Otherwise we can end up with uninitialized values, this fixes following
valgrind warning:
==71705== Uninitialised byte(s) found during client check request
==71705== at 0x73B6DB8: util_bitpack_uint (bitpack_helpers.h:55)
==71705== by 0x73B6DB8: GFX11_PIPE_CONTROL_pack (gen11_pack.h:19885)
==71705== by 0x73B6DB8: iris_emit_raw_pipe_control (iris_state.c:10022)
==71705== by 0x6F93386: iris_emit_pipe_control_write (iris_pipe_control.c:97)
==71705== by 0x6FBCCAA: iris_resource_update_indirect_color (iris_resolve.c:1241)
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Rohan Garg <rohan.garg@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30990>
Otherwise we can end up with uninitialized values, this fixes following
valgrind warning:
==31283== Uninitialised byte(s) found during client check request
==31283== at 0x503E4DE: anv_batch_bo_finish (anv_batch_chain.c:345)
==31283== by 0x504220A: anv_cmd_buffer_end_batch_buffer (anv_batch_chain.c:1103)
==31283== by 0x55A0E4F: end_command_buffer (genX_cmd_buffer.c:3455)
==31283== by 0x55A0E82: gfx11_EndCommandBuffer (genX_cmd_buffer.c:3466)
==31283== by 0x11233A: ??? (in /usr/bin/vkcube)
==31283== by 0x10BDEE: ??? (in /usr/bin/vkcube)
==31283== by 0x49B5149: (below main) (in /usr/lib64/libc.so.6)
==31283== Address 0xc10c4d8 is 1,240 bytes inside a block of size 8,192 client-defined
==31283== at 0x5036EF6: anv_bo_pool_alloc (anv_allocator.c:1284)
==31283== by 0x503E0E1: anv_batch_bo_create (anv_batch_chain.c:262)
==31283== by 0x5040D3F: anv_cmd_buffer_init_batch_bo_chain (anv_batch_chain.c:868)
==31283== by 0x504F9C1: anv_create_cmd_buffer (anv_cmd_buffer.c:147)
==31283== by 0x6B718C4: vk_common_AllocateCommandBuffers (vk_command_pool.c:206)
==31283== by 0x4FB06B2: vkAllocateCommandBuffers (trampoline.c:1996)
==31283== by 0x111E6B: ??? (in /usr/bin/vkcube)
==31283== by 0x10BDEE: ??? (in /usr/bin/vkcube)
==31283== by 0x49B5149: (below main) (in /usr/lib64/libc.so.6)
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Rohan Garg <rohan.garg@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30990>
If the next renderpass uses the same depth attachment, clears it
with generic clear - ZPASS_DONE may somehow read stale values that
are apparently invalidated by CCU_INVALIDATE_DEPTH.
Fixes:
dEQP-VK.fragment_operations.early_fragment.sample_count_early_fragment_tests_depth_alpha_to_coverage_samples_2_maintenance5
dEQP-VK.fragment_operations.early_fragment.sample_count_early_fragment_tests_depth_alpha_to_coverage_samples_4_maintenance5
dEQP-VK.fragment_operations.early_fragment.sample_count_early_fragment_tests_depth_samples_2_maintenance5
dEQP-VK.fragment_operations.early_fragment.sample_count_early_fragment_tests_depth_samples_4_maintenance5
When running them with TU_DEBUG=sysmem
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30858>
Per description of VkPipelineShaderStageCreateFlags
```
VK_PIPELINE_SHADER_STAGE_CREATE_REQUIRE_FULL_SUBGROUPS_BIT specifies
that the subgroup sizes must be launched with all invocations active in
the task, mesh, or compute stage.
```
Future CTS tests will use that.
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31023>
Certain AOSP targets feature a glibc or musl-based build,
where (__ANDROID__) isn't defined, but:
- ANDROID is defined
- the relevant headers are present.
For such builds, buffer_handle_t shouldn't be defined
internally.
Reviewed-by: Roman Stratiienko <r.stratiienko@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31030>
This will allocate less LDS for LS outputs if there are holes between
varyings when we have monolithic merged LS+TCS. (it removes the holes)
There are 2 steps to this:
- add helper si_shader_lshs_vertex_stride and use it everywhere
- pass the TCS inputs_read bitmask instead of the "map" callback
to si_lower_ls_outputs_mem
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30962>
Right now we don't allocate LDS for HS inputs when all HS inputs are passed
via VGPRs.
This changes it to skip allocating exactly the HS inputs passed via VGPRs
by reducing the inputs_read mask to remove holes.
radeonsi changes to the LDS allocation will be in a different MR.
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30962>
For most runtime usages, e.g. apitrace via zink on venus, the sfb cmds
normally don't exceed 3. So a limit of 5 cmds would be enough. This
would avoid that dEQP-VK.synchronization.basic.timeline_semaphore.chain
can easily leave 700+ free cmds in the cache.
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31035>
Zink currently requires VK_EXT_queue_family_foreign to set
PIPE_CAP_DMABUF, which is one of the requirements to create a
gbm context.
v3dv already conditionally supported this extension for Android.
As it is now required for Zink in Mesa, move it to the driver's
common set.
This allows v3dv to create gbm contexts with Zink again since
this was made a stricter requirement as a side effect of
ab08b79ef7 ("gbm: use driver check for dmabuf export").
Tested with Zink on a gbm EGL application as well as sway with
the wlroots Vulkan backend, which also requires this extension.
Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30938>
