| <<<Back 1 day (to 2012/10/14) | 2012/10/15 |
ray_laptop | well, after a bit of digging, I determined that painting patterns using the clist when the pattern has an interpolated image tends to be EXPENSIVE! (not much surprise when you think about it) | 08:20.25 |
| in this case the image that makes up the pattern filled (most of) the page and was scaled up by 8:1, so it was playing back the clist for the image multiple times and interpolating and converting colors and painting pixels MANY times | 08:22.04 |
chrisl | ray_laptop: from your e-mail, though, it sounds like you have a change that's potentially a big win for that? | 08:22.49 |
ray_laptop | chrisl: yes -- one is to increase the MaxPatternBitmap (for their older code base, it is a compile time value). The other is to use the 'default' image handling in the pattern clist when the image is interpolated. | 08:25.48 |
| avoiding the clist based patterns is the biggest win | 08:26.18 |
chrisl | Fingers crossed the customer finds one of those acceptable...... | 08:26.44 |
ray_laptop | chrisl: I TOTALLY agree ! | 08:27.01 |
| chrisl: and the other fingers I have crossed is to hope that their PPC processor sees similar perfromance improvements. | 08:28.13 |
| the target CPU has disappointed me in the past. | 08:29.22 |
Robin_Watts | Late night ray. | 08:29.56 |
chrisl | ray_laptop: well, the best optimisation for slow code is to avoid executing it (often), so..... | 08:29.59 |
Robin_Watts | An alternative idea would be to cached the interpolated image if we can spot that it's used in a pattern? | 08:30.43 |
ray_laptop | Robin_Watts: yeah. I've been moving house, and tonight my daugher had a math homework assignment that it turns out wasn't done earlier in the weekend | 08:31.10 |
Robin_Watts | ray_laptop: Ah. One of those weekends. | 08:31.34 |
kens | wonders why ray_laptop is not in bed... | 08:32.10 |
Robin_Watts | I've just had the people replacing some windows phone up and say "can we come in 20 minutes rather than next week?" | 08:32.13 |
ray_laptop | Robin_Watts: well, using the 'default' (copy color) output from the image processing when interpolating wn image into an upscaled image in a pattern does what we want (I theorized) | 08:33.31 |
| but avoiding using clist pattern playback is OBVIOUSLY the best solution | 08:34.44 |
| and kens is spot on -- time for bed ... | 08:35.05 |
| g'nite | 08:36.41 |
chrisl | g'nite Ray | 08:37.30 |
kens | chrisl ping | 09:49.07 |
| Looking for some help with Linux. Installed a 64-bit Linux on VMWare. TOld me I should download vmware tools, which I did. mounted the iso, extracted the .tgz contents to my home folder. When I try to run vmware-install.pl I get a permission denied error, even when I su to root. ASN | 09:50.36 |
| Any ideas wghat I'm missing ? | 09:50.45 |
chrisl | kens: it may not be executable? | 09:58.31 |
| chmod +x vmware-install.pl | 09:59.00 |
kens | I think that may be it. | 09:59.00 |
| Well its different :-) | 10:00.07 |
| Now I get a 'bad interpreter: No such file or directory' from /usr/bin/perl | 10:00.31 |
| I'm guessing I don't have a perl interpreter installed... | 10:00.42 |
chrisl | Could be, seems odd though, because so much depends on it. "which perl" will tell you if it's installed elsewhere | 10:01.17 |
Robin_Watts | kens: What linux distro ? | 10:01.28 |
kens | Fedora 14 | 10:01.43 |
| And no, there's no perl | 10:01.51 |
chrisl | Well, should be easy enough to install....... | 10:02.15 |
kens | Yeah, if I can remember the magic incantation ofr the package manager | 10:03.03 |
chrisl | There should be a nice gui for the package manager | 10:03.25 |
| or (guessing) "yum install perl" | 10:03.44 |
kens | Ah that could be, I'm currently looking at a very confusing GUI thing | 10:04.15 |
| yum is better :-) | 10:04.36 |
chrisl | Rememer, you need to run the command as root | 10:04.41 |
| s/Rememer/Remember | 10:04.54 |
kens | Not a problem, still root from a minute or two back | 10:05.44 |
kens | wonders what else will need installed.... | 10:05.59 |
| Well, perl works now :-) | 10:06.38 |
chrisl | You'll probably want git(-gui), x11 dev libs, maybe gtk dev libs....... | 10:07.27 |
kens | Yes, probably. If I can get to the point of running a test I'll be happy | 10:12.12 |
| Hmm, still getting a permission denied.... | 10:12.45 |
chrisl | From? | 10:12.58 |
kens | As I was about to say, it looks like various scripts are being created without the executable bit | 10:14.53 |
chrisl | Hmmm, I haven't used VMWare for years, so I can't remember the details..... | 10:15.48 |
kens | If I go around adding the bit manually, it gets further and further. | 10:16.39 |
| But its dead tedious | 10:16.45 |
chrisl | Ahem, "find ./ | xargs -I {} chmod +x {}" | 10:17.30 |
| (not recommended in general, but.....) | 10:18.02 |
kens | Well I've got as far as gcc not being installed.... | 10:19.50 |
sebras | kens: could it be that your umask is not set appropriately so that you don't get the executable bit? | 10:56.16 |
kens | sebras, no idea what umask is | 10:58.11 |
| presumably user mask | 10:58.26 |
Robin_Watts | I wonder if for some security reason tar unpacked without x bits. | 10:58.27 |
kens | However, I also can't execute the gs binary I just downloaded | 10:58.43 |
Robin_Watts | umask is the default permissions given to new files. | 10:58.56 |
| kens: But that's to be expected. | 10:59.02 |
kens | Why ? | 10:59.15 |
Robin_Watts | If you download a file from the net it should NOT have it's x bit set. | 10:59.19 |
kens | It does now, I chmod it | 10:59.26 |
Robin_Watts | Right, that's the correct thing to do. | 10:59.34 |
| Thus you have made a determination that that particular file is safe to execute. | 10:59.53 |
kens | Yes indeed, still doens't work though | 11:00.10 |
Robin_Watts | Ah, why not? | 11:00.17 |
kens | one moment and I'll quote it | 11:00.27 |
Robin_Watts | different runtime libs? | 11:00.29 |
kens | 'cannot execute binary file' | 11:01.01 |
Robin_Watts | file gs | 11:01.16 |
kens | THat's gs-906-linux_x86_64 | 11:01.21 |
| ELD 64-bit LSB executabel, x86-64, version 1 (SYSV), dynamically linked (uses shared libs) | 11:02.02 |
| Bet I don't have the requisite shared libs | 11:02.13 |
| s/ELD/ELF/ | 11:02.22 |
Robin_Watts | or your system isn't set up for version 1 (SYSV) binaries. | 11:02.36 |
kens | I guess htat's possible, don't know how to tell though | 11:03.01 |
Robin_Watts | Pick an exe that works, and file <that> | 11:03.19 |
kens | No gcc runs and it is also a version 1(SYSV) binary | 11:03.26 |
| And also uses shared libs (though possibly not the same ones) | 11:03.55 |
Robin_Watts | OK, so it could be the shared libs, yes. | 11:04.01 |
| Dumb idea... | 11:04.10 |
| why not use the package manager to get ghostscript? | 11:04.24 |
kens | Robin_Watts : I'm trying to reproduce what a free user did | 11:04.36 |
Robin_Watts | That should hopefully pull in all the prerequesite tools etc. | 11:04.49 |
kens | I guess I could get the package, and then get the executable and run it locally | 11:04.58 |
Robin_Watts | kens: Right. Then you can do your own build. | 11:05.11 |
kens | gs works, ./gs-906-linux_x86_64 doesn't | 11:07.12 |
| same error | 11:07.19 |
Robin_Watts | gs-906-linux.... being one that we ship? | 11:07.43 |
kens | Yes, its from our downloads site. | 11:08.03 |
Robin_Watts | I'm not entirely surprised by there being problems. | 11:08.06 |
kens | Oh the package is 8.71.... | 11:08.09 |
| That's probably not new enough | 11:08.18 |
Robin_Watts | Shipping systems from one unix box and expecting them to just work on another is always slightly hairy. | 11:08.48 |
kens | Well yes, Like I said I was trying to reproduce hte users situation | 11:09.08 |
| OK well I guess I will leave it until I get this VM working properly | 11:09.32 |
chrisl | kens: you're almost certainly either missing shared libs, or have ones that are not ABI compatible with the pre-built binary - hence why I strongly objected to supplying them (but was overruled) | 11:27.49 |
kens | chrisl I'm sure you are right. So I'll leave the problem until I can get my installation working properly | 11:43.49 |
kens | goes to download the source and build that | 11:45.49 |
| well at least configure seems to work :-) | 11:49.14 |
| Well, that is interesting. If I run a PDF->PDF on WIndows, then the Author and so forth is carried thourgh. If I run it on (64-bit) LInux, it isn't. | 11:58.41 |
| Of course, that's not the current source, but still | 11:58.58 |
| chrisl would you check that for me please ?TO make sure I'm not being totally stupid | 12:00.30 |
| Input file is : http://freeshell.de/~mpfusion/input.pdf | 12:00.44 |
| just run ./gs -sDEVICE=pdfwrite -o <output> input.pdf | 12:01.03 |
| For me, on Windows, that produces a file with teh /Author in it, on Linux it doesn't | 12:01.30 |
chrisl | kens: on the current code? Or a release? | 12:07.38 |
kens | I treied 9.06 release, but either would be good | 12:12.54 |
chrisl | 9.06 loses author etc, but the current code seems to maintain it | 12:13.52 |
kens | ah, then it was a transient bug thanks | 12:18.57 |
chrisl | Do you want me to bisect to see where it was fixed? | 12:19.20 |
kens | Nope, I'll tell user to build from current source, thanks | 12:27.05 |
henrys | happy monday | 13:39.30 |
Robin_Watts | hands henrys a tamborine and a spliff and names him Bez. | 13:41.00 |
henrys | goes to google | 13:41.19 |
| so it's interesting you were working on page fade, do you know about our new project request? | 13:42.45 |
Robin_Watts | henrys: Yes, that's exactly why. | 13:43.45 |
Robin_Watts | wonders if henrys has been copied in on the emails... | 13:44.08 |
| No. Let me forward you some emails. | 13:44.34 |
henrys | I have not but I just spoke to the customer about this. | 13:44.36 |
| at chicago | 13:44.43 |
| right okay that makes sense. | 13:46.20 |
Robin_Watts | so, same customer? | 13:47.48 |
henrys | yes | 13:48.07 |
| just miles and I aren't communicating properly. | 13:58.11 |
| I was studying it, there is a standard for signage and a few companies doing it. The standard group calls for svg but I thought that was near dead. | 13:59.13 |
| Ah reports out, it's official I have more customer bugs than everyone else combined. damn. | 14:07.35 |
| does ray_laptop ever sleep? | 14:31.15 |
Robin_Watts | Only during daylight. Muhahaha | 14:38.08 |
ray_laptop | With moving houses, I've been burning the candle at both ends (and then middle) | 15:45.02 |
| and my daughter tells me last night at 11:45 saying she needs help with her math h/w :-( Then the day starts over when my alarm goes off at 6:15 | 15:46.29 |
| I have to run over to the bank. be back in about 30 min. | 15:47.48 |
mvrhel | good morning | 16:13.57 |
Robin_Watts | Morning. | 16:14.04 |
ray_laptop | ok, I'm back. | 16:32.01 |
henrys | hi mvrhel | 16:37.49 |
mvrhel | hi henrys | 16:37.59 |
| cool. my simulated overprinting seems to work which means we can get simulated overprinting for our CMYK device | 17:16.22 |
| s | 17:16.24 |
| that is, for the spot colors | 17:17.00 |
| these devices already do overprinting when the colorants are CMYK | 17:17.16 |
| but spot color overprint simulation only worked for the separation devices | 17:17.36 |
| I think this is the last thing for customer 330 | 17:18.10 |
| oops. looks like I have an issue in the clist case though | 17:20.35 |
| ok and now the clist case is working | 17:52.04 |
| nice. Visual studio crash.... | 17:52.55 |
| need to check case with transparency.... | 18:02.59 |
ray_laptop | mvrhel: great -- so this does the overprint sim using the tint transform ? | 18:04.12 |
mvrhel | yes | 18:04.31 |
| ray_laptop: essentially, for these devices the separation and deviceN colors are converted to CMYK with the alternate tint transform (and the output ICC profile if needed) and then blended in the op device | 18:06.22 |
| it seems to match AR nicely | 18:06.28 |
| the icc profile comes into play if the alternate color space is ICC, lab, or RGB | 18:08.15 |
| and sure enough, with the pdf14 device in the way things dont work.... | 18:09.07 |
| a bit more work to do | 18:09.14 |
| hmm this maybe problematic | 18:10.11 |
marcosw_ | henrys: we may have an issue even if the customer who is having performance problems with 8.64 that they didn't have with 8.54 upgrades to 9.06 | 18:25.20 |
| gs9.06 not as fast as 8.54 (50 minutes for 9.05 vs 20 minutes for 8.54). | 18:25.52 |
| at least 9.06 doesn't have the memory issue that 8.54 does. | 18:26.27 |
henrys | knowing the customer there is little reason to think they'll upgrade to 9.06 but we should have a bug report. | 18:27.26 |
| is that just aix? | 18:28.28 |
| I sent you email about mountain lion - I've upgrade macpro so feel free to poke around. | 18:29.20 |
| I'm curious if anyone has given thought to ghostscript or mupdf going Affero GPL? We do have one SaaS customer. I think Miles and Scott should go after SaaS gpl users as if they were regular software companies abusing the gpl, even though the legal loophole exists. For now we can at least threaten them with Affero in future releases. | 18:36.59 |
Robin_Watts | googles. | 18:37.48 |
henrys | not an important thing just throwin' it out there might be meeting stuff in the future. | 18:38.41 |
STE | hello, Iam looking for advice on how to speedup ghostscript on EC2? | 18:45.44 |
| I have a micro instance and converting to JPEG takes almost 20 minutes or so | 18:46.19 |
| how can I speed it up? or should I be looking at something other than EC2? | 18:47.18 |
Robin_Watts | STE: Without us seeing a file/command line in question, we cannot make any statements with any accuracy. | 18:51.53 |
STE | Robin_Watts : ah! ok! well I'm running ImageMagick's convert command : "convert -density 300 file.pdf files.jpg" | 18:53.53 |
Robin_Watts | Do you really need 300dpi? | 18:54.30 |
| And we generally recommend people drive gs directly rather than going through imagemagick and reporting a bug, as it's quite possible that it's imagemagick causing the problem. | 18:55.51 |
| gs -o file%d.jpg -sDEVICE=jpeg -r300 file.pdf | 18:56.37 |
| That will take each page of file.pdf and output it as a file1.jpg, file2.jpg etc... | 18:56.57 |
mvrhel | bbiab | 18:57.04 |
STE | oh! | 18:57.07 |
| and is there a way to limit memory consumption? | 18:57.19 |
Robin_Watts | STE: There are various memory tuning options, yes. | 18:57.50 |
STE | the 300dpi was because the text was not looking to great at 72 and 150, I haven't tried more as of yet | 18:58.24 |
Robin_Watts | gs can either run in 'page mode', where it holds a whole image of the page in memory, or it can run in 'banded' mode where it holds a band at a time. | 18:58.41 |
| STE: OK. You can possibly improve that by doing: | 18:58.54 |
| gs -o file%d.jpg -sDEVICE=jpeg -r300 -dTextAlphaBits=4 file.pdf | 18:59.17 |
| BUT... if you are generating pages with text in them, then JPEG is a poor choice. | 18:59.34 |
STE | what should I use? then end result is to present images on a website of converted pdfs\ | 19:00.04 |
Robin_Watts | JPEG is very poor at representing areas of high frequency data, such as the edges of black/white text. You'll get horrid ringing effects. | 19:00.08 |
henrys | bbiab lunch | 19:00.11 |
Robin_Watts | PNG. | 19:00.11 |
STE | won't they be heavy? | 19:00.27 |
Robin_Watts | gs -o file%d.png -sDEVICE=png16m -r72 -dTextAlphaBits=4 file.pdf | 19:00.28 |
| quality or size. Your choice. | 19:00.55 |
| Personally, I don't believe pngs will be too big. | 19:01.04 |
STE | well then I will try both! | 19:01.44 |
Robin_Watts | You may also want to consider using MuPDF. | 19:02.36 |
STE | I hope it will take less time to convert, from some readings it would seem my micro instance is not made for document conversions | 19:02.46 |
| MuPDF? | 19:02.52 |
Robin_Watts | Whenever this comes up, it seems to me like MuPDF is a better bet. | 19:02.55 |
| An alternative PDF renderer, from the developers of ghostscript. | 19:03.09 |
| More screen orientated than gs. | 19:03.26 |
STE | ok, but not for headless use? | 19:04.41 |
Robin_Watts | mudraw is one component of mupdf, that is a command line rendering tool. | 19:05.12 |
| mudraw -o file%d.png -r72 file.pdf | 19:05.29 |
STE | thanks, so mupdf takes less memory and cpu ressources? | 19:07.00 |
Robin_Watts | Depends. | 19:08.14 |
| MuPDF is a set of core libraries, that we have wrapped into various tools. | 19:08.44 |
STE | ok | 19:09.01 |
Robin_Watts | mudraw drives mupdf in a "page view" mode, rather than a banded mode. | 19:09.19 |
| but you get AA for free. | 19:10.02 |
STE | banded vs page view means what? | 19:10.15 |
Robin_Watts | In page mode we hold an image the size of the page in memory to draw to. | 19:11.01 |
| In banded mode, you can pass over the data several times, and draw 'bands' of the image. Thus you trade CPU time for memory use. | 19:11.35 |
| You could drive the mupdf libs in banded mode, but mudraw does not do this by default. | 19:11.59 |
| Back in a bit. | 19:12.08 |
STE | Sorry had connection troubles | 19:13.13 |
| Robin_Watts : might have missed your last post | 19:13.38 |
| thank you Robin_Watts I'll go and try the commands | 19:21.17 |
| Forward 1 day (to 2012/10/16)>>> | |