This a new matrix inversion test written from scratch to be suitable for
running in CI: quick to run and automatically detects success/failure.
This all is a result of what I learnt while working on
https://gitlab.freedesktop.org/pq/fourbyfour
Computing the residual error with infinity norm comes straight from
fourbyfour documentation on how to evaluate matrix inversion error.
Most of the hard-coded test matrices have been generated with fourbyfour
project as well, as it contains the generator code. The matrices are
hard-coded here also to make testing faster, but primarily because the
generator code needs BLAS and LAPACK, and having those as Weston
dependencies would be far too much just for this.
Now, if someone wants to modify weston_matrix stuff, we should at least
detect matrix inversion and multiplication bugs.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
This test program was useful a decade ago when weston_matrix_invert()
was being developed. It was a manual test program that ran for a certain
number of seconds and required human interpretation of numbers to see if
results were acceptable or not. Hence it was foundamentally unsuitable
for CI.
The way it generated random matrices for inversion testing was also very
naive, and it used the determinant value to determine invertability
which is completely bogus. This made it also a bad test for correctness.
Much better speed and correctness testing is implemented in
https://gitlab.freedesktop.org/pq/fourbyfour
with documented testing procedures. It has a copy of the weston_matrix
implementation.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
matrix.h is a public installed header and even used by libweston.h.
See "Rename compositor.h to libweston/libweston.h" for rationale.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Using the parent '../' path component in #include statements makes
the codebase more rigid and is redundant due to proper -I use.
Signed-off-by: Jon A. Cruz <jonc@osg.samsung.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
There was an issue recently in screen-share.c where config.h was not
being included, resulting in the wrong definition for off_t being used on
32 bit systems. I checked and I don't think this problem is happening
elsewhere, but to help avoid this sort of problem in the future, I went
through and made sure that config.h is included first whenever system
headers are included.
The config.h header should be included before any system headers, failing
to do this can result in the wrong type sizes being defined on certain
systems, e.g. off_t from sys/types.h
Signed-off-by: Andrew Wedgbury <andrew.wedgbury@realvnc.com>
We just set the input region to the bounding box of the window frame
and set the opaque region to be the opaque rectangle inside the window
if the child widget is opaque.
The compositor will likely do an order of magnitude less matrix
inversions than point transformations with an inverse, hence we do not
really need the optimised path for single-shot invert-and-transform.
Expose only the computing of the explicit inverse matrix in the API.
However, the matrix inversion tests need access to the internal
functions. Designate a unit test build by #defining UNIT_TEST, and
export the internal functions in that case.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Add a new directory tests/ for unit test applications. This directory
will be built only if --enable-tests is given to ./configure.
Add matrix-test application. It excercises especially the
weston_matrix_invert() and weston_matrix_inverse_transform() functions.
It has one test for correctness and precision, and other tests for
measuring the speed of various matrix operations.
For the record, the correctness test prints:
a random matrix:
1.112418e-02 2.628150e+00 8.205844e+02 -1.147526e-04
4.943677e-04 -1.117819e-04 -9.158849e-06 3.678122e-02
7.915063e-03 -3.093254e-04 -4.376583e+02 3.424706e-02
-2.504038e+02 2.481788e+03 -7.545445e+01 1.752909e-03
The matrix multiplied by its inverse, error:
0.000000e+00 -0.000000e+00 -0.000000e+00 -0.000000e+00
0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
-0.000000e+00 -0.000000e+00 0.000000e+00 -0.000000e+00
0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
max abs error: 0, original determinant 11595.2
Running a test loop for 10 seconds...
test fail, det: -0.00464805, error sup: inf
test fail, det: -0.0424053, error sup: 1.30787e-06
test fail, det: 5.15191, error sup: 1.15956e-06
tests: 6791767 ok, 1 not invertible but ok, 3 failed.
Total: 6791771 iterations.
These results are expected with the current precision thresholds in
src/matrix.c and tests/matrix-test.c. The random number generator is
seeded with a constant, so the random numbers should be the same on
every run. Machine speed and scheduling affect how many iterations are
run.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Pekka Paalanen
Jon Cruz
Bryce Harrington
Derek Foreman
Andrew Wedgbury
Rob Bradford
John Kåre Alsaker
Kristian Høgsberg
Pekka Paalanen