This operation could be implemented in the TP cores, but this operation
tends to be added by convertors that export to TFLite from frameworks
with different channel order, and end up being no-ops.
Once we move to NIR for tensor operations, we can support this operation
and then remove it when we have an explicit transpose operation that is
negated by a consequent transpose operation.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34629>
This old code was needed to get the backend assembler to do the
right thing when emitting index and address registers, but sfn
is handling this now so we can drop this.
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36101>
This change is inspired by 1021d6fe62 ("dri: deal
with ARGB1555")
This issue is now mostly fixed with
https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36081
Anyway, the dri3_cpp_for_fourcc entry is still missing and should
be added.
This change is useful for instance with r600 which
can handle this format.
Note: this mode was generated at the "glx visuals" level
on r600 by default before the commit d709b42180.
This change was tested on r600 palm and cayman with X11
loaded with a version of mesa generating this very mode:
glx/glx-visuals-depth -pixmap: fail pass
glx/glx-visuals-stencil -pixmap: fail pass
Fixes: 00aa095d53 ("dri: Support 1555/4444 formats")
Signed-off-by: Patrick Lerda <patrick9876@free.fr>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34294>
When the rasterizer state is updated, we only need to update
the scissoring state if the rasterizer scissor state has changed.
This avoids re-sending the same scissor state any time the rasterizer
is changed.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36352>
We need to update the CFG_BITS packet if the early_fragment_test status
changed vs previous draw call. But we don't need to update it every
time the FS is changed, we only need to update it when disable_ez
value is different from previous FS.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36352>
The programming model matches very closely to that of NVIDIA's NVDLA.
Enough is implemented to run SSDLite MobileDet with roughly the same
performance as the blob (when running on a single NPU core).
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29698>
On Rockchip, we need a tolerance of 8 to pass all tests (especifically
the whole MobileNetV1 model).
Though all other tests pass with a tolerance of just 2, 8 is still not
that high that we would risk letting bugs slip in.
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29698>
While these GUIDs make their way through the SDK, add them for now
directly, to be able to perform Input QP related development.
Reviewed-by: Pohsiang (John) Hsu <pohhsu@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36379>
For samplers the type_size() callback can return 0, which triggers
a NIR validation error.
In this case set range to ~0 which means the range is unknown.
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36263>
This traces back to c688f8f8c5, but the
shape of the fix would be different if against that. So we do the
optimal for the current code flow and only port to stable.
Cc: mesa-stable
Reviewed-by: Lucas Fryzek <lfryzek@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36335>
Add lvp_image_bind helper so that bind status is robustly handled
outside the core bind, which simplifies the exit paths. Meanwhile, the
spec doesn't require to proceed binding of image internal planes if a
prior plane binding has failed. So this change further simplifies the
non-disjoint image binding result handling.
Reviewed-by: Lucas Fryzek <lfryzek@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36335>
Per spec of VkBindImageMemorySwapchainInfoKHR:
> If swapchain is not NULL, the swapchain and imageIndex are used to
determine the memory that the image is bound to, instead of memory and
memoryOffset.
Meanwhile, common wsi is doing dedicated allocation for swapchain image
memory, so it's required to use zero memoryOffset by the spec. Then here
we can safely set to zero memoryOffset before passing to the actual
binding call.
In practice, when the struct is initialized with proper sType and memory
being VK_NULL_HANDLE, the memoryOffset is most likely left being zero
initialized. Not a critical must fix but still a bug.
Fixes: ace49d9e52 ("lavapipe: adopt wsi_common_get_memory")
Reviewed-by: Lucas Fryzek <lfryzek@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36335>
Set to true everywhere except:
- spirv_to_nir used by Vulkan
- bindless handles in GLSL
- some internal shaders and driver-specific code
Acked-by: Job Noorman <job@noorman.info>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36099>
In the dri3_handle_present_event() function, it uses an event enum
and two macros. So I suggest to change these all to use enums for
consistency.
Signed-off-by: Luc Ma <luc@sietium.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30461>
The compiled image function is derived from the image view, which is
supposed to match what is declared in the shader, however we did fixups
for array targets that only had one layer, since pipe_shader_image
actually doesn't have a target (derived from the resource instead).
But this is not correct at least for vulkan, since then the layer
coord is completely ignored, so we never got OOB behavior wrt the
layer coord.
Use single_layer_view field in pipe_shader_image to denote non-array
targets (the GL state tracker uses this in a similar way already,
although llvmpipe ignored it).
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Brian Paul <brian.paul@broadcom.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36304>
As we are already updating tmu_dirty_rcl based on the the
shader usage of tmu writes at v3d_emit_gl_shader_state we
can avoid setting it everytime we have a SSBO or image
attached.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36343>
Replace incorrect MIN2 clamping with proper 5.8 signed fixed point
encoding. The hardware expects LOD values in 5.8 format with a range
of -16.0 to +15.99609375. Clamp input values to this valid range
before conversion to handle overflow correctly.
Passes dEQP-GLES3.functional.texture.mipmap.*.max_lod.* on GC7000.
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/36303>
