mirrors
/
vkd3d-proton
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

vkd3d: Serialize all queue submissions. 

Gets rid of the full barrier on command buffer end.
Instead, do what D3D12 wants, which is to serialize all
ExecuteCommandLists. Simplify the existing timeline sempahore setup for
sparse queues and use it for all submissions.

Signed-off-by: Hans-Kristian Arntzen <post@arntzen-software.no>
This commit is contained in:
Hans-Kristian Arntzen 2020-06-09 11:12:40 +02:00
parent bc7426ceb0
commit d8ca2cef6d
2 changed files with 67 additions and 58 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

122
libs/vkd3d/command.c
View File

@ -2579,7 +2579,6 @@ static HRESULT STDMETHODCALLTYPE d3d12_command_list_Close(d3d12_command_list_ifa
{
struct d3d12_command_list *list = impl_from_ID3D12GraphicsCommandList(iface);
const struct vkd3d_vk_device_procs *vk_procs;
VkMemoryBarrier memory_barrier;
VkResult vr;
TRACE("iface %p.\n", iface);
@ -2596,18 +2595,6 @@ static HRESULT STDMETHODCALLTYPE d3d12_command_list_Close(d3d12_command_list_ifa
if (list->is_predicated)
VK_CALL(vkCmdEndConditionalRenderingEXT(list->vk_command_buffer));
// TODO We should emit a barrier between submissions,
// not between command lists within the same submission.
memory_barrier.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER;
memory_barrier.pNext = NULL;
memory_barrier.srcAccessMask = VK_ACCESS_MEMORY_WRITE_BIT;
memory_barrier.dstAccessMask = VK_ACCESS_MEMORY_READ_BIT;
VK_CALL(vkCmdPipelineBarrier(list->vk_command_buffer,
VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
0, 1, &memory_barrier, 0, NULL, 0, NULL));
if ((vr = VK_CALL(vkEndCommandBuffer(list->vk_command_buffer))) < 0)
{
WARN("Failed to end command buffer, vr %d.\n", vr);
@ -6611,8 +6598,7 @@ static ULONG STDMETHODCALLTYPE d3d12_command_queue_Release(ID3D12CommandQueue *i
pthread_mutex_destroy(&command_queue->queue_lock);
pthread_cond_destroy(&command_queue->queue_cond);
VK_CALL(vkDestroySemaphore(device->vk_device, command_queue->sparse_binding_signal.vk_semaphore, NULL));
VK_CALL(vkDestroySemaphore(device->vk_device, command_queue->sparse_binding_wait.vk_semaphore, NULL));
VK_CALL(vkDestroySemaphore(device->vk_device, command_queue->submit_timeline.vk_semaphore, NULL));
vkd3d_free(command_queue->submissions);
vkd3d_free(command_queue);
@ -7069,11 +7055,17 @@ static const struct ID3D12CommandQueueVtbl d3d12_command_queue_vtbl =
static void d3d12_command_queue_wait(struct d3d12_command_queue *command_queue,
struct d3d12_fence *fence, UINT64 value)
{
static const VkPipelineStageFlagBits wait_stage_mask = VK_PIPELINE_STAGE_ALL_COMMANDS_BIT;
static const VkPipelineStageFlagBits wait_stage_mask[2] = {
VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT,
VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT,
};
VkTimelineSemaphoreSubmitInfoKHR timeline_submit_info;
const struct vkd3d_vk_device_procs *vk_procs;
VkSemaphore wait_semaphores[2];
struct vkd3d_queue *queue;
VkSubmitInfo submit_info;
uint64_t wait_counts[2];
uint64_t signal_value;
VkQueue vk_queue;
VkResult vr;
@ -7100,15 +7092,31 @@ static void d3d12_command_queue_wait(struct d3d12_command_queue *command_queue,
vk_procs = &command_queue->device->vk_procs;
queue = command_queue->vkd3d_queue;
command_queue->sparse_binding_wait.last_signaled += 1;
wait_counts[0] = value;
wait_semaphores[0] = fence->timeline_semaphore;
/* Serialize submissions. */
wait_counts[1] = command_queue->submit_timeline.last_signaled;
wait_semaphores[1] = command_queue->submit_timeline.vk_semaphore;
signal_value = command_queue->submit_timeline.last_signaled + 1;
assert(fence->timeline_semaphore);
timeline_submit_info.sType = VK_STRUCTURE_TYPE_TIMELINE_SEMAPHORE_SUBMIT_INFO_KHR;
timeline_submit_info.pNext = NULL;
timeline_submit_info.waitSemaphoreValueCount = 1;
timeline_submit_info.pWaitSemaphoreValues = &value;
timeline_submit_info.waitSemaphoreValueCount = 2;
timeline_submit_info.pWaitSemaphoreValues = wait_counts;
timeline_submit_info.signalSemaphoreValueCount = 1;
timeline_submit_info.pSignalSemaphoreValues = &command_queue->sparse_binding_wait.last_signaled;
timeline_submit_info.pSignalSemaphoreValues = &signal_value;
submit_info.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO;
submit_info.pNext = &timeline_submit_info;
submit_info.waitSemaphoreCount = 2;
submit_info.pWaitSemaphores = wait_semaphores;
submit_info.pWaitDstStageMask = wait_stage_mask;
submit_info.commandBufferCount = 0;
submit_info.pCommandBuffers = NULL;
submit_info.signalSemaphoreCount = 1;
submit_info.pSignalSemaphores = &command_queue->submit_timeline.vk_semaphore;
if (!(vk_queue = vkd3d_queue_acquire(queue)))
{
@ -7116,16 +7124,6 @@ static void d3d12_command_queue_wait(struct d3d12_command_queue *command_queue,
return;
}
submit_info.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO;
submit_info.pNext = &timeline_submit_info;
submit_info.waitSemaphoreCount = 1;
submit_info.pWaitSemaphores = &fence->timeline_semaphore;
submit_info.pWaitDstStageMask = &wait_stage_mask;
submit_info.commandBufferCount = 0;
submit_info.pCommandBuffers = NULL;
submit_info.signalSemaphoreCount = 1;
submit_info.pSignalSemaphores = &command_queue->sparse_binding_wait.vk_semaphore;
vr = VK_CALL(vkQueueSubmit(vk_queue, 1, &submit_info, VK_NULL_HANDLE));
vkd3d_queue_release(queue);
@ -7135,6 +7133,7 @@ static void d3d12_command_queue_wait(struct d3d12_command_queue *command_queue,
ERR("Failed to submit wait operation, vr %d.\n", vr);
}
command_queue->submit_timeline.last_signaled = signal_value;
/* We should probably trigger DEVICE_REMOVED if we hit any errors in the submission thread. */
}
@ -7144,8 +7143,10 @@ static void d3d12_command_queue_signal(struct d3d12_command_queue *command_queue
static const VkPipelineStageFlagBits wait_stage_mask = VK_PIPELINE_STAGE_ALL_COMMANDS_BIT;
VkTimelineSemaphoreSubmitInfoKHR timeline_submit_info;
const struct vkd3d_vk_device_procs *vk_procs;
VkSemaphore signal_semaphores[2];
struct vkd3d_queue *vkd3d_queue;
struct d3d12_device *device;
uint64_t signal_values[2];
VkSubmitInfo submit_info;
VkQueue vk_queue;
VkResult vr;
@ -7165,24 +7166,29 @@ static void d3d12_command_queue_signal(struct d3d12_command_queue *command_queue
TRACE("queue %p, fence %p, value %#"PRIx64".\n", command_queue, fence, value);
signal_semaphores[0] = fence->timeline_semaphore;
signal_values[0] = value;
signal_semaphores[1] = command_queue->submit_timeline.vk_semaphore;
signal_values[1] = command_queue->submit_timeline.last_signaled + 1;
/* Need to hold the fence lock while we're submitting, since another thread could come in and signal the semaphore
* to a higher value before we call vkQueueSubmit, which creates a non-monotonically increasing value. */
timeline_submit_info.sType = VK_STRUCTURE_TYPE_TIMELINE_SEMAPHORE_SUBMIT_INFO_KHR;
timeline_submit_info.pNext = NULL;
timeline_submit_info.waitSemaphoreValueCount = 1;
timeline_submit_info.pWaitSemaphoreValues = &command_queue->sparse_binding_signal.last_signaled;
timeline_submit_info.signalSemaphoreValueCount = 1;
timeline_submit_info.pSignalSemaphoreValues = &value;
timeline_submit_info.pWaitSemaphoreValues = &command_queue->submit_timeline.last_signaled;
timeline_submit_info.signalSemaphoreValueCount = 2;
timeline_submit_info.pSignalSemaphoreValues = signal_values;
submit_info.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO;
submit_info.pNext = &timeline_submit_info;
submit_info.waitSemaphoreCount = 1;
submit_info.pWaitSemaphores = &command_queue->sparse_binding_signal.vk_semaphore;
submit_info.pWaitSemaphores = &command_queue->submit_timeline.vk_semaphore;
submit_info.pWaitDstStageMask = &wait_stage_mask;
submit_info.commandBufferCount = 0;
submit_info.pCommandBuffers = NULL;
submit_info.signalSemaphoreCount = 1;
submit_info.pSignalSemaphores = &fence->timeline_semaphore;
submit_info.signalSemaphoreCount = 2;
submit_info.pSignalSemaphores = signal_semaphores;
if (!(vk_queue = vkd3d_queue_acquire(vkd3d_queue)))
{
@ -7205,6 +7211,8 @@ static void d3d12_command_queue_signal(struct d3d12_command_queue *command_queue
return;
}
command_queue->submit_timeline.last_signaled = signal_values[1];
if (FAILED(hr = vkd3d_enqueue_timeline_semaphore(&device->fence_worker, fence, value, vkd3d_queue)))
{
/* In case of an unexpected failure, try to safely destroy Vulkan objects. */
@ -7221,30 +7229,33 @@ static void d3d12_command_queue_execute(struct d3d12_command_queue *command_queu
const struct vkd3d_vk_device_procs *vk_procs = &command_queue->device->vk_procs;
VkTimelineSemaphoreSubmitInfoKHR timeline_submit_info;
VkSubmitInfo submit_desc;
uint64_t signal_value;
uint64_t wait_value;
VkQueue vk_queue;
VkResult vr;
TRACE("queue %p, command_list_count %u, command_lists %p.\n",
command_queue, count, cmd);
command_queue->sparse_binding_wait.last_signaled += 1;
wait_value = command_queue->submit_timeline.last_signaled;
signal_value = wait_value + 1;
timeline_submit_info.sType = VK_STRUCTURE_TYPE_TIMELINE_SEMAPHORE_SUBMIT_INFO_KHR;
timeline_submit_info.pNext = NULL;
timeline_submit_info.waitSemaphoreValueCount = 1;
timeline_submit_info.pWaitSemaphoreValues = &command_queue->sparse_binding_signal.last_signaled;
timeline_submit_info.pWaitSemaphoreValues = &wait_value;
timeline_submit_info.signalSemaphoreValueCount = 1;
timeline_submit_info.pSignalSemaphoreValues = &command_queue->sparse_binding_wait.last_signaled;
timeline_submit_info.pSignalSemaphoreValues = &signal_value;
submit_desc.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO;
submit_desc.pNext = &timeline_submit_info;
submit_desc.waitSemaphoreCount = 1;
submit_desc.pWaitSemaphores = &command_queue->sparse_binding_signal.vk_semaphore;
submit_desc.pWaitSemaphores = &command_queue->submit_timeline.vk_semaphore;
submit_desc.pWaitDstStageMask = &wait_stage_mask;
submit_desc.commandBufferCount = count;
submit_desc.pCommandBuffers = cmd;
submit_desc.signalSemaphoreCount = 1;
submit_desc.pSignalSemaphores = &command_queue->sparse_binding_wait.vk_semaphore;
submit_desc.pSignalSemaphores = &command_queue->submit_timeline.vk_semaphore;
if (!(vk_queue = vkd3d_queue_acquire(command_queue->vkd3d_queue)))
{
@ -7256,6 +7267,7 @@ static void d3d12_command_queue_execute(struct d3d12_command_queue *command_queu
ERR("Failed to submit queue(s), vr %d.\n", vr);
vkd3d_queue_release(command_queue->vkd3d_queue);
command_queue->submit_timeline.last_signaled = signal_value;
}
static void d3d12_command_queue_bind_sparse(struct d3d12_command_queue *command_queue,
@ -7275,7 +7287,9 @@ static void d3d12_command_queue_bind_sparse(struct d3d12_command_queue *command_
struct vkd3d_queue *queue;
VkDeviceMemory vk_memory;
VkDeviceSize vk_offset;
uint64_t signal_value;
unsigned int i, j, k;
uint64_t wait_value;
VkQueue vk_queue;
VkResult vr;
@ -7284,19 +7298,20 @@ static void d3d12_command_queue_bind_sparse(struct d3d12_command_queue *command_
vk_procs = &command_queue->device->vk_procs;
command_queue->sparse_binding_signal.last_signaled += 1;
wait_value = command_queue->submit_timeline.last_signaled;
signal_value = wait_value + 1;
timeline_submit_info.sType = VK_STRUCTURE_TYPE_TIMELINE_SEMAPHORE_SUBMIT_INFO_KHR;
timeline_submit_info.pNext = NULL;
timeline_submit_info.waitSemaphoreValueCount = 1;
timeline_submit_info.pWaitSemaphoreValues = &command_queue->sparse_binding_wait.last_signaled;
timeline_submit_info.pWaitSemaphoreValues = &wait_value;
timeline_submit_info.signalSemaphoreValueCount = 1;
timeline_submit_info.pSignalSemaphoreValues = &command_queue->sparse_binding_signal.last_signaled;
timeline_submit_info.pSignalSemaphoreValues = &signal_value;
bind_sparse_info.sType = VK_STRUCTURE_TYPE_BIND_SPARSE_INFO;
bind_sparse_info.pNext = &timeline_submit_info;
bind_sparse_info.waitSemaphoreCount = 1;
bind_sparse_info.pWaitSemaphores = &command_queue->sparse_binding_wait.vk_semaphore;
bind_sparse_info.pWaitSemaphores = &command_queue->submit_timeline.vk_semaphore;
bind_sparse_info.bufferBindCount = 0;
bind_sparse_info.pBufferBinds = NULL;
bind_sparse_info.imageOpaqueBindCount = 0;
@ -7304,7 +7319,7 @@ static void d3d12_command_queue_bind_sparse(struct d3d12_command_queue *command_
bind_sparse_info.imageBindCount = 0;
bind_sparse_info.pImageBinds = NULL;
bind_sparse_info.signalSemaphoreCount = 1;
bind_sparse_info.pSignalSemaphores = &command_queue->sparse_binding_signal.vk_semaphore;
bind_sparse_info.pSignalSemaphores = &command_queue->submit_timeline.vk_semaphore;
first_packed_tile = dst_resource->sparse.tile_count;
@ -7431,6 +7446,7 @@ static void d3d12_command_queue_bind_sparse(struct d3d12_command_queue *command_
ERR("Failed to perform sparse binding, vr %d.\n", vr);
vkd3d_queue_release(queue);
command_queue->submit_timeline.last_signaled = signal_value;
cleanup:
vkd3d_free(memory_binds);
@ -7574,19 +7590,15 @@ static HRESULT d3d12_command_queue_init(struct d3d12_command_queue *queue,
queue->drain_count = 0;
queue->queue_drain_count = 0;
if (FAILED(hr = vkd3d_create_timeline_semaphore(device, 0, &queue->sparse_binding_wait.vk_semaphore)))
if (FAILED(hr = vkd3d_create_timeline_semaphore(device, 0, &queue->submit_timeline.vk_semaphore)))
goto fail;
if (FAILED(hr = vkd3d_create_timeline_semaphore(device, 0, &queue->sparse_binding_signal.vk_semaphore)))
goto fail_signal_semaphore;
queue->sparse_binding_wait.last_signaled = 0;
queue->sparse_binding_signal.last_signaled = 0;
queue->submit_timeline.last_signaled = 0;
if ((rc = pthread_mutex_init(&queue->queue_lock, NULL)) < 0)
{
hr = hresult_from_errno(rc);
goto fail_pthread_mutex;
goto fail_signal_semaphore;
}
if ((rc = pthread_cond_init(&queue->queue_cond, NULL)) < 0)
@ -7622,10 +7634,8 @@ fail_pthread_create:
pthread_cond_destroy(&queue->queue_cond);
fail_pthread_cond:
pthread_mutex_destroy(&queue->queue_lock);
fail_pthread_mutex:
VK_CALL(vkDestroySemaphore(device->vk_device, queue->sparse_binding_signal.vk_semaphore, NULL));
fail_signal_semaphore:
VK_CALL(vkDestroySemaphore(device->vk_device, queue->sparse_binding_wait.vk_semaphore, NULL));
VK_CALL(vkDestroySemaphore(device->vk_device, queue->submit_timeline.vk_semaphore, NULL));
fail:
return hr;
}

3
libs/vkd3d/vkd3d_private.h
View File

@ -1345,8 +1345,7 @@ struct d3d12_command_queue
uint64_t drain_count;
uint64_t queue_drain_count;
struct vkd3d_timeline_semaphore sparse_binding_wait;
struct vkd3d_timeline_semaphore sparse_binding_signal;
struct vkd3d_timeline_semaphore submit_timeline;
struct vkd3d_private_store private_store;
};