The latency is from LLVM's SISchedule.td
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Acked-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29162>
Similar to the prior fix for msm. On the dmabuf import path, tu_bo_init
can be outside of the vma lock, but left inside for code simplicity.
Fixes: f17c5297d7 ("tu: Add virtgpu support")
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29093>
There's a success path on found dmabuf while the iova won't be cleaned
up. This change defers iova alloc till lookup miss and also to prepare
for later racy dmabuf re-import fix.
Also documented a potential leak on error path due to unable to tell
whether a gem handle should be closed or not without refcounting.
Fixes: f17c5297d7 ("tu: Add virtgpu support")
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29093>
This should be similar to the previous ctx_cs logic that was
implemented in order to reduce context rolls but this also applies
to shader objects.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28983>
This allows us to unify emitting monolithic graphics pipelines and
shader objects.
Though, this temporarily reduces performance in some games due to more
context rolls. But this will be fixed in the following commits by
using the recent mechanism to track context register writes.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28983>
Signed-off-by: tarsin <yuanqingxiang233@163.com>
[rob: various fixes for android-cts failures]
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Roman Stratiienko <r.stratiienko@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29090>
Signed-off-by: tarsin <yuanqingxiang233@163.com>
[rob: improve comments about modifier taking precedence over image
create flags, and fix bogus assert]
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Roman Stratiienko <r.stratiienko@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29090>
Signed-off-by: Roman Stratiienko <r.stratiienko@gmail.com>
Tested-by: tarsin <yuanqingxiang233@163.com>
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Roman Stratiienko <r.stratiienko@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29090>
This change is required to enable common ANB functionality,
which the following patch adds.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Roman Stratiienko <r.stratiienko@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29090>
GetPhysicalDeviceImageFormatProperties() must that an image {format,
flags, usage} combo is unsupported if gralloc will not be able to
perform the allocation. The practical way to test this is to do a
small test allocation.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Roman Stratiienko <r.stratiienko@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29090>
If p_start_linear_vgpr allocates a VGPR that is already blocked, RA
will try moving the blocking VGPR somewhere else. If
p_start_linear_vgpr is inserted right before the branch, that move will
be inserted after exec has been overwritten, which might cause the move
to be skipped for some threads.
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28041>
The layout for dyn UBOs changed but we were still cleaning right after
UBOs.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Fixes: 9c553bda9c ("panvk: Prepare dynamic buffer descriptors at bind time")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29191>
so that users don't have to shift it at every use. It was supposed to be
like this from the beginning.
Fixes: fb994f44d9 - util: make BITSET_TEST_RANGE_INSIDE_WORD take a value to compare with
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29187>
Tiled scan-out buffer works only for those platforms supporting
set_tiling/get_tiling ioctl, which is not used for newer platforms
(e.g., dGPU). This change switch to querying modifier reliably with
gralloc API.
Signed-off-by: Weifeng Liu <weifeng.liu@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29185>
This matches anv and specification behavior, change makes us toggle the
field in tests/application that may use layered framebuffer but do not
assign any value to gl_Layer.
Example of such a test:
KHR-GL46.geometry_shader.layered_rendering_boundary_condition.layered_rendering_boundary_condition_no_default_layer
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29064>
