Commit graph

56 commits

Author SHA1 Message Date
2e06b84e27
vulkan: do not reinvent a queue context struct
We recently introduced a public field which was a superset
of the queue context we used to have.

Switch to using it entirely.

This also allows us to get rid of the NIH function which was
valid only for video queues.
2024-12-23 04:25:09 +09:00
157cd820ad
vulkan: remove pointless mutex locks
This code was simply incorrect through and through. It did not
protect what actually has to be protected in a multi-threaded setup.
Perhaps it was used to silence threading errors?

Either way, remove it, and document the correct way to use execution
pools in a threaded environment.
2024-12-23 04:25:09 +09:00
187fd52864
vulkan: fix use of atomics for the current context index
The code used to use atomic, but over time, this got broken.
This commit also remmoves the is-the-last-submission-ready
shortcut, which rarely did anything.
There's also value in relying on the fact that contexts
always carry their frames in a strictly incremental order
with no gaps.
2024-11-28 01:29:21 +09:00
8ac49ae9e0
vulkan: add opaque field to execution contexts
Helps keep track of what dispatch happened where.
2024-11-26 14:14:14 +01:00
1876026f83
vulkan: add ff_vk_exec_add_dep_sw_frame
Some software frames may be mapped, and we'd like to have
them as proper dependencies.
2024-11-18 07:54:20 +01:00
c918b42dcd
vulkan: retrieve Vulkan 1.1 properties
Required to know the subgroup size.
2024-11-18 07:45:46 +01:00
eb8f3b8460
hwcontext_vulkan: fix planar RGB images
They were non-working for quite a while.
2024-11-18 07:45:41 +01:00
4b128de44a
vulkan: enable selecting a compatible representation of format
When using **integer** images inside shaders, it turns out
that conversion doesn't automatically happen, but we need to
explicitly use the imageviews to get the image exposed as
a suitable representation for the shader.

Finally enables bitexact image representations.
2024-10-16 12:48:15 +02:00
c982376076
vulkan: extend ff_vk_shader_rep_fmt to be useful for bitexactness
The original either reported 8 or 16-bit conversion from the
original, rather than being able to return the actual original.

This makes it usable in a situation where preserving exactness
is required.
2024-10-11 23:54:18 +02:00
37d5cb84e8
vulkan: check if current buffer has finished execution before picking another
This saves resources, as dependencies are freed/reclaimed with a lower latency,
and provies a speedup.
2024-10-04 10:10:42 +02:00
877c5a9692
vulkan: use shader objects if supported
Shader objects finally allow completely independent shaders.
2024-10-04 10:10:41 +02:00
d233d9902f
vulkan: move shader data execution state to execution pools
This finally permits using fully compiled shaders across
multiple execution contexts.
2024-10-04 10:10:41 +02:00
0a37d5a3b1
vulkan: merge FFVkSPIRVShader and FFVkPipeline into FFVkShader
Pipelines are just shaders. There's no reason to treat them
differently.
This also lets us implement shader objects and is an overall
cleanup.
2024-10-04 10:10:36 +02:00
bc36fe6f1f
vulkan: use push descriptors where possible
Push descriptors are in theory slightly faster, but come with
limitations for which we have to check.

Either way, they're not difficult to implement, so even though
no one should be using peasant-tier descriptors, do it anyway.
2024-09-23 13:41:07 +02:00
8a7af4aa49
vulkan: add support for regular descriptor pools
This permits:
 - The use of Vulkan filtering on many more devices
 - Better debugging due to lack of descriptor buffer support in layers

Much of the changes here are due to a requirement that updates to
descriptors must happen between the command buffer being waited on,
and the pipeline not being bound.

We routinely did it the other way around, by updating only after
we bind the pipeline.
2024-09-23 13:40:38 +02:00
3d75ba7495
vulkan: separate out descriptor layouts from sets
Just avoids a single temporary allocation.
2024-09-23 13:40:38 +02:00
2395444c80
vulkan: add ff_vk_exec_add_dep_bool_sem
This function simply takes in a binary semaphore as a dependency
to an execution.
2024-09-22 02:11:08 +02:00
5e9845f11e
vulkan(_decode): fix, simplify and improve queries
The old query code never worked properly, and did some hideous
heuristics to read the status bit, and work that into a return
code.
This is all best left to callers to do, which simplifies
our code a lot.

This also fixes minor validation errors regarding calling queries
which are not in their active state.
2024-09-09 07:05:46 +02:00
620822c0c5
vulkan: add a ff_vk_init function
This function sets the class correctly, and calls functions
that all users have to call anyway.
2024-09-09 07:05:45 +02:00
ef11a6456d
hwcontext_vulkan: do not chain structs of unsupported extensions in vkCreateDevice
Fixes:

vkCreateDevice(): pCreateInfo->pNext<VkPhysicalDeviceOpticalFlowFeaturesNV> includes a
pointer to a VkPhysicalDeviceOpticalFlowFeaturesNV, but when creating VkDevice, the
parent extension (VK_NV_optical_flow) was not included in ppEnabledExtensionNames.
The Vulkan spec states: Each pNext member of any structure (including this one) in
the pNext chain must be either NULL or a pointer to a valid struct for extending
VkDeviceCreateInfo.
2024-08-11 05:13:17 +02:00
a30b7c0158
hwcontext_vulkan: initialize optical flow queues if available
Lets us implement FPS conversion.
2024-08-11 05:13:10 +02:00
bedfabc437
vulkan: use the new queue family mechanism 2024-08-11 05:13:09 +02:00
80ddc72717
vulkan: rename read_only to singular
There's nothing stopping users from writing to such buffers.
Its more accurate to say they are singular, i.e. not duplicated
between multiple submissions.

This can be helpful for global statistics, or error propagation
purposes.
2024-07-14 18:33:56 +02:00
Andreas Rheinhardt
1b7308b61d avutil/vulkan: Remove unused ff_vk_set_descriptor_sampler()
Reviewed-by: Lynne <dev@lynne.ee>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-03-03 22:56:25 +01:00
Andreas Rheinhardt
c346c097af avutil/vulkan: Make ff_vk_set_descriptor_image() static
Only used in vulkan.c.

Reviewed-by: Lynne <dev@lynne.ee>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-03-03 22:56:21 +01:00
Andreas Rheinhardt
e429b0fdb7 avutil/vulkan: Don't autoinclude vulkan_loader.h
Only include it where necessary.

Reviewed-by: Lynne <dev@lynne.ee>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-03-03 22:55:26 +01:00
Zhao Zhili
f6f06fb42a avutil/vulkan: avoid overreads in ff_vk_count_images
Signed-off-by: Zhao Zhili <zhilizhao@tencent.com>
2024-03-01 17:20:35 +08:00
1a8e766984
vulkan: return VK_NOT_READY when no queries are available
Fixes a validation issue.
The issue is that the function gets called before we've sumitted a frame
for decoding to that context. However, we cannot run queries before
they've been reset, which happens at submission time.
As we'd need to otherwise run a command queue at init-time, just check
if submissions have happened.
2023-10-28 21:16:15 +02:00
358919506d
vulkan: enable VK_KHR_cooperative_matrix
It's of interest to API users, and of interest to us,
as a DCT/DST can be implemented via matrix multiplies.
2023-08-26 23:14:53 +02:00
3e3d46309b
lavu/vulkan: remove unused field from the execution pool structure 2023-07-21 20:04:21 +02:00
97890c2b55
lavu/vulkan: remove threadsafe buffer index load and fix a signed overflow
It's not needed anymore.
2023-07-21 20:04:20 +02:00
975cd48bb3
vulkan: synchronize access to execution pool fences
vkResetFences is specified as being user-synchronized
(yet vkWaitFences, is not).
2023-06-07 23:59:16 +02:00
dfff3877b7
vulkan: add support for the atomic float ops extension 2023-05-29 00:42:01 +02:00
33fc919bb7
hwcontext_vulkan: remove duplicate code, port to use generic vulkan utils
The temporary AVFrame on staack enables us to use the common
dependency/dispatch code in prepare_frame().
The prepare_frame() function is used for both frame initialization
and frame import/export queue family transfer operations.
In the former case, no AVFrame exists yet, so, as this is purely
libavutil code, we create a temporary frame on stack. Otherwise,
we'd need to allocate multiple frames somewhere, one for each
possible command buffer dispatch.
2023-05-29 00:41:51 +02:00
a4d63b46d9
vulkan: make GLSL macro functions semicolumn-safe 2023-05-29 00:41:49 +02:00
83024beec2
vulkan: enable forcing of full subgroups 2023-05-29 00:41:49 +02:00
758f8b26b9
vulkan: add ff_vk_count_images() 2023-05-29 00:41:48 +02:00
b5eaeb1f13
vulkan: rewrite to support all necessary features
This commit rewrites the majority of vulkan.c to enable its use
as a general-purpose high-level utility code, usable for decoding,
encoding, and filtering of video frames.

The dependency system was rewritten to simplify management of
execution.

The image handling system was rewritten to accomodate multiplane
images.

Due to how related all the new features were, this is a single
commit.
2023-05-29 00:41:48 +02:00
721b71da4a
vulkan: return current queue index from ff_vk_qf_rotate() 2023-05-29 00:41:48 +02:00
b15104ed97
vulkan: add support for retrieving queue, query and video properties 2023-05-29 00:41:47 +02:00
6eaf3fe69c
vulkan: add support for queries 2023-05-29 00:41:47 +02:00
f3fb1b50bb
vulkan: minor indent fix, add support for synchronous submission/waiting 2023-05-29 00:41:46 +02:00
d386988c39
vulkan: use device properties 2 and add a convenience loader function 2023-05-29 00:41:46 +02:00
bf69a64135
vulkan: add size tracking to buffer structs 2023-05-29 00:41:46 +02:00
15de0af8f0
vulkan: allow alloc pNext in ff_vk_create_buf 2023-05-29 00:41:45 +02:00
3c2f43d8ee
vulkan: expose ff_vk_alloc_mem() 2023-05-29 00:41:44 +02:00
fa67ccee37
vulkan: add ff_vk_image_create() 2023-05-29 00:41:44 +02:00
e8fce74abf
vulkan: add ff_vk_qf_fill() 2023-05-29 00:41:43 +02:00
b5e333bba7
vulkan: add pNext argument to ff_vk_create_buf() 2023-05-29 00:41:43 +02:00
a0d47a2ad9
vulkan: fix comment statement about exec_queue blocking 2023-05-29 00:41:43 +02:00