path: root/DOCS
diff options
authorNiklas Haas <>2017-09-24 15:05:24 +0200
committerMartin Herkt <>2017-12-25 00:47:53 +0100
commitbded247fb53558dd5cba26560d1f24e9234ae24e (patch)
tree1e2c1819fc009acf9eea0d481a003799f7ffda8c /DOCS
parenta3c9685257e60e32646bb54a895ef7574a945f69 (diff)
vo_gpu: vulkan: support split command pools
Instead of using a single primary queue, we generate multiple vk_cmdpools and pick the right one dynamically based on the intent. This has a number of immediate benefits: 1. We can use async texture uploads 2. We can use the DMA engine for buffer updates 3. We can benefit from async compute on AMD GPUs Unfortunately, the major downside is that due to the lack of QF ownership tracking, we need to use CONCURRENT sharing for all resources (buffers *and* images!). In theory, we could try figuring out a way to get rid of the concurrent sharing for buffers (which is only needed for compute shader UBOs), but even so, the concurrent sharing mode doesn't really seem to have a significant impact over here (nvidia). It's possible that other platforms may disagree. Our deadlock-avoidance strategy is stupidly simple: Just flush the command every time we need to switch queues, and make sure all submission and callbacks happen in FIFO order. This required lifting the cmds_pending and cmds_queued out from vk_cmdpool to mpvk_ctx, and some functions died/got moved as a result, but that's a relatively minor change. On my hardware this is a fairly significant performance boost, mainly due to async transfers. (Nvidia doesn't expose separate compute queues anyway). On AMD, this should be a performance boost as well due to async compute.
Diffstat (limited to 'DOCS')
1 files changed, 2 insertions, 1 deletions
diff --git a/DOCS/man/options.rst b/DOCS/man/options.rst
index a92a8df962..f50f67a203 100644
--- a/DOCS/man/options.rst
+++ b/DOCS/man/options.rst
@@ -4295,7 +4295,8 @@ The following video options are currently all specific to ``--vo=gpu`` and
Controls the number of VkQueues used for rendering (limited by how many
your device supports). In theory, using more queues could enable some
parallelism between frames (when using a ``--swapchain-depth`` higher than
- 1). (Default: 1)
+ 1), but it can also slow things down on hardware where there's no true
+ parallelism between queues. (Default: 1)
Use WARP (Windows Advanced Rasterization Platform) with the D3D11 GPU