version works through rgb_bitdist, which I'll bet a nickel is a very
2586 powerful test indeed.
2587
2588 ------------------------------------------------------------------------
2589 r108 | rgb | 2004-11-19 12:26:42 -0500 (Fri, 19 Nov 2004) | 4 lines
2590
2591 This is a "permanent" checkin. I think that this fixes rgb_bitdist
2592 nicely to use sample() and provides a prototype for doing other
2593 tests.
2594
2595 ------------------------------------------------------------------------
2596 r106 | rgb | 2004-11-18 21:09:25 -0500 (Thu, 18 Nov 2004) | 11 lines
2597
2598 This is a fairly major fix -- I was truncating blen in bits.c at
2599 the sizeof(uint), not 8*sizeof(uint). One lesson is that this
2600 truncation isn't right anyway. We rather need to just punt/die.
2601
2602 I'm wondering now if the apparent failure that is still present
2603 (although not nearly as bad) for the larger ntuples is because fewer
2604 bins pass the cutoff in the formation of the primary sample pvalue(s).
2605
2606 We might just try lowering this cutoff a bit. I don't know exactly what
2607 a "degree of freedom" is, but we do need to be pretty careful with it.
2608
2609 ------------------------------------------------------------------------
2610 r105 | rgb | 2004-11-18 18:43:21 -0500 (Thu, 18 Nov 2004) | 6 lines
2611
2612 Now it REALLY looks like it works, and even the best rng's look like
2613 they FAIL the test in fairly short order. Now we're cookin' with gas,
2614 although I've got to see the details of the failure soon enough.
2615
2616 Hmmm, maybe I need to have a lot more bins...
2617
2618 ------------------------------------------------------------------------
2619 r104 | rgb | 2004-11-18 18:25:52 -0500 (Thu, 18 Nov 2004) | 3 lines
2620
2621 Just to verify that it APPEARS to work to quite high precision through
2622 triplets. We could just keep adding things, I suppose...
2623
2624 ------------------------------------------------------------------------
2625 r103 | rgb | 2004-11-18 18:17:39 -0500 (Thu, 18 Nov 2004) | 22 lines
2626
2627 This is now working. Working amazingly well, actually. Well enough
2628 to double the size of the bits in rgb_bitdist_test().
2629
2630 The one MAJOR remaining problem is that I cannot use samples for tests
2631 that return a vector of pvalues. Oh, and it is fairly difficult to
2632 pass arguments to the testing function in when it is an argument TO
2633 samples().
2634
2635 This means that I have two itty bitty problems to solve -- one is to
2636 pass in parameters (possibly by making them global variables). This
2637 makes sense IF I want to be able to control them from the command line
2638 anyway. The other is to return a vector of pvalues. The only way I
2639 can think of doing this is to make pvalues[] a global vector as well
2640 of length (say) 1K. This puts an upper bound on the number of pvalues
2641 that can be returned by a test, but that SHOULDN'T be much of a problem,
2642 as it is really a question of what granularity one wishes to evaluate p
2643 at.
2644
2645 Anyway, just a BIT more work and rgb_bitdist should be production ready,
2646 AND I should be perfectly ready to clean up p-sampling and testing as
2647 separate entities in the other tests.
2648
2649 ------------------------------------------------------------------------
2650 r102 | rgb | 2004-11-18 17:28:50 -0500 (Thu, 18 Nov 2004) | 2 lines
2651
2652 This MIGHT be working.
2653
2654 ------------------------------------------------------------------------
2655 r101 | rgb | 2004-11-18 15:42:41 -0500 (Thu, 18 Nov 2004) | 2 lines
2656
2657 I sent this home, I did, I did.
2658
2659 ------------------------------------------------------------------------
2660 r100 | rgb | 2004-11-18 15:08:32 -0500 (Thu, 18 Nov 2004) | 14 lines
2661
2662 This is still fairly screwed up, at least in the sense that
2663 it doesn't look like rgb_bitdist works. Curiously, it LOOKS like
2664 it WORKS -- walking through the code, it looks VERY much like it
2665 is collecting two bits at a time and correctly incrementing the
2666 correct bit count in the correct vector.
2667
2668 The final histogram, however, comes out wrong, wrong, wrong.
2669
2670 I may have to make this simpler. Or maybe I'm doing something else
2671 wrong -- come to think of it, the totals in the histograms shouldn't
2672 equal the number of samples for EACH value of the ntuple, only the
2673 total should sum to the number of samples. Maybe this is what is
2674 wrong...
2675
2676 ------------------------------------------------------------------------
2677 r99 | rgb | 2004-11-18 11:27:27 -0500 (Thu, 18 Nov 2004) | 13 lines
2678
2679 We have to go into Duke, but we are very much ready to finish off bits.c
2680 and rgb_bitdist.c to where we can eliminate BOTH rgb_binomial AND
2681 rgb_bit2.c AND at least one, maybe 3-4 diehard tests AND a couple or
2682 three sts tests as being equivalent to this test, for particular call
2683 values. I have great hope that this rgb_bitdist will become "the" bit
2684 frequency test for all random bit sequences. There may still be a point
2685 to tests that look at intervals >>between<< bit sequences thought.
2686
2687 In fact, I suspect that the best way to proceed with the latter is to
2688 test lagged correlation for arbitrary lags in a long bitstring. This
2689 SAME TEST applied with arbitrary displacements between samples might be
2690 revealing...
2691
2692 ------------------------------------------------------------------------
2693 r98 | rgb | 2004-11-17 14:57:04 -0500 (Wed, 17 Nov 2004) | 5 lines
2694
2695 This is HIGHLY BROKEN but is absolutely necessary. We have to break
2696 this code up to unify the replicated pieces and streamline the
2697 testing processes now that we know how a test "works" in the
2698 abstract.
2699
2700 ------------------------------------------------------------------------
2701 r97 | rgb | 2004-11-16 22:40:28 -0500 (Tue, 16 Nov 2004) | 16 lines
2702
2703 A bugfix commit. The sanity check in get_bit() is broken and is
2704 commented out -- if I'm going to allocate rand_int[] vectors other than
2705 size in length, I cannot test on a global size to see if get_bit() is
2706 out of bounds. This is STILL broken in that there is a risk with no
2707 warning, but there is also functionality for the moment (and I have
2708 to write a bunch of new bitlevel functions and can rewrite get_bits
2709 at the same time).
2710
2711 More important, I found a real bug in Btest where I was initializing
2712 btest->chisq to zero before accumulation but was accumulating in chisq.
2713 Curiously, it worked a lot of the time the old way, but only for certain
2714 rng's. I may have memory management problem, which isn't surprising
2715 given the slovenliness of the code at this moment...;-)
2716
2717 rgb
2718
2719 ------------------------------------------------------------------------
2720 r96 | rgb | 2004-11-16 18:46:17 -0500 (Tue, 16 Nov 2004) | 2 lines
2721
2722 This is a tagged checkin, about to push.
2723
2724 ------------------------------------------------------------------------
2725 r94 | rgb | 2004-11-16 18:45:43 -0500 (Tue, 16 Nov 2004) | 2 lines
2726
2727 This appears ready for a checkin.
2728
2729 ------------------------------------------------------------------------
2730 r93 | rgb | 2004-11-16 18:19:54 -0500 (Tue, 16 Nov 2004) | 4 lines
2731
2732 This appears to FINALLY fix rgb_binomial so that it reliably works.
2733 It remains to be seen whether or not it is any more senstitive than e.g.
2734 sts_monobit.
2735
2736 ------------------------------------------------------------------------
2737 r92 | rgb | 2004-11-16 15:11:18 -0500 (Tue, 16 Nov 2004) | 3 lines
2738
2739 This is STILL broken as far as the Ntests are concerned. I've really
2740 got to figure this out...
2741
2742 ------------------------------------------------------------------------
2743 r91 | rgb | 2004-11-16 10:51:23 -0500 (Tue, 16 Nov 2004) | 2 lines
2744
2745 Send this to Duke to finish this morning.
2746
2747 ------------------------------------------------------------------------
2748 r90 | rgb | 2004-11-14 12:40:14 -0500 (Sun, 14 Nov 2004) | 61 lines
2749
2750 This is a pretty serious bugfix -- probably need to update the website.
2751 Basically, my kstest was simply wrong last night; now it is working,
2752 I've also added the Kuiper form of the KS test, and will program
2753 Anderson and Darling's version (the one used, apparently, in Diehard)
2754 when I get around to it. However, for tests involving more than a very
2755 few p-values in a vector, it shouldn't really matter -- Kuiper KS and the
2756 regular KS and Anderson-Davis KS should all GENERALLY generate similar
2757 p-distributions -- different perhaps where they don't matter, but very
2758 similar at the ends. The key question is just whether one has a
2759 tendency to pass a vector of p-values where the other would consistently
2760 fail it. So far it looks like USUALLY if one fails the other fails.
2761
2762 I think I'm still going to want to do a histogram picture of binned
2763 pvalues and do a Pearson chisq p-test on the result. This should really
2764 be pretty easy... maybe today, maybe not.
2765
2766 We're getting close to being ready to go BACK and mess with the Ntest
2767 and Xtest stuff. I think that now that I understand Pearson vs the
2768 alternatives, I can PROBABLY arrange things so that I can use a single
2769 set of common tools to do all the test assessment.
2770
2771 One thing, for example, would be to make each test return a p-value,
2772 period, and put the samples loop in rand_rate_work, to ALWAYS fill a
2773 vector of p-values and ALWAYS do KS tests, confidence interval tests,
2774 and histogram tests. This would have a number of advantages -- being
2775 able to produce a really pretty, really standardized picture of results
2776 for one.
2777
2778 A second thing that would make this tool relatively interesting to the
2779 mass unwashed would be to put a nice little GUI onto it. There are two
2780 generic ways to do this. One is to leave it a command line tool but
2781 REALLY clean up the output result so that it is just a single line
2782 per test with ks test scores for the various forms of test with three
2783 lines of # delimited frame, period. Then I can make a perl-Gtk app to
2784 call the binary, parse the result, and (e.g.) plot histograms or do
2785 other nifty graphical things. The other is to use Gtk directly, but
2786 perhaps have the GUI only come up if there is an appropriate command
2787 line flag (or not).
2788
2789 A third thing to work on is clearly going to be splitting the source
2790 into distinct components in distinct directories. We will need a common
2791 library containing the kstest, chisq, Ntest, Xtest etc code, the input,
2792 the output, etc. We will need a directory containing extensions to the
2793 GSL random number library, e.g. /dev/random, /dev/urandom, empty, and
2794 shuffled because the one thing that is absolutely true is that we need
2795 to add a shuffling/mixing random number generator, one that permits us
2796 to set up a shuffling list and refill it from a secondary LIST of rng's!
2797
2798 A fourth thing (noted already elsewhere) is to do the simplest of tests
2799 -- apply a KS test to the GSL distribution-specific generators
2800 themselves. If a "test" is generically generating a known distribution
2801 presuming randomness and then seeing if the result is indeed the
2802 targeted distribution, then EVERY distribution generator in the GSL can
2803 simultaneously be the target of a test for algorithmic purposes AND a
2804 test component for the GSL rng's.
2805
2806 Beyond that, I need to implement spectral tests and tests for
2807 hyperplanes in N dimensions and uniformity tests. Sigh. I think that I
2808 DO need to write a grant proposal for this -- I think there is enough
2809 work to justify it.
2810
2811 ------------------------------------------------------------------------
2812 r89 | rgb | 2004-11-14 04:38:33 -0500 (Sun, 14 Nov 2004) | 3 lines
2813
2814 Tagged and on the repository as 0.4.3, with diehard 3d spheres and a
2815 MAYBE working KS test.
2816
2817 ------------------------------------------------------------------------
2818 r87 | rgb | 2004-11-14 04:37:35 -0500 (Sun, 14 Nov 2004) | 2 lines
2819
2820 Sort of playing with KS -- I'm not done here yet...
2821
2822 ------------------------------------------------------------------------
2823 r86 | rgb | 2004-11-14 03:04:37 -0500 (Sun, 14 Nov 2004) | 8 lines
2824
2825 OK, found my REALLY stupid bug. I was computing the absolute
2826 length of r from the origin, not the distance between point pairs.
2827 No, I wasn't even doing that well -- I was computing the dot product
2828 of the random vectors. Now things look nearly correct.
2829
2830 All I need is a KS test and life would be, if not complete, well, worth
2831 living.
2832
2833 ------------------------------------------------------------------------
2834 r85 | rgb | 2004-11-14 02:22:29 -0500 (Sun, 14 Nov 2004) | 5 lines
2835
2836 This "works". Except that it doesn't. It's very odd, but although
2837 it works perfectly as far as I can tell by any measure, r is simply not
2838 as small as it should be in order to make the pvalue come out between
2839 0 and 1.
2840
2841 ------------------------------------------------------------------------
2842 r84 | rgb | 2004-11-13 20:19:53 -0500 (Sat, 13 Nov 2004) | 2 lines
2843
2844 THis is on the way to being another test.
2845
2846 ------------------------------------------------------------------------
2847 r83 | rgb | 2004-11-13 17:19:55 -0500 (Sat, 13 Nov 2004) | 2 lines
2848
2849 This is hopefully a tagged snapshot with a new test!
2850
2851 ------------------------------------------------------------------------
2852 r81 | rgb | 2004-11-13 16:50:54 -0500 (Sat, 13 Nov 2004) | 8 lines
2853
2854 OK, time to bump the revision number, as birthdays is home and even
2855 works tolerably, as far as I can tell.
2856
2857 SOON I'm going to do the KS test on vectors of p's. SOON I'm going to
2858 really clean up the code so that chisq -> p is consistently computed,
2859 and so that a set of p's is consistently evaluated for the
2860 random/nonrandom decision.
2861
2862 ------------------------------------------------------------------------
2863 r80 | rgb | 2004-11-13 12:17:57 -0500 (Sat, 13 Nov 2004) | 2 lines
2864
2865 This is it, ready to proceed.
2866
2867 ------------------------------------------------------------------------
2868 r79 | rgb | 2004-11-13 12:12:24 -0500 (Sat, 13 Nov 2004) | 3 lines
2869
2870 This checks in a placeholder for a Kolmogorov-Smirnov test, likely to
2871 be applied to a vector of p-values.
2872
2873 ------------------------------------------------------------------------
2874 r78 | rgb | 2004-11-13 12:06:36 -0500 (Sat, 13 Nov 2004) | 10 lines
2875
2876 We'll commit this for the moment. I think the sensible thing to do is
2877 to create as general as possible a tool for generating Pearson's chisq
2878 for discretely binned data, in particular and immediately for the
2879 Poissonian birthday histogram but also for other purposes. Note that
2880 these routines should not only generate chisq, but when possible go
2881 ahead and compute goodness of fit p-values, ideally in a vector
2882 associated with independent trials. This vector of p-values can itself
2883 then be subjected to a kolmogorov-smirnov analysis and transformed into
2884 a conclusion for the generator being tested.
2885
2886 ------------------------------------------------------------------------
2887 r77 | rgb | 2004-11-13 04:34:29 -0500 (Sat, 13 Nov 2004) | 3 lines
2888
2889 Just added output of lambda, which is indeed 2 with the parameters
2890 given...
2891
2892 ------------------------------------------------------------------------
2893 r76 | rgb | 2004-11-13 04:32:46 -0500 (Sat, 13 Nov 2004) | 7 lines
2894
2895 This is REALLY CLOSE to having diehard birthdays finished. We just
2896 need to add a chisq test for Poisson distributions sampled samples times
2897 with known (per sample) lambda, and a loop to convert a table of chisq into
2898 a table of p-values. I'm tempted to bump minor and tag, but I shouldn't
2899 need to -- I've been really careful and things really look like they're
2900 working so far.
2901
2902 ------------------------------------------------------------------------
2903 r75 | rgb | 2004-11-12 21:02:55 -0500 (Fri, 12 Nov 2004) | 2 lines
2904
2905 This is a simple checkin prior to doing diehard birthdays test.
2906
2907 ------------------------------------------------------------------------
2908 r74 | rgb | 2004-11-12 20:32:58 -0500 (Fri, 12 Nov 2004) | 2 lines
2909
2910 This splits off the confidence interval test from STS docs.
2911
2912 ------------------------------------------------------------------------
2913 r73 | rgb | 2004-11-12 20:28:07 -0500 (Fri, 12 Nov 2004) | 3 lines
2914
2915 This is a small adjustment (still in 0.4.1 I guess). Let's try another
2916 diehard, I think.
2917
2918 ------------------------------------------------------------------------
2919 r72 | rgb | 2004-11-12 20:16:48 -0500 (Fri, 12 Nov 2004) | 10 lines
2920
2921 This is actually a fairly functional diehard test!
2922
2923 I think that we can actually implement a test for the uniformity of
2924 p-values as suggested by NIST to run on TOP of the existing confidence
2925 interval test. This would actually break the p-distribution down by
2926 interval and return a p-value of its own computed against the assumption
2927 of uniformity. Or I could get fancier and try kolmogorov-smirnov, if
2928 GSL doesn't have one and I work hard enough to program one. If this is
2929 really distinct -- it isn't clear that it is.
2930
2931 ------------------------------------------------------------------------
2932 r70 | rgb | 2004-11-12 17:32:03 -0500 (Fri, 12 Nov 2004) | 4 lines
2933
2934 This is actually sort of semi-functional. What I >>really<< need now
2935 is canned Kolmogorov-Smirnov code. Could it be that this is in the
2936 GSL already? I'll be it is...
2937
2938 ------------------------------------------------------------------------
2939 r69 | rgb | 2004-11-11 10:59:28 -0500 (Thu, 11 Nov 2004) | 2 lines
2940
2941 Continuing to hack this up.
2942
2943 ------------------------------------------------------------------------
2944 r68 | rgb | 2004-11-10 17:19:21 -0500 (Wed, 10 Nov 2004) | 3 lines
2945
2946 This is a nearly functional diehard_runs -- I just need to figure out
2947 what the expected values and sigmas are...
2948
2949 ------------------------------------------------------------------------
2950 r67 | rgb | 2004-11-10 00:32:44 -0500 (Wed, 10 Nov 2004) | 4 lines
2951
2952 This is simply lovely. A nice litte addition to the Makefile that
2953 automatically indicates the current version in the abstract. I actually
2954 have things fairly distributable!
2955
2956 ------------------------------------------------------------------------
2957 r66 | rgb | 2004-11-10 00:24:18 -0500 (Wed, 10 Nov 2004) | 10 lines
2958
2959 OK, added a few minor changes to manage the bits issue yet another way.
2960
2961 Really, I'm going to have to figure out a consistent way of indicating
2962 whether a test can have size OR bits OR both OR neither specified.
2963
2964 Also, it would be really lovely to have another outer loop and to
2965 present the lowest p in a set of (say) ten runs of a test combo.
2966 Although in many cases running with -s 10x larger should do the same
2967 thing, really.
2968
2969 ------------------------------------------------------------------------
2970 r65 | rgb | 2004-11-09 23:53:49 -0500 (Tue, 09 Nov 2004) | 2 lines
2971
2972 This is it and running, version 0.4.0 as published. Seems to work.
2973
2974 ------------------------------------------------------------------------
2975 r63 | rgb | 2004-11-09 23:53:15 -0500 (Tue, 09 Nov 2004) | 2 lines
2976
2977 One last checkin, then a tag, then a checkin as published.
2978
2979 ------------------------------------------------------------------------
2980 r61 | rgb | 2004-11-09 20:43:14 -0500 (Tue, 09 Nov 2004) | 3 lines
2981
2982 This is a tagged release, mostly bugfixes. At the moment it all looks
2983 like it works.
2984
2985 ------------------------------------------------------------------------
2986 r59 | rgb | 2004-11-09 20:37:58 -0500 (Tue, 09 Nov 2004) | 9 lines
2987
2988 This seems to work perfectly, for the very short moment. It is by no
2989 means perfect or mutually exclusive. We very definitely need to
2990 generalize the bitdist test to handle bit ntuples of arbitrary length,
2991 where the length is a variable.
2992
2993 I think I'll retag this. It is also probably time to think about
2994 putting this up on the website, especially if I'm going to write a
2995 proposal on it.
2996
2997 ------------------------------------------------------------------------
2998 r58 | rgb | 2004-11-09 17:14:57 -0500 (Tue, 09 Nov 2004) | 2 lines
2999
3000 Because it wasn't checked in!
3001
3002 ------------------------------------------------------------------------
3003 r57 | rgb | 2004-11-09 17:14:36 -0500 (Tue, 09 Nov 2004) | 2 lines
3004
3005 Why didn't bit2.c go home?
3006
3007 ------------------------------------------------------------------------
3008 r56 | rgb | 2004-11-09 15:09:36 -0500 (Tue, 09 Nov 2004) | 3 lines
3009
3010 This is going home with a split out routine and some nice changes that
3011 will make it easier to add new tests with arbitrary numbers.
3012
3013 ------------------------------------------------------------------------
3014 r55 | rgb | 2004-11-09 14:30:32 -0500 (Tue, 09 Nov 2004) | 2 lines
3015
3016 Just checking repository Root.
3017
3018 ------------------------------------------------------------------------
3019 r54 | rgb | 2004-11-09 14:30:05 -0500 (Tue, 09 Nov 2004) | 2 lines
3020
3021 Let's send this home...
3022
3023 ------------------------------------------------------------------------
3024 r53 | rgb | 2004-11-09 09:51:31 -0500 (Tue, 09 Nov 2004) | 7 lines
3025
3026 OK, fixing Makefile to actually get this home, AND adding the URL
3027 of the web reference from the last checkin:
3028
3029 http://world.std.com/~franl/crypto/random-numbers.html
3030
3031 (we need to implement some of its hyperplane tests).
3032
3033 ------------------------------------------------------------------------
3034 r52 | rgb | 2004-11-09 09:48:40 -0500 (Tue, 09 Nov 2004) | 5 lines
3035
3036 We're actually working on this once again. I need to get my own "runs"
3037 test working, as it will replace a whole RANGE of STS, and I need to
3038 implement a spectral distribution test with bins as is done in the
3039 nice web reference I found.
3040
3041 ------------------------------------------------------------------------
3042 r51 | rgb | 2004-11-08 09:52:52 -0500 (Mon, 08 Nov 2004) | 2 lines
3043
3044 This adds yet another built-in device to GSL.
3045
3046 ------------------------------------------------------------------------
3047 r50 | rgb | 2003-06-10 11:21:15 -0400 (Tue, 10 Jun 2003) | 3 lines
3048
3049 This adds an "empty" generator to help us determine gsl call overhead
3050 separately.
3051
3052 ------------------------------------------------------------------------
3053 r49 | rgb | 2003-01-30 17:16:15 -0500 (Thu, 30 Jan 2003) | 4 lines
3054
3055 This is broken as shit. I see what I did -- I made the ntest evaluation
3056 and presentation routines use n+1 bits (because in rgb_binomial I needed
3057 to do the end of the binomial). However, I have to fix it later...
3058
3059 ------------------------------------------------------------------------
3060 r48 | rgb | 2003-01-30 15:12:20 -0500 (Thu, 30 Jan 2003) | 2 lines
3061
3062 Forgot to send this...
3063
3064 ------------------------------------------------------------------------
3065 r47 | rgb | 2003-01-29 14:19:55 -0500 (Wed, 29 Jan 2003) | 4 lines
3066
3067 Not obviously broken, and time to add bitpair counters. Should be
3068 really easy -- left-shift in two bits at a time to creat the int index
3069 of the counter, then increment it.
3070
3071 ------------------------------------------------------------------------
3072 r46 | rgb | 2003-01-29 09:27:09 -0500 (Wed, 29 Jan 2003) | 2 lines
3073
3074 Some NOTES on future work.
3075
3076 ------------------------------------------------------------------------
3077 r45 | rgb | 2003-01-29 09:14:38 -0500 (Wed, 29 Jan 2003) | 8 lines
3078
3079 This checks in a whole new test, which should probably be combined with
3080 sts_monobit (it generates monobit stats as it goes)
3081 rgb_persist (one can easily generate a bitmask as one goes)
3082 rgb_binomial (one can generate binomial stats on top of monobit as one
3083 goes).
3084
3085 and possibly with more tests.
3086
3087 ------------------------------------------------------------------------
3088 r44 | rgb | 2003-01-26 02:54:43 -0500 (Sun, 26 Jan 2003) | 7 lines
3089
3090 This last little pair of changes causes measure_rate to use its own,
3091 fixed, number of samples ("more than enough"). It also installs a
3092 "summary report" mode that isn't horribly useful because of conflict
3093 between e.g. -b, -n, -s definitions here and there. Also, different
3094 tests need to be run in different ways to demonstrate failure (or a
3095 lack thereof).
3096
3097 ------------------------------------------------------------------------
3098 r43 | rgb | 2003-01-26 02:23:28 -0500 (Sun, 26 Jan 2003) | 9 lines
3099
3100 OK, we haven't done TOO much, but we have definitely learned that
3101 all the rng's that are weak in rgb_persist will definitely fail the
3102 monobit test (for obvious reasons). Furthermore, when a generator
3103 is weak in certain bits and we evaluate the bits from the other end
3104 (whichever end that might be!) it can often PASS the monobit test.
3105
3106 Bits that repeat, random_max's that aren't powers of two-1 (and probably
3107 EVEN powers of two at that) are going to be trouble!
3108
3109 ------------------------------------------------------------------------
3110 r42 | rgb | 2003-01-26 00:00:04 -0500 (Sun, 26 Jan 2003) | 4 lines
3111
3112 This is a VERY IMPORTANT new test, rgb_persist(), and a very useful
3113 new routine, dumpbits(). Read NOTES (and inline comments and output)
3114 to see a bit of what it does and why it is important.
3115
3116 ------------------------------------------------------------------------
3117 r41 | rgb | 2003-01-25 16:55:32 -0500 (Sat, 25 Jan 2003) | 3 lines
3118
3119 This actually works. In fact, it works fabulously. I can directly
3120 and fairly powerfully look for bitlevel correlations in the output.
3121
3122 ------------------------------------------------------------------------
3123 r40 | rgb | 2003-01-25 15:53:55 -0500 (Sat, 25 Jan 2003) | 16 lines
3124
3125 OK, we've learned the hard way that some bits in e.g. boroshi don't
3126 change AT ALL, EVER. Which makes it pretty hard to be random, of
3127 course.
3128
3129 So we're going to invent a new tool -- rgb_persist(), which doesn't
3130 (yet) to a formal statistical test. It just is going to dump successive
3131 unsigned ints from the rng (bitwise) AND maybe run a string of &'s on
3132 the string of ints returned. If they share any 1 bits, the successive
3133 &'s will preserve them a LOT longer than permitted by binary flips on
3134 the slots.
3135
3136 This could be made into a fairly powerful bitlevel sequential
3137 correlation test in several ways. We'll investigate them as we go, but
3138 one reason to write this now is that I'm not quite convinced that what
3139 I'm seeing isn't some sort of bug in the get_bit() routine or the like.
3140
3141 ------------------------------------------------------------------------
3142 r39 | rgb | 2003-01-25 10:54:02 -0500 (Sat, 25 Jan 2003) | 3 lines
3143
3144 This is well on the way to being MUCH better, and ready to
3145 systematically extend.
3146
3147 ------------------------------------------------------------------------
3148 r38 | rgb | 2003-01-24 19:50:32 -0500 (Fri, 24 Jan 2003) | 2 lines
3149
3150 And now we send the tagged package to Duke.
3151
3152 ------------------------------------------------------------------------
3153 r36 | rgb | 2003-01-24 19:50:07 -0500 (Fri, 24 Jan 2003) | 2 lines
3154
3155 Checking in the notes.
3156
3157 ------------------------------------------------------------------------
3158 r35 | rgb | 2003-01-24 17:52:03 -0500 (Fri, 24 Jan 2003) | 20 lines
3159
3160 This is worth a minor bump. First, we fixed get_bit(). Second, we
3161 completed sts_runs (for what it is worth, which isn't a whole lot as
3162 nearly everything that fails it also fails monobit and binomial as
3163 expected). However, working through it suggests how to make binomial
3164 work better.
3165
3166 Next (to make it easier to check results relative to the sts documents)
3167 I need to implement -b (get_bit(0 permits this pretty much
3168 transparently, at least in the sts routines) and implement a -f filename
3169 filled with e.g. raw bitstrings or ascii floats or binary numbers
3170 in xmlish wrappers that indicate the storage mechanism? Thus I can test
3171 explicit short bitstrings against the explicit sts numbers to be sure
3172 that my erfc and conversions (and sometimes slightly different
3173 implementation) yield the same answers as theirs, except where I don't
3174 care because I think theirs are (basically) wrong.
3175
3176 See also NOTES (about to be checked in) for a fairly detailed beginning
3177 critique of sts, which I don't think is particularly strong or useful,
3178 really.
3179
3180 ------------------------------------------------------------------------
3181 r34 | rgb | 2003-01-24 16:43:14 -0500 (Fri, 24 Jan 2003) | 6 lines
3182
3183 This is my home-grown version of sts_runs. It is no better than the
3184 actual sts version, really, but the sts version is not terribly good.
3185 I'm going to add a (hopefully vastly improved) binomial version of the
3186 test to rgb_binomial, where I can do all the tests at once with a
3187 single set of code and multiple trials (random number seeds).
3188
3189 ------------------------------------------------------------------------
3190 r33 | rgb | 2003-01-23 00:51:56 -0500 (Thu, 23 Jan 2003) | 3 lines
3191
3192 Just adding some notes, and preparing to add the next sts test,
3193 TOMORROW.
3194
3195 ------------------------------------------------------------------------
3196 r32 | rgb | 2003-01-22 23:52:47 -0500 (Wed, 22 Jan 2003) | 2 lines
3197
3198 I have no idea why the tag went down into fitany...
3199
3200 ------------------------------------------------------------------------
3201 r30 | rgb | 2003-01-22 23:52:07 -0500 (Wed, 22 Jan 2003) | 14 lines
3202
3203 This ups the minor revision number to 0.3.0. Worthwhile because now I
3204 have BOTH an erfc AND a Q evaluation of p-value. I could certainly
3205 prettify sts_monobit, but since I generally think that it isn't that
3206 great a test (although it does indicate how starkly many rng's FAIL to
3207 be even this random) I won't do so right away.
3208
3209 Next (after tagging and resync'ing) is going to be adding more tests.
3210 At this point adding a test should be pretty easy, given the hopefully
3211 reusable routines I have written to do the pre- and post- processing.
3212 All I really have to do is input the expected values, write a loop to
3213 generate the "experimental" statistic, and pass everything on to a
3214 standard set of tools for outputting the results and deciding on the
3215 quality of the results.
3216
3217 ------------------------------------------------------------------------
3218 r29 | rgb | 2003-01-22 23:46:08 -0500 (Wed, 22 Jan 2003) | 7 lines
3219
3220 All right, this LOOKS like it correctly implements the STS monobit
3221 frequency test. I would still claim that anything that fails this test
3222 will also fail the binomial test, and that in addition things that pass it
3223 (e.g. the vax rng) FAIL the binomial test, so the monobit test is
3224 a waste of time and more prone to error. However, mine is not to reason
3225 why...
3226
3227 ------------------------------------------------------------------------
3228 r28 | rgb | 2003-01-22 13:04:47 -0500 (Wed, 22 Jan 2003) | 7 lines
3229
3230 This is working incredibly well, and I've split off nearly everything
3231 required to make further n-point chisq tests trivial to implement and
3232 assess. All that remains is to do a 1-point (normal) test such as the
3233 sts_monobit test (which should really be done internal to the
3234 rgb_binomial test and may one day be, but for the moment we'll just do
3235 it directly).
3236
3237 ------------------------------------------------------------------------
3238 r27 | rgb | 2003-01-22 08:33:21 -0500 (Wed, 22 Jan 2003) | 2 lines
3239
3240 Just making SURE this is at Duke...
3241
3242 ------------------------------------------------------------------------
3243 r26 | rgb | 2003-01-21 18:56:15 -0500 (Tue, 21 Jan 2003) | 18 lines
3244
3245 This works just lovely!
3246
3247 HOWEVER, it is also clear that running it once, twice, three times,
3248 for EACH generator looking for good ones is a PITA. We'll have to
3249 eventually rearrange this so that there is a "search mode" that runs
3250 a loop through all known generators, identifying the ones that pass at
3251 least at the 1% or higher level.
3252
3253 BTW, I'm now prepared to bet a nickel that the rgb binomial test has a
3254 great deal of sensitivity, since it fails all but literally three or
3255 four of the available RNG's for absurdly short data strings. As in they
3256 aren't even APPROXIMATELY random...NONE of them. If one used them to
3257 generate a humble binomial distribution numerically it would be in
3258 significant error.
3259
3260 I do need to alter this test so that I can run it for arbitrary bit
3261 string lengths, but for the moment I'm not going to worry about it.
3262
3263 ------------------------------------------------------------------------
3264 r25 | rgb | 2003-01-21 16:40:07 -0500 (Tue, 21 Jan 2003) | 3 lines
3265
3266 This is now VERY CLOSE. I should be able to determine chisq in a matter
3267 of minutes when I return...
3268
3269 ------------------------------------------------------------------------
3270 r24 | rgb | 2003-01-21 14:05:50 -0500 (Tue, 21 Jan 2003) | 2 lines
3271
3272 This is considerably cleaner and more decrufted...
3273
3274 ------------------------------------------------------------------------
3275 r23 | rgb | 2003-01-21 13:35:01 -0500 (Tue, 21 Jan 2003) | 3 lines
3276
3277 This finishes the split off of list_rand and list_rngs from the code.
3278 I do need to "fix" the Usage() routine to reflect the change.
3279
3280 ------------------------------------------------------------------------
3281 r22 | rgb | 2003-01-21 13:08:34 -0500 (Tue, 21 Jan 2003) | 3 lines
3282
3283 Breaking things up into subroutines a bit better to clarify the program
3284 structure.
3285
3286 ------------------------------------------------------------------------
3287 r21 | rgb | 2003-01-17 15:37:14 -0500 (Fri, 17 Jan 2003) | 7 lines
3288
3289 This is coming along, although I'm silly for not just finishing the
3290 monobit test before introducing a binomial test. Still, all very
3291 instructive.
3292
3293 I need to get all this on my laptop and take it with me, along with the
3294 notebook.
3295
3296 ------------------------------------------------------------------------
3297 r20 | rgb | 2003-01-17 14:28:36 -0500 (Fri, 17 Jan 2003) | 10 lines
3298
3299 Fixes a nasty bug in sts_monobit, which I think I'm gonna rename
3300 rgb_binomial (and screw sts's monobit test, which is immensely sloppy
3301 compared to actually systematically exploring the binomial distribution
3302 of 1's and 0's in the overall bit strings generated by different seeds.
3303
3304 Actually, a better thing still is to leave sts_monobit, but add
3305 rgb_binomial and document that it is more sensitive (in particular, that
3306 e.g. alternating series that easily pass monobit fail binomial, and that
3307 NOTHING that fails monobit will PASS binomial).
3308
3309 ------------------------------------------------------------------------
3310 r19 | rgb | 2003-01-17 01:06:32 -0500 (Fri, 17 Jan 2003) | 2 lines
3311
3312 Sending off the tag
3313
3314 ------------------------------------------------------------------------
3315 r17 | rgb | 2003-01-17 01:06:16 -0500 (Fri, 17 Jan 2003) | 8 lines
3316
3317 OK, this is good for a full minor number bump to 0.2.0. We have
3318 basically installed the guts of the STS monobit test. All that we
3319 lack is the computation of the statistics and p-value, which should be
3320 fairly straightforward, especially with the gsl handy. I SHOULD be
3321 able to just cumulate the one-count (e.g.) in a vector and hand it to
3322 the gsl stats routines and have mean, stddev, skew, kurtosis, and
3323 anything else I might like just handed back to me...
3324
3325 ------------------------------------------------------------------------
3326 r16 | rgb | 2003-01-17 00:16:59 -0500 (Fri, 17 Jan 2003) | 4 lines
3327
3328 Just a bit of cleanup, and some moderately important additions.
3329
3330 Now we REALLY need to think about tests.
3331
3332 ------------------------------------------------------------------------
3333 r15 | rgb | 2003-01-16 23:23:32 -0500 (Thu, 16 Jan 2003) | 2 lines
3334
3335 Tagged.
3336
3337 ------------------------------------------------------------------------
3338 r13 | rgb | 2003-01-16 23:23:04 -0500 (Thu, 16 Jan 2003) | 7 lines
3339
3340 This is about ready for a semipermanent snapshot, so I bumped the
3341 minor version number. I'd say that we are now "good" with the ability
3342 to add sw rng's, including interfaces to hw rng's square within the gsl
3343 format.
3344
3345 Now to give de old tests a try...
3346
3347 ------------------------------------------------------------------------
3348 r12 | rgb | 2003-01-16 22:54:25 -0500 (Thu, 16 Jan 2003) | 6 lines
3349
3350 Hot diggity dawg! It works! However, I don't need types.c. All I
3351 need is to follow the dev_random.c template and call a routine
3352 add_my_rngs() (to be defined) before working with gsl's rng's, and
3353 keep track (crudely) of which ones are which. So this can be decrufted
3354 a bit and then reorganized now that I know how it works.
3355
3356 ------------------------------------------------------------------------
3357 r11 | rgb | 2003-01-16 21:54:49 -0500 (Thu, 16 Jan 2003) | 4 lines
3358
3359 We'll try these as the basic wrappers required. With luck we'll
3360 override the types subroutine in gsl itself, although I do have my
3361 doubts...
3362
3363 ------------------------------------------------------------------------
3364 r10 | rgb | 2003-01-16 21:38:52 -0500 (Thu, 16 Jan 2003) | 6 lines
3365
3366 This significantly improves the Usage and cl parsing, and pre-structures
3367 it for addition of sts/diehard tests.
3368
3369 We still need to see if we can gsl-wrap our own tests without a full
3370 gsl recompile.
3371
3372 ------------------------------------------------------------------------
3373 r9 | rgb | 2003-01-16 15:40:56 -0500 (Thu, 16 Jan 2003) | 2 lines
3374
3375 Sending the tagged copy home...
3376
3377 ------------------------------------------------------------------------
3378 r7 | rgb | 2003-01-16 15:40:38 -0500 (Thu, 16 Jan 2003) | 2 lines
3379
3380 This is now going to be v_0_1_0.
3381
3382 ------------------------------------------------------------------------
3383 r6 | rgb | 2003-01-16 15:39:55 -0500 (Thu, 16 Jan 2003) | 15 lines
3384
3385 This is now functional UP TO all the gsl rngs, not any of the add-ons.
3386
3387 Which is fine, as we'll probably completely change how the add-ons work.
3388
3389 Next, we need to do all of the following, in some order:
3390
3391 a) figure out how to wrap up new gsl_rngs, preferrably without
3392 recompiling the whole damn library.
3393 b) decruft all the command line options and no-longer-used variables.
3394 c) add back command line options for doing quality tests. Start with
3395 the very simplest test -- something from diehard or the bits test from
3396 sts.
3397 d) In the meantime, increment revision, tag, and consider "publishing"
3398 as we go.
3399
3400 ------------------------------------------------------------------------
3401 r5 | rgb | 2003-01-16 15:14:54 -0500 (Thu, 16 Jan 2003) | 3 lines
3402
3403 This SHOULD split rand_rate off so it has its own CVS tree outside of
3404 the "random" project overall, which I think is for the best.
3405
3406 ------------------------------------------------------------------------
3407 r4 | rgb | 2003-01-16 15:13:05 -0500 (Thu, 16 Jan 2003) | 4 lines
3408
3409 This actually works, and needs to be saved in snapshot form. I'm not
3410 at ALL certain that I'm getting accurate measurements in terms of the
3411 number of rands per second I can generate, but this too, we shall see...
3412
3413 ------------------------------------------------------------------------
3414 r3 | rgb | 2003-01-13 17:12:27 -0500 (Mon, 13 Jan 2003) | 2 lines
3415
3416 This is a fair amount of progress to having something working...
3417
3418 ------------------------------------------------------------------------
3419 r2 | rgb | 2003-01-11 19:07:56 -0500 (Sat, 11 Jan 2003) | 4 lines
3420
3421 This is basically the original checkin for my lookin-major random number
3422 project. By the time this is done, I'd doggone better have a paper or
3423 two out of it, if not more.
3424
3425 ------------------------------------------------------------------------