10500+ entries in 0.019s
mod6: had to hardpower off the fucking cocksucker
mod6: infact, this thing crashed my entire computer. pff
mod6: any other graph tool suggestions?
mod6: drawing of the graph is crashing LibreOffice Calc.
mod6: im having trouble making a graph with nearly a million datapoints in it.
☟︎ mod6: asciilifeform: np all tall, my pleasure
mod6: <+asciilifeform> not a big deal, read+write ~= total ProcessBlock time << makes sense. stangely, i overlooked that last patch
mod6: ok, i'll see what I can do about getting those two curves put together today.
mod6: <+asciilifeform> anyway looks like you don't have this << ahh, sadly, no.
mod6: that was in the odometer?
mod6: accept block, process block, and db write wait.
mod6: my checkblock stats are mean/median of all of those values
mod6: (btw, the full log is available via the blog), but here's a snippit of one block eaten
mod6: I don't see a "stall time".
mod6: lemme take a quick look
mod6: <+mircea_popescu> o look at that slim checkblock << thought it might be cool to do some perl, parse out the stats, do some calcs & provide.
☟︎ mod6: did never started on fire... so that was good.
mod6: save your lulz for the charts
mod6: yah. this thing is like single thread.
mod6: 2x Xeon 5650 (24 core) / 64Gb RAM / Kingston 1Tb SSD
mod6: 9181:40 = ~6.375 days
mod6: aight, lemme work on getting a post put up here.
mod6: I think I just reached the end of my eatblock test...
mod6: i'd like to have the discussion with rgb himself and see what the intent was there.
mod6: even if, misguided.
mod6: wonder if the intent is to provide greater coverage of the pool
mod6: i wonder if that was the one from osx.
mod6: <+asciilifeform>
http://archive.li/SnwEH << recall these ? << illustration of the rewind thing. << ah, interesting. who ran that, anyone recall?
mod6: (see posted output files on my site)
mod6: version listed at the site is 3.31.1, which is the same version I'm using.
mod6: copyright at this www.phy.duke.edu/~rgb/General/dieharder.php page is denoted as 2017, so he must still be ~somewhat~ active.
mod6: lemme see if I can dig him up
mod6: I'm using the USB-TTL cable that I bought, separate from the one shipped with the fg. There's a picture of the connection up close in the blog post. It is plugged directly into the computer.
mod6: I'm going to collect one more 1Gb+ from this 1st FG, run the ent & dh again, and then move on to the next one. Results should be done in ~24hrs.
mod6: running ent & dh on the output file. this was produced from the same fg unit that the blog was covering.
mod6: also, collected another 1.1Gb of entropy from fg since yesterday.
mod6: the offline eatblock of entire chain (or at least what I had before was getting blackholed)
mod6: should be done soon. will publish the results complete with nmon charts.
mod6: update: 453K+ blocks, into blk0050... the last one I had.
mod6: looks like 7655 minutes running so far. ~5.3 days.
mod6: update on offline-eatblock: 431K+ and into blk0041
mod6: gonna see what those results look like... then will probably continue on and test the remainder of the fgs I've got.
mod6: not sure if I mentioned yesterday, but I did start a collection of another ~1Gb of entropy from the same fg as I wrote up the blog post about.
mod6: Looks pretty solid to me.
mod6: asciilifeform: aha. ok, just read about that on the `ent' page. mine was: Monte Carlo value for Pi is 3.141730891 (error 0.00 percent).
mod6: i'd like to get more familiar with the types of statistical tests from `ent` and `dieharder`.
mod6: asciilifeform: werd. anyway, it's exciting to test these.
mod6: <+asciilifeform> NOW if it happens with all of the units in mod6 's parcel, and ~every~ run -- then problem. << this is key too. i'll perhaps collect another 1Gb+ a few more times from that initial first one and report back on the ent/dh results too.
mod6: i'll be testing the other ones and will report back how the ent/dh results look on each.
mod6: <+asciilifeform> incidentally i'd be curious to hear if anybody's ever seen a dieharder output on something that isn't a prng or prng-whitened trng, that consistently produced no 'weak'. << werd.
mod6: <+asciilifeform> and spiffy scope, mod6 << thx!
mod6: apparently 157385 seconds
mod6: I'll test all of my other units in this same fashion too.
mod6: anyway, just wanted to be /sure/ that I wasn't getting some sort of varient test results from the dh.
mod6: The only entire difference in the two runs was the rands/second at the top.
mod6: I re-ran the dh on the same fg.bin file (~1.2Gb) and the 'WEAK' tests were exactly the same ones, with exactly the same output values.
mod6: yeah, just one randomly selected unit.
mod6: i may dig into that more at some point here.
mod6: So in total from the dieharder: 112 PASSED, 2 WEAK
mod6: Ima run the dieharder again as `dieharder -a -g 201 -f fg.bin`, and see if I get the same results.
mod6: # The file file_input_raw was rewound 194 times rgb_lagged_sum| 31| 1000000| 100|0.00000479| WEAK
mod6: # The file file_input_raw was rewound 118 times rgb_lagged_sum| 23| 1000000| 100|0.00063764| WEAK
mod6: <+mircea_popescu> lol the 4-5 places where dieharder reports weakness prolly need some review of dieharder code. anyone feel like emailing the maintainers ? << I should note, I hope the pics of this arn't too misleading. if you like at the actual text output, there are only 2 'WEAK' tests, which are:
mod6: diana_coman: thanks, yw!
mod6: thanks mircea_popescu
mod6: also, 406K+ blocks eaten, and into blk0032 nao
mod6: well, i even added the unboxing pics in there too.