Let's be a little less conservative then QEMU -- fence only
on draw calls and possible fence creations.
Can we do even better, and fence only when Gallium requests
VIRGL_BIND_CUSTOM? Someone should look into this, but let's
get this landed to unblock the CQ.
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Signed-off-by: Jakob Bornecrantz <jakob@collabora.com>
Include stdint.h in virgl_hw.h, since it uses uint32_t.
Remove now duplicate structure virgl_box in vtest_renderer.c
Reviewed-by: Jakob Bornecrantz <jakob@collabora.com>
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Signed-off-by: Jakob Bornecrantz <jakob@collabora.com>
Refactor the main loop into smaller functions so that it is easier to extend
with more functionality.
There should be no functional change with the exception of some socket
listening code would return error codes as fds, where they now instead call
perrno and exit(1).
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Jakob Bornecrantz <jakob@collabora.com>
Turns out some files managed to have a mix between 2, 3, 4 and tabs for
indenting. The only way to really fix this was to do just a single commit
that fixes all of them.
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>
Signed-off-by: Jakob Bornecrantz <jakob@collabora.com>
run_renderer now uses an array to get the corresponding function
pointer for a command.
Some vtest_* function headers had to be uniformized.
Reviewed-by: Jakob Bornecrantz <jakob@collabora.com>
Signed-off-by: Nathan Gauer <nathan@gauer.org>
Signed-off-by: Jakob Bornecrantz <jakob@collabora.com>
With the main aim of making vtest's usage of virglrenderer more similar
to that of QEMU, this patch sets a backing store to each resource and
reads from or writes to at each transfer operation.
v2: - Instead of adding a callback on resource destruction, free the
iovec when we unref a resource as we know that we won't be doing
anything else with it. (Dave Airlie)
v3: - Store pointers to iovecs in the global to also release them
v4: - Add new commands for resource creation and transfers, to be used
when protocol version >= 1.
v5: - Replaced the global array of iovecs with a hash table because some
es31 tests create thousands of resources and reach any sane limit.
Tested-by: Jakob Bornecrantz <jakob@collabora.com>
Reviewed-by: Jakob Bornecrantz <jakob@collabora.com>
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Signed-off-by: Jakob Bornecrantz <jakob@collabora.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This commit negotiates a common protocol version with the clients,
allowing for backwards compatibility on both sides.
v2: - Negotiate the version in a new command, so the client can probe
for its existence and stay compatible with older servers. (Dave
Airlie)
Tested-by: Jakob Bornecrantz <jakob@collabora.com>
Reviewed-by: Jakob Bornecrantz <jakob@collabora.com>
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Signed-off-by: Jakob Bornecrantz <jakob@collabora.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
As we don't have yet a proper way of negotiating a common protocol
version to use between clients and the server, we introduce a command
without reply arguments that allows us to figure out if the server is
new enough to handle negotiation of the protocol version.
Tested-by: Jakob Bornecrantz <jakob@collabora.com>
Reviewed-by: Jakob Bornecrantz <jakob@collabora.com>
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Signed-off-by: Jakob Bornecrantz <jakob@collabora.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
1) We use the preprocessor flag HAVE_EPOXY_GLX_H, but never
set it.
2) WITH_GLX diverges with HAVE_EPOXY_EGL_H.
3) virgl_test_server needs to link with X11 when using a GLX
backend.
To fix this, query epoxy for GLX support and allow the user to
disable it as well.
TEST=./autogen.sh --disable-egl
Make sure the next renderer is a cleaned up one.
Reviewed-by: Jakob Bornecrantz <jakob@collabora.com>
Signed-off-by: Jakob Bornecrantz <jakob@collabora.com>
This is usefull for running vtest under apitrace.
Signed-off-by: Jakob Bornecrantz <jakob@collabora.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Adds a VIRGL_RENDERER_USE_GLES flag for specifying the API to request
when creating the EGL context.
Users can enable this flag when using vtest with the VTEST_USE_GLES
environment variable.
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
To make it easier to run the test suite on environments without graphics
hardware, add an environment variable VIRGL_EGL_SURFACELESS that will
force the use of the surfaceless platform.
Combined with the following flags, we can run virglrenderer and its
tests without any graphics hardware or windowing system present:
LIBGL_ALWAYS_SOFTWARE=true GALLIUM_DRIVER=llvmpipe VIRGL_EGL_SURFACELESS=yes
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
I've come up with a workaround on the mesa side that should be
backwards compatible with old vtests and vice-versa.
It involves sending both caps 2 and 1 queries back to back,
and taking the first response an indication of what to expect,
v1.1: Check max_size is > 0 before mallocing
Reviewed-by: Jakob Bornecrantz <jakob@collabora.com>
This is useful to do some quite accurate comparative
benchmarks/profiling, and to run fuzzers, such as american fuzzy lop.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Just as bad as waiting only for the last fence, after all.
Signed-off-by: Marc-André Lureau <marcandre.lureau@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
For the same glmark2 "build" test, perf reports about half
instructions.
Signed-off-by: Marc-André Lureau <marcandre.lureau@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Gurchetan Singh
Gert Wollny
Elie Tournier
Jakob Bornecrantz
Dave Airlie
Nathan Gauer
Tomeu Vizoso
Marc-André Lureau
Marc-André Lureau
Dave Airlie