This requires GLIBC 2.23+, plus either gcc 6+ or clang 14+.
- Provides build-time feature detection
- Use with (un)premultiply for ~10% perf gain on AVX CPUs
- Slightly increases binary size, so best to use sparingly
arrayjoin was making a region on every input image during startup, and
repeating for each thread (!!) so large arrays could be very expensive
to join.
Instead, make input regions on demand, and computre set of required
input images rather than searching for them.
See https://github.com/libvips/libvips/discussions/3247
* ADded optional output parameter to smartcrop to track the center of
attention.
* Added unit test.
* Added docs & updated changelog.
* Updated argument labels & types in documentation.
* reimplement threadpool
just a set of threads that get recycled
"ninja test" passes and dzsave seems to work, though pytest fails for
some reason
* clean up threading code a bit
move base stuff into thread.c, so we have a simple
thread -> threadset -> threadpool
layering
* start trying to revert g_threadpool
based on the original commit
* working!
nice low systime again
still need to repply cnages to threadpool.c since 80e0cc3d1
* reapply fixes from master
so threadpool.c is now up to date
* rename VipsThread as VipsWorker
a bit less confusing
* use a semaphore to count workers in a pool
* tidy up
* formatting
* dynamic threadpool sizing
based on counting the number of blocked threads in each pool
it works, but the improvement is not great :(
* add "concurrency" metadata item
so operators can hint threadpool size (dzsave especially)
* don't use thinstrip for small images
* add RGB mode to openslide
since flatten was taking 20% of CPU time for dzsave
* fix up rgb mode
now actually works
* make the tile buffer per thread
in the new openslideload rgb mode
* fix dynamic pool downsize
* mild refactoring
* fix the buffer system
oops, turned it off by mistake
* all done!
* revise changelog
* Update libvips/iofuncs/threadset.c
Co-authored-by: Kleis Auke Wolthuizen <github@kleisauke.nl>
* Update libvips/iofuncs/threadset.c
Co-authored-by: Kleis Auke Wolthuizen <github@kleisauke.nl>
* LSan: add libMagickCore to suppression file
* Revert "Remove mutex lock for VipsThreadStartFn"
This reverts commit 41440491.
* add VIPS_MAX_THREADS
to set a hard limit on the threadset size
* Revert "Revert "Remove mutex lock for VipsThreadStartFn""
This reverts commit 77e8520966ba79194fff3b4e648bbd295cd5c260.
* remove sslock from sink.c
* move fixed threadpool build to init
not first use
* add some doc comments
* revert test suite threshold change
* add a test for MAX_THREADS
and move the test tmp/ area into the builddir
* limit VIPS_MAX_THREADS to sane values
* use tabs rather than spaces
Co-authored-by: Kleis Auke Wolthuizen <github@kleisauke.nl>
We were avoiding /0 by testing for alpha==0, however, this will still
allow very small values of alpha to generate +/- Inf. Instead, check for
abs(alpha)<epsilon.
Fixes some artifacts after unpremul.
The `*_class_intern_init` functions have already been patched within
wasm-vips to include the missing `class_data` member, so this adapter
function was not needed.
- Remove `HAVE_LCMS` definition in favor of `HAVE_LCMS2`.
- Remove `HAVE_WINDOWS_H` definition in favor of `G_OS_WIN32`.
- Remove stray `vips_text_get_type` in `conversion.c`.
- Remove duplicated `unistd.h` include.
- Remove redundant `strcasecmp` definition, we use `g_ascii_strcasecmp` everywhere.
- Remove unnecessary header checks in `configure.ac` and `meson.build`.
- Ensure `unistd.h` include is guarded with `HAVE_UNISTD_H`.
- Fail early when `-Dfontconfig=enabled` and `pangoft2` is not found.
Use GLib's i18n support instead of copying and pasting that
logic into its own header. This deprecates the vips/intl.h
header in favour of glib/gi18n.h.
* Ensure that double asterisk characters are only used for gtk-doc comments
This triggers warnings when parsing the files with the introspection scanner
and gtk-doc
* Enhance the introspection support by fixing annotations
Also refine the .metadata file
Co-authored-by: John Cupitt <jcupitt@gmail.com>
This patch makes "insert" issue minimise signals for inputs in sequential
mode. This can drop memory use in some programs, for example:
```python
import sys
import random
import pyvips
image = pyvips.Image.black(20000, 20000)
for filename in sys.argv[2:]:
tile = pyvips.Image.new_from_file(filename, access='sequential')
x = random.randint(0, image.width - tile.width)
y = random.randint(0, image.height - tile.height)
image = image.insert(tile, x, y)
image.write_to_file(sys.argv[1])
```
Before this patch:
```
$ for i in {0..1000}; do cp ~/pics/k2.jpg $i.jpg; done
$ /usr/bin/time -f %M:%e ../manyjoin.py ../x.jpg *.jpg
5456256:4.34
```
With this patch:
```
$ /usr/bin/time -f %M:%e ../manyjoin.py ../x.jpg *.jpg
2475324:3.38
```
This patch makes arrayjoin leave a bigger margin off the end of each
input image. This can help some loaders which have large output caches.
See https://github.com/libvips/libvips/issues/2440
* fic gtk-doc typenames in cgif
* fix flatten clipping
flatten could produce out of range values if max_alpha was less than
the limit of the numeric range of the format
https://github.com/libvips/libvips/issues/2431
* Ensure UBSan exits with a non-zero code on error
* Avoid misaligned member access in mosaic_fuzzer
* Add missing VIPS_CLIP in scRGB2sRGB/scRGB2BW
* Fix UBSan error in flatten
By using saturated casts for the int types (copied from vips_cast).
* CI: ensure fuzzer log is printed on error
* Avoid UB in heifload
* Revert flatten change
I could no longer reproduce this with clang 12 locally.
* Indentation fixes [skip ci]
By making the sequential line cache non-persistent, and only minimising
when the read point is well past the image.
On large arrayjoin operations, this saves many GB of memory.
See https://github.com/kleisauke/net-vips/issues/135
arrayjoin with a sequential pipeline will now minimise inputs when they
are no longer being used. This drops the number of open file handles
needed to construct large arrays.
Sadly, memory savings are minimal.
see https://github.com/kleisauke/net-vips/issues/135