| <<<Back 1 day (to 2012/10/30) | 2012/10/31 |
mvrhel_laptop | henrys: I meant to send you an email that I was not going to be at the meeting. Sorry I did not do that | 00:05.45 |
| off to the last session for the day. worn out from all the talk of C++, C++ Amp, WinRT, Direct X. No one has said the word Metro though | 00:26.18 |
| but the desktop app is dead | 00:26.32 |
Robin_Watts | mvrhel_laptop: ping | 00:46.39 |
| e86db1f has the -i and -o stuff. | 00:47.16 |
| 4df939d has my work in progress stuff. | 00:47.28 |
kens | chrisl did you have a test file that creates a PDF file > 4 Gb ? | 09:10.58 |
chrisl | I used the one from the temp file bug, and changed the number of iterations through the loop to 5000 | 09:12.15 |
kens | OK I'll try that thanks | 09:12.47 |
| ALso eneed to check ps2write of course | 09:13.08 |
chrisl | kens: and I used this command line: http://pastebin.com/gVv6ZcBF | 09:13.28 |
kens | pdfwrite uses stell *everywhere* but I think a lot of them are safe as they are. I hope so because changing some of them would be very difficult. | 09:13.55 |
chrisl | Hmm, could be a problem - I see a lot of them storing in a long int :-( | 09:18.55 |
kens | A lot of them are temporary streams, and are therefore short (atrings for indexed spaces and colour spaces | 09:19.30 |
| Problem is, a number of them are passing values to other routines such as cos_add_* and those are *very* extensively used.]# | 09:20.05 |
| I may end up putting this aside until I have some more time | 09:20.20 |
| I've already had to rework a lot of stuff this morning and I'm by no means convinced I've gone far enough | 09:20.49 |
sebras | Icelands! see you soon! :) | 09:23.35 |
chrisl | Right, I'm off to stumble hopelessly around a squash court for while - bbiaw...... | 09:31.48 |
kens | bye | 09:31.55 |
chrisl_away | bye! | 09:32.03 |
chrisl | Huh, the cluster doesn't test PCL/PXL @ 300dpi..... | 13:48.30 |
henrys | chrisl:we agreed on 600 and 75 I thought. | 14:33.02 |
chrisl | henrys: fair enough - I just assumed as there were only two "approved" resolutions we'd be testing them both - it's not important | 14:34.00 |
Robin_Watts | OK, I must be being dim. What is wrong with http://ghostscript.com/~robin/out.pam ? | 14:38.01 |
henrys | yes I think 300 and 600 would probably be a good decision too but I like the 75 dpi speed for local testing | 14:38.10 |
Robin_Watts | sheesh. Now it works. | 14:38.50 |
chrisl | henrys: yeh, in this case I wanted higher res than 75dpi, but quicker than 600dpi - 600 is fine, just takes up more cluster time. | 14:39.36 |
Robin_Watts | OK, what is wrong with http://ghostscript.com/~robin/out.pam ? | 14:42.40 |
| Morning mvrhel_laptop | 14:44.43 |
| I have the ETS test code updated to read and write pams so we can try cmyk | 14:45.07 |
| BUT... the output pam is being spat out by convert and I can't see why. | 14:45.23 |
chrisl | Is CMYK a standard tupltype? | 14:46.12 |
Robin_Watts | I have another P7 one with an identical header, and convert takes that fine. | 14:46.42 |
chrisl | OKay, the docs don't list CMYK as standard. Strange. Does convert give an error? | 14:47.37 |
Robin_Watts | gs -sDEVICE=pamcmyk32 -o out.pam gs/examples/tiger.eps | 14:47.43 |
| That will give you an out.pam that convert likes. | 14:47.50 |
| $ convert out.pam out.png | 14:48.13 |
| Magick: unable to read image data `out.pam' @ error/pnm.c/ReadPNMImage/1166. | 14:48.15 |
| Magick: missing an image filename `out.png' @ error/convert.c/ConvertImageCommand/3016. | 14:48.17 |
mvrhel_laptop | Robin_Watts: good deal. I will update later today | 14:49.02 |
| heading out the door now | 14:49.10 |
Robin_Watts | mvrhel_laptop: Have fun. | 14:49.19 |
henrys | my morning gl/2 conversation... I can just see his requests are going to be endless. | 15:11.20 |
kens | I saw the email :-( | 15:11.40 |
henrys | I should be positive at least the emails aren't bug trying to do new stuff. | 15:15.54 |
| s/bug/bug fixing | 15:16.02 |
Robin_Watts | ETS' implementation of FS dithering is very strange. | 15:25.04 |
| Either I'm reading it wrong, or it's formulated (at best) badly, or (at worst) wrong. | 15:25.49 |
henrys | what and the patent clerk didn't catch it :-^ ? | 15:45.41 |
Robin_Watts | I think it's the implementation, not the patent. | 15:47.53 |
| The central part of FS is that for each pixel you calculate the error between the desired pixel value, and the actual pixel value, and you propagate that e,se,s, and w. | 15:48.55 |
| with varying weights. | 15:49.06 |
| Firstly, I think he has the weights wrong. | 15:49.21 |
| Secondly, when you propagate the error, you have to be careful how you subdivide the error. | 15:49.45 |
| In theory 7/16ths go e, 1/16 goes se, 5/16 go s, and 3/16 goes w. | 15:50.24 |
| But if you take a value 'x' and work the errors out as 7x/16, 1x/16, 5x/16, 3x/16 the odds are that adding those up won't come back to x due to rounding. | 15:51.09 |
henrys | I have to read it but this sounds exactly like error diffusion. | 15:51.23 |
Robin_Watts | Right, I'm describing exactly FS error diffusion. | 15:51.46 |
henrys | oh okay | 15:52.03 |
Robin_Watts | So you calculate 3 of the weights, and get the 4th one by x - (sum of those 3). | 15:52.14 |
| That way you can be sure the overall error propagated is exactly correct. | 15:52.30 |
| In ETS, he doesn't work like that. | 15:52.50 |
| He stores a single error value for each pixel, and sums it with weightings later, hence losing the insurance against error accumulation. | 15:53.45 |
paulgardiner | That could work provided you take care how you work out the value on the last occasion that you use a past-pixel's error. | 15:55.22 |
Robin_Watts | The fact that he has the weights wrong too makes me doubt his "we tried alternating left -> right, right -> left passes and found that it gave no benefit" statement. | 15:55.38 |
| paulgardiner: No such care is taken, afaict. | 15:56.09 |
chrisl | Are the stored values fixed point/scaled, possibly making the rounding problems small enough to ignore? | 15:57.32 |
Robin_Watts | The stored values are fixed point, yes. | 15:57.53 |
henrys | then it is possible it is not visually significant in practice. | 15:58.21 |
Robin_Watts | It is indeed. | 15:58.56 |
henrys | I assume your concern is theoretical and you haven't traced artifacts back to this problem. | 15:59.13 |
Robin_Watts | henrys: Yes. | 15:59.31 |
| I'm still trying to understand the code. | 15:59.40 |
| For some reason he run length encodes lines before processing them. | 15:59.56 |
| And part of the processing involves unrunlength encoding the line. | 16:00.10 |
chrisl | Yeh, I thought that looked a pointless exercise | 16:00.29 |
Robin_Watts | plus he has a wierdass thing for 'counting white' that I really don't follow. Possibly it's to do with the multiplane optimisation stuff, but I just haven't understood it yet. | 16:01.17 |
| So I'm still looking at the code trying to follow it step by step - and the weights being wrong is my latest confusion. | 16:01.42 |
chrisl | Is there a way to get nmake to spit out the targets it's following? | 16:02.45 |
Robin_Watts | -d, I believe. | 16:02.56 |
| but be warned it's VERY verbose. | 16:03.02 |
chrisl | Cool, thanks, that's what I want | 16:03.29 |
henrys | I think halloween should be an artifex holiday. | 16:06.29 |
paulgardiner | tor8, Robin_Watts: a few more commits on paulg/master when you have some time. | 16:07.24 |
henrys | it's today in the state. | 16:07.24 |
| states | 16:07.33 |
| tor8 are you about? | 16:07.46 |
tor8 | henrys: yes? | 16:11.52 |
henrys | I missed you at the meeting yesterday and wanted to ask obligatory viewer questions. How's that going? | 16:12.27 |
tor8 | henrys: progressing. I'm about to tackle text selection and copying | 16:13.09 |
| Robin_Watts: I have a patch on tor/viewer (the text search one) that could do with some input; I'll want to rebase it for master | 16:13.57 |
| Robin_Watts: basically it's moving the common text search core into fz_text_page | 16:14.22 |
| Robin_Watts: when you have time, it's no rush | 16:14.39 |
Robin_Watts | henrys: halloween here too. Just put the pumpkin out. | 16:15.40 |
tor8 | henrys: I split the viewer logic into common core and gui-specific parts with a fairly simple interface between the two | 16:15.42 |
Robin_Watts | Best thing is, we have building work going on, so there are large holes in the front lawn. How many kids can we trap? | 16:16.08 |
kens | You participating in trick-or-treat Robin ? | 16:16.09 |
tor8 | henrys: once I've got the text selection and copying implemented I'm planning on digging into win32 :( | 16:16.28 |
Robin_Watts | kens: We hand out sweets, yeah. | 16:16.29 |
kens | Yes, the pumpkin is a signal round here | 16:16.39 |
Robin_Watts | Helen teaches many of the local kids piano, so... | 16:16.48 |
kens | ah | 16:16.56 |
henrys | tor8:great news - well for us, maybe not you. | 16:17.48 |
paulgardiner | Here, we give out sweets to the front of the house and accept thrown eggs to the back. I assume it's different kids. | 16:19.40 |
Robin_Watts | http://ghostscript.com/~robin/out.png or http://ghostscript.com/~robin/out2.png ? | 16:40.42 |
chrisl | Hmm, not to keen on either | 16:42.42 |
kens | IMO out.png is slightly better | 16:43.31 |
| marginally less 'leakage' into the white areas | 16:43.51 |
chrisl | There seems to be a lot of ~45 degree "chains" of pixels in both | 16:43.59 |
kens | Yes indeed | 16:44.06 |
Robin_Watts | chrisl: Yes. Raph doesn't do the serpentine stuff that would potentially stop that. | 16:44.22 |
chrisl | I don't remember getting it to that extent with just plain FS - I could be mistaken, it's been a while | 16:45.26 |
Robin_Watts | out.png is with the weights corrected. | 16:45.54 |
chrisl | Yep, I think out.png looks better, too. There are some disatracting colour artifacts in out2.png that would bother me | 16:47.52 |
kens | TBH its marginal, but for me out.png is better | 16:48.32 |
Robin_Watts | Yes, I don't think it's a massive difference, but then ETS doesn't give a *massive* difference over vanilla FS anyway. | 16:49.13 |
| I'm trying to think if there is a cheap and cheesy way I can test boustrophedonic operation... | 16:50.16 |
kens | thninks that's an awfully big word | 16:50.53 |
Robin_Watts | It's one of those words that you only ever see in very limited contexts :) | 16:51.20 |
| mvrhel_laptop: Hey. | 16:51.36 |
mvrhel_laptop | Robin_Watts: hello | 16:51.53 |
Robin_Watts | mvrhel_laptop: Just pushed a new version with fixed weights. I'm going to separate that out into its own commit though, so expect a new push shortly. | 16:55.36 |
mvrhel_laptop | Robin_Watts: ok great | 16:55.47 |
| Robin_Watts: so are you seeing the issue with the black channel that Max mentioned? | 16:56.11 |
Robin_Watts | mvrhel_laptop: I'm still not that far through. | 16:56.34 |
mvrhel_laptop | ok. I see your comment about serpentine. I was surprised to read in Raph's paper that he did not do serpentine | 16:57.11 |
| I do realize he has the backward distance working in there | 16:57.25 |
| which should help | 16:57.28 |
| but still I would think serpentine would be good. | 16:57.40 |
| also, the random noise thing is a bit of a concern | 16:57.49 |
Robin_Watts | I'm hoping that he wasn't seeing a benefit from it, because the weights were wrong. | 16:57.53 |
| random noise? | 16:58.06 |
| random noise is an integral part of the algorithm. | 16:58.15 |
mvrhel_laptop | in tests that I have done with noise in the past I have found it to not be helpful | 16:58.23 |
| Robin_Watts: it is described in the paper | 16:58.32 |
| and I see it in the code | 16:58.38 |
| he has a lut that allows a bit of perturbation on the thresholds | 16:59.10 |
Robin_Watts | Indeed, yes. That's what I meant by "integral part of the algorithm". | 16:59.31 |
mvrhel_laptop | in an attempt to break up patterns at 1/3, 1/2 and 2/3 | 16:59.36 |
| oh I read a NOT in front of your sentence | 16:59.51 |
| sorry | 16:59.53 |
| so you agree with my questioning this | 17:00.12 |
Robin_Watts | Sorry, I thought you were meaning that Max was seeing some "random noise" in the output or something. | 17:00.14 |
mvrhel_laptop | no | 17:00.17 |
| I am suspicious of the use of noise in the algorithm | 17:00.34 |
Robin_Watts | I understand why he's added it. I can't (yet) claim to have any strong feelings about it. | 17:00.41 |
| Actually, that's not true. I *hate* his implementation of it :) | 17:00.53 |
| Magic tables, with magic shifts, all exposed in the api. urgh... | 17:01.13 |
| I suspect that there is a process that should be followed when adapting this code for a new printer. | 17:01.48 |
mvrhel_laptop | no noise to start with... | 17:01.58 |
Robin_Watts | 1) Get the code working without any random noise, no multiplane optimisations etc. Adjust the luts etc so you get the best possible results. | 17:02.21 |
| 2) then introduce the strengths so multiplane optimisation works. | 17:02.41 |
mvrhel_laptop | yes. I agree with this | 17:02.50 |
| I wish there was some way for us to easily break up the work between us | 17:03.05 |
Robin_Watts | 3) then when you have got that working as well as possible, try introducing some random noise. | 17:03.10 |
| mvrhel_laptop: Well, at the moment, you're busy with this conference, right? | 17:03.27 |
| I suspect I'll be burned out on this by the time you come back from that and more than happy to pass it over :) | 17:03.48 |
mvrhel_laptop | ok. sounds good. I just feel a bit useless right now. I will be burned out from all this MS stuff by the end of the week | 17:04.14 |
Robin_Watts | Is this a WIndows 8 conference, or a Windows Phone 8 conference? or both ? | 17:04.44 |
mvrhel_laptop | all of the above | 17:04.52 |
| they are pushing app development for the phone, tablet and PC (non-desktop apps) | 17:05.16 |
Robin_Watts | "desktop is dead". yeah, right. | 17:05.28 |
kens | Yet another 'conde once deply anywhere' | 17:05.31 |
mvrhel_laptop | the C++ Amp stuff is interesting | 17:06.36 |
| the compiler will push these code sections to the GPU | 17:06.59 |
| or if the GPU is not available it will end up doing SSE2 optimizations | 17:07.15 |
Robin_Watts | It's halloween so I'm going to assume your positive reference to C++ is influenced by that :) | 17:07.15 |
mvrhel_laptop | hehe | 17:07.20 |
| that is what they are pushing here | 17:07.31 |
| at least all ISO C++ is OK | 17:07.43 |
| any libraries (like mupdf) would need to be wrapped up | 17:08.04 |
| I am thinking that I should, when I get done with this, create a simple viewer app for the windows phone and surface | 17:08.35 |
| I don't think it would be too difficult | 17:08.51 |
| I am going to try to talk to some of the phone vendors here today | 17:09.10 |
Robin_Watts | mvrhel_laptop: Can you hear that? | 17:09.22 |
mvrhel_laptop | I had heard rumors that the windows phone 8 has native pdf support but not sure | 17:09.34 |
| what Robin_Watts :) | 17:09.39 |
Robin_Watts | I think I can hear tor8 screaming for joy from here :) | 17:09.41 |
mvrhel_laptop | hehe | 17:09.45 |
| it might make sense for me to do it | 17:09.51 |
Robin_Watts | actually, I fear you doing a viewer for TIFKAM will not save tor8 from having to do a Win32 one. | 17:10.20 |
kens | nope can'g ignore old versions | 17:10.32 |
mvrhel_laptop | well there were some desktop win32 questions from the audience at some of the sessions, and a lot of the speakers did not know the answers | 17:11.22 |
| it is like the desktop stuff is going to bit rot..... | 17:11.39 |
kens | So MS hopes.... | 17:11.49 |
mvrhel_laptop | I do hate that with windows 8 there is no longer an xp mode | 17:11.50 |
| like there is with windows 7 | 17:12.03 |
Robin_Watts | XP mode being a compatibility mode ? | 17:12.18 |
mvrhel_laptop | no it is like a virtual machine | 17:12.31 |
| you can install your old apps | 17:12.39 |
| it is quite nice | 17:12.45 |
kens | not in all versions of 7 | 17:12.49 |
Robin_Watts | hmm. Never seen/needed that with my windows 7. | 17:13.02 |
mvrhel_laptop | oh all your applications from xp days run in windows 7? | 17:13.19 |
Robin_Watts | Yes. | 17:13.27 |
mvrhel_laptop | lucky you | 17:13.31 |
Robin_Watts | What have you found that doesn't ? | 17:13.44 |
mvrhel_laptop | I had a drawing package (vectorwoks) and an old version of MATLAB that did not work | 17:13.54 |
| the version of MATLAB included all the toolboxes | 17:14.09 |
Robin_Watts | How old was the version of matlab ? | 17:14.17 |
Robin_Watts | is a Xara boy. | 17:14.38 |
mvrhel_laptop | It is from 2001 | 17:14.41 |
| I have a newer version that is fine on windows 7 but without all the toolboxes | 17:14.53 |
| just the optimization toolbox and image processing toolbox | 17:15.03 |
| anyway, with windows 8 I think there are going to be a lot of stuff that is no longer going to work | 17:16.24 |
| I would like to get VS 2012. going to a talk today about the new features | 17:16.55 |
Robin_Watts | mvrhel_laptop: brace yourself for how much VS2012 ultimate edition will cost though :( | 17:17.23 |
mvrhel_laptop | yes | 17:17.29 |
Robin_Watts | Cos the ultimate edition is the one with the reverse debugging in. | 17:17.41 |
mvrhel_laptop | I can probably get an employee cost here | 17:17.43 |
| I will check on that today | 17:18.02 |
henrys | are they talking about cloud and SaaS stuff or is this primarily mobile devices and such? | 17:18.03 |
mvrhel_laptop | Oh there is a lot of talk about integration of the application with the cloud | 17:18.24 |
| and how to do it | 17:18.29 |
| of course this stuff is coming fast and furious and all in c++ with classes and operators and wacky notation that it is hard to keep up | 17:19.12 |
| at least I have not heard the word c# or dotnet from any of the talks | 17:20.05 |
kens | No, because those were last decades special sauce.... | 17:20.30 |
mvrhel_laptop | yes | 17:20.35 |
henrys | IMH analysis the cloud is the problem MS needs to deal with. Competing cloud apps to word and excel could make MS a much smaller company. | 17:20.39 |
mvrhel_laptop | today it is all XAML for UI, DirectX for fast graphics, and C++ | 17:20.56 |
| henrys: they want to make a cloud version of those that requires a yearly subscription | 17:21.30 |
| which to me seems unlikely to succeed | 17:21.39 |
henrys | so they're sticking with DirectX on the surface and phones not opengl es | 17:22.06 |
| figures | 17:22.20 |
mvrhel_laptop | DirectX on the surface and phones | 17:22.32 |
| Xbox music sound cool. Ballmer said it was free in his talk yesterday. Turns out that it is free for the first 30 days. Then $99/year | 17:23.13 |
kens | Free, for sufficiently small values of... | 17:23.32 |
henrys | well it will really be interesting to get the full report from you, I think it is important that you went (Artifex-wise) | 17:23.39 |
mvrhel_laptop | maybe that is eseentially free to him | 17:23.43 |
henrys | no chair throwing ? | 17:24.05 |
Robin_Watts | mvrhel_laptop: Have you heard "managed code"? That's their current euphemism for .net I think. | 17:24.19 |
mvrhel_laptop | henrys: not yet. They did make a poke about having maps that work in the windows phone | 17:24.22 |
kens | I hear Uncle Fester's beentold not to do that any more | 17:24.27 |
mvrhel_laptop | No mentiond of managed code | 17:24.31 |
| ISO C++ and a bit of what they call C++ CX for WinRT interfacing | 17:24.56 |
Robin_Watts | The last version of windows phone only supported managed code. | 17:25.08 |
mvrhel_laptop | plus the C++ AMP stuff for getting big speedups with GPUs and SSE | 17:25.12 |
| Robin_Watts: not now. | 17:25.35 |
| Robin_Watts: supposedly you can reuse a lot of code between the devices | 17:26.04 |
Robin_Watts | mvrhel_laptop: Yeah, I think the last release of windows phone was such an utter failure, that the message actually got through on that one. | 17:26.05 |
mvrhel_laptop | yes | 17:26.11 |
Robin_Watts | mvrhel_laptop: Yeah, the new API is designed to be similar everywhere, AFAICT. | 17:26.33 |
mvrhel_laptop | I need to find out about the native pdf support in the phone (or the lack there of) | 17:26.36 |
Robin_Watts | mvrhel_laptop: If they don't have it, would they like to have it? :) | 17:27.14 |
mvrhel_laptop | well exactly | 17:27.29 |
| that is what miles wanted me to find out | 17:27.35 |
| the trick is finding the right person to talk to | 17:27.49 |
| these guys only know there little bit of the puzzel | 17:28.00 |
| puzzle | 17:28.03 |
Robin_Watts | yeah. | 17:28.03 |
mvrhel_laptop | ok. they just opened up the phone vendor area. I am going to head over there | 17:28.35 |
Robin_Watts | Remember to laugh at Nokia. They love that. | 17:28.57 |
henrys | just ask them to bring up a pdf in the phone's browser. | 17:29.11 |
kens | OK I'm off, night all | 17:29.13 |
mvrhel_laptop | oh good idea henrys | 17:29.19 |
| ok bbiaw | 17:29.30 |
henrys | good luck | 17:29.35 |
mvrhel_laptop | thanks | 17:29.39 |
henrys | well I'm in this race: http://www.challenge-penticton.com/ - if only I could pronounce it I'd tell folks. | 17:30.38 |
tor8 | mvrhel_laptop: henrys: https://code.google.com/p/angleproject/ might save us there (it's a GLES to DirectX translation layer) if we ever do intend to attemtp h/w acceleration on tablets | 17:31.18 |
henrys | cool and BSD | 17:31.58 |
Robin_Watts | henrys: challenge pentagram, eh? Very halloween themed. | 17:34.34 |
henrys | maybe I'll just call it that and people will know what I mean. | 17:37.50 |
| I though mvrhel_laptop might know, it is up in his neck of the woods. | 17:38.30 |
| surprising alexcher had power through this ordeal from what I've been reading, lots of outages. | 17:39.21 |
Robin_Watts | So... the purpose of the RLL stuff is to allow the code to cope with different src_width and dst_widths. | 17:39.30 |
chrisl | Huh? Why? | 17:39.58 |
Robin_Watts | But surely it'd be better to interpolate rather than to do that. | 17:40.13 |
| chrisl: beats me. | 17:40.19 |
chrisl | Surely if you're just screening, you want in and out raster dimensions to match | 17:40.58 |
Robin_Watts | I mean, it doesn't solve the problem of wanting to change in the Y direction. | 17:41.12 |
| chrisl: You'd think. | 17:41.21 |
| henrys: Do we have a record of which customers are using which features? | 17:41.44 |
henrys | I'm pretty sure we don't but you could check with rayjj | 17:42.09 |
chrisl | Robin_Watts: How much hassle would it be to add the ETS "core" to the tiffscaled devices? | 17:42.22 |
Robin_Watts | chrisl: Probably not stupidly hard. | 17:43.22 |
| but I'd probably want to kick this code a bit more before I did that. | 17:43.35 |
chrisl | If it were optional, if would let people run "real" jobs, and have a direct, back to back comparison with "plain" FS | 17:44.25 |
Robin_Watts | chrisl: yeah, this has occurred to me. | 17:44.42 |
| at the moment, I'm running real jobs to pamcmyk32 and then feeding those through ets. | 17:45.04 |
henrys | I do recall getting positive feedback and I'm quite certain that wouldn't have happened with the png results you had so something is awry | 17:45.41 |
chrisl | There would be other benefits - like TIFF output, and easy to get both composite and separated output. | 17:45.48 |
Robin_Watts | chrisl: tiff output is NOT a benefit. | 17:46.31 |
| and convert will go from .pam to anything else we need. | 17:46.46 |
| The thing that stops us being able to do a direct comparison is that we don't have a tiffscaled device that produces cmyk. | 17:47.28 |
chrisl | I thought tiffsep1 could use the tiffscaled stuff | 17:48.04 |
Robin_Watts | oh, hmm... maybe. | 17:48.29 |
| It can use the downscaler certainly. | 17:48.40 |
chrisl | I was thinking that folks evaluating would probably want composite to start with, and then separations when they start throwing stuff at a real printer, and we really only have the tiff devices for separations. | 17:49.54 |
Robin_Watts | Well, they can run jobs to pamcmyk32 then feed that to ETS to get pam's out at required bitdepths. | 17:52.24 |
| Those can then either be 'convert'ed to separations or to composite. | 17:52.41 |
| So there is a route there already, but I see your point that it would be nice to have it all in one go. | 17:53.17 |
chrisl | That wouldn't work for spot colours | 17:53.46 |
Robin_Watts | ok, tiffsep1 is not honouring DownScaleFactor. | 17:55.24 |
chrisl | Ah well, just a thought | 17:55.41 |
Robin_Watts | But tiffsep does. | 17:55.54 |
| But that produces 8bpp output. | 17:56.30 |
| 8bpc even. | 17:56.36 |
chrisl | Oh boy, Disney in charge of Star Wars Ep.7 - that can't be good..... :-( | 17:57.13 |
Robin_Watts | chrisl: It's got to be MUCH better than Lucas being in charge. | 17:58.00 |
chrisl | I dunno, I can see it being even more annoying than Ep.1 | 17:58.38 |
Robin_Watts | http://ghostscript.com/~robin/out.png or http://ghostscript.com/~robin/out2.png or http://ghostscript.com/~robin/out3.png | 18:10.15 |
| oh, wait, ignore that. | 18:10.51 |
tor8 | chrisl: Episode VII: The Galaxy Explodes. Directed by Michael Bay. | 18:11.14 |
chrisl | tor8: even that would be better than "Episode VII: Mickey Mouse Goes Jedi", or something...... | 18:14.15 |
henrys | out3.png gets close to rosettes but I have to think there is a math problem or this has all been some sort of sham | 18:14.35 |
| well closer than the others | 18:15.08 |
Robin_Watts | OK, better out3.png there now. | 18:15.25 |
henrys | now it has vertical lines that can't be right worse than the other out3.png | 18:16.19 |
chrisl | Hmm, don't like the vertical lines - I think I prefer out2.png, the diagonal (rather than vertical lines) seem less pronounced | 18:17.35 |
Robin_Watts | The difference between out2 and out3 is that out3 has the serpentine stuff in. | 18:18.09 |
| (I'm doing it by flipping lines/errors/distance metrics etc) | 18:18.28 |
chrisl | I prefer the flat gray backdrop in out3, but not the vertical lines in the tiger's nose area | 18:19.48 |
henrys | why would there be any lines - diagonal or vertical, seems like a bug. | 18:19.58 |
| ? | 18:20.00 |
| I wonder if there is a math difference raph undoubtedly did this work 32 bit and you are probably using a 64 bit machine. | 18:21.10 |
Robin_Watts | henrys: 32bit ints though, this is windows. | 18:21.29 |
| there being lines is not unexpected. | 18:21.50 |
| They aren't so much lines as 'holes in the dithering'. | 18:22.20 |
| The hole claims it's first small child. Muhahah! | 18:25.23 |
| s/it's/its/ | 18:25.42 |
henrys | jesus robin_watts | 18:26.10 |
Robin_Watts | (no injury to small child, and I had specifically warned him). | 18:27.41 |
| or is it my inability to use apostrophes correctly? | 18:28.02 |
henrys | when they fall in do you dirt to shovel in on top of them. | 18:28.10 |
| ? | 18:28.17 |
Robin_Watts | I'd have to put my shoes on first, and he scrambled out too fast. | 18:28.46 |
henrys | they are nimble | 18:29.26 |
| we could also check the code on linux for possible differences, that would have been where the code was developed. | 18:31.22 |
Robin_Watts | henrys: Does linux have signed or unsigned chars? | 18:31.48 |
| signed if memory serves, same as windows. | 18:32.00 |
henrys | yes signed | 18:32.11 |
| I'm wondering if some math call is different but he probably doesn't use -lm much. | 18:32.45 |
Robin_Watts | I am bothered by the yellow band at the top of the tigers head before the orange. | 18:33.53 |
| Updated out3.png, plus new out4.png and out5.png | 18:53.49 |
| out3.png has the multilevel opts turned off. | 18:53.59 |
| out4.png has that enabled again. | 18:54.09 |
| out5 has the random noise turned off. | 18:54.16 |
| I much prefer out5, but that may be a sign that the test code has the noise turned up too high. | 18:56.28 |
| and it may be a screen vs paper issue. | 18:56.39 |
henrys | power back on. | 19:03.45 |
Robin_Watts | henrys: No maths calls in the core of the code. | 19:05.47 |
| (see my previous wibblings for more images) | 19:06.10 |
henrys | since my dog bit the meter man they asked me to get a meter that transmits wirelessly, I thought I should do my part to unemploy the meter man so I complied, but they didn't tell me they'd shut off power installing it. No big deal, don't warn me, who need those lines of code. | 19:06.17 |
Robin_Watts | I think the results can vary wildly according to the params you pick. | 19:06.49 |
| henrys: Hmm. Did you have to pay for that? | 19:07.09 |
| and how wirelessly does it transmit? Live via the internet? or does it send readings back up the mains using similar techiques to power line networking ? | 19:08.07 |
henrys | it's a radio transmittor actually. | 19:09.10 |
| yeah 100.00 | 19:09.55 |
Robin_Watts | so the meter man still has to get within range of that, right? | 19:10.13 |
| And can anyone read your meter ? | 19:10.20 |
henrys | yes but you can rifle down the street in a car and get everyone's reading very quickly it significantly cuts into work hours. | 19:11.51 |
Robin_Watts | henrys: right. | 19:12.11 |
henrys | I wonder if it is encrypted | 19:14.09 |
| I'll look at the logs | 19:15.04 |
| tor8 don't you remember raph showing us ets stuff that looked much different/better than this? | 19:17.26 |
| for the logs | 19:17.42 |
| bbiaw | 19:53.53 |
Robin_Watts | mvrhel_laptop: Hey | 20:04.03 |
mvrhel_laptop | hi Robin_Watts | 20:04.11 |
Robin_Watts | dunno if you can read the logs from where you are, but there are various bitmaps online for people to pass judgement on. | 20:04.26 |
mvrhel_laptop | oh ok let me look | 20:04.34 |
| none of them look particularly sensational. We should also include a standard FS with serpentine for comparison | 20:08.11 |
| then we know we are doing better than that | 20:12.13 |
| off to see if I can find a phone developer from MS | 20:13.34 |
| Forward 1 day (to 2012/11/01)>>> | |