Instead of having genX(emit_apply_pipe_flushes) doing the clearing,
ask genX(emit_apply_pipe_flushes) for the emitted bits and do the
clearing using a helper.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Rohan Garg <rohan.garg@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23675>
In the following sequence :
- write buffer B with a shader
- barrier on buffer from shader-write to transfer
- vkCmdCopyQueryPoolResults to buffer B
The barrier should take care of ordering things between the shader
writes and vkCmdCopyQueryPoolResults.
The problem is that vkCmdCopyQueryPoolResults runs on the command
streamer and that is not coherent or synchronized in the same way as
shaders.
This change marks the barrier has potentially containing pending
buffer writes for queries so that we can insert the necessary flush
for vkCmdCopyQueryPoolResults later.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9013
Cc: mesa-stable
Reviewed-by: Rohan Garg <rohan.garg@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23675>
In C the size of a struct { uin32_t a; uint8_t b; } is 8, not 5, so we have
to account for the biggest alignment across all struct members.
Funny that the OpenCL CTS doesn't catch that.
Fixes: 44d32e62fb ("glsl: add cl_size and cl_alignment")
Signed-off-by: Karol Herbst <git@karolherbst.de>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23701>
As c11 already provided timespec_get, to avoid symbol conflict, use c23_timespec_get as the function name
and define timespec_get c23_timespec_get to achieve that
Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23733>
src/freedreno/vulkan/tu_pipeline.c:4723:25: error: passing argument 1 of ‘pthread_mutex_unlock’ from incompatible pointer type [-Werror=incompatible-pointer-types]
4723 | pthread_mutex_unlock(&dev->pipeline_mutex);
| ^~~~~~~~~~~~~~~~~~~~
| |
| mtx_t *
In file included from ../../src/freedreno/vulkan/tu_common.h:14,
from ../../src/freedreno/vulkan/tu_pipeline.h:13,
from ../../src/freedreno/vulkan/tu_pipeline.c:10:
/usr/include/pthread.h:835:51: note: expected ‘pthread_mutex_t *’ but argument is of type ‘mtx_t *’
835 | extern int pthread_mutex_unlock (pthread_mutex_t *__mutex)
Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Acked-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23733>
This happens if we pass our own validator options. It's nothing we can
control, SPIRV-Tools just passes NULL instead of "input".
Signed-off-by: Karol Herbst <git@karolherbst.de>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23700>
The whole point of vkGetDeviceImageMemoryRequirementsKHR is to avoid
creating an image so we should completely avoid any allocation like
the private binding.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: 4075dd16ab ("anv: implement vkGetDeviceImageMemoryRequirementsKHR")
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23720>
Also dropping the 3D image write special case, because there is nothing in
the spec treating it differently.
Signed-off-by: Karol Herbst <git@karolherbst.de>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Nora Allen <blackcatgames@protonmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23714>
This moves the expensive to_pipe_format call to image creation time.
Signed-off-by: Karol Herbst <git@karolherbst.de>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Nora Allen <blackcatgames@protonmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23714>
This pass uses a safe iterator so it can't lower new instructions that are
injected as part of the lowering, which is exactly what lower_tg4_offsets
does, and if lower_tg4_broadcom_swizzle is also set then we need to lower
these new instructions. Handle this by running the pass twice when both
are set: the first pass will only handle lower_tg4_offsets and
the second pass (which will see the new tg4 instructions produced with
lower_tg4_offsets) will process the remaining options, including
lower_tg4_broadcom_swizzle.
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23616>
With RADV_DEBUG=noibs (aka no chaining) the number of IBs to submit
depends on the number of old IB buffers of every CS.
This fixes a stack smashing error.
Fixes: 53b439d24f ("radv/amdgpu: Use STACK_ARRAY for IB array to reduce stack usage."
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23669>
There is no need to split the CF if only the register ID
in a previous write is the same, we should look at the actual
slots instead, ut we have also to take writes of 0 and 1 into
account.
Cc: mesa-stable
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23710>
The way it's used is unacceptable. Here's the change it suggested:
diff --git a/src/amd/vulkan/radv_device.c b/src/amd/vulkan/radv_device.c
index 05b9cb0a3f1..e58abfa885f 100644
--- a/src/amd/vulkan/radv_device.c
+++ b/src/amd/vulkan/radv_device.c
@@ -783,8 +783,7 @@ radv_CreateDevice(VkPhysicalDevice physicalDevice, const VkDeviceCreateInfo *pCr
device->overallocation_disallowed = overallocation_disallowed;
mtx_init(&device->overallocation_mutex, mtx_plain);
- if (physical_device->rad_info.register_shadowing_required ||
- device->instance->debug_flags & RADV_DEBUG_SHADOW_REGS)
+ if (physical_device->rad_info.register_shadowing_required || device->instance->debug_flags & RADV_DEBUG_SHADOW_REGS)
device->uses_shadow_regs = true;
/* Create one context per queue priority. */
This is the dumbest reason to prevent merging a MR. I don't want to see this
in projects I work on.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23687>
