[Mesa-dev] [PATCH v2 20/32] vulkan/wsi: Set a proper pWaitDstStageMask on the dummy submit

Chad Versace chadversary at chromium.org
Sat Dec 2 01:58:09 UTC 2017


On Tue 28 Nov 2017, Jason Ekstrand wrote:
> Neither mesa driver really cares, but we should set it none the less for
> the sake of correctness.
> ---
>  src/vulkan/wsi/wsi_common.c | 17 +++++++++++++++++
>  1 file changed, 17 insertions(+)
> 
> diff --git a/src/vulkan/wsi/wsi_common.c b/src/vulkan/wsi/wsi_common.c
> index 4f6648f..e5a9a28 100644
> --- a/src/vulkan/wsi/wsi_common.c
> +++ b/src/vulkan/wsi/wsi_common.c
> @@ -542,14 +542,31 @@ wsi_common_queue_present(const struct wsi_device *wsi,
>           .sType = VK_STRUCTURE_TYPE_SUBMIT_INFO,
>           .pNext = NULL,
>        };
> +      VkPipelineStageFlags *stage_flags = NULL;
>        if (i == 0) {
>           /* We only need/want to wait on semaphores once.  After that, we're
>            * guaranteed ordering since it all happens on the same queue.
>            */
>           submit_info.waitSemaphoreCount = pPresentInfo->waitSemaphoreCount,
>           submit_info.pWaitSemaphores = pPresentInfo->pWaitSemaphores,
> +
> +         /* Set up the pWaitDstStageMasks */
> +         stage_flags = vk_alloc(&swapchain->alloc,
> +                                sizeof(VkPipelineStageFlags) *
> +                                pPresentInfo->waitSemaphoreCount,
> +                                8,
> +                                VK_SYSTEM_ALLOCATION_SCOPE_COMMAND);
> +         if (!stage_flags) {
> +            result = VK_ERROR_OUT_OF_HOST_MEMORY;
> +            goto fail_present;
> +         }
> +         for (uint32_t s = 0; s < pPresentInfo->waitSemaphoreCount; s++)
> +            stage_flags[s] = VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT;
> +
> +         submit_info.pWaitDstStageMask = stage_flags;

Since VkSwapchain is required to be externally synchronized, you could
embed stage_flags directly in struct wsi_swapchain, doubling its size
when needed. But meh.

Reviewed-by: Chad Versace <chadversary at chromium.org>


More information about the mesa-dev mailing list