struct list_head running_command_buffer;
};
+struct overlay_draw {
+ struct list_head link;
+
+ VkCommandBuffer command_buffer;
+
+ VkSemaphore semaphore;
+ VkFence fence;
+
+ VkBuffer vertex_buffer;
+ VkDeviceMemory vertex_buffer_mem;
+ VkDeviceSize vertex_buffer_size;
+
+ VkBuffer index_buffer;
+ VkDeviceMemory index_buffer_mem;
+ VkDeviceSize index_buffer_size;
+};
+
/* Mapped from VkSwapchainKHR */
struct swapchain_data {
struct device_data *device;
VkCommandPool command_pool;
- struct {
- VkCommandBuffer command_buffer;
-
- VkBuffer vertex_buffer;
- VkDeviceMemory vertex_buffer_mem;
- VkDeviceSize vertex_buffer_size;
-
- VkBuffer index_buffer;
- VkDeviceMemory index_buffer_mem;
- VkDeviceSize index_buffer_size;
- } frame_data[2];
+ struct list_head draws; /* List of struct overlay_draw */
bool font_uploaded;
VkImage font_image;
VkBuffer upload_font_buffer;
VkDeviceMemory upload_font_buffer_mem;
- VkSemaphore submission_semaphore;
-
/**/
ImGuiContext* imgui_context;
ImVec2 window_size;
}
#define HKEY(obj) ((uint64_t)(obj))
-#define FIND_SWAPCHAIN_DATA(obj) ((struct swapchain_data *)find_object_data(HKEY(obj)))
-#define FIND_CMD_BUFFER_DATA(obj) ((struct command_buffer_data *)find_object_data(HKEY(obj)))
-#define FIND_DEVICE_DATA(obj) ((struct device_data *)find_object_data(HKEY(obj)))
-#define FIND_QUEUE_DATA(obj) ((struct queue_data *)find_object_data(HKEY(obj)))
-#define FIND_PHYSICAL_DEVICE_DATA(obj) ((struct instance_data *)find_object_data(HKEY(obj)))
-#define FIND_INSTANCE_DATA(obj) ((struct instance_data *)find_object_data(HKEY(obj)))
+#define FIND(type, obj) ((type *)find_object_data(HKEY(obj)))
+
static void *find_object_data(uint64_t obj)
{
simple_mtx_lock(&vk_object_to_data_mutex);
/**/
-static void check_vk_result(VkResult err)
-{
- if (err != VK_SUCCESS)
- printf("ERROR!\n");
-}
-
static struct instance_data *new_instance_data(VkInstance instance)
{
struct instance_data *data = rzalloc(NULL, struct instance_data);
VkFenceCreateInfo fence_info = {};
fence_info.sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO;
fence_info.flags = VK_FENCE_CREATE_SIGNALED_BIT;
- VkResult err = device_data->vtable.CreateFence(device_data->device,
- &fence_info,
- NULL,
- &data->queries_fence);
- check_vk_result(err);
+ VK_CHECK(device_data->vtable.CreateFence(device_data->device,
+ &fence_info,
+ NULL,
+ &data->queries_fence));
if (data->flags & VK_QUEUE_GRAPHICS_BIT)
device_data->graphic_queue = data;
data->device = device_data;
data->swapchain = swapchain;
data->window_size = ImVec2(instance_data->params.width, instance_data->params.height);
+ list_inithead(&data->draws);
map_object(HKEY(data->swapchain), data);
return data;
}
ralloc_free(data);
}
+struct overlay_draw *get_overlay_draw(struct swapchain_data *data)
+{
+ struct device_data *device_data = data->device;
+ struct overlay_draw *draw = list_empty(&data->draws) ?
+ NULL : list_first_entry(&data->draws, struct overlay_draw, link);
+
+ VkSemaphoreCreateInfo sem_info = {};
+ sem_info.sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO;
+
+ if (draw && device_data->vtable.GetFenceStatus(device_data->device, draw->fence) == VK_SUCCESS) {
+ list_del(&draw->link);
+ VK_CHECK(device_data->vtable.ResetFences(device_data->device,
+ 1, &draw->fence));
+ list_addtail(&draw->link, &data->draws);
+ return draw;
+ }
+
+ draw = rzalloc(data, struct overlay_draw);
+
+ VkCommandBufferAllocateInfo cmd_buffer_info = {};
+ cmd_buffer_info.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO;
+ cmd_buffer_info.commandPool = data->command_pool;
+ cmd_buffer_info.level = VK_COMMAND_BUFFER_LEVEL_PRIMARY;
+ cmd_buffer_info.commandBufferCount = 1;
+ VK_CHECK(device_data->vtable.AllocateCommandBuffers(device_data->device,
+ &cmd_buffer_info,
+ &draw->command_buffer));
+ VK_CHECK(device_data->set_device_loader_data(device_data->device,
+ draw->command_buffer));
+
+
+ VkFenceCreateInfo fence_info = {};
+ fence_info.sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO;
+ VK_CHECK(device_data->vtable.CreateFence(device_data->device,
+ &fence_info,
+ NULL,
+ &draw->fence));
+
+ VK_CHECK(device_data->vtable.CreateSemaphore(device_data->device, &sem_info,
+ NULL, &draw->semaphore));
+
+ list_addtail(&draw->link, &data->draws);
+
+ return draw;
+}
+
static const char *param_unit(enum overlay_param_enabled param)
{
switch (param) {
case OVERLAY_PARAM_ENABLED_frame_timing:
case OVERLAY_PARAM_ENABLED_acquire_timing:
+ case OVERLAY_PARAM_ENABLED_present_timing:
return "(us)";
case OVERLAY_PARAM_ENABLED_gpu_timing:
return "(ns)";
if (s == OVERLAY_PARAM_ENABLED_frame_timing ||
s == OVERLAY_PARAM_ENABLED_acquire_timing ||
+ s == OVERLAY_PARAM_ENABLED_present_timing ||
s == OVERLAY_PARAM_ENABLED_gpu_timing) {
double min_time = data->stats_min.stats[s] / data->time_dividor;
double max_time = data->stats_max.stats[s] / data->time_dividor;
*buffer_size = new_size;
}
-static void render_swapchain_display(struct swapchain_data *data,
- const VkSemaphore *wait_semaphores,
- unsigned n_wait_semaphores,
- unsigned image_index)
+static struct overlay_draw *render_swapchain_display(struct swapchain_data *data,
+ const VkSemaphore *wait_semaphores,
+ unsigned n_wait_semaphores,
+ unsigned image_index)
{
ImDrawData* draw_data = ImGui::GetDrawData();
if (draw_data->TotalVtxCount == 0)
- return;
+ return NULL;
struct device_data *device_data = data->device;
- uint32_t idx = data->n_frames % ARRAY_SIZE(data->frame_data);
- VkCommandBuffer command_buffer = data->frame_data[idx].command_buffer;
+ struct overlay_draw *draw = get_overlay_draw(data);
- device_data->vtable.ResetCommandBuffer(command_buffer, 0);
+ device_data->vtable.ResetCommandBuffer(draw->command_buffer, 0);
VkRenderPassBeginInfo render_pass_info = {};
render_pass_info.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO;
VkCommandBufferBeginInfo buffer_begin_info = {};
buffer_begin_info.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
- device_data->vtable.BeginCommandBuffer(command_buffer, &buffer_begin_info);
+ device_data->vtable.BeginCommandBuffer(draw->command_buffer, &buffer_begin_info);
- ensure_swapchain_fonts(data, command_buffer);
+ ensure_swapchain_fonts(data, draw->command_buffer);
/* Bounce the image to display back to color attachment layout for
* rendering on top of it.
imb.subresourceRange.layerCount = 1;
imb.srcQueueFamilyIndex = device_data->graphic_queue->family_index;
imb.dstQueueFamilyIndex = device_data->graphic_queue->family_index;
- device_data->vtable.CmdPipelineBarrier(command_buffer,
+ device_data->vtable.CmdPipelineBarrier(draw->command_buffer,
VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT,
VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT,
0, /* dependency flags */
0, nullptr, /* buffer memory barriers */
1, &imb); /* image memory barriers */
- device_data->vtable.CmdBeginRenderPass(command_buffer, &render_pass_info,
+ device_data->vtable.CmdBeginRenderPass(draw->command_buffer, &render_pass_info,
VK_SUBPASS_CONTENTS_INLINE);
/* Create/Resize vertex & index buffers */
size_t vertex_size = draw_data->TotalVtxCount * sizeof(ImDrawVert);
size_t index_size = draw_data->TotalIdxCount * sizeof(ImDrawIdx);
- if (data->frame_data[idx].vertex_buffer_size < vertex_size) {
+ if (draw->vertex_buffer_size < vertex_size) {
CreateOrResizeBuffer(device_data,
- &data->frame_data[idx].vertex_buffer,
- &data->frame_data[idx].vertex_buffer_mem,
- &data->frame_data[idx].vertex_buffer_size,
+ &draw->vertex_buffer,
+ &draw->vertex_buffer_mem,
+ &draw->vertex_buffer_size,
vertex_size, VK_BUFFER_USAGE_VERTEX_BUFFER_BIT);
}
- if (data->frame_data[idx].index_buffer_size < index_size) {
+ if (draw->index_buffer_size < index_size) {
CreateOrResizeBuffer(device_data,
- &data->frame_data[idx].index_buffer,
- &data->frame_data[idx].index_buffer_mem,
- &data->frame_data[idx].index_buffer_size,
+ &draw->index_buffer,
+ &draw->index_buffer_mem,
+ &draw->index_buffer_size,
index_size, VK_BUFFER_USAGE_INDEX_BUFFER_BIT);
}
/* Upload vertex & index data */
- VkBuffer vertex_buffer = data->frame_data[idx].vertex_buffer;
- VkDeviceMemory vertex_mem = data->frame_data[idx].vertex_buffer_mem;
- VkBuffer index_buffer = data->frame_data[idx].index_buffer;
- VkDeviceMemory index_mem = data->frame_data[idx].index_buffer_mem;
ImDrawVert* vtx_dst = NULL;
ImDrawIdx* idx_dst = NULL;
- VK_CHECK(device_data->vtable.MapMemory(device_data->device, vertex_mem,
+ VK_CHECK(device_data->vtable.MapMemory(device_data->device, draw->vertex_buffer_mem,
0, vertex_size, 0, (void**)(&vtx_dst)));
- VK_CHECK(device_data->vtable.MapMemory(device_data->device, index_mem,
+ VK_CHECK(device_data->vtable.MapMemory(device_data->device, draw->index_buffer_mem,
0, index_size, 0, (void**)(&idx_dst)));
for (int n = 0; n < draw_data->CmdListsCount; n++)
{
}
VkMappedMemoryRange range[2] = {};
range[0].sType = VK_STRUCTURE_TYPE_MAPPED_MEMORY_RANGE;
- range[0].memory = vertex_mem;
+ range[0].memory = draw->vertex_buffer_mem;
range[0].size = VK_WHOLE_SIZE;
range[1].sType = VK_STRUCTURE_TYPE_MAPPED_MEMORY_RANGE;
- range[1].memory = index_mem;
+ range[1].memory = draw->index_buffer_mem;
range[1].size = VK_WHOLE_SIZE;
VK_CHECK(device_data->vtable.FlushMappedMemoryRanges(device_data->device, 2, range));
- device_data->vtable.UnmapMemory(device_data->device, vertex_mem);
- device_data->vtable.UnmapMemory(device_data->device, index_mem);
+ device_data->vtable.UnmapMemory(device_data->device, draw->vertex_buffer_mem);
+ device_data->vtable.UnmapMemory(device_data->device, draw->index_buffer_mem);
/* Bind pipeline and descriptor sets */
- device_data->vtable.CmdBindPipeline(command_buffer, VK_PIPELINE_BIND_POINT_GRAPHICS, data->pipeline);
+ device_data->vtable.CmdBindPipeline(draw->command_buffer, VK_PIPELINE_BIND_POINT_GRAPHICS, data->pipeline);
VkDescriptorSet desc_set[1] = { data->descriptor_set };
- device_data->vtable.CmdBindDescriptorSets(command_buffer, VK_PIPELINE_BIND_POINT_GRAPHICS,
+ device_data->vtable.CmdBindDescriptorSets(draw->command_buffer, VK_PIPELINE_BIND_POINT_GRAPHICS,
data->pipeline_layout, 0, 1, desc_set, 0, NULL);
/* Bind vertex & index buffers */
- VkBuffer vertex_buffers[1] = { vertex_buffer };
+ VkBuffer vertex_buffers[1] = { draw->vertex_buffer };
VkDeviceSize vertex_offset[1] = { 0 };
- device_data->vtable.CmdBindVertexBuffers(command_buffer, 0, 1, vertex_buffers, vertex_offset);
- device_data->vtable.CmdBindIndexBuffer(command_buffer, index_buffer, 0, VK_INDEX_TYPE_UINT16);
+ device_data->vtable.CmdBindVertexBuffers(draw->command_buffer, 0, 1, vertex_buffers, vertex_offset);
+ device_data->vtable.CmdBindIndexBuffer(draw->command_buffer, draw->index_buffer, 0, VK_INDEX_TYPE_UINT16);
/* Setup viewport */
VkViewport viewport;
viewport.height = draw_data->DisplaySize.y;
viewport.minDepth = 0.0f;
viewport.maxDepth = 1.0f;
- device_data->vtable.CmdSetViewport(command_buffer, 0, 1, &viewport);
+ device_data->vtable.CmdSetViewport(draw->command_buffer, 0, 1, &viewport);
/* Setup scale and translation through push constants :
float translate[2];
translate[0] = -1.0f - draw_data->DisplayPos.x * scale[0];
translate[1] = -1.0f - draw_data->DisplayPos.y * scale[1];
- device_data->vtable.CmdPushConstants(command_buffer, data->pipeline_layout,
+ device_data->vtable.CmdPushConstants(draw->command_buffer, data->pipeline_layout,
VK_SHADER_STAGE_VERTEX_BIT,
sizeof(float) * 0, sizeof(float) * 2, scale);
- device_data->vtable.CmdPushConstants(command_buffer, data->pipeline_layout,
+ device_data->vtable.CmdPushConstants(draw->command_buffer, data->pipeline_layout,
VK_SHADER_STAGE_VERTEX_BIT,
sizeof(float) * 2, sizeof(float) * 2, translate);
scissor.offset.y = (int32_t)(pcmd->ClipRect.y - display_pos.y) > 0 ? (int32_t)(pcmd->ClipRect.y - display_pos.y) : 0;
scissor.extent.width = (uint32_t)(pcmd->ClipRect.z - pcmd->ClipRect.x);
scissor.extent.height = (uint32_t)(pcmd->ClipRect.w - pcmd->ClipRect.y + 1); // FIXME: Why +1 here?
- device_data->vtable.CmdSetScissor(command_buffer, 0, 1, &scissor);
+ device_data->vtable.CmdSetScissor(draw->command_buffer, 0, 1, &scissor);
// Draw
- device_data->vtable.CmdDrawIndexed(command_buffer, pcmd->ElemCount, 1, idx_offset, vtx_offset, 0);
+ device_data->vtable.CmdDrawIndexed(draw->command_buffer, pcmd->ElemCount, 1, idx_offset, vtx_offset, 0);
idx_offset += pcmd->ElemCount;
}
vtx_offset += cmd_list->VtxBuffer.Size;
}
- device_data->vtable.CmdEndRenderPass(command_buffer);
- device_data->vtable.EndCommandBuffer(command_buffer);
-
- if (data->submission_semaphore) {
- device_data->vtable.DestroySemaphore(device_data->device,
- data->submission_semaphore,
- NULL);
- }
- /* Submission semaphore */
- VkSemaphoreCreateInfo semaphore_info = {};
- semaphore_info.sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO;
- VK_CHECK(device_data->vtable.CreateSemaphore(device_data->device, &semaphore_info,
- NULL, &data->submission_semaphore));
+ device_data->vtable.CmdEndRenderPass(draw->command_buffer);
+ device_data->vtable.EndCommandBuffer(draw->command_buffer);
VkSubmitInfo submit_info = {};
VkPipelineStageFlags stage_wait = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT;
submit_info.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO;
submit_info.commandBufferCount = 1;
- submit_info.pCommandBuffers = &command_buffer;
+ submit_info.pCommandBuffers = &draw->command_buffer;
submit_info.pWaitDstStageMask = &stage_wait;
submit_info.waitSemaphoreCount = n_wait_semaphores;
submit_info.pWaitSemaphores = wait_semaphores;
submit_info.signalSemaphoreCount = 1;
- submit_info.pSignalSemaphores = &data->submission_semaphore;
+ submit_info.pSignalSemaphores = &draw->semaphore;
+
+ device_data->vtable.QueueSubmit(device_data->graphic_queue->queue, 1, &submit_info, draw->fence);
- device_data->vtable.QueueSubmit(device_data->graphic_queue->queue, 1, &submit_info, VK_NULL_HANDLE);
+ return draw;
}
static const uint32_t overlay_vert_spv[] = {
NULL, &data->framebuffers[i]));
}
- /* Command buffer */
+ /* Command buffer pool */
VkCommandPoolCreateInfo cmd_buffer_pool_info = {};
cmd_buffer_pool_info.sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO;
cmd_buffer_pool_info.flags = VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT;
VK_CHECK(device_data->vtable.CreateCommandPool(device_data->device,
&cmd_buffer_pool_info,
NULL, &data->command_pool));
-
- VkCommandBuffer cmd_bufs[ARRAY_SIZE(data->frame_data)];
-
- VkCommandBufferAllocateInfo cmd_buffer_info = {};
- cmd_buffer_info.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO;
- cmd_buffer_info.commandPool = data->command_pool;
- cmd_buffer_info.level = VK_COMMAND_BUFFER_LEVEL_PRIMARY;
- cmd_buffer_info.commandBufferCount = 2;
- VK_CHECK(device_data->vtable.AllocateCommandBuffers(device_data->device,
- &cmd_buffer_info,
- cmd_bufs));
- for (uint32_t i = 0; i < ARRAY_SIZE(data->frame_data); i++) {
- VK_CHECK(device_data->set_device_loader_data(device_data->device,
- cmd_bufs[i]));
-
- data->frame_data[i].command_buffer = cmd_bufs[i];
- }
}
static void shutdown_swapchain_data(struct swapchain_data *data)
{
struct device_data *device_data = data->device;
+ list_for_each_entry_safe(struct overlay_draw, draw, &data->draws, link) {
+ device_data->vtable.DestroySemaphore(device_data->device, draw->semaphore, NULL);
+ device_data->vtable.DestroyFence(device_data->device, draw->fence, NULL);
+ device_data->vtable.DestroyBuffer(device_data->device, draw->vertex_buffer, NULL);
+ device_data->vtable.DestroyBuffer(device_data->device, draw->index_buffer, NULL);
+ device_data->vtable.FreeMemory(device_data->device, draw->vertex_buffer_mem, NULL);
+ device_data->vtable.FreeMemory(device_data->device, draw->index_buffer_mem, NULL);
+ }
+
for (uint32_t i = 0; i < data->n_images; i++) {
device_data->vtable.DestroyImageView(device_data->device, data->image_views[i], NULL);
device_data->vtable.DestroyFramebuffer(device_data->device, data->framebuffers[i], NULL);
device_data->vtable.DestroyRenderPass(device_data->device, data->render_pass, NULL);
- for (uint32_t i = 0; i < ARRAY_SIZE(data->frame_data); i++) {
- device_data->vtable.FreeCommandBuffers(device_data->device,
- data->command_pool,
- 1, &data->frame_data[i].command_buffer);
- if (data->frame_data[i].vertex_buffer)
- device_data->vtable.DestroyBuffer(device_data->device, data->frame_data[i].vertex_buffer, NULL);
- if (data->frame_data[i].index_buffer)
- device_data->vtable.DestroyBuffer(device_data->device, data->frame_data[i].index_buffer, NULL);
- if (data->frame_data[i].vertex_buffer_mem)
- device_data->vtable.FreeMemory(device_data->device, data->frame_data[i].vertex_buffer_mem, NULL);
- if (data->frame_data[i].index_buffer_mem)
- device_data->vtable.FreeMemory(device_data->device, data->frame_data[i].index_buffer_mem, NULL);
- }
device_data->vtable.DestroyCommandPool(device_data->device, data->command_pool, NULL);
- if (data->submission_semaphore)
- device_data->vtable.DestroySemaphore(device_data->device, data->submission_semaphore, NULL);
-
device_data->vtable.DestroyPipeline(device_data->device, data->pipeline, NULL);
device_data->vtable.DestroyPipelineLayout(device_data->device, data->pipeline_layout, NULL);
ImGui::DestroyContext(data->imgui_context);
}
-static void before_present(struct swapchain_data *swapchain_data,
- const VkSemaphore *wait_semaphores,
- unsigned n_wait_semaphores,
- unsigned imageIndex)
+static struct overlay_draw *before_present(struct swapchain_data *swapchain_data,
+ const VkSemaphore *wait_semaphores,
+ unsigned n_wait_semaphores,
+ unsigned imageIndex)
{
struct instance_data *instance_data = swapchain_data->device->instance;
+ struct overlay_draw *draw = NULL;
snapshot_swapchain_frame(swapchain_data);
if (!instance_data->params.no_display && swapchain_data->n_frames > 0) {
compute_swapchain_display(swapchain_data);
- render_swapchain_display(swapchain_data, wait_semaphores, n_wait_semaphores, imageIndex);
+ draw = render_swapchain_display(swapchain_data,
+ wait_semaphores, n_wait_semaphores,
+ imageIndex);
}
+
+ return draw;
}
static VkResult overlay_CreateSwapchainKHR(
const VkAllocationCallbacks* pAllocator,
VkSwapchainKHR* pSwapchain)
{
- struct device_data *device_data = FIND_DEVICE_DATA(device);
+ struct device_data *device_data = FIND(struct device_data, device);
VkResult result = device_data->vtable.CreateSwapchainKHR(device, pCreateInfo, pAllocator, pSwapchain);
if (result != VK_SUCCESS) return result;
VkSwapchainKHR swapchain,
const VkAllocationCallbacks* pAllocator)
{
- struct swapchain_data *swapchain_data = FIND_SWAPCHAIN_DATA(swapchain);
+ struct swapchain_data *swapchain_data =
+ FIND(struct swapchain_data, swapchain);
shutdown_swapchain_data(swapchain_data);
swapchain_data->device->vtable.DestroySwapchainKHR(device, swapchain, pAllocator);
VkQueue queue,
const VkPresentInfoKHR* pPresentInfo)
{
- struct queue_data *queue_data = FIND_QUEUE_DATA(queue);
+ struct queue_data *queue_data = FIND(struct queue_data, queue);
struct device_data *device_data = queue_data->device;
struct instance_data *instance_data = device_data->instance;
uint32_t query_results[OVERLAY_QUERY_COUNT];
/* Before getting the query results, make sure the operations have
* completed.
*/
- VkResult err = device_data->vtable.ResetFences(device_data->device,
- 1, &queue_data->queries_fence);
- check_vk_result(err);
- err = device_data->vtable.QueueSubmit(queue, 0, NULL, queue_data->queries_fence);
- check_vk_result(err);
- err = device_data->vtable.WaitForFences(device_data->device,
- 1, &queue_data->queries_fence,
- VK_FALSE, UINT64_MAX);
- check_vk_result(err);
+ VK_CHECK(device_data->vtable.ResetFences(device_data->device,
+ 1, &queue_data->queries_fence));
+ VK_CHECK(device_data->vtable.QueueSubmit(queue, 0, NULL, queue_data->queries_fence));
+ VK_CHECK(device_data->vtable.WaitForFences(device_data->device,
+ 1, &queue_data->queries_fence,
+ VK_FALSE, UINT64_MAX));
/* Now get the results. */
list_for_each_entry_safe(struct command_buffer_data, cmd_buffer_data,
if (cmd_buffer_data->pipeline_query_pool) {
memset(query_results, 0, sizeof(query_results));
- err =
- device_data->vtable.GetQueryPoolResults(device_data->device,
- cmd_buffer_data->pipeline_query_pool,
- cmd_buffer_data->query_index, 1,
- sizeof(uint32_t) * OVERLAY_QUERY_COUNT,
- query_results, 0, VK_QUERY_RESULT_WAIT_BIT);
- check_vk_result(err);
+ VK_CHECK(device_data->vtable.GetQueryPoolResults(device_data->device,
+ cmd_buffer_data->pipeline_query_pool,
+ cmd_buffer_data->query_index, 1,
+ sizeof(uint32_t) * OVERLAY_QUERY_COUNT,
+ query_results, 0, VK_QUERY_RESULT_WAIT_BIT));
for (uint32_t i = OVERLAY_PARAM_ENABLED_vertices;
i <= OVERLAY_PARAM_ENABLED_compute_invocations; i++) {
}
if (cmd_buffer_data->timestamp_query_pool) {
uint64_t gpu_timestamps[2] = { 0 };
- err =
- device_data->vtable.GetQueryPoolResults(device_data->device,
- cmd_buffer_data->timestamp_query_pool,
- cmd_buffer_data->query_index * 2, 2,
- 2 * sizeof(uint64_t), gpu_timestamps, sizeof(uint64_t),
- VK_QUERY_RESULT_WAIT_BIT | VK_QUERY_RESULT_64_BIT);
- check_vk_result(err);
+ VK_CHECK(device_data->vtable.GetQueryPoolResults(device_data->device,
+ cmd_buffer_data->timestamp_query_pool,
+ cmd_buffer_data->query_index * 2, 2,
+ 2 * sizeof(uint64_t), gpu_timestamps, sizeof(uint64_t),
+ VK_QUERY_RESULT_WAIT_BIT | VK_QUERY_RESULT_64_BIT));
gpu_timestamps[0] &= queue_data->timestamp_mask;
gpu_timestamps[1] &= queue_data->timestamp_mask;
if (instance_data->params.no_display) {
for (uint32_t i = 0; i < pPresentInfo->swapchainCount; i++) {
VkSwapchainKHR swapchain = pPresentInfo->pSwapchains[i];
- struct swapchain_data *swapchain_data = FIND_SWAPCHAIN_DATA(swapchain);
+ struct swapchain_data *swapchain_data =
+ FIND(struct swapchain_data, swapchain);
before_present(swapchain_data,
pPresentInfo->pWaitSemaphores,
pPresentInfo->waitSemaphoreCount,
pPresentInfo->pImageIndices[i]);
+
+ VkPresentInfoKHR present_info = *pPresentInfo;
+ present_info.swapchainCount = 1;
+ present_info.pSwapchains = &swapchain;
+
+ uint64_t ts0 = os_time_get();
+ result = queue_data->device->vtable.QueuePresentKHR(queue, &present_info);
+ uint64_t ts1 = os_time_get();
+ swapchain_data->frame_stats.stats[OVERLAY_PARAM_ENABLED_present_timing] += ts1 - ts0;
}
- result = queue_data->device->vtable.QueuePresentKHR(queue, pPresentInfo);
} else {
for (uint32_t i = 0; i < pPresentInfo->swapchainCount; i++) {
VkSwapchainKHR swapchain = pPresentInfo->pSwapchains[i];
- struct swapchain_data *swapchain_data = FIND_SWAPCHAIN_DATA(swapchain);
+ struct swapchain_data *swapchain_data =
+ FIND(struct swapchain_data, swapchain);
VkPresentInfoKHR present_info = *pPresentInfo;
present_info.swapchainCount = 1;
present_info.pSwapchains = &swapchain;
- before_present(swapchain_data,
- pPresentInfo->pWaitSemaphores,
- pPresentInfo->waitSemaphoreCount,
- pPresentInfo->pImageIndices[i]);
+ uint32_t image_index = pPresentInfo->pImageIndices[i];
+
+ struct overlay_draw *draw = before_present(swapchain_data,
+ pPresentInfo->pWaitSemaphores,
+ pPresentInfo->waitSemaphoreCount,
+ image_index);
+
/* Because the submission of the overlay draw waits on the semaphores
* handed for present, we don't need to have this present operation
* wait on them as well, we can just wait on the overlay submission
* semaphore.
*/
- present_info.pWaitSemaphores = &swapchain_data->submission_semaphore;
+ present_info.pWaitSemaphores = &draw->semaphore;
present_info.waitSemaphoreCount = 1;
+ uint64_t ts0 = os_time_get();
VkResult chain_result = queue_data->device->vtable.QueuePresentKHR(queue, &present_info);
+ uint64_t ts1 = os_time_get();
+ swapchain_data->frame_stats.stats[OVERLAY_PARAM_ENABLED_present_timing] += ts1 - ts0;
if (pPresentInfo->pResults)
pPresentInfo->pResults[i] = chain_result;
if (chain_result != VK_SUCCESS && result == VK_SUCCESS)
VkFence fence,
uint32_t* pImageIndex)
{
- struct swapchain_data *swapchain_data = FIND_SWAPCHAIN_DATA(swapchain);
+ struct swapchain_data *swapchain_data =
+ FIND(struct swapchain_data, swapchain);
struct device_data *device_data = swapchain_data->device;
uint64_t ts0 = os_time_get();
const VkAcquireNextImageInfoKHR* pAcquireInfo,
uint32_t* pImageIndex)
{
- struct swapchain_data *swapchain_data = FIND_SWAPCHAIN_DATA(pAcquireInfo->swapchain);
+ struct swapchain_data *swapchain_data =
+ FIND(struct swapchain_data, pAcquireInfo->swapchain);
struct device_data *device_data = swapchain_data->device;
uint64_t ts0 = os_time_get();
uint32_t firstVertex,
uint32_t firstInstance)
{
- struct command_buffer_data *cmd_buffer_data = FIND_CMD_BUFFER_DATA(commandBuffer);
+ struct command_buffer_data *cmd_buffer_data =
+ FIND(struct command_buffer_data, commandBuffer);
cmd_buffer_data->stats.stats[OVERLAY_PARAM_ENABLED_draw]++;
struct device_data *device_data = cmd_buffer_data->device;
device_data->vtable.CmdDraw(commandBuffer, vertexCount, instanceCount,
int32_t vertexOffset,
uint32_t firstInstance)
{
- struct command_buffer_data *cmd_buffer_data = FIND_CMD_BUFFER_DATA(commandBuffer);
+ struct command_buffer_data *cmd_buffer_data =
+ FIND(struct command_buffer_data, commandBuffer);
cmd_buffer_data->stats.stats[OVERLAY_PARAM_ENABLED_draw_indexed]++;
struct device_data *device_data = cmd_buffer_data->device;
device_data->vtable.CmdDrawIndexed(commandBuffer, indexCount, instanceCount,
uint32_t drawCount,
uint32_t stride)
{
- struct command_buffer_data *cmd_buffer_data = FIND_CMD_BUFFER_DATA(commandBuffer);
+ struct command_buffer_data *cmd_buffer_data =
+ FIND(struct command_buffer_data, commandBuffer);
cmd_buffer_data->stats.stats[OVERLAY_PARAM_ENABLED_draw_indirect]++;
struct device_data *device_data = cmd_buffer_data->device;
device_data->vtable.CmdDrawIndirect(commandBuffer, buffer, offset, drawCount, stride);
uint32_t drawCount,
uint32_t stride)
{
- struct command_buffer_data *cmd_buffer_data = FIND_CMD_BUFFER_DATA(commandBuffer);
+ struct command_buffer_data *cmd_buffer_data =
+ FIND(struct command_buffer_data, commandBuffer);
cmd_buffer_data->stats.stats[OVERLAY_PARAM_ENABLED_draw_indexed_indirect]++;
struct device_data *device_data = cmd_buffer_data->device;
device_data->vtable.CmdDrawIndexedIndirect(commandBuffer, buffer, offset, drawCount, stride);
uint32_t maxDrawCount,
uint32_t stride)
{
- struct command_buffer_data *cmd_buffer_data = FIND_CMD_BUFFER_DATA(commandBuffer);
+ struct command_buffer_data *cmd_buffer_data =
+ FIND(struct command_buffer_data, commandBuffer);
cmd_buffer_data->stats.stats[OVERLAY_PARAM_ENABLED_draw_indirect_count]++;
struct device_data *device_data = cmd_buffer_data->device;
device_data->vtable.CmdDrawIndirectCountKHR(commandBuffer, buffer, offset,
uint32_t maxDrawCount,
uint32_t stride)
{
- struct command_buffer_data *cmd_buffer_data = FIND_CMD_BUFFER_DATA(commandBuffer);
+ struct command_buffer_data *cmd_buffer_data =
+ FIND(struct command_buffer_data, commandBuffer);
cmd_buffer_data->stats.stats[OVERLAY_PARAM_ENABLED_draw_indexed_indirect_count]++;
struct device_data *device_data = cmd_buffer_data->device;
device_data->vtable.CmdDrawIndexedIndirectCountKHR(commandBuffer, buffer, offset,
uint32_t groupCountY,
uint32_t groupCountZ)
{
- struct command_buffer_data *cmd_buffer_data = FIND_CMD_BUFFER_DATA(commandBuffer);
+ struct command_buffer_data *cmd_buffer_data =
+ FIND(struct command_buffer_data, commandBuffer);
cmd_buffer_data->stats.stats[OVERLAY_PARAM_ENABLED_dispatch]++;
struct device_data *device_data = cmd_buffer_data->device;
device_data->vtable.CmdDispatch(commandBuffer, groupCountX, groupCountY, groupCountZ);
VkBuffer buffer,
VkDeviceSize offset)
{
- struct command_buffer_data *cmd_buffer_data = FIND_CMD_BUFFER_DATA(commandBuffer);
+ struct command_buffer_data *cmd_buffer_data =
+ FIND(struct command_buffer_data, commandBuffer);
cmd_buffer_data->stats.stats[OVERLAY_PARAM_ENABLED_dispatch_indirect]++;
struct device_data *device_data = cmd_buffer_data->device;
device_data->vtable.CmdDispatchIndirect(commandBuffer, buffer, offset);
VkPipelineBindPoint pipelineBindPoint,
VkPipeline pipeline)
{
- struct command_buffer_data *cmd_buffer_data = FIND_CMD_BUFFER_DATA(commandBuffer);
+ struct command_buffer_data *cmd_buffer_data =
+ FIND(struct command_buffer_data, commandBuffer);
switch (pipelineBindPoint) {
case VK_PIPELINE_BIND_POINT_GRAPHICS: cmd_buffer_data->stats.stats[OVERLAY_PARAM_ENABLED_pipeline_graphics]++; break;
case VK_PIPELINE_BIND_POINT_COMPUTE: cmd_buffer_data->stats.stats[OVERLAY_PARAM_ENABLED_pipeline_compute]++; break;
VkCommandBuffer commandBuffer,
const VkCommandBufferBeginInfo* pBeginInfo)
{
- struct command_buffer_data *cmd_buffer_data = FIND_CMD_BUFFER_DATA(commandBuffer);
+ struct command_buffer_data *cmd_buffer_data =
+ FIND(struct command_buffer_data, commandBuffer);
struct device_data *device_data = cmd_buffer_data->device;
+ memset(&cmd_buffer_data->stats, 0, sizeof(cmd_buffer_data->stats));
+
/* We don't record any query in secondary command buffers, just make sure
* we have the right inheritance.
*/
return result;
}
- /* Primary command buffers with no queries. */
- if (!cmd_buffer_data->pipeline_query_pool && cmd_buffer_data->timestamp_query_pool)
- return device_data->vtable.BeginCommandBuffer(commandBuffer, pBeginInfo);
-
/* Otherwise record a begin query as first command. */
VkResult result = device_data->vtable.BeginCommandBuffer(commandBuffer, pBeginInfo);
static VkResult overlay_EndCommandBuffer(
VkCommandBuffer commandBuffer)
{
- struct command_buffer_data *cmd_buffer_data = FIND_CMD_BUFFER_DATA(commandBuffer);
+ struct command_buffer_data *cmd_buffer_data =
+ FIND(struct command_buffer_data, commandBuffer);
struct device_data *device_data = cmd_buffer_data->device;
if (cmd_buffer_data->timestamp_query_pool) {
VkCommandBuffer commandBuffer,
VkCommandBufferResetFlags flags)
{
- struct command_buffer_data *cmd_buffer_data = FIND_CMD_BUFFER_DATA(commandBuffer);
+ struct command_buffer_data *cmd_buffer_data =
+ FIND(struct command_buffer_data, commandBuffer);
struct device_data *device_data = cmd_buffer_data->device;
memset(&cmd_buffer_data->stats, 0, sizeof(cmd_buffer_data->stats));
uint32_t commandBufferCount,
const VkCommandBuffer* pCommandBuffers)
{
- struct command_buffer_data *cmd_buffer_data = FIND_CMD_BUFFER_DATA(commandBuffer);
+ struct command_buffer_data *cmd_buffer_data =
+ FIND(struct command_buffer_data, commandBuffer);
struct device_data *device_data = cmd_buffer_data->device;
/* Add the stats of the executed command buffers to the primary one. */
for (uint32_t c = 0; c < commandBufferCount; c++) {
- struct command_buffer_data *sec_cmd_buffer_data = FIND_CMD_BUFFER_DATA(pCommandBuffers[c]);
+ struct command_buffer_data *sec_cmd_buffer_data =
+ FIND(struct command_buffer_data, pCommandBuffers[c]);
for (uint32_t s = 0; s < OVERLAY_PARAM_ENABLED_MAX; s++)
cmd_buffer_data->stats.stats[s] += sec_cmd_buffer_data->stats.stats[s];
const VkCommandBufferAllocateInfo* pAllocateInfo,
VkCommandBuffer* pCommandBuffers)
{
- struct device_data *device_data = FIND_DEVICE_DATA(device);
+ struct device_data *device_data = FIND(struct device_data, device);
VkResult result =
device_data->vtable.AllocateCommandBuffers(device, pAllocateInfo, pCommandBuffers);
if (result != VK_SUCCESS)
pAllocateInfo->commandBufferCount,
overlay_query_flags,
};
- VkResult err =
- device_data->vtable.CreateQueryPool(device_data->device, &pool_info,
- NULL, &pipeline_query_pool);
- check_vk_result(err);
+ VK_CHECK(device_data->vtable.CreateQueryPool(device_data->device, &pool_info,
+ NULL, &pipeline_query_pool));
}
if (device_data->instance->params.enabled[OVERLAY_PARAM_ENABLED_gpu_timing]) {
VkQueryPoolCreateInfo pool_info = {
pAllocateInfo->commandBufferCount * 2,
0,
};
- VkResult err =
- device_data->vtable.CreateQueryPool(device_data->device, &pool_info,
- NULL, ×tamp_query_pool);
- check_vk_result(err);
+ VK_CHECK(device_data->vtable.CreateQueryPool(device_data->device, &pool_info,
+ NULL, ×tamp_query_pool));
}
for (uint32_t i = 0; i < pAllocateInfo->commandBufferCount; i++) {
uint32_t commandBufferCount,
const VkCommandBuffer* pCommandBuffers)
{
- struct device_data *device_data = FIND_DEVICE_DATA(device);
+ struct device_data *device_data = FIND(struct device_data, device);
for (uint32_t i = 0; i < commandBufferCount; i++) {
struct command_buffer_data *cmd_buffer_data =
- FIND_CMD_BUFFER_DATA(pCommandBuffers[i]);
+ FIND(struct command_buffer_data, pCommandBuffers[i]);
+
+ /* It is legal to free a NULL command buffer*/
+ if (!cmd_buffer_data)
+ continue;
+
uint64_t count = (uintptr_t)find_object_data(HKEY(cmd_buffer_data->pipeline_query_pool));
if (count == 1) {
unmap_object(HKEY(cmd_buffer_data->pipeline_query_pool));
const VkSubmitInfo* pSubmits,
VkFence fence)
{
- struct queue_data *queue_data = FIND_QUEUE_DATA(queue);
+ struct queue_data *queue_data = FIND(struct queue_data, queue);
struct device_data *device_data = queue_data->device;
device_data->frame_stats.stats[OVERLAY_PARAM_ENABLED_submit]++;
for (uint32_t s = 0; s < submitCount; s++) {
for (uint32_t c = 0; c < pSubmits[s].commandBufferCount; c++) {
struct command_buffer_data *cmd_buffer_data =
- FIND_CMD_BUFFER_DATA(pSubmits[s].pCommandBuffers[c]);
+ FIND(struct command_buffer_data, pSubmits[s].pCommandBuffers[c]);
/* Merge the submitted command buffer stats into the device. */
for (uint32_t st = 0; st < OVERLAY_PARAM_ENABLED_MAX; st++)
const VkAllocationCallbacks* pAllocator,
VkDevice* pDevice)
{
- struct instance_data *instance_data = FIND_PHYSICAL_DEVICE_DATA(physicalDevice);
+ struct instance_data *instance_data =
+ FIND(struct instance_data, physicalDevice);
VkLayerDeviceCreateInfo *chain_info =
get_device_chain_info(pCreateInfo, VK_LAYER_LINK_INFO);
VkDevice device,
const VkAllocationCallbacks* pAllocator)
{
- struct device_data *device_data = FIND_DEVICE_DATA(device);
+ struct device_data *device_data = FIND(struct device_data, device);
device_unmap_queues(device_data);
device_data->vtable.DestroyDevice(device, pAllocator);
destroy_device_data(device_data);
VkInstance instance,
const VkAllocationCallbacks* pAllocator)
{
- struct instance_data *instance_data = FIND_INSTANCE_DATA(instance);
+ struct instance_data *instance_data = FIND(struct instance_data, instance);
instance_data_map_physical_devices(instance_data, false);
instance_data->vtable.DestroyInstance(instance, pAllocator);
destroy_instance_data(instance_data);
if (dev == NULL) return NULL;
- struct device_data *device_data = FIND_DEVICE_DATA(dev);
+ struct device_data *device_data = FIND(struct device_data, dev);
if (device_data->vtable.GetDeviceProcAddr == NULL) return NULL;
return device_data->vtable.GetDeviceProcAddr(dev, funcName);
}
if (instance == NULL) return NULL;
- struct instance_data *instance_data = FIND_INSTANCE_DATA(instance);
+ struct instance_data *instance_data = FIND(struct instance_data, instance);
if (instance_data->vtable.GetInstanceProcAddr == NULL) return NULL;
return instance_data->vtable.GetInstanceProcAddr(instance, funcName);
}