The new version of the code is more obviously correct. Hopefully
the compiler is smart enough to turn it into something fast.
--HG--
extra : rebase_source : bdf2f00c5eae5d5a1bf6dff223b1b05f561384ab
Cairo is a 3rd party project and we decided in bug that silencing the
warnings was easier than patching.
--HG--
extra : rebase_source : aa7a22b208e62d0d84d1b4a6e5c58d366d957bb9
This does a 2x2 ordered dither in the same way that Skia does.
One of the things I'm currently unhappy with is that it duplicates
the gradient walker code for 16 bits. We could turn it into a large
macro that does the appropriate things for 16 bit and 32 bit
versions, but that's not particularly appealing.
This does a 2x2 ordered dither in the same way that Skia does.
One of the things I'm currently unhappy with is that it duplicates
the gradient walker code for 16 bits. We could turn it into a large
macro that does the appropriate things for 16 bit and 32 bit
versions, but that's not particularly appealing.
--HG--
extra : rebase_source : ef89cd53d68166db825bb993e1262db342ba9b5d
Skia uses a value of 0.5, however using such a high value causes some reftest failures.
Switch to 0.3 because we can do that without changing the reftests too much. We
can still use a higher value if we need to later. This should give us sqrt(3)
reduction in the number of segments.
--HG--
extra : rebase_source : 39f1892a61f7e242460deb1bc21adb30778147e3
This takes a trick from skia to go faster. For now, it's only
used on non-NEON devices.
--HG--
extra : rebase_source : c2ddce91cc8cf71ee609420107b0922d1166db43
This will allows us to scale and repeat in a single pass instead
of doing it in two.
--HG--
extra : rebase_source : 8f40aec620494d9c584a3c58cd70ac39723041a4
This takes the bilinear interpolation code from Skia. It
uses 4 bits of precision instead of 8. This lets it interpolate
two components at a time because the results fit in 16 bits.
The lower precision code is only used in the fallback code
and not in any of the specialized code for NEON. This means
pixman gives different results depending on the cpu which isn't
great. However, this was easiest and the NEON code doesn't
gain as much from using lower precision.
Skia actually uses even lower interpolation when working with
565 but that's harder to plug in right now, and this gives
a reasonable improvement.
It seems like we sometimes end up using EXTEND_NONE unintentionally, this is
likely caused by different rounding between cairo and pixman. Pixman already
does a similar reduction to EXTEND_NONE so there's not point in duplicating it
here.
--HG--
extra : rebase_source : 5649a2583a8a01084567bfb06cb3c788437210f5
Alan Coopersmith (2):
Change MMX ldq_u to return _m64 instead of forcing all callers to cast
Make mmx code compatible with Solaris Studio 12.3 compilers
Andrea Canciani (13):
Include noop in win32 builds
Silence autoconf warnings
radial: Improve documentation and naming
radial: Fix typos and trailing whitespace
win32: Build benchmarks
Workaround bug in llvm-gcc
test: Fix compilation on win32
build: Reuse sources and pixman-combine build rules
build: Reuse test sources
build-win32: Share targets and variables across win32 makefiles
build-win32: Add root Makefile.win32
test: Do not include config.h unless HAVE_CONFIG_H is defined
build-win32: Add 'check' target
Benjamin Otte (2):
tests: Add PNG_CFLAGS/LIBS to tests
build: replace @VAR@ with $(VAR) in makefiles
Bobby Salazar (1):
Android Runtime Detection Support For ARM NEON
Chris Wilson (1):
bits: optimise fetching width==1 repeats
Colin Walters (1):
autogen.sh: Support GNOME Build API
Jeff Muizelaar (1):
Add definitions of INT64_MIN and INT64_MAX
Jon TURNEY (1):
Fix build on cygwin after commit efdf65c0c4fff551fb3cd9104deda9adb6261e22
Matt Turner (11):
lowlevel-blt: add over_x888_8_8888
mmx: fix formats in commented code
mmx: convert while (w) to if (w) when possible
mmx: rename USE_MMX to USE_X86_MMX
mmx: wrap x86/MMX inline assembly in ifdef USE_X86_MMX
mmx: fix unaligned accesses
mmx: prepare pixman-mmx.c to be compiled for ARM/iwmmxt
mmx: compile on ARM for iwmmxt optimizations
mmx: optimize unaligned 64-bit ARM/iwmmxt loads
Make sure iwMMXt is only detected on ARM
Correct the minimum gcc version needed for iwmmxt
Naohiro Aota (1):
Don't use non-POSIX test
Nis Martensen (1):
Fix a few typos in pixman-combine.c.template
Siarhei Siamashka (4):
C fast path for scaled src_x888_8888 with nearest filter
ARM: workaround binutils bug #12931 (code sections alignment)
test: better coverage for BILINEAR->NEAREST filter optimization
BILINEAR->NEAREST filter optimization for simple rotation and translation
Søren Sandmann (5):
mmx: Delete some unused variables
sse2: Delete some unused variables
demos: Comment out some unused variables
ARM: Fix two bugs in neon_composite_over_n_8888_0565_ca().
test: Make fuzzer-find-diff.pl executable
Søren Sandmann Pedersen (63):
Add a "noop" implementation.
Add a noop composite function for the DST operator
Move noop dest fetching to noop implementation
Add a noop src iterator
Move NULL iterator into pixman-noop.c
Move NOP src iterator into noop implementation.
Replace instances of "dst_*" with "dest_*"
In pixman-general.c rename image_parameters to {src, mask, dest}_image
Replace argumentxs to composite functions with a pointer to a struct
blitters-test: Make common formats more likely to be tested.
Pre-release version bump to 0.23.2
Post-release version bump to 0.23.3
Makefile.am: Add pixman@lists.freedesktop.org to RELEASE_ANNOUNCE_LIST
Fix lcg_rand_u32() to return 32 random bits.
New test of pixman_region_contains_{rectangle,point}
Speed up pixman_region{,32}_contains_rectangle()
Use find_box_for_y() in pixman_region_contains_point() too
Don't include stdint.h in lowlevel-blt-bench.c
In pixman_image_create_bits() allow images larger than 2GB
Rename pixman-fast-path.h to pixman-inlines.h
Use repeat() function from pixman-inlines.h in pixman-bits-image.c
Move bilinear interpolation to pixman-inlines.h
Pre-release version bump to 0.23.4
Post-release version bump to 0.23.5
test: New function to save a pixman image to .png
Use pkg-config to determine the flags to use with libpng
test: Use smaller boxes in region_contains_test()
A few tweaks to a comment in pixman-combine.c.template
Add a generic unorm_to_unorm() conversion utility
Add general pixel converter
Add initial version of the MAKE_ACCESSORS() macro
Use MAKE_ACCESSORS() to generate all the 32 bit accessors
Use MAKE_ACCESSORS() to generate accessors for all the 16bpp formats
Use MAKE_ACCESSORS() to generate accessors for 8bpp RGB formats
Use MAKE_ACCESSORS() to generate accessors for 4 bpp RGB formats
Use MAKE_ACCESSORS() to generate accessors for 24bpp formats
Use MAKE_ACCESSORS() to generate accessors for the a1 format.
Use MAKE_ACCESSORS() to generate accessors for paletted formats
Remove x and y coordinates from analyze_extents() and compute_sample_extents()
Split computation of sample area into own function
Eliminate compute_sample_extents() function
test: Occasionally use a BILINEAR filter in blitters-test
Strength-reduce BILINEAR filter to NEAREST filter for identity transforms
Don't include stdint.h in scaling-helpers-test.
Add src, mask, and dest flags to the composite args struct.
Pre-release version bump to 0.23.6
Post-release version bump to 0.23.7
Add stable release / development snapshot to draft release notes
gradient walker: Correct types and fix formatting
Use sentinels instead of special casing first and last stops
Simplify gradient_walker_reset()
Merge branch 'gradients'
Fix use of uninitialized fields reported by valgrind
Pre-release version bump to 0.23.8
Post-release version bump to 0.23.9
Pre-release version bump to 0.24.0
Post-release version bump to 0.24.1
gradient-walker: For NONE repeats, when x < 0 or x > 1, set both colors to 0
pixman-image.c: Fix typo in pixman_image_set_transform()
Fix some signed overflow bugs
Reject trapezoids where top (botttom) is above (below) the edges
Fix bugs with alpha maps
Pre-release version bump to 0.24.2
Taekyun Kim (25):
Replace boolean arguments with flags for bilinear fast path template
REPEAT_NORMAL support for bilinear fast path template
sse2: Declare bilinear src_8888_8888 REPEAT_NORMAL composite function
ARM: Add REPEAT_NORMAL functions to bilinear BIND macros
Enable REPEAT_NORMAL bilinear fast path entries
Bilinear REPEAT_NORMAL source line extension for too short src_width
ARM NEON: Standard fast path out_reverse_8_8888
ARM: NEON better instruction scheduling of over_n_8_8888
ARM: NEON better instruction scheduling of over_n_8888
sse2: Macros for assembling bilinear interpolation code fractions
sse2: Bilinear scaled over_8888_8888
sse2: Bilinear scaled over_8888_8_8888
init/fini functions for pixman_image_t
Add new fast path flag FAST_PATH_BITS_IMAGE
Move _pixman_lookup_composite_function() to pixman-utils.c
Simple repeat fast path
Simple repeat: Extend too short source scanlines into temporary buffer
ARM: NEON: Some cleanup of bilinear scanline functions
ARM: NEON: Bilinear macro template for instruction scheduling
ARM: NEON: Replace old bilinear scanline generator with new template
ARM: NEON: Instruction scheduling of bilinear over_8888_8888
ARM: NEON: Instruction scheduling of bilinear over_8888_8_8888
ARM: NEON: Standard fast path src_n_8_8888
ARM: NEON: Standard fast path src_n_8_8
ARM: NEON: Fix assembly typo error in src_n_8_8888
There's no need to call through _cairo_surface_check_span_renderer() because
we can make the decision directly.
--HG--
extra : rebase_source : 610b624d398665bc1ca9b47c35bd129d601c232b
Currently fixup_unbounded_boxes assumes that the extents are tight when we only
have one box. This is not always true. e.g. when we tesselate boxes we can end
up with only one box which is smaller than the extents and the extents are not
updated.
--HG--
extra : rebase_source : 5e01da7d845c1cd34f9e87d5c7de22dafe9df7a4
Cairo requires image surfaces to have a particular stride alignment.
Sometimes the texture we get back from Direct3D doesn't meet these
requirements. In that case, copy to a different buffer that meets
these requirments.
--HG--
extra : rebase_source : b4a09d5dd33c921d36158f33fc2cd386ebe57883
The glyph cache uses a pointer for a hash value, and uses
that hash value as a pointer. We need to ensure that the
hash type is large enough to fit a pointer.
This will need to be reverted once our build systems are upgraded to a newer
version of FreeType. In the long run, we need to increase our runtime
dependency to this newer version.
Andrea Canciani (1):
test: Fix compilation on win32
Dave Yeo (1):
Check for working mmap()
Gilles Espinasse (2):
Fix missing AC_MSG_RESULT value from Werror test
Fix OpenMP not supported case
Siarhei Siamashka (7):
ARM: tweaked horizontal weights update in NEON bilinear scaling code
ARM: use aligned memory writes in NEON bilinear scaling code
ARM: support for software pipelining in bilinear macros
ARM: use less ARM instructions in NEON bilinear scaling code
ARM: support different levels of loop unrolling in bilinear scaler
ARM: pipelined NEON implementation of bilinear scaled 'src_8888_8888'
ARM: pipelined NEON implementation of bilinear scaled 'src_8888_0565'
Søren Sandmann Pedersen (8):
Makefile.am: Put development releases in "snapshots" directory
ARM: Tiny improvement in over_n_8888_8888_ca_process_pixblock_head
ARM: Add 'neon_composite_over_n_8888_0565_ca' fast path
Offset rendering in pixman_composite_trapezoids() by (x_dst, y_dst)
Pre-release version bump to 0.21.8
Post-release version bump to 0.21.9
Pre-release version bump to 0.22.0
Post-release version bump to 0.23.1
Taekyun Kim (3):
ARM: Common macro for scaled bilinear scanline function with A8 mask
ARM: NEON scanline functions for bilinear scaling
ARM: Enable bilinear fast paths using scanline functions in pixman-arm-neon-asm-bilinear.S
--HG--
extra : rebase_source : aff729718af3e7d25577603241d204bf02ecfb90
Pixman upgrade from bug 640250 resulted in ARM cpu features
not being detected properly in Android. As a result, NEON
optimizations were not used at all.
Alan Coopersmith (1):
Sun's copyrights belong to Oracle now
Alexandros Frantzis (2):
Add simple support for the r8g8b8a8 and r8g8b8x8 formats.
Add support for the r8g8b8a8 and r8g8b8x8 formats to the tests.
Andrea Canciani (14):
Improve precision of linear gradients
Make classification consistent with rasterization
Remove unused enum value
Fix an overflow in the new radial gradient code
Remove unused stop_range field
Fix opacity check
Improve conical gradients opacity check
Improve handling of tangent circles
Add a test for radial gradients
Fix compilation on Win32
test: Fix tests for compilation on Windows
test: Add Makefile for Win32
Do not include unused headers
test: Silence MSVC warnings
Cyril Brulebois (2):
Fix argument quoting for AC_INIT.
Fix linking issues when HAVE_FEENABLEEXCEPT is set.
Jon TURNEY (2):
Plug another leak in alphamap test
Remove stray #include <fenv.h>
Rolland Dudemaine (4):
test: Fix for mismatched 'fence_malloc' prototype/implementation
Correct the initialization of 'max_vx'
test: Use the right enum types instead of int to fix warnings
Fix "variable was set but never used" warnings
Scott McCreary (1):
Added check to find pthread on Haiku.
Siarhei Siamashka (62):
Fixed broken configure check for __thread support
Do CPU features detection from 'constructor' function when compiled with gcc
ARM: fix 'vld1.8'->'vld1.32' typo in add_8888_8888 NEON fast path
ARM: NEON: source image pixel fetcher can be overrided now
ARM: nearest scaling support for NEON scanline compositing functions
ARM: macro template in C code to simplify using scaled fast paths
ARM: performance tuning of NEON nearest scaled pixel fetcher
ARM: NEON optimization for scaled over_8888_8888 with nearest filter
ARM: NEON optimization for scaled over_8888_0565 with nearest filter
ARM: NEON optimization for scaled src_8888_0565 with nearest filter
ARM: NEON optimization for scaled src_0565_8888 with nearest filter
ARM: optimization for scaled src_0565_0565 with nearest filter
C fast path for a1 fill operation
ARM: added 'neon_composite_over_n_8_8' fast path
ARM: introduced 'fetch_mask_pixblock' macro to simplify code
ARM: better NEON instructions scheduling for over_n_8_0565
ARM: added 'neon_composite_over_8888_n_0565' fast path
ARM: reuse common NEON code for over_{n_8|8888_n|8888_8}_0565
ARM: added 'neon_composite_over_0565_n_0565' fast path
ARM: added 'neon_composite_add_8888_8_8888' fast path
ARM: better NEON instructions scheduling for add_8888_8888_8888
ARM: added 'neon_composite_add_n_8_8888' fast path
ARM: added 'neon_composite_add_8888_n_8888' fast path
ARM: added flags parameter to some asm fast path wrapper macros
ARM: added 'neon_composite_in_n_8' fast path
ARM: added 'neon_src_rpixbuf_8888' fast path
Fix for potential unaligned memory accesses
COPYING: added Nokia to the list of copyright holders
Revert "Fix "syntax error: empty declaration" warnings."
Fix for "syntax error: empty declaration" Solaris Studio warnings
Workaround for a preprocessor issue in old Sun Studio
Bugfix for a corner case in 'pixman_transform_is_inverse'
Make 'fast_composite_scaled_nearest_*' less suspicious
A new configure option --enable-static-testprogs
ARM: do /proc/self/auxv based cpu features detection only in linux
The code in 'bitmap_addrect' already assumes non-null 'reg->data'
test: affine-test updated to stress 90/180/270 degrees rotation more
New flags for 90/180/270 rotation
C fast paths for a simple 90/270 degrees rotation
Use const modifiers for source buffers in nearest scaling fast paths
test: Extend scaling-test to support a8/solid mask and ADD operation
Support for a8 and solid mask in nearest scaling main loop template
Better support for NONE repeat in nearest scaling main loop template
ARM: new macro template for using scaled fast paths with a8 mask
ARM: NEON optimization for nearest scaled over_8888_8_0565
ARM: NEON optimization for nearest scaled over_0565_8_0565
SSE2 optimization for nearest scaled over_8888_n_8888
Ensure that tests run as the last step of a build for 'make check'
Main loop template for fast single pass bilinear scaling
test: check correctness of 'bilinear_pad_repeat_get_scanline_bounds'
SSE2 optimization for bilinear scaled 'src_8888_8888'
ARM: NEON optimization for bilinear scaled 'src_8888_8888'
ARM: use prefetch in nearest scaled 'src_0565_0565'
ARM: common macro for nearest scaling fast paths
ARM: assembly optimized nearest scaled 'src_8888_8888'
ARM: new bilinear fast path template macro in 'pixman-arm-common.h'
ARM: NEON: common macro template for bilinear scanline scalers
ARM: use common macro template for bilinear scaled 'src_8888_8888'
ARM: NEON optimization for bilinear scaled 'src_8888_0565'
ARM: NEON optimization for bilinear scaled 'src_0565_x888'
ARM: NEON optimization for bilinear scaled 'src_0565_0565'
ARM: a bit faster NEON bilinear scaling for r5g6b5 source images
Søren Sandmann Pedersen (79):
Remove the class field from source_image_t
Pre-release version bump to 0.19.6
Post-release version bump to 0.19.7
Pre-release version bump to 0.20.0
Post-release version bump to 0.20.1
Version bump 0.21.1.
COPYING: Stop saying that a modification is currently under discussion.
Remove workaround for a bug in the 1.6 X server.
[mmx] Mark some of the output variables as early-clobber.
Delete the source_image_t struct.
Generate {a,x}8r8g8b8, a8, 565 fetchers for nearest/affine images
Pre-release version bump
Post-release version bump to 0.21.3
test: Make composite test use some existing macros instead of defining its own
Add enable_fp_exceptions() function in utils.[ch]
Extend gradient-crash-test
test: Move palette initialization to utils.[ch]
test/utils.c: Initialize palette->rgba to 0.
Make the argument to fence_malloc() an int64_t
Add a stress-test program.
Add a test compositing with the various PDF operators.
Fix divide-by-zero in set_lum().
sse2: Skip src pixels that are zero in sse2_composite_over_8888_n_8888()
Add iterators in the general implementation
Move initialization of iterators for bits images to pixman-bits-image.c
Eliminate the _pixman_image_store_scanline_32/64 functions
Move iterator initialization to the respective image files
Virtualize iterator initialization
Use an iterator in pixman_image_get_solid()
Move get_scanline_32/64 to the bits part of the image struct
Allow NULL property_changed function
Consolidate the various get_scanline_32() into get_scanline_narrow()
Linear: Optimize for horizontal gradients
Get rid of the classify methods
Add direct-write optimization back
Skip fetching pixels when possible
Turn on testing for destination transformation
Fix destination fetching
Fix dangling-pointer bug in bits_image_fetch_bilinear_no_repeat_8888().
Pre-release version bump to 0.21.4
Post-release version bump to 0.21.5
Print a warning when a development snapshot is being configured.
Move fallback decisions from implementations into pixman-cpu.c.
Add a test for over_x888_8_0565 in lowlevel_blt_bench().
Add SSE2 fetcher for x8r8g8b8
Add SSE2 fetcher for a8
Improve performance of sse2_combine_over_u()
Add SSE2 fetcher for 0565
Add pixman-conical-gradient.c to Makefile.win32.
Move all the GTK+ based test programs to a new subdir, "demos"
Add @TESTPROGS_EXTRA_LDFLAGS@ to AM_LDFLAGS
test/Makefile.am: Move all the TEST_LDADD into a new global LDADD.
Add pixman_composite_trapezoids().
Add a test program for pixman_composite_trapezoids().
Add support for triangles to pixman.
Add a test program, tri-test
Optimize adding opaque trapezoids onto a8 destination.
Add new public function pixman_add_triangles()
Avoid marking images dirty when properties are reset
In pixman_image_set_transform() allow NULL for transform
Coding style: core_combine_in_u_pixelsse2 -> core_combine_in_u_pixel_sse2
sse2: Convert all uses of MMX registers to use SSE2 registers instead.
sse2: Delete unused MMX functions and constants and all _mm_empty()s
sse2: Don't compile pixman-sse2.c with -mmmx anymore
sse2: Remove all the core_combine_* functions
sse2: Delete obsolete or redundant comments
sse2: Remove pixman-x64-mmx-emulation.h
sse2: Minor coding style cleanups.
Delete pixman-x64-mmx-emulation.h from pixman/Makefile.am
Minor fix to the RELEASING file
Pre-release version bump to 0.21.6
Post-release version bump to 0.21.7
test: In image_endian_swap() use pixman_image_get_format() to get the bpp.
test: Do endian swapping of the source and destination images.
In delegate_{src,dest}_iter_init() call delegate directly.
Fill out parts of iters in _pixman_implementation_{src,dest}_iter_init()
Simplify the prototype for iterator initializers.
test: Randomize some tests if PIXMAN_RANDOMIZE_TESTS is set
test: Fix infinite loop in composite
Previously we missed a decrement at the end of the recursion.
This caused us to accumulate over time, causing us to misreport.
--HG--
extra : rebase_source : 4129933528b3c40aaefad7f8f8a4c5a76f35a9b1
