Currently the number of bits per pixel is used instead of the
number of bytes per pixel when calculating image strides. This
does not cause any real problems, but the gaps between scanlines
are excessively large.
This patch actually converts bits to bytes and rounds up the result
to the nearest byte boundary.
Signed-off-by: Siarhei Siamashka <siarhei.siamashka@gmail.com>
Reviewed-by: soren.sandmann@gmail.com
__builtin_shuffle was removed in clang 5.0.
Build log says:
test/utils-prng.c:207:27: error: use of unknown builtin '__builtin_shuffle' [-Wimplicit-function-declaration]
randdata.vb = __builtin_shuffle (randdata.vb, bswap_shufflemask);
^
test/utils-prng.c:207:25: error: assigning to 'uint8x16' (vector of 16 'uint8_t' values) from incompatible type 'int'
randdata.vb = __builtin_shuffle (randdata.vb, bswap_shufflemask);
^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2 errors generated
Link to original discussion:
http://lists.llvm.org/pipermail/cfe-dev/2017-August/055140.html
It's possible to build pixman if attached patch is applied. Basically
patch adds check for __builtin_shuffle support and in case there is
none, falls back to clang-specific __builtin_shufflevector that do the
same but have different API.
Bugzilla: https://bugs.gentoo.org/646360
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=104886
Tested-by: Philip Chimento <philip.chimento@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Just builds on Fedora 28 for x86_64 at the moment, but it's a start.
Credit to Daniel Stone for eliminating the nested docker image.
Signed-off-by: Adam Jackson <ajax@redhat.com>
Use vector intrinsic for loading possibly unaligned data instead of a
typecast.
Bugzilla: https://bugzilla.redhat.com/1572540
Signed-off-by: Dan Horák <dan@danny.cz>
Signed-off-by: Adam Jackson <ajax@redhat.com>
Tested-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Siarhei Siamashka <siarhei.siamashka@gmail.com>
Expanded the size slightly (from ~4.25 to 5) to make the cutoff less
noticable. Previouly the value at the cutoff was
gaussian_filter(sqrt(2)*3/2) = 0.00626 which is larger than the
difference between 8-bit pixels (1/255 = 0.003921). New cutoff is
gaussian_filter(2.5) = 0.001089 which is smaller.
v11: added some math to commit message
v14: left SIGMA in there
Signed-off-by: Bill Spitzak <spitzak@gmail.com>
Acked-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Søren Sandmann <soren.sandmann@gmail.com>
There are a few bugs in the current normalization code
(1) The normalization is based on the sum of the *floating point*
values generated by integral(). But in order to get the sum to be
close to pixman_fixed_1, the sum of the rounded fixed point values
should be used.
(2) The multiplications in the normalization loops often round the
same way, so the residual error can fairly large.
(3) The residual error is added to the sample located at index
(width - width / 2), which is not the midpoint for odd widths (and
for width 1 is in fact outside the array).
This patch fixes these issues by (1) using the sum of the fixed point
values as the total to divide by, (2) doing error diffusion in the
normalization loop, and (3) putting any residual error (which is now
guaranteed to be less than pixman_fixed_e) at the first sample, which
is the only one that didn't get any error diffused into it.
Signed-off-by: Søren Sandmann <soren.sandmann@gmail.com>
The convolution of two BOX filters is simply the length of the
interval where both are non-zero, so we can simply return width from
the integral() function because the integration region has already
been restricted to be such that both functions are non-zero on it.
This is both faster and more accurate than doing numerical integration.
This patch is based on one by Bill Spitzak
https://lists.freedesktop.org/archives/pixman/2016-March/004446.html
with these changes:
- Rebased to not assume any changes in the arguments to integral().
- Dropped the multiplication by scale
- Added more details in the commit message.
Signed-off-by: Søren Sandmann <soren.sandmann@gmail.com>
Reviewed-by: Bill Spitzak <spitzak@gmail.com>
Only the triangle is discontinuous at 0. The other filters resemble a
cubic closely enough that Simpsons integration works without
splitting.
Changes by Søren: Rebase without the changes to the integral function,
update comment to match the new code.
Signed-off-by: Bill Spitzak <spitzak@gmail.com>
Signed-off-by: Søren Sandmann <soren.sandmann@gmail.com>
Reviewed-by: Søren Sandmann <soren.sandmann@gmail.com>
Simpsons uses cubic curve fitting, with 3 samples defining each
cubic. This makes the weights of the samples be in a pattern of
1,4,2,4,2...4,1, and then dividing the result by 3.
The previous code was using weights of 1,2,0,6,0,6...,2,1.
With this fix the integration is accurate enough that the number of
samples could be reduced a lot. Multiples of 12 seem to work best.
v7: Merged with patch to reduce from 128 samples to 16
v9: Changed samples from 16 to 12
v10: Fixed rebase error that made it not compile
v11: minor whitespace change
v14: more whitespace changes
Signed-off-by: Bill Spitzak <spitzak@gmail.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Søren Sandmann <soren.sandmann@gmail.com>
Rearranged so that the entire block of memory for the filter pair
is allocated first, and then filled in. Previous version allocated
and freed two temporary buffers for each filter and did an extra
memcpy.
v8: small refactor to remove the filter_width function
v10: Restored filter_width function but with arguments changed to
match later patches
v11: Removed unused arg and pointer from filter_width function
Whitespace fixes.
Signed-off-by: Bill Spitzak <spitzak@gmail.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Acked-by: Søren Sandmann <soren.sandmann@gmail.com>
If enable-gnuplot is configured, then you can pipe the output of a
pixman-using program to gnuplot and get a continuously-updated plot of
the horizontal filter. This works well with demos/scale to test the
filter generation.
The plot is all the different subposition filters shuffled
together. This is misleading in a few cases:
IMPULSE.BOX - goes up and down as the subfilters have different
numbers of non-zero samples
IMPULSE.TRIANGLE - somewhat crooked for the same reason
1-wide filters - looks triangular, but a 1-wide box would be more
accurate
Changes by Søren: Rewrote the pixman-filter.c part to
- make it generate correct coordinates
- add a comment on how coordinates are generated
- in rounding.txt, add a ceil() variant of the first-sample
formula
- make the gnuplot output slightly prettier
v7: First time this ability was included
v8: Use config option
Moved code to the filter generator
Modified scale demo to not call filter generator a second time.
v10: Only print if successful generation of plots
Use #ifdef, not #if
v11: small whitespace fixes
v12: output range from -width/2 to width/2 and include y==0, to avoid misleading plots
for subsample_bits==0 and for box filters which may have no small values.
Signed-off-by: Bill Spitzak <spitzak@gmail.com>
This is very useful for comparing the results of SEPARABLE_CONVOLUTION
with BILINEAR and NEAREST.
v14: Removed good/best items
v15: Skip filter generation so gnuplot output continues showing previous value
Signed-off-by: Bill Spitzak <spitzak@gmail.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
It now shows the initial value of 4 when the demo is started
Signed-off-by: Bill Spitzak <spitzak@gmail.com>
Reviewed-by: Søren Sandmann <soren.sandmann@gmail.com>
Instead of using the boundary of xformed rectangle, use the boundary
of xformed ellipse. This is much more accurate and less blurry. In
particular the filtering does not change as the image is rotated.
Signed-off-by: Bill Spitzak <spitzak@gmail.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Soren Sandmann <soren.sandmann@gmail.com>
Generalize and simplify the code that reduces BILINEAR to NEAREST so
that the reduction happens for all affine transformations where
t00...t12 are integers and (t00 + t01) and (t10 + t11) are both
odd. This is a sufficient condition for the resulting transformed
coordinates to be exactly at the center of a pixel so that BILINEAR
becomes identical to NEAREST.
V2: Address some comments by Bill Spitzak
Signed-off-by: Søren Sandmann <soren.sandmann@gmail.com>
Reviewed-by: Bill Spitzak <spitzak@gmail.com>
This new test tests a bunch of bilinear downscalings, where many have
a transformation such that the BILINEAR filter can be reduced to
NEAREST (and many don't).
A CRC32 is computed for all the resulting images and compared to a
known-good value for both 4-bit and 7-bit interpolation.
V2: Remove leftover comment, some minor formatting fixes, use a
timestamp as the PRNG seed.
Signed-off-by: Søren Sandmann <soren.sandmann@gmail.com>
Reviewed-by: Bill Spitzak <spitzak@gmail.com>
When a BILINEAR filter is reduced to NEAREST, it is possible for both
types of fast paths to run; in this case, the NEAREST ones should be
preferred as that is the simpler filter.
Signed-off-by: Soren Sandmann <soren.sandmann@gmail.com>
Reviewed-by: Bill Spitzak <spitzak@gmail.com>
<float.h> is included unconditionally by pixman-private.h, which in
turn gets included by assembler files. Unfortunately, with certain C
libraries (like the musl C library), <float.h> cannot be included in
assembler files:
CCLD libpixman-arm-simd.la
/home/test/buildroot/output/host/usr/arm-buildroot-linux-musleabihf/sysroot/usr/include/float.h: Assembler messages:
/home/test/buildroot/output/host/usr/arm-buildroot-linux-musleabihf/sysroot/usr/include/float.h:8: Error: bad instruction `int __flt_rounds(void)'
/home/test/buildroot/output/host/usr/arm-buildroot-linux-musleabihf/sysroot/usr/include/float.h: Assembler messages:
/home/test/buildroot/output/host/usr/arm-buildroot-linux-musleabihf/sysroot/usr/include/float.h:8: Error: bad instruction `int __flt_rounds(void)'
It turns out however that <float.h> is not needed by assembly files,
so we move its inclusion within the #ifndef __ASSEMBLER__ condition,
which solves the problem.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: Siarhei Siamashka <siarhei.siamashka@gmail.com>
The `check` target in test/Makefile.win32 assumed that any non-0 exit
code from the tests was an error, but the testsuite is currently using
77 as a SKIP exit code (based on the convention used in autotools).
Fixes fence-image-self-test and cover-test (now reported as SKIP).
Signed-off-by: Andrea Canciani <ranma42@gmail.com>
Acked-by: Oded Gabbay <oded.gabbay@gmail.com>
The `rm` command is not usually available when running on Win32 in a
`cmd.exe` shell. Instead the shell provides the `del` builtin, which
has somewhat more limited wildcars expansion and error handling.
This makes all of the Makefile targets work on Win32 both using
`cmd.exe` and using the MSYS environment.
Signed-off-by: Simon Richter <Simon.Richter@hogyros.de>
Signed-off-by: Andrea Canciani <ranma42@gmail.com>
Acked-by: Oded Gabbay <oded.gabbay@gmail.com>
When the build is performed using `cmd.exe` as shell, the `mkdir`
command does not support the `-p` flag. The ability to create multiple
netsted folder is not used, hence it can be easily replaced by only
creating the directory if it does not exist.
This makes the build work on the `cmd.exe` shell, except for the
`clean` targets.
Signed-off-by: Andrea Canciani <ranma42@gmail.com>
Acked-by: Oded Gabbay <oded.gabbay@gmail.com>
Instead of explicitly depending on "pixman" for the "all" and "check"
targets, rely on the dependency to the .lib file
Signed-off-by: Andrea Canciani <ranma42@gmail.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Since 3d81d89c29 BUILT_SOURCES is not
used anymore, but it was unintentionally left in Win32 Makefiles.
Signed-off-by: Andrea Canciani <ranma42@gmail.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
This patch modifies the SSE2 & SSSE3 tests in configure.ac to use a
global variable to initialize vector variables. In addition, we now
return the value of the computation instead of 0.
This is done so gcc 4.9 (and lower) won't optimize the SSE assembly
instructions (when using -O1 and higher), because then the configure test
might incorrectly pass even though the assembler doesn't support the
SSE instructions (the test will pass because the compiler does support
the intrinsics).
v2: instead of using volatile, use a global variable as input
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Older versions of clang emitted an error on the "K" constraint, but at
least since version 3.7 it is supported. Just like gcc, this
constraint is only allowed for constants, but apparently clang
requires them to be known before inlining.
Using the macro definition _mm_shuffle_pi16(A, N) ensures that the "K"
constraint is always applied to a literal constant, independently from
the compiler optimizations and allows building pixman-mmx on modern
clang.
Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Andrea Canciani <ranma42@gmail.com>
