OS Version: 16.04 Ghostscript Version: 9.18 We find a problem about ghostscript: Print file on Ubuntu 14.04, Ghostscript 9.10 is fast. But, after I update to Ubuntu 16.04, Ghostscript 9.18, I print the same file with the same driver is slow. I figure out that all the print jobs is much slower 60%(Or more slower) on Ubuntu 16.04 than Unbuntu 14.04. I view the "CUPS Error Log"(Log.zig), it shows that Ghostscript may be the key process that makes the print job slowly. By counting the time from " Starting process "renderer"" to " Closing renderer", we figure out that ghostscript of 16.04 spend more time than 14.04 on "renderer" process. So we assuming that:"renderer" is the key process that makes the print job slowly. So it seems that the newer Ghostscript version used by 16.04 is processed much slower by your printer than the PostScript of 14.04's Ghostscript version. By the way, more and more our printer driver users are affect by this problem. So, we are eagerly want to know that you can dealing with this problem ASAP. And we hope that you can share the investigation with us. Attach file introduction: ---------------------------------------------------------- Test log: Ubuntu14.04,Job621-Job630: \log\Ubuntu_14.04_error_log_Job621_630 Ubuntu16.04,Job621-Job630: \log\Ubuntu_16.04_error_log_Job621_630 ---------------------------------------------------------- Test Driver: Driver carry by Ubuntu14.04 and Ubuntu16.04 itself :”Ricoh MP C5503 PXL” Print Output: file:/test.prn Print file: OS Testpage. We test many commercial’s drivers (below), it shows that other drivers(use ghostsrcipt) have the same slower phenomenon: HP Color LaserJet 2500 Foomatic/pxlcolor HP Color LaserJet 3600 Foomatic/pxljr Brother HL-7050 Foomatic/pxlmono Brother MFC9840CDW Foomatic/pxlcolor
Created attachment 16576 [details] Add Log file:
First, try an up to date Ghostscript: the latest release is 9.26. Secondly, if that doesn't help, we cannot investigate without an example file to reproduce the problem. A cups log is not really any use to us.
I should have also said: an example file which is the input to Ghostscript (so Postscript or PDF).
Dear Chris, >First, try an up to date Ghostscript: the latest release is 9.26. We had tried to use ghostscript 9.26 to print the same jobs, the investigation result shows that the newest version 9.26 is still have the same slower phenomenon. >an example file which is the input to Ghostscript (so Postscript or PDF). We use a PDF file(sample.pdf) to do the investigation, which we attach in sample.zip. Best Regards,
Created attachment 16594 [details] Sample file
Well, running locally using the command line options: time gs -q -dNOPAUSE -dBATCH -dSAFER -sDEVICE=ps2write -sOUTPUTFILE=/dev/null -dLanguageLevel=3 -r600 -dCompressFonts=false -dNoT3CCITT -dNOINTERPOLATE -c "save pop" -f sample.pdf (writing to /dev/null to avoid i/o variations). I get 9.16 giving: real 0m0.835s user 0m0.819s sys 0m0.016s and 9.26 giving: real 0m0.453s user 0m0.441s sys 0m0.012s So, as far as I can see, 9.26 is considerably *faster* than 9.16.
Dear Chris, First, we found that the gs version 9.10(not 9.16) command execute faster than version 9.18 / 9.26. Second, the "sDEVICE" we point out is "pxlcolor". The following gs command in our log file slow down with version 9.18 / 9.26 than version 9.10. gs -q -dBATCH -dPARANOIDSAFER -dNOPAUSE -dNOINTERPOLATE -sDEVICE=pxlcolor -r600x600 -sPAPERSIZE=a4 -sOutputFile=- -
Okay, will using: time ./gs -q -dBATCH -dPARANOIDSAFER -dNOPAUSE -dNOINTERPOLATE -sDEVICE=pxlcolor -r600x600 -sPAPERSIZE=a4 -sOutputFile=/dev/null sample.pdf I get, from 9.10: real 0m0.271s user 0m0.263s sys 0m0.008s And from 9.26, I get: real 0m0.316s user 0m0.308s sys 0m0.008s So there is difference in the favour of 9.10, but it's pretty hard to see how that translates into an observable slow down within the cups infrastructure. Perhaps if you supply a Ghostscript command line, and the results you see from running that command with "time", it might help narrow down what's going on.
Dear Chris, Thanks a lot for your suggestion. We'll try the gs command with "time". And if there is new phenomenon, contact you later.
Dear Chris, We tried the gs command as below, and found the gs version 9.26(on Ubuntu 16.04) is about 30%(0.1s) slower than version 9.10(on Ubuntu 14.04). time ./gs -q -dBATCH -dPARANOIDSAFER -dNOPAUSE -dNOINTERPOLATE -sDEVICE=pxlcolor -r600x600 -sPAPERSIZE=a4 -sOutputFile=/dev/null sample.pdf gs version 9.10(on Ubuntu 14.04): real 0m0.743s user 0m0.321s sys 0m0.036s gs version 9.26(on Ubuntu 16.04): real 0m0.782s user 0m0.419s sys 0m0.021s *The time is changing when try again, but still the "user time" is about 30% (0.1s)slower with gs version 9.26 than with 9.10. *Our Ubuntu 14.04 64bit and Ubuntu 16.04 64bit machines have the same hardware configuration. I saw your "user time" with gs version 9.26 is only about 17%(0.04s) slower than 9.10. Would you mind to tell me if your two machines which run gs version 9.26 and 9.10 have the same hardware configuration? Although 0.1 second is tiny in one print job, but when hundreds of thousands jobs be print in the meantime, the slowing down became obvious. Would you mind to do some analysis to find out the reason of the slowing down phenomenon?
Both my tests were run on the same machine. I also ran the tests 5 times each, to ensure I wasn't seeing an outlier result from a single execution - in this case the times were so close, it wasn't worth averaging them. Sorry, but frankly, no, I won't be doing analysis to find a 0.04 (or less) second slow down. If you take a look at our git history between 9.10 and 9.18 you'll find over 1000 commits, many of which were to cope with broken or out spec PDF files, and most of which required extra work to deal with such cases. Any one, or combination of those could be the culprit. If you can identify a small (<5 pages) job that exhibits a more pronounced slow down (>2 seconds difference, ideally), then feel free to attach it here, and reopen this, and then I (or one of us) will look at it.