Commit Graph

1522 Commits

Author SHA1 Message Date
John Cupitt
88148318eb fix performance regression
the extra check on bandfmt in sizeof() in 8.3.2 was causing some performance
problems ... move the check to file read, so we only do it once
per image, not once per pixel or scanline

thanks Lovell!
2016-07-30 10:51:54 +01:00
John Cupitt
379b69fda8 note on test suite failure 2016-05-08 10:02:34 +01:00
John Cupitt
d46a57cf6b update notes 2016-04-11 13:27:04 +01:00
John Cupitt
269cbb8641 better mask sizing for lanczos 2016-04-11 13:19:00 +01:00
John Cupitt
511a0a7da5 better rounding in reducev orc path 2016-04-10 17:52:51 +01:00
John Cupitt
bf1c8707c7 better draw docs
add a caching discussion to the start of draw

see https://github.com/jcupitt/libvips/issues/411
2016-04-06 10:26:53 +01:00
John Cupitt
8aa4566c87 update notes 2016-04-02 20:24:54 +01:00
John Cupitt
bc320f1090 much better int mask creation
removes some banding from the orc path for reducev
2016-04-02 18:46:10 +01:00
John Cupitt
5f24468dba tiny code cleanup, note a bug 2016-04-02 11:25:27 +01:00
John Cupitt
95a74cbf28 tiny help polishing 2016-03-25 16:35:28 +00:00
John Cupitt
ddd7ab7459 doc fixups 2016-03-25 10:56:59 +00:00
John Cupitt
1a25c7a083 better rounding in vipsthumbnail 2016-03-25 10:42:21 +00:00
John Cupitt
2996947b0d start fixing up test_thumbnail
since --interpolator no longer does anything
2016-03-25 06:11:01 +00:00
John Cupitt
213f9e23b4 count colour pixels
more overcomputation testing
2016-03-24 17:41:05 +00:00
John Cupitt
3c98ba78a0 small experiment 2016-03-22 14:19:32 +00:00
John Cupitt
e8b5cb6c23 add VIPS_COUNT_PIXELS, overcomputation tracking
added VIPS_COUNT_PIXELS(), use like this:

	static int vips_shrinkh_gen( VipsRegion *or, ... )
	{
        	VIPS_COUNT_PIXELS( or, "vips_shrinkh_gen" );
	}

and on image close, if more than 100% of the pixels have been
calculated, you get a warning

only if you enable debugging, since this hurts perf slightly
2016-03-22 12:26:45 +00:00
John Cupitt
4f8aca3a93 sync 2016-03-21 16:40:55 +00:00
John Cupitt
385f8a167f notes 2016-03-21 11:36:15 +00:00
John Cupitt
05156edd15 add orc path for reducev
works, but still not quite right, see some banding
2016-03-18 17:41:51 +00:00
John Cupitt
c20eb9dca0 remove old unrolled case for reduce
not used any more
2016-03-15 17:34:31 +00:00
John Cupitt
ff88087a28 try making the reduce mask larger with residual 2016-03-15 14:59:59 +00:00
John Cupitt
45c3fea6ac more resize tuning 2016-03-15 09:49:30 +00:00
John Cupitt
773b474116 Merge branch 'master' of github.com:jcupitt/libvips 2016-03-14 13:35:47 +00:00
John Cupitt
82cecf3d9f bilinear interp. supports complex types
in line with cubic etc.
2016-03-14 10:06:45 +00:00
John Cupitt
fba2ac2f85 fix sizing issues with new reduce code
a couple of bugs and some changed conventions
2016-03-13 17:44:24 +00:00
John Cupitt
f12fef4aa9 fix compile 2016-03-13 11:35:35 +00:00
John Cupitt
dc7902b12f Merge branch 'master' into add-lanczos3 2016-03-13 11:23:03 +00:00
John Cupitt
70db9aa77b better rounding in resize 2016-03-12 13:14:34 +00:00
John Cupitt
493482c8a7 hey ho, line cache goes back too 2016-03-11 14:56:06 +00:00
John Cupitt
a832a0c254 notes 2016-03-11 14:42:12 +00:00
John Cupitt
4f41617d89 TODO notes 2016-03-11 13:47:19 +00:00
John Cupitt
2ca7b29214 better gauss mat generation 2016-03-10 22:01:55 +00:00
John Cupitt
4974a1ed9c better rounding for vips_resize()
we were getting off by one size errors
2016-03-10 19:53:05 +00:00
John Cupitt
9209fb25c5 update notes 2016-03-10 17:40:19 +00:00
John Cupitt
d351e07a6b finish up 2016-03-10 17:33:34 +00:00
John Cupitt
e9728e389b first commit 2016-03-09 12:56:30 +00:00
John Cupitt
2ea5c5f7e5 make python bandrank() work like bandjoin()
we had a custom wrapper for bandjoin(), but bandrank(), a function with
an almost identical interface, did not ... this was confusing

bandrank() now has a custom wrapper too ... this breaks API
unfortunately, but hopefully very few people were using this thing and
it's better to make this change as soon as possible
2016-03-01 12:12:51 +00:00
John Cupitt
25c3c49d1c better handling of cast+shift for non-int formats
before, im.cast(uchar, shift = true) where im was float and tagged as
rgb16 would not shift the image, since it's unclear how much to shift a float
type by

now we do two casts: first, we guess the numeric range from the
interpretation, so rgb16 would be ushort, so we cast float->ushort;
second, we cast to the target type and do the shift on the way

see https://github.com/jcupitt/libvips/issues/397

thanks apacheark
2016-03-01 11:31:54 +00:00
John Cupitt
b197773095 oop, overflow in sharpen
plus another tweak of the defaults
2016-02-28 11:24:35 +00:00
John Cupitt
35650c2244 reduce the number of copy() ops
reduce the number of copy() ops issued during colour conversion
2016-02-26 14:05:35 +00:00
John Cupitt
c95bbe43b5 oops in sharpen 2016-02-26 12:26:12 +00:00
John Cupitt
d33306836f better rgb16 -> scrgb handling
helps new sharpen tests
2016-02-26 09:08:42 +00:00
John Cupitt
04cc13e321 notes 2016-02-25 11:47:53 +00:00
John Cupitt
c1f1a7c1b3 improve sharpen for screen work
it was a bit granular (since it was designed for offset work), make sharpen
easier to adjust for smaller effects

- new 'sigma' parameter
- greater parameter range
- more self-tests
2016-02-25 11:16:24 +00:00
John Cupitt
789e154cd7 phew, done 2016-02-13 14:04:29 +00:00
John Cupitt
fdf0c349ae almost there
seeing transparency tags when we shouldn't, strange
2016-02-13 11:51:50 +00:00
John Cupitt
c2be771342 fix load from buffer, update APIs 2016-02-12 20:04:46 +00:00
John Cupitt
0dbd64a122 fix gifload memleak 2016-02-12 17:53:15 +00:00
John Cupitt
b10f332404 very close now
just a leak to nail down and some small stuff
2016-02-12 17:30:56 +00:00
John Cupitt
9262d672b7 almost done
a valgrind failure still
2016-02-12 10:53:04 +00:00