Surprising no one, the insane usage rule has a catch. Only happens in the layered_dpb path, which was poorly tested.
Fixes: "[AVHWDeviceContext @ 0x557f4b87dc00] Validation Error: [ VUID-VkImageMemoryBarrier2-srcQueueFamilyIndex-07122 ] Object 0: handle = 0xaf31600000000ce, type = VK_OBJECT_TYPE_IMAGE; | MessageID = 0xfe471484 | vkCmdPipelineBarrier2(): pDependencyInfo->pImageMemoryBarriers[0].oldLayout (VK_IMAGE_LAYOUT_VIDEO_DECODE_DPB_KHR) is not compatible with VkImage 0xaf31600000000ce[] usage flags 0x405. The Vulkan spec states: If srcQueueFamilyIndex and dstQueueFamilyIndex define a queue family ownership transfer or oldLayout and newLayout define an image layout transition, and oldLayout or newLayout is VK_IMAGE_LAYOUT_VIDEO_DECODE_DPB_KHR then image must have been created with VK_IMAGE_USAGE_VIDEO_DECODE_DPB_BIT_KHR (https://www.khronos.org/registry/vulkan/specs/1.3-extensions/html/vkspec.html#VUID-VkImageMemoryBarrier2-srcQueueFamilyIndex-07122) " Patch attached.
>From 8594a95ca1c0819fdf7b04f57542787c5f574223 Mon Sep 17 00:00:00 2001 From: Lynne <d...@lynne.ee> Date: Wed, 25 Oct 2023 01:32:20 +0200 Subject: [PATCH 2/2] vulkan_decode: fix another validation issue Surprising no one, the insane usage rule has a catch. --- libavcodec/vulkan_decode.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/libavcodec/vulkan_decode.c b/libavcodec/vulkan_decode.c index c01eeb9cdc..7f575d1283 100644 --- a/libavcodec/vulkan_decode.c +++ b/libavcodec/vulkan_decode.c @@ -449,7 +449,8 @@ int ff_vk_decode_frame(AVCodecContext *avctx, .srcAccessMask = VK_ACCESS_2_NONE, .dstAccessMask = VK_ACCESS_2_VIDEO_DECODE_WRITE_BIT_KHR, .oldLayout = vkf->layout[0], - .newLayout = vp->dpb_frame ? VK_IMAGE_LAYOUT_VIDEO_DECODE_DST_KHR : + .newLayout = (dec->layered_dpb && vp->dpb_frame) ? + VK_IMAGE_LAYOUT_VIDEO_DECODE_DST_KHR : VK_IMAGE_LAYOUT_VIDEO_DECODE_DPB_KHR, /* Spec, 07252 utter madness */ .srcQueueFamilyIndex = vkf->queue_family[0], .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, -- 2.42.0
_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".