r3wp [groups: 83 posts: 189283]
  • Home
  • Script library
  • AltME Archive
  • Mailing list
  • Articles Index
  • Site search
 

World: r3wp

[Postscript] Emitting Postscript from REBOL

Henrik
24-Feb-2008
[1432x2]
wow, that took a long time to generate.
but it works
Geomol
24-Feb-2008
[1434]
! Great! :)
Henrik
24-Feb-2008
[1435x2]
15 seconds at 100% CPU
when adding DeviceRGB in your previous example, that example works 
too.
Geomol
24-Feb-2008
[1437]
Maybe because it had to load image from web also!? Try with a local 
version of the image.
Henrik
24-Feb-2008
[1438]
and that took less than 1 second to generate.
Geomol
24-Feb-2008
[1439]
cool
Henrik
24-Feb-2008
[1440]
well, it would be bad if it ate that much CPU, just waiting for network 
stuff.
Geomol
24-Feb-2008
[1441x2]
I load the image with
load-image/update
to be sure to always get latest version.
Maybe your system had a hickup? Try again!
Henrik
24-Feb-2008
[1443]
tried several times, same result.
Geomol
24-Feb-2008
[1444]
ok
Henrik
24-Feb-2008
[1445]
will test with a local file soon
Geomol
24-Feb-2008
[1446]
Try take a local copy of postscript.r and change the load-image to 
not do a /update.
Henrik
24-Feb-2008
[1447x4]
writing palms.jpg locally took 1 second
testing postscript generation
perhaps 12 seconds with a local palms.jpg file
what do you do to translate the data? is it in mezzanine form?
Geomol
24-Feb-2008
[1451x2]
Images are implemented in postscript.r using ASCIIHexDecode filter, 
which double the size of image data. It would be better to implement 
ASCII85, which does a 4:5 increase only.


Also images are handled in something called dictionaries in PostScript. 
This require PostScript v. 2.0.
If you look in postscript.r, I encode the image data in line 211-220. 
It's just normal hex written as ASCII, so you get '0'-'9' + 'A'-'F'.
Henrik
24-Feb-2008
[1453x2]
when building the image data, how do you do that? or does ASCIIHexDecode 
do that?
ok
Geomol
24-Feb-2008
[1455]
So this is very simple encoding. ASCII85 encoding is a bit more difficult, 
but will take up less space.
Henrik
24-Feb-2008
[1456x3]
perhaps it's equal in speed, I don't know. odd though, I can't see 
what's so slow about that code, other than the newline insertion 
thing
perhaps inserting in images is very slow?
does it need to be done?
Geomol
24-Feb-2008
[1459x2]
It seems to work without the newlines, but then the ps file become 
difficult to enter with e.g. vim. That part is fast, I think. It 
only put in a newline for every 80 chars.
My guess is, that it's the load-image/update that give you performance 
hit. Try change it!
Henrik
24-Feb-2008
[1461x3]
it does not, I tried a local file. 12 seconds.
will try to add some profiling hints
nope, it's the inserting part. the rest is taking no time at all.
Geomol
24-Feb-2008
[1464]
You say, you tried a local file, but did you change the load-image 
in postscript.r?
Henrik
24-Feb-2008
[1465]
so I think there should be an option to leave that out, or by default 
not have it. I'm guessing most people want to generate and print 
PS file, rather than read and edit them by hand.
Geomol
24-Feb-2008
[1466]
ok, the inserting of the newlines is the problem?
Henrik
24-Feb-2008
[1467x2]
didn't try that with /update yet. will have to be sure
yep, same thing without /update.
Geomol
24-Feb-2008
[1469]
If the inserting is the problem, then I might be able to fix that 
with copy to another series in stead. Problem with insert is, when 
you have large series, it has to move the content down the memory.
Henrik
24-Feb-2008
[1470x2]
I'm trying to remove that code right now to see if it improves performance.
now GS won't read the PS file, but the generation is much much faster. 
less than 0.5 secs
Geomol
24-Feb-2008
[1472]
Just uncomment after the trim in line 213.
Henrik
24-Feb-2008
[1473]
did that
Geomol
24-Feb-2008
[1474x3]
ok :)
Ok, I'll see, if I can make a faster insert...
Henrik, try with new version now:
http://home.tiscali.dk/john.niclasen/postscript/postscript.r
Henrik
24-Feb-2008
[1477x4]
odd, GS won't eat it
oh, error 40 :-)
much faster now and it works
tried a much bigger image. no problems at all.
Geomol
24-Feb-2008
[1481]
Great! That's the version to use for now then. Maybe I should put 
an announce in that group?