Saved time is around 3 - 4 minutes per job,
decreasing runtime from ~ 19 minutes to ~ 15 minutes.
Acked-by: Daniel Stone <daniels@collabora.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23964>
Instead of manually indexing a single-dimensional array as 2-dimensional
(and using the wrong stride for the outer array) just actually make it
a 2-dimensional array.
Fixes: 7edae456 ("d3d12: Track up to 16 contexts worth of batch references locally in bos")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24041>
According to Bspec, COMPCS0_CCS_AUX_INV register offset
is 042C8h and COMPCS0_AUX_TABLE_BASE_ADDR is defined to 042C0h.
Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23958>
According to Bspec, COMPCS0_CCS_AUX_INV register offset
is 042C8h and COMPCS0_AUX_TABLE_BASE_ADDR is defined to 042C0h.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23958>
nir_builder_at requires a block to chase back to the function impl,
but for an empty function impl, the previous code produced a cursor
with a null pointer. It was also just extra complicated.
While I'm here, use the new foreach helper since this code needs an impl.
Fixes: 12a268ea ("microsoft: Use nir_builder_at")
Reviewed-by: Eric Engestrom <eric@igalia.com>
Acked-by: Daniel Stone <daniels@collabora.com>
Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24042>
This is a regex, not a glob, so `dEQP-VK.ssbo.phys.layout.3*` matched
all of `dEQP-VK.ssbo.phys.layout`.
Signed-off-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Juan A. Suarez <jasuarez@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24018>
This is a regex, not a glob, so `spec@!opengl es 3*` matched all of
`spec@!opengl es `.
Signed-off-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Juan A. Suarez <jasuarez@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24018>
Avoids the generation of sampler views as they are unnecessary
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23159>
Use a lightweight container for ping, ssh, curl and bash support.
Also use an image located at fd.o infrastructure, since we are having
some issues with Collabora's gitlab one lately.
Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23534>
The pipe cap isn't really accurate since shaders created by lavapipe take
a different codepath compared to GL. The format information is part of
the descriptor.
This allows vkd3d-proton to advertise typed UAV loads. (Whatever that
means)
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24014>
Some drivers might return 0 indicating they haven't written any data. But
treating it as 0 for all current `COMPUTE_CAP`s is fine, so no need to
panic at runtime here.
Signed-off-by: Karol Herbst <git@karolherbst.de>
Acked-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Nora Allen <blackcatgames@protonmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22893>
This will be required for `cl_intel_required_subgroup_size`, but it
already helps implementing OpenCL subgroups as this allows us to check
with every subgroup size when implementing
`CL_KERNEL_LOCAL_SIZE_FOR_SUB_GROUP_COUNT`.
Signed-off-by: Karol Herbst <git@karolherbst.de>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22893>
Same as DIV_ROUND_UP
Signed-off-by: Karol Herbst <git@karolherbst.de>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Nora Allen <blackcatgames@protonmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22893>
This will be required by drivers supporting multiple subgroup sizes with
a given CSO to properly implement OpenCL subgroups.
Signed-off-by: Karol Herbst <git@karolherbst.de>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22893>
OpenCL 3.0 core requires __opencl_c_subgroups to be set, the OpenCL
cl_khr_subgroups extenions can only be enabled if and only if the driver
guarentees independent forward progress between subgroups.
See CL_DEVICE_SUB_GROUP_INDEPENDENT_FORWARD_PROGRESS for more information.
Signed-off-by: Karol Herbst <git@karolherbst.de>
Reviewed-by: Nora Allen <blackcatgames@protonmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22893>
The BLT blit currently assumes that all blits with src samples > 1
are downsampling blits. This isn't correct, as we can also blit
between two multisampled resources with src and dst samples > 1.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <cgmeiner@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23595>
When blitting between two multisampled resources, the blit region needs
to be scaled by the number of samples. Currently this is only done
correctly for downsampling blits where src samples > dst samples. Fix
it to also work as expected when src samples == dst samples.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <cgmeiner@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23595>
Width and height of the resource level are stored in pixels, while
the padded dimensions are stored in samples. Move the members to
make them line up with the comments.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <cgmeiner@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23595>
If an array is accessed indirectly, add a buffer live range of
one cycle before (for write) or after (for read), instead of pinning
the array from the start.
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24036>
If we try to re-use the AR register after a new CF was started
we will get into trouble, because it is no longer loaded, so
clear the AR register handle when we hit a non-ALU instruction
while scanning the shader to split AR loads. With that
dependencies will be set accodingly.
Fixes: d21054b4bc
r600/sfn: Add pass to split address and index register loads
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24036>
On Pre-Evergreen hardware we have a flag
"Force Clamp X,Y policy to wrap for CubeMaps"
but it doesn't seem to affect how border clamping is done. With
bf3027 this is set to PIPE_TEX_WRAP_CLAMP_TO_EDGE for cube maps,
and results in the regression reported in #9028.
Forcing repeat mode fixes the issue.
Fixes: bf3027c391
mesa/st: Normalize wrap modes for seamless cubes
Related: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9028
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23848>
