Wladimir J. van der Laan authored 7 years ago and Robert Foss committed 7 years ago

Experiemental NEON implementation of tiling/untiling that adds
specialized NEON function for tiling 8, 16, 32 bit per element
4x4 tiles. To optimize memory read/write sizes, there are functions that
process multiple horizontically adjactent tiles as well. These
are automatically used for the part that consists of whole tiles.

To make this work, Mesa needs to be compiled with CFLAGS -mfpu=neon,
otherwise a compile error will be thrown.

Signed-off-by: Wladimir J. van der Laan <laanwj@gmail.com>

Aleksander Morgado authored 7 years ago and Robert Foss committed 7 years ago

Despite being a member of the etna_screen struct, 'refcnt' is used by
the winsys-specific logic to track the reference count of the object
managed in a hash table. When the count reaches zero, the pipe screen
is removed from the table and destroyed.

Fix the logic by initializing the refcnt to 1 when screen created.
This initialization is done in etna_screen_create(), to follow the
same logic as in freedreno and virgl.

Signed-off-by: Aleksander Morgado <aleksander@aleksander.es>

Aleksander Morgado authored 7 years ago and Robert Foss committed 7 years ago

The 'opencl_func_mutex' is destroyed in dri_destroy_screen_helper()
unconditionally, so make sure it is always initialized to avoid any
undefined behaviour.

Signed-off-by: Aleksander Morgado <aleksander@aleksander.es>

Aleksander Morgado authored 7 years ago and Robert Foss committed 7 years ago

The check for the pointer being non-NULL was being done too late.

Signed-off-by: Aleksander Morgado <aleksander@aleksander.es>

Signed-off-by: Robert Foss <robert.foss@collabora.com>

Add support for buffer object modifiers to the Android platform.

Signed-off-by: Robert Foss <robert.foss@collabora.com>

Signed-off-by: Robert Foss <robert.foss@collabora.com>

Signed-off-by: Robert Foss <robert.foss@collabora.com>

This addition fixes broken local copies of the definitions gbm_gralloc structs.

Signed-off-by: Robert Foss <robert.foss@collabora.com>

Rob Herring authored 9 years ago and Robert Foss committed 7 years ago

gralloc_drm_get_gem_handle has been removed from AOSP drm_gralloc.
Remove the dependency on it and just access the struct
gralloc_drm_handle_t directly. Maybe there's a better way than accessing
the gralloc private data.

Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Robert Foss <robert.foss@collabora.com>

Rob Herring authored 8 years ago and Robert Foss committed 7 years ago

Signed-off-by: Rob Herring <robh@kernel.org>

WuZhen authored 9 years ago and Robert Foss committed 7 years ago

Redirect logs printed to stderr to logcat.

NO_REF_TASK
Tested: local run

Change-Id: I58e3966a608af361b86c54b4c95a92561b711968
Signed-off-by: Chih-Wei Huang <cwhuang@linux.org.tw>

Rob Herring authored 9 years ago and Robert Foss committed 7 years ago

This is required by freedreno at least for GLES3 support.

See docs/patents.txt for information about turning this on for s/w
renderers.

Signed-off-by: Rob Herring <robh@kernel.org>

Rob Herring authored 9 years ago and Robert Foss committed 7 years ago

Add support for 32-bit RGBX/RGBA formats which are preferred for Android.

Note: This is reverted in mesa as it breaks X11.

Signed-off-by: Rob Herring <robh@kernel.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Eric Anholt <eric@anholt.net>

Tomasz Figa authored 8 years ago and Robert Foss committed 7 years ago

Make the code at least compile when being built without drm_gralloc
headers.

v2: Replaced #ifdefs with stubs for gralloc_drm_get_gem_handle()
    and GRALLOC_MODULE_PERFORM_GET_DRM_FD.
    Removed explicit render node probing code.

Signed-off-by: Tomasz Figa <tfiga@chromium.org>
Signed-off-by: Rob Herring <robh@kernel.org>

Lucas Stach authored 7 years ago and Robert Foss committed 7 years ago

Push this format to the pipe driver unchanged.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
(cherry picked from commit 667554dc)

Lucas Stach authored 7 years ago and Robert Foss committed 7 years ago

For resources that need fast CPU access it is beneficial to keep a shadow
copy in cached memory. We don't switch the BO to cached, as benchmarks
have shown that the needed cache maintenance on cached BOs is marginally
slower than the current unoptimized tile upload. With some more
optimizations to the upload path, the performance advantage of the copy
will likely be even bigger.

Currently this is only done for sampler resources, so we don't have to deal
with buffer content changed by the GPU, which would require even more seqno
tracking.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>

Lucas Stach authored 7 years ago and Robert Foss committed 7 years ago

This allows the pipe driver to optimize the resource allocation by providing
accurate binding and expected CPU access pattern.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>

Lucas Stach authored 7 years ago and Robert Foss committed 7 years ago

So the DRI driver can optimize the buffer allocation.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>

Lucas Stach authored 7 years ago and Robert Foss committed 7 years ago

For now all DRI images implicitly assume both render and texture usage,
as well as no (or slow) CPU access. Creating DRI images is the only way
for GBM to allocate buffers. With gbm_gralloc the usage of the buffer
might be something wildly different from the current default, so DRI
image need to grow some flags to describe those.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>

Lucas Stach authored 7 years ago and Robert Foss committed 7 years ago

With the advent of gbm_gralloc, GBM buffers aren't exclusively used as
render/scanout/cursor BOs anymore. The current flags are insufficient
to describe the real usage, so add some flags to capture gralloc
use-cases.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>

Lucas Stach authored 7 years ago and Robert Foss committed 7 years ago

Simply advertise all supported modifiers, independent of the format.
Special formats, like compressed, which don't support all those modifiers
are already culled from the dmabuf format list, as we don't support
the render target binding for them.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
---
v2: don't advertise split tiled formats on single pipe/buffer GPUs

Lucas Stach authored 7 years ago and Robert Foss committed 7 years ago

This allows to create buffers with a specific tiling layout, which is primarily
used by GBM to allocate the EGL back buffers with the correct tiling/modifier
for use with the scanout engines.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>

Lucas Stach authored 7 years ago and Robert Foss committed 7 years ago

This allows the state trackers to know the tiling layout of the
resource and pass this through the various userspace protocols.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Wladimir J. van der Laan <laanwj@gmail.com>

Lucas Stach authored 7 years ago and Robert Foss committed 7 years ago

There is no point in keeping this indirection. Makes the code easier to
follow.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Wladimir J. van der Laan <laanwj@gmail.com> (v1)

Lucas Stach authored 7 years ago and Robert Foss committed 7 years ago

This implements resource import with modifier, deriving the correct
internal layout from the modifier and constructing a render compatible
base resource if needed.

This removes the special cases for DDX and renderonly scanout allocated
buffers, as the linear modifier is enough to trigger correct handling
of those buffers.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Wladimir J. van der Laan <laanwj@gmail.com>

Lucas Stach authored 7 years ago and Robert Foss committed 7 years ago

This reworks the logic in etna_update_sampler_source to select the
newest resource view for updating the texture view. This should make
the logic easier to follow and fixes texture updates from imported
dma-bufs.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Wladimir J. van der Laan <laanwj@gmail.com>

Lucas Stach authored 7 years ago and Robert Foss committed 7 years ago

If we import a dma-buf with a sampler/pixel pipe incompatible modifier,
the imported buffer will end up in an external resource view. As
resource_changed signals the change of the imported resource, we need
to update the external view seqno, instead of the base resource seqno.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Wladimir J. van der Laan <laanwj@gmail.com>

Lucas Stach authored 7 years ago and Robert Foss committed 7 years ago

This fixes failures to import the scanout buffer with screen resolutions
that don't satisfy the RS alignment restrictions, like 1680x1050.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Wladimir J. van der Laan <laanwj@gmail.com>

Lucas Stach authored 7 years ago and Robert Foss committed 7 years ago

The minimum RS alignment calculation is needed in various places.
Extract a helper to avoid open-coding the calcuation at every site.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>

Lucas Stach authored 7 years ago and Robert Foss committed 7 years ago

The current way of importing the resource from renderonly after allocation
is opaque and is taking away control from the driver, which it needs in
order to implement more advanced scenarios, than the simple linear
scanout with matching stride alignments.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>