The i915 is able to special case gradients with just 2 color stops to
avoid creating temporary gradient textures, so add a 3 stop linear
gradient to compare the speed difference.
Cover the similar source with min/mag scale factors as well, so we can
compare the performance impact with scaled image sources. This is useful
to distinguish between transport overhead and transform cost.
Performing the unicode to index is quite expensive, the
FcFreeTypeCharIndex() taking over 12% in the cairo-perf text benchmarks.
By adding a simple cache of translated unicode indices, we save around 25%
of the lookups during benchmarks, with a relative reduction in runtime.
A filled equivalent of stroke-image, that checks that the pattern
matrices are applied correctly during fills - useful with the
segregation between fills and strokes introduced by spans.
This test attempts to trigger failures in those backend clone_similar
methods that have size restrictions on the resulting image. It also
triggers errors in scaling down large image surfaces as the image
backend also fails this test.
RTLD_DEFAULT is a gnu-ism (at least according to the manpage on my linux
system) so declare _GNU_SOURCE before including dlfcn.h and failing that
provide our own definition.
If we cannot test the xlib backend simply because there is no Display,
just report UNTESTED and do not clutter the output with superfluous
warnings [see the output from the buildbots for an example]. However,
keep the warnings around so that a developer can re-enable them
and so simply move them to a new "lower priority" macro.
As reported in https://bugs.freedesktop.org/show_bug.cgi?id=19283, the
fallback freetype version compare is broken inside the configure script as
the $1-$3 arguments are interpreted as the script is constructed. To avoid
making that awk comparison any more complicated, we import a version compare
from the autoconf archives - such that we have a reusable macro for the
furture.
In order to have locale-independent output of decimal values, we need to
manually transform such numbers into strings. As this is a solved problem
for cairo, we adopt _cairo_output_stream_printf() and in particular the
_cairo_dtostr() routine for our own printf processing.
Profiling a silly video renderer that called set-source; rectangle; fill;
for each pixel, we can shave 5% off the cairo overhead by introducing an
early return if we attempt to reset the current colour.
This interferes with the application being traced. It is not clear from
printf(3) whether "%.f" is locale dependent or not - but until we have a
failure do not break applications unnecessarily!
Handle the case of tracing an application that spawns it own graphical
children but using the autonaming facility within cairo-trace. Currently
the traced process tree would all attempt to write to the same file,
creating a broken trace. This means sacrificing the display of the output
name, but allows use for a wider range of applications.
