Shaders used in set_buffer_layer will apply colorspace conversion,
so we need to set correct matrix depending on the input and output
surface formats.
Use BT.601 (default) for YUV to RGB and identity for RGB to RGB.
Reviewed-by: Leo Liu <leo.liu@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24626>
We don't need the path, not at all when we use external kernel.
Acked-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Christian Gmeiner <cgmeiner@igalia.com>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24646>
curl is already installed in these images, drop it.
Acked-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Christian Gmeiner <cgmeiner@igalia.com>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24646>
In different traces for different GPU models I see the same pattern:
0x00c80000, /* [00A00] PA.VIEWPORT_SCALE_X = 200.000000 */
0xff880000, /* [00A04] PA.VIEWPORT_SCALE_Y = -120.000000 */
etna_f32_to_fixp16(..) handles the negative case differently then blob. In the
concrete example -120 gets converted to 0xff880001. Switch to S_FIXED(..) to
simplify our code and to get the same results as blob.
Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24364>
Lets have a look at trace from blob (GC3000):
0x000b000b, /* [10080] NTE.SAMPLER[0].SIZE := WIDTH=11,HEIGHT=11 */
0x2001b86e, /* [10100] NTE.SAMPLER[0].LOG_SIZE := WIDTH=3.437500,HEIGHT=3.437500,ASTC=0,INT_FILTER=1,SRGB=0 */
If we ignore the WIDTH and HEIGHT outputs and only look at the raw value (0x2001b86e) we can see that 0x6e is used for WIDTH and HEIGHT.
Lets have a look at an other texutre size 180x180:
0x00b400b4, /* [10080] NTE.SAMPLER[0].SIZE := WIDTH=180,HEIGHT=180 */
0x2003bcef, /* [10100] NTE.SAMPLER[0].LOG_SIZE := WIDTH=7.468750,HEIGHT=7.468750,ASTC=0,INT_FILTER=1,SRGB=0 */
Lets use the same test on a different GPU that uses texture descriptors (GC7000):
[44] SIZE: 0x000b000b WIDTH=11,HEIGHT=11
...
[74] LOG_SIZE_EXT: 0x03750375 WIDTH=3.457031,HEIGHT=3.457031
If we ignore the WIDTH and HEIGHT outputs and only look at the raw value (0x03750375) we can see that 0x375 is used for WIDTH and HEIGHT.
Lets have a look at an other texutre size 180x180:
[44] SIZE: 0x00b400b4 WIDTH=180,HEIGHT=180
..
[74] LOG_SIZE_EXT: 0x077d077d WIDTH=7.488281,HEIGHT=7.488281
etnaviv creates the following values for the two texture sizes:
11 -> 0x6f
180 -> 0xf0
Lets switch to U_FIXED(..) which results in equal values.
Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24364>
Drop our custom calculation of the logarithm base 2 and switch
to log2f(..).
Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24364>
With certain build configuration that value can be a non empty string and
needs to be used.
This will also require distributions to rebuild mesa if and only if
CLANG_RESOURCE_DIR changes between clang rebuilds or updates.
Signed-off-by: Karol Herbst <git@karolherbst.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23981>
Even with direct descriptors, our current implementation works by
doing A64 messages to read/write dynamic buffers. This is so that we
can apply the sliding range view of the descriptor while having
robustness support.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: 06dfd216d3 ("anv: add direct descriptor support to apply_layout")
Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24256>
This is to inform you of some planned downtime in the LAVA lab as follows:
* Start: 2023-08-14 08:00 BST (UTC+1)
* End: 2023-08-14 12:00 BST (UTC+1)
Signed-off-by: Sergi Blanch Torne <sergi.blanch.torne@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24424>
This both gets rid of its use of nir_ssa_dest_init() but also will make
it easier to mechanically remove nir_dest entirely later.
Acked-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24658>
This both gets rid of its use of nir_ssa_dest_init() but also will make
it easier to mechanically remove nir_dest entirely later.
Acked-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24658>
Catching the cases that Coccinelle missed. Mostly in C++ files using
gtest which causes Coccinelle to just give up and walk away.
Acked-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24658>
The buffer data is not directly accessible to application and it's
internally used to only store VACodedBufferSegment struct.
Ignore the size requested by application and instead allocate
sizeof(VACodedBufferSegment). Use calloc to zero out the struct.
This can save significant amount of memory, for example FFmpeg
will request up to tens of MB for single buffer.
Cc: mesa-stable
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6462
Reviewed-by: Thong Thai <thong.thai@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24410>
GL uses this for adding clip distances to shaders that are missing them,
but the vulkan spec guarantees "A shader must write a single clip distance
for each enabled clip half-space"
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24657>
