vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
/*
|
|
|
|
|
* Copyright © 2024 Intel Corporation
|
|
|
|
|
*
|
|
|
|
|
* Permission is hereby granted, free of charge, to any person obtaining a
|
|
|
|
|
* copy of this software and associated documentation files (the "Software"),
|
|
|
|
|
* to deal in the Software without restriction, including without limitation
|
|
|
|
|
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
|
|
|
|
|
* and/or sell copies of the Software, and to permit persons to whom the
|
|
|
|
|
* Software is furnished to do so, subject to the following conditions:
|
|
|
|
|
*
|
|
|
|
|
* The above copyright notice and this permission notice (including the next
|
|
|
|
|
* paragraph) shall be included in all copies or substantial portions of the
|
|
|
|
|
* Software.
|
|
|
|
|
*
|
|
|
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
|
|
|
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
|
|
|
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
|
|
|
|
|
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
|
|
|
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
|
|
|
|
|
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
|
|
|
|
|
* IN THE SOFTWARE.
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Copyright (C) 2015-2021 Valve Corporation
|
|
|
|
|
* Copyright (C) 2015-2021 LunarG, Inc.
|
|
|
|
|
*
|
|
|
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
|
|
|
* you may not use this file except in compliance with the License.
|
|
|
|
|
* You may obtain a copy of the License at
|
|
|
|
|
*
|
|
|
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
|
*
|
|
|
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
|
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
|
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
|
|
|
* See the License for the specific language governing permissions and
|
|
|
|
|
* limitations under the License.
|
|
|
|
|
*
|
|
|
|
|
* Author: Cody Northrop <cody@lunarg.com>
|
|
|
|
|
* Author: David Pinedo <david@lunarg.com>
|
|
|
|
|
* Author: Jon Ashburn <jon@lunarg.com>
|
|
|
|
|
* Author: Tony Barbour <tony@lunarg.com>
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
#include <string.h>
|
|
|
|
|
#include <stdlib.h>
|
|
|
|
|
#include <assert.h>
|
|
|
|
|
#include <pthread.h>
|
|
|
|
|
#include <png.h>
|
|
|
|
|
#include <time.h>
|
|
|
|
|
|
|
|
|
|
#include <vulkan/vulkan_core.h>
|
|
|
|
|
#include <vulkan/vk_layer.h>
|
|
|
|
|
|
|
|
|
|
#include "git_sha1.h"
|
|
|
|
|
|
|
|
|
|
#include "screenshot_params.h"
|
|
|
|
|
|
|
|
|
|
#include "util/u_debug.h"
|
|
|
|
|
#include "util/hash_table.h"
|
|
|
|
|
#include "util/list.h"
|
|
|
|
|
#include "util/ralloc.h"
|
|
|
|
|
#include "util/os_time.h"
|
|
|
|
|
#include "util/os_socket.h"
|
|
|
|
|
#include "util/simple_mtx.h"
|
|
|
|
|
#include "util/u_math.h"
|
|
|
|
|
|
|
|
|
|
#include "vk_enum_to_str.h"
|
|
|
|
|
#include "vk_dispatch_table.h"
|
|
|
|
|
#include "vk_util.h"
|
|
|
|
|
|
|
|
|
|
typedef pthread_mutex_t loader_platform_thread_mutex;
|
|
|
|
|
static inline void loader_platform_thread_create_mutex(loader_platform_thread_mutex *pMutex) { pthread_mutex_init(pMutex, NULL); }
|
|
|
|
|
static inline void loader_platform_thread_lock_mutex(loader_platform_thread_mutex *pMutex) { pthread_mutex_lock(pMutex); }
|
|
|
|
|
static inline void loader_platform_thread_unlock_mutex(loader_platform_thread_mutex *pMutex) { pthread_mutex_unlock(pMutex); }
|
|
|
|
|
static inline void loader_platform_thread_delete_mutex(loader_platform_thread_mutex *pMutex) { pthread_mutex_destroy(pMutex); }
|
|
|
|
|
|
|
|
|
|
static int globalLockInitialized = 0;
|
|
|
|
|
static loader_platform_thread_mutex globalLock;
|
|
|
|
|
|
|
|
|
|
/* Mapped from VkInstace/VkPhysicalDevice */
|
|
|
|
|
struct instance_data {
|
|
|
|
|
struct vk_instance_dispatch_table vtable;
|
|
|
|
|
struct vk_physical_device_dispatch_table pd_vtable;
|
|
|
|
|
VkInstance instance;
|
|
|
|
|
|
|
|
|
|
struct screenshot_params params;
|
|
|
|
|
|
|
|
|
|
int control_client;
|
|
|
|
|
int socket_fd;
|
|
|
|
|
|
|
|
|
|
/* Enabling switch for taking screenshot */
|
|
|
|
|
bool screenshot_enabled;
|
|
|
|
|
|
vulkan/screenshot-layer: Add region command option
Adds an optional region selection, based off percentages of the
starting/ending of an image's X & Y values.
This is intended as a performance enhancement tradeoff for smaller
images to be created.
With a smaller image size, the screenshotting layer will change the
region boundaries on the GPU side, which will decrease the amount of
time it takes to copy the image over to CPU-accessible memory.
Using vkcube as an example, the original image size is 500x500:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 123530 nanoseconds
Then, by cropping the area to a 100x100 image, we get the following:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Using region: startX = 40% (200), startY = 40% (200), endX = 60% (300), endY = 60% (300)
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 12679 nanoseconds
For this example, this is a ~90% time reduction improvement!
Overall, this option reduces the copy time to a point where it can
become negligible, relative to the frame time of the application.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Reviewed-by: Felix DeGrood felix.j.degrood@intel.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32016>
2024-11-06 12:00:02 -08:00
|
|
|
/* Region switch for enabling region use on a per-frame basis */
|
|
|
|
|
bool region_enabled;
|
|
|
|
|
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
/* Enabling switch for socket communications */
|
|
|
|
|
bool socket_enabled;
|
|
|
|
|
bool socket_setup;
|
|
|
|
|
|
|
|
|
|
const char *filename;
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
pthread_cond_t ptCondition = PTHREAD_COND_INITIALIZER;
|
|
|
|
|
pthread_mutex_t ptLock = PTHREAD_MUTEX_INITIALIZER;
|
|
|
|
|
|
|
|
|
|
VkFence copyDone;
|
|
|
|
|
const VkPipelineStageFlags dstStageWaitBeforeSubmission = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
|
|
|
|
|
const VkSemaphore *pSemaphoreWaitBeforePresent;
|
|
|
|
|
uint32_t semaphoreWaitBeforePresentCount;
|
|
|
|
|
VkSemaphore semaphoreWaitAfterSubmission;
|
|
|
|
|
|
|
|
|
|
/* Mapped from VkDevice */
|
|
|
|
|
struct queue_data;
|
|
|
|
|
struct device_data {
|
|
|
|
|
struct instance_data *instance;
|
|
|
|
|
|
|
|
|
|
PFN_vkSetDeviceLoaderData set_device_loader_data;
|
|
|
|
|
|
|
|
|
|
struct vk_device_dispatch_table vtable;
|
|
|
|
|
VkPhysicalDevice physical_device;
|
|
|
|
|
VkDevice device;
|
|
|
|
|
|
|
|
|
|
VkPhysicalDeviceProperties properties;
|
|
|
|
|
|
|
|
|
|
struct queue_data *graphic_queue;
|
2025-02-10 22:11:59 -08:00
|
|
|
struct queue_data* queue_data_head;
|
|
|
|
|
struct queue_data* queue_data_tail;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
/* Mapped from VkQueue */
|
|
|
|
|
struct queue_data {
|
|
|
|
|
struct device_data *device;
|
2025-02-10 22:11:59 -08:00
|
|
|
struct queue_data *next;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
VkQueue queue;
|
2025-02-10 22:11:59 -08:00
|
|
|
uint32_t familyIndex;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
uint32_t index;
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
/* Mapped from VkSwapchainKHR */
|
|
|
|
|
struct swapchain_data {
|
|
|
|
|
struct device_data *device;
|
|
|
|
|
|
|
|
|
|
VkSwapchainKHR swapchain;
|
|
|
|
|
VkExtent2D imageExtent;
|
|
|
|
|
VkFormat format;
|
|
|
|
|
|
|
|
|
|
VkImage image;
|
2024-10-19 15:28:28 -07:00
|
|
|
uint32_t imageListSize;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
static struct hash_table_u64 *vk_object_to_data = NULL;
|
|
|
|
|
static simple_mtx_t vk_object_to_data_mutex = SIMPLE_MTX_INITIALIZER;
|
|
|
|
|
|
|
|
|
|
static inline void ensure_vk_object_map(void)
|
|
|
|
|
{
|
|
|
|
|
if (!vk_object_to_data)
|
|
|
|
|
vk_object_to_data = _mesa_hash_table_u64_create(NULL);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#define HKEY(obj) ((uint64_t)(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);
|
|
|
|
|
ensure_vk_object_map();
|
|
|
|
|
void *data = _mesa_hash_table_u64_search(vk_object_to_data, obj);
|
|
|
|
|
simple_mtx_unlock(&vk_object_to_data_mutex);
|
|
|
|
|
return data;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void map_object(uint64_t obj, void *data)
|
|
|
|
|
{
|
|
|
|
|
simple_mtx_lock(&vk_object_to_data_mutex);
|
|
|
|
|
ensure_vk_object_map();
|
|
|
|
|
_mesa_hash_table_u64_insert(vk_object_to_data, obj, data);
|
|
|
|
|
simple_mtx_unlock(&vk_object_to_data_mutex);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void unmap_object(uint64_t obj)
|
|
|
|
|
{
|
|
|
|
|
simple_mtx_lock(&vk_object_to_data_mutex);
|
|
|
|
|
_mesa_hash_table_u64_remove(vk_object_to_data, obj);
|
|
|
|
|
simple_mtx_unlock(&vk_object_to_data_mutex);
|
|
|
|
|
}
|
|
|
|
|
|
2024-10-19 15:28:28 -07:00
|
|
|
void map_images(swapchain_data *data, VkImage *imageList, uint32_t size) {
|
|
|
|
|
data->imageListSize = size;
|
|
|
|
|
VkImage *image;
|
2024-10-23 18:47:11 -07:00
|
|
|
image = (VkImage *)malloc(sizeof(VkImage) * size);
|
2024-10-19 15:28:28 -07:00
|
|
|
for (uint32_t index = 0; index < size; index++) {
|
2024-10-23 18:47:11 -07:00
|
|
|
image[index] = imageList[index];
|
|
|
|
|
map_object(HKEY(index), &image[index]);
|
2024-10-19 15:28:28 -07:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void select_image_from_map(swapchain_data *data, uint32_t index) {
|
|
|
|
|
data->image = *(FIND(VkImage, index));
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void unmap_images(swapchain_data *data) {
|
2024-10-23 18:47:11 -07:00
|
|
|
VkImage *image, *first;
|
|
|
|
|
first = nullptr;
|
2024-10-19 15:28:28 -07:00
|
|
|
for (uint32_t index = 0; index < data->imageListSize; index++) {
|
|
|
|
|
image = FIND(VkImage, index);
|
2024-10-23 18:47:11 -07:00
|
|
|
if (!first)
|
|
|
|
|
first = image;
|
2024-10-19 15:28:28 -07:00
|
|
|
unmap_object(HKEY(index));
|
|
|
|
|
}
|
2024-10-23 18:47:11 -07:00
|
|
|
free(first);
|
2024-10-19 15:28:28 -07:00
|
|
|
data->imageListSize = 0;
|
|
|
|
|
}
|
|
|
|
|
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
#define VK_CHECK(expr) \
|
|
|
|
|
do { \
|
|
|
|
|
VkResult __result = (expr); \
|
|
|
|
|
if (__result != VK_SUCCESS) { \
|
|
|
|
|
LOG(ERROR, "'%s' line %i failed with %s\n", \
|
|
|
|
|
#expr, __LINE__, vk_Result_to_str(__result)); \
|
|
|
|
|
} \
|
|
|
|
|
} while (0)
|
|
|
|
|
|
|
|
|
|
static VkLayerInstanceCreateInfo *get_instance_chain_info(const VkInstanceCreateInfo *pCreateInfo,
|
|
|
|
|
VkLayerFunction func)
|
|
|
|
|
{
|
|
|
|
|
vk_foreach_struct_const(item, pCreateInfo->pNext) {
|
|
|
|
|
if (item->sType == VK_STRUCTURE_TYPE_LOADER_INSTANCE_CREATE_INFO &&
|
|
|
|
|
((VkLayerInstanceCreateInfo *) item)->function == func)
|
|
|
|
|
return (VkLayerInstanceCreateInfo *) item;
|
|
|
|
|
}
|
2025-07-23 09:17:35 +02:00
|
|
|
UNREACHABLE("instance chain info not found");
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
return NULL;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static VkLayerDeviceCreateInfo *get_device_chain_info(const VkDeviceCreateInfo *pCreateInfo,
|
|
|
|
|
VkLayerFunction func)
|
|
|
|
|
{
|
|
|
|
|
vk_foreach_struct_const(item, pCreateInfo->pNext) {
|
|
|
|
|
if (item->sType == VK_STRUCTURE_TYPE_LOADER_DEVICE_CREATE_INFO &&
|
|
|
|
|
((VkLayerDeviceCreateInfo *) item)->function == func)
|
|
|
|
|
return (VkLayerDeviceCreateInfo *)item;
|
|
|
|
|
}
|
2025-07-23 09:17:35 +02:00
|
|
|
UNREACHABLE("device chain info not found");
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
return NULL;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**/
|
|
|
|
|
|
|
|
|
|
static struct instance_data *new_instance_data(VkInstance instance)
|
|
|
|
|
{
|
|
|
|
|
struct instance_data *data = rzalloc(NULL, struct instance_data);
|
|
|
|
|
data->instance = instance;
|
|
|
|
|
data->control_client = -1;
|
|
|
|
|
data->socket_fd = -1;
|
|
|
|
|
map_object(HKEY(data->instance), data);
|
|
|
|
|
return data;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void destroy_instance_data(struct instance_data *data)
|
|
|
|
|
{
|
|
|
|
|
destroy_frame_list(data->params.frames);
|
|
|
|
|
if (data->socket_fd >= 0)
|
|
|
|
|
os_socket_close(data->socket_fd);
|
|
|
|
|
unmap_object(HKEY(data->instance));
|
|
|
|
|
ralloc_free(data);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void instance_data_map_physical_devices(struct instance_data *instance_data,
|
|
|
|
|
bool map)
|
|
|
|
|
{
|
|
|
|
|
uint32_t physicalDeviceCount = 0;
|
|
|
|
|
instance_data->vtable.EnumeratePhysicalDevices(instance_data->instance,
|
|
|
|
|
&physicalDeviceCount,
|
|
|
|
|
NULL);
|
|
|
|
|
|
|
|
|
|
VkPhysicalDevice *physicalDevices = (VkPhysicalDevice *) malloc(sizeof(VkPhysicalDevice) * physicalDeviceCount);
|
|
|
|
|
instance_data->vtable.EnumeratePhysicalDevices(instance_data->instance,
|
|
|
|
|
&physicalDeviceCount,
|
|
|
|
|
physicalDevices);
|
|
|
|
|
|
|
|
|
|
for (uint32_t i = 0; i < physicalDeviceCount; i++) {
|
|
|
|
|
if (map)
|
|
|
|
|
map_object(HKEY(physicalDevices[i]), instance_data);
|
|
|
|
|
else
|
|
|
|
|
unmap_object(HKEY(physicalDevices[i]));
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
free(physicalDevices);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**/
|
|
|
|
|
static struct device_data *new_device_data(VkDevice device, struct instance_data *instance)
|
|
|
|
|
{
|
|
|
|
|
struct device_data *data = rzalloc(NULL, struct device_data);
|
|
|
|
|
data->instance = instance;
|
|
|
|
|
data->device = device;
|
2025-02-10 22:11:59 -08:00
|
|
|
data->graphic_queue = VK_NULL_HANDLE;
|
|
|
|
|
data->queue_data_head = VK_NULL_HANDLE;
|
|
|
|
|
data->queue_data_tail = VK_NULL_HANDLE;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
map_object(HKEY(data->device), data);
|
|
|
|
|
return data;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static struct queue_data *new_queue_data(VkQueue queue,
|
|
|
|
|
struct device_data *device_data,
|
2025-02-10 22:11:59 -08:00
|
|
|
uint32_t index,
|
|
|
|
|
uint32_t familyIndex)
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
{
|
|
|
|
|
struct queue_data *data = rzalloc(device_data, struct queue_data);
|
|
|
|
|
data->device = device_data;
|
|
|
|
|
data->queue = queue;
|
|
|
|
|
data->index = index;
|
2025-02-10 22:11:59 -08:00
|
|
|
data->familyIndex = familyIndex;
|
|
|
|
|
data->next = VK_NULL_HANDLE;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
map_object(HKEY(data->queue), data);
|
2025-02-10 22:11:59 -08:00
|
|
|
if (device_data->queue_data_head == VK_NULL_HANDLE) {
|
|
|
|
|
device_data->queue_data_head = data;
|
|
|
|
|
device_data->queue_data_tail = data;
|
|
|
|
|
} else {
|
|
|
|
|
device_data->queue_data_tail->next = data;
|
|
|
|
|
device_data->queue_data_tail = data;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
}
|
|
|
|
|
return data;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void destroy_queue(struct queue_data *data)
|
|
|
|
|
{
|
|
|
|
|
struct device_data *device_data = data->device;
|
|
|
|
|
unmap_object(HKEY(data->queue));
|
|
|
|
|
ralloc_free(data);
|
|
|
|
|
}
|
|
|
|
|
|
2025-02-10 22:11:59 -08:00
|
|
|
static void device_destroy_queues(struct device_data *data)
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
{
|
2025-02-10 22:11:59 -08:00
|
|
|
struct queue_data *tmp_queue = VK_NULL_HANDLE;
|
|
|
|
|
for (auto it = data->queue_data_head; it != VK_NULL_HANDLE;) {
|
|
|
|
|
tmp_queue = it->next;
|
|
|
|
|
destroy_queue(it);
|
|
|
|
|
it = tmp_queue;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void destroy_device_data(struct device_data *data)
|
|
|
|
|
{
|
|
|
|
|
loader_platform_thread_lock_mutex(&globalLock);
|
|
|
|
|
unmap_object(HKEY(data->device));
|
|
|
|
|
ralloc_free(data);
|
|
|
|
|
loader_platform_thread_unlock_mutex(&globalLock);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static struct swapchain_data *new_swapchain_data(VkSwapchainKHR swapchain,
|
|
|
|
|
struct device_data *device_data)
|
|
|
|
|
{
|
|
|
|
|
struct instance_data *instance_data = device_data->instance;
|
|
|
|
|
struct swapchain_data *data = rzalloc(NULL, struct swapchain_data);
|
|
|
|
|
data->device = device_data;
|
|
|
|
|
data->swapchain = swapchain;
|
|
|
|
|
map_object(HKEY(data->swapchain), data);
|
|
|
|
|
return data;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void destroy_swapchain_data(struct swapchain_data *data)
|
|
|
|
|
{
|
2024-10-23 18:47:11 -07:00
|
|
|
unmap_images(data);
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
unmap_object(HKEY(data->swapchain));
|
|
|
|
|
ralloc_free(data);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void parse_command(struct instance_data *instance_data,
|
|
|
|
|
const char *cmd, unsigned cmdlen,
|
|
|
|
|
const char *param, unsigned paramlen)
|
|
|
|
|
{
|
|
|
|
|
/* parse string (if any) from capture command */
|
|
|
|
|
if (!strncmp(cmd, "capture", cmdlen)) {
|
|
|
|
|
instance_data->screenshot_enabled = true;
|
|
|
|
|
if (paramlen > 1) {
|
|
|
|
|
instance_data->filename = param;
|
|
|
|
|
} else {
|
|
|
|
|
instance_data->filename = NULL;
|
|
|
|
|
}
|
vulkan/screenshot-layer: Add region command option
Adds an optional region selection, based off percentages of the
starting/ending of an image's X & Y values.
This is intended as a performance enhancement tradeoff for smaller
images to be created.
With a smaller image size, the screenshotting layer will change the
region boundaries on the GPU side, which will decrease the amount of
time it takes to copy the image over to CPU-accessible memory.
Using vkcube as an example, the original image size is 500x500:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 123530 nanoseconds
Then, by cropping the area to a 100x100 image, we get the following:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Using region: startX = 40% (200), startY = 40% (200), endX = 60% (300), endY = 60% (300)
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 12679 nanoseconds
For this example, this is a ~90% time reduction improvement!
Overall, this option reduces the copy time to a point where it can
become negligible, relative to the frame time of the application.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Reviewed-by: Felix DeGrood felix.j.degrood@intel.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32016>
2024-11-06 12:00:02 -08:00
|
|
|
} else if (!strncmp(cmd, "region", cmdlen)) {
|
|
|
|
|
instance_data->params.region = getRegionFromInput(param);
|
|
|
|
|
instance_data->region_enabled = instance_data->params.region.useImageRegion;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#define BUFSIZE 4096
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* This function will process commands through the control file.
|
|
|
|
|
*
|
|
|
|
|
* A command starts with a colon, followed by the command, and followed by an
|
|
|
|
|
* option '=' and a parameter. It has to end with a semi-colon. A full command
|
|
|
|
|
* + parameter looks like:
|
|
|
|
|
*
|
|
|
|
|
* :cmd=param;
|
|
|
|
|
*/
|
|
|
|
|
static void process_char(struct instance_data *instance_data, char c)
|
|
|
|
|
{
|
|
|
|
|
static char cmd[BUFSIZE];
|
|
|
|
|
static char param[BUFSIZE];
|
|
|
|
|
|
|
|
|
|
static unsigned cmdpos = 0;
|
|
|
|
|
static unsigned parampos = 0;
|
|
|
|
|
static bool reading_cmd = false;
|
|
|
|
|
static bool reading_param = false;
|
|
|
|
|
|
|
|
|
|
switch (c) {
|
|
|
|
|
case ':':
|
|
|
|
|
cmdpos = 0;
|
|
|
|
|
parampos = 0;
|
|
|
|
|
reading_cmd = true;
|
|
|
|
|
reading_param = false;
|
|
|
|
|
break;
|
vulkan/screenshot-layer: Add region command option
Adds an optional region selection, based off percentages of the
starting/ending of an image's X & Y values.
This is intended as a performance enhancement tradeoff for smaller
images to be created.
With a smaller image size, the screenshotting layer will change the
region boundaries on the GPU side, which will decrease the amount of
time it takes to copy the image over to CPU-accessible memory.
Using vkcube as an example, the original image size is 500x500:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 123530 nanoseconds
Then, by cropping the area to a 100x100 image, we get the following:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Using region: startX = 40% (200), startY = 40% (200), endX = 60% (300), endY = 60% (300)
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 12679 nanoseconds
For this example, this is a ~90% time reduction improvement!
Overall, this option reduces the copy time to a point where it can
become negligible, relative to the frame time of the application.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Reviewed-by: Felix DeGrood felix.j.degrood@intel.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32016>
2024-11-06 12:00:02 -08:00
|
|
|
case ',':
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
case ';':
|
|
|
|
|
if (!reading_cmd)
|
|
|
|
|
break;
|
|
|
|
|
cmd[cmdpos++] = '\0';
|
|
|
|
|
param[parampos++] = '\0';
|
|
|
|
|
parse_command(instance_data, cmd, cmdpos, param, parampos);
|
vulkan/screenshot-layer: Add region command option
Adds an optional region selection, based off percentages of the
starting/ending of an image's X & Y values.
This is intended as a performance enhancement tradeoff for smaller
images to be created.
With a smaller image size, the screenshotting layer will change the
region boundaries on the GPU side, which will decrease the amount of
time it takes to copy the image over to CPU-accessible memory.
Using vkcube as an example, the original image size is 500x500:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 123530 nanoseconds
Then, by cropping the area to a 100x100 image, we get the following:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Using region: startX = 40% (200), startY = 40% (200), endX = 60% (300), endY = 60% (300)
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 12679 nanoseconds
For this example, this is a ~90% time reduction improvement!
Overall, this option reduces the copy time to a point where it can
become negligible, relative to the frame time of the application.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Reviewed-by: Felix DeGrood felix.j.degrood@intel.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32016>
2024-11-06 12:00:02 -08:00
|
|
|
if (c == ';') {
|
|
|
|
|
reading_cmd = false;
|
|
|
|
|
} else {
|
|
|
|
|
cmdpos = 0;
|
|
|
|
|
parampos = 0;
|
|
|
|
|
}
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
reading_param = false;
|
|
|
|
|
break;
|
|
|
|
|
case '=':
|
|
|
|
|
if (!reading_cmd)
|
|
|
|
|
break;
|
|
|
|
|
reading_param = true;
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
if (!reading_cmd)
|
|
|
|
|
break;
|
|
|
|
|
|
|
|
|
|
if (reading_param) {
|
|
|
|
|
/* overflow means an invalid parameter */
|
|
|
|
|
if (parampos >= BUFSIZE - 1) {
|
|
|
|
|
reading_cmd = false;
|
|
|
|
|
reading_param = false;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
param[parampos++] = c;
|
|
|
|
|
} else {
|
|
|
|
|
/* overflow means an invalid command */
|
|
|
|
|
if (cmdpos >= BUFSIZE - 1) {
|
|
|
|
|
reading_cmd = false;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
cmd[cmdpos++] = c;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void control_send(struct instance_data *instance_data,
|
|
|
|
|
const char *cmd, unsigned cmdlen,
|
|
|
|
|
const char *param, unsigned paramlen)
|
|
|
|
|
{
|
|
|
|
|
unsigned msglen = 0;
|
|
|
|
|
char buffer[BUFSIZE];
|
|
|
|
|
|
|
|
|
|
assert(cmdlen + paramlen + 3 < BUFSIZE);
|
|
|
|
|
|
|
|
|
|
buffer[msglen++] = ':';
|
|
|
|
|
|
|
|
|
|
memcpy(&buffer[msglen], cmd, cmdlen);
|
|
|
|
|
msglen += cmdlen;
|
|
|
|
|
|
|
|
|
|
if (paramlen > 0) {
|
|
|
|
|
buffer[msglen++] = '=';
|
|
|
|
|
memcpy(&buffer[msglen], param, paramlen);
|
|
|
|
|
msglen += paramlen;
|
|
|
|
|
buffer[msglen++] = ';';
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
os_socket_send(instance_data->control_client, buffer, msglen, 0);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void control_send_connection_string(struct device_data *device_data)
|
|
|
|
|
{
|
|
|
|
|
struct instance_data *instance_data = device_data->instance;
|
|
|
|
|
|
|
|
|
|
const char *controlVersionCmd = "MesaScreenshotControlVersion";
|
|
|
|
|
const char *controlVersionString = "1";
|
|
|
|
|
|
|
|
|
|
control_send(instance_data, controlVersionCmd, strlen(controlVersionCmd),
|
|
|
|
|
controlVersionString, strlen(controlVersionString));
|
|
|
|
|
|
|
|
|
|
const char *deviceCmd = "DeviceName";
|
|
|
|
|
const char *deviceName = device_data->properties.deviceName;
|
|
|
|
|
|
|
|
|
|
control_send(instance_data, deviceCmd, strlen(deviceCmd),
|
|
|
|
|
deviceName, strlen(deviceName));
|
|
|
|
|
|
|
|
|
|
const char *mesaVersionCmd = "MesaVersion";
|
|
|
|
|
const char *mesaVersionString = "Mesa " PACKAGE_VERSION MESA_GIT_SHA1;
|
|
|
|
|
|
|
|
|
|
control_send(instance_data, mesaVersionCmd, strlen(mesaVersionCmd),
|
|
|
|
|
mesaVersionString, strlen(mesaVersionString));
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void control_client_check(struct device_data *device_data)
|
|
|
|
|
{
|
|
|
|
|
struct instance_data *instance_data = device_data->instance;
|
|
|
|
|
|
|
|
|
|
/* Already connected, just return. */
|
|
|
|
|
if (instance_data->control_client >= 0)
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
int socket_fd = os_socket_accept(instance_data->socket_fd);
|
|
|
|
|
if (socket_fd == -1) {
|
|
|
|
|
if (errno != EAGAIN && errno != EWOULDBLOCK && errno != ECONNABORTED)
|
|
|
|
|
LOG(ERROR, "socket error: %s\n", strerror(errno));
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (socket_fd >= 0) {
|
|
|
|
|
os_socket_block(socket_fd, false);
|
|
|
|
|
instance_data->control_client = socket_fd;
|
|
|
|
|
control_send_connection_string(device_data);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void control_client_disconnected(struct instance_data *instance_data)
|
|
|
|
|
{
|
|
|
|
|
os_socket_close(instance_data->control_client);
|
|
|
|
|
instance_data->control_client = -1;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void process_control_socket(struct instance_data *instance_data)
|
|
|
|
|
{
|
|
|
|
|
const int client = instance_data->control_client;
|
|
|
|
|
if (client >= 0) {
|
|
|
|
|
char buf[BUFSIZE];
|
|
|
|
|
|
|
|
|
|
while (true) {
|
|
|
|
|
ssize_t n = os_socket_recv(client, buf, BUFSIZE, 0);
|
|
|
|
|
|
|
|
|
|
if (n == -1) {
|
|
|
|
|
if (errno == EAGAIN || errno == EWOULDBLOCK) {
|
|
|
|
|
/* nothing to read, try again later */
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (errno != ECONNRESET)
|
|
|
|
|
LOG(ERROR, "Connection failed: %s\n", strerror(errno));
|
|
|
|
|
|
|
|
|
|
control_client_disconnected(instance_data);
|
|
|
|
|
} else if (n == 0) {
|
|
|
|
|
/* recv() returns 0 when the client disconnects */
|
|
|
|
|
control_client_disconnected(instance_data);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
for (ssize_t i = 0; i < n; i++) {
|
|
|
|
|
process_char(instance_data, buf[i]);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* If we try to read BUFSIZE and receive BUFSIZE bytes from the
|
|
|
|
|
* socket, there's a good chance that there's still more data to be
|
|
|
|
|
* read, so we will try again. Otherwise, simply be done for this
|
|
|
|
|
* iteration and try again on the next frame.
|
|
|
|
|
*/
|
|
|
|
|
if (n < BUFSIZE)
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2025-02-10 22:11:59 -08:00
|
|
|
static void screenshot_GetDeviceQueue(VkDevice device, uint32_t queueFamilyIndex, uint32_t queueIndex, VkQueue *pQueue) {
|
|
|
|
|
struct device_data *device_data = FIND(struct device_data, device);
|
|
|
|
|
device_data->vtable.GetDeviceQueue(device, queueFamilyIndex, queueIndex, pQueue);
|
|
|
|
|
loader_platform_thread_lock_mutex(&globalLock);
|
|
|
|
|
struct queue_data *it = device_data->queue_data_head;
|
|
|
|
|
while (it != VK_NULL_HANDLE) {
|
|
|
|
|
if (it->queue == *pQueue) {
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
it = it->next;
|
|
|
|
|
}
|
|
|
|
|
if (it == VK_NULL_HANDLE) {
|
|
|
|
|
new_queue_data(*pQueue, device_data, queueIndex, queueFamilyIndex);
|
|
|
|
|
} else {
|
|
|
|
|
it->familyIndex = queueFamilyIndex;
|
|
|
|
|
it->index = queueIndex;
|
|
|
|
|
}
|
|
|
|
|
loader_platform_thread_unlock_mutex(&globalLock);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void screenshot_GetDeviceQueue2(VkDevice device, const VkDeviceQueueInfo2 *pQueueInfo, VkQueue *pQueue) {
|
|
|
|
|
if (pQueueInfo) screenshot_GetDeviceQueue(device, pQueueInfo->queueFamilyIndex, pQueueInfo->queueIndex, pQueue);
|
|
|
|
|
}
|
|
|
|
|
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
static VkResult screenshot_CreateSwapchainKHR(
|
|
|
|
|
VkDevice device,
|
|
|
|
|
const VkSwapchainCreateInfoKHR* pCreateInfo,
|
|
|
|
|
const VkAllocationCallbacks* pAllocator,
|
|
|
|
|
VkSwapchainKHR* pSwapchain)
|
|
|
|
|
{
|
|
|
|
|
struct device_data *device_data = FIND(struct device_data, device);
|
|
|
|
|
|
|
|
|
|
// Turn on transfer src bit for image copy later on.
|
|
|
|
|
VkSwapchainCreateInfoKHR createInfo = *pCreateInfo;
|
|
|
|
|
createInfo.imageUsage |= VK_IMAGE_USAGE_TRANSFER_SRC_BIT;
|
|
|
|
|
VkResult result = device_data->vtable.CreateSwapchainKHR(device, &createInfo, pAllocator, pSwapchain);
|
|
|
|
|
if (result != VK_SUCCESS) return result;
|
|
|
|
|
|
|
|
|
|
loader_platform_thread_lock_mutex(&globalLock);
|
|
|
|
|
|
|
|
|
|
struct swapchain_data *swapchain_data = new_swapchain_data(*pSwapchain, device_data);
|
|
|
|
|
swapchain_data->imageExtent = pCreateInfo->imageExtent;
|
|
|
|
|
swapchain_data->format = pCreateInfo->imageFormat;
|
|
|
|
|
loader_platform_thread_unlock_mutex(&globalLock);
|
|
|
|
|
return result;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static VkResult screenshot_GetSwapchainImagesKHR(
|
|
|
|
|
VkDevice device,
|
|
|
|
|
VkSwapchainKHR swapchain,
|
|
|
|
|
uint32_t* pCount,
|
|
|
|
|
VkImage* pSwapchainImages)
|
|
|
|
|
{
|
|
|
|
|
struct swapchain_data *swapchain_data = FIND(struct swapchain_data, swapchain);
|
|
|
|
|
struct vk_device_dispatch_table *vtable = &(swapchain_data->device->vtable);
|
|
|
|
|
VkResult result = vtable->GetSwapchainImagesKHR(device, swapchain, pCount, pSwapchainImages);
|
|
|
|
|
|
|
|
|
|
loader_platform_thread_lock_mutex(&globalLock);
|
2024-10-23 18:47:11 -07:00
|
|
|
LOG(DEBUG, "Swapchain size: %d\n", *pCount);
|
2024-10-19 15:28:28 -07:00
|
|
|
if (swapchain_data->imageListSize > 0)
|
|
|
|
|
unmap_images(swapchain_data);
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
if (result == VK_SUCCESS) {
|
2024-10-19 15:28:28 -07:00
|
|
|
// Save the images produced from the swapchain in a hash table
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
if (*pCount > 0) {
|
|
|
|
|
if(pSwapchainImages){
|
2024-10-19 15:28:28 -07:00
|
|
|
map_images(swapchain_data, pSwapchainImages, *pCount);
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
loader_platform_thread_unlock_mutex(&globalLock);
|
|
|
|
|
return result;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void screenshot_DestroySwapchainKHR(
|
|
|
|
|
VkDevice device,
|
|
|
|
|
VkSwapchainKHR swapchain,
|
|
|
|
|
const VkAllocationCallbacks* pAllocator)
|
|
|
|
|
{
|
|
|
|
|
if (swapchain == VK_NULL_HANDLE) {
|
|
|
|
|
struct device_data *device_data = FIND(struct device_data, device);
|
|
|
|
|
device_data->vtable.DestroySwapchainKHR(device, swapchain, pAllocator);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
struct swapchain_data *swapchain_data =
|
|
|
|
|
FIND(struct swapchain_data, swapchain);
|
|
|
|
|
|
|
|
|
|
swapchain_data->device->vtable.DestroySwapchainKHR(device, swapchain, pAllocator);
|
|
|
|
|
destroy_swapchain_data(swapchain_data);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Convert long int to string */
|
|
|
|
|
static void itoa(uint32_t integer, char *dest_str)
|
|
|
|
|
{
|
|
|
|
|
// Our sizes are limited to uin32_t max value: 4,294,967,295 (10 digits)
|
|
|
|
|
sprintf(dest_str, "%u", integer);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static bool get_mem_type_from_properties(
|
|
|
|
|
VkPhysicalDeviceMemoryProperties* mem_properties,
|
|
|
|
|
uint32_t bits_type,
|
|
|
|
|
VkFlags requirements_mask,
|
|
|
|
|
uint32_t* type_index)
|
|
|
|
|
{
|
|
|
|
|
for (uint32_t i = 0; i < 32; i++) {
|
|
|
|
|
if ((bits_type & 1) == 1) {
|
|
|
|
|
if ((mem_properties->memoryTypes[i].propertyFlags & requirements_mask) == requirements_mask) {
|
|
|
|
|
*type_index = i;
|
|
|
|
|
return true;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
bits_type >>= 1;
|
|
|
|
|
}
|
|
|
|
|
return false;
|
|
|
|
|
}
|
|
|
|
|
|
2025-02-10 22:11:59 -08:00
|
|
|
VkQueue getQueueForScreenshot(struct device_data *device_data,
|
|
|
|
|
struct instance_data *instance_data) {
|
|
|
|
|
// Find a queue that we can use for taking a screenshot
|
|
|
|
|
VkQueue queue = VK_NULL_HANDLE;
|
|
|
|
|
VkBool32 presentCapable = VK_FALSE;
|
|
|
|
|
uint32_t n_family_props;
|
|
|
|
|
instance_data->pd_vtable.GetPhysicalDeviceQueueFamilyProperties(device_data->physical_device,
|
|
|
|
|
&n_family_props,
|
|
|
|
|
NULL);
|
|
|
|
|
if (n_family_props > 0) {
|
|
|
|
|
VkQueueFamilyProperties *family_props =
|
|
|
|
|
(VkQueueFamilyProperties *)malloc(sizeof(VkQueueFamilyProperties) * n_family_props);
|
|
|
|
|
instance_data->pd_vtable.GetPhysicalDeviceQueueFamilyProperties(device_data->physical_device,
|
|
|
|
|
&n_family_props,
|
|
|
|
|
family_props);
|
|
|
|
|
|
|
|
|
|
// Iterate over all queues for this device, searching for a queue that is graphics capable
|
|
|
|
|
for (auto it = device_data->queue_data_head; it != VK_NULL_HANDLE; it = it->next) {
|
|
|
|
|
queue = it->queue;
|
|
|
|
|
if((family_props[it->familyIndex].queueFlags & VK_QUEUE_GRAPHICS_BIT) != 0) {
|
|
|
|
|
break;
|
|
|
|
|
} else {
|
|
|
|
|
// Clear the queue if it's not graphics capable
|
|
|
|
|
queue = VK_NULL_HANDLE;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
free(family_props);
|
|
|
|
|
}
|
|
|
|
|
return queue;
|
|
|
|
|
}
|
|
|
|
|
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
// Track allocated resources in writeFile()
|
|
|
|
|
// and clean them up when they go out of scope.
|
|
|
|
|
struct WriteFileCleanupData {
|
|
|
|
|
device_data *dev_data;
|
|
|
|
|
VkImage image2;
|
|
|
|
|
VkImage image3;
|
|
|
|
|
VkDeviceMemory mem2;
|
|
|
|
|
VkDeviceMemory mem3;
|
|
|
|
|
bool mem2mapped;
|
|
|
|
|
bool mem3mapped;
|
|
|
|
|
VkCommandBuffer commandBuffer;
|
|
|
|
|
VkCommandPool commandPool;
|
|
|
|
|
~WriteFileCleanupData();
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
WriteFileCleanupData::~WriteFileCleanupData() {
|
|
|
|
|
if (mem2mapped) dev_data->vtable.UnmapMemory(dev_data->device, mem2);
|
|
|
|
|
if (mem2) dev_data->vtable.FreeMemory(dev_data->device, mem2, NULL);
|
|
|
|
|
if (image2) dev_data->vtable.DestroyImage(dev_data->device, image2, NULL);
|
|
|
|
|
|
|
|
|
|
if (mem3mapped) dev_data->vtable.UnmapMemory(dev_data->device, mem3);
|
|
|
|
|
if (mem3) dev_data->vtable.FreeMemory(dev_data->device, mem3, NULL);
|
|
|
|
|
if (image3) dev_data->vtable.DestroyImage(dev_data->device, image3, NULL);
|
|
|
|
|
|
|
|
|
|
if (commandBuffer) dev_data->vtable.FreeCommandBuffers(dev_data->device, commandPool, 1, &commandBuffer);
|
|
|
|
|
if (commandPool) dev_data->vtable.DestroyCommandPool(dev_data->device, commandPool, NULL);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static uint64_t get_time() {
|
|
|
|
|
if (LOG_TYPE == DEBUG) {
|
|
|
|
|
struct timespec tspec;
|
|
|
|
|
long BILLION = 1000000000;
|
|
|
|
|
clock_gettime(CLOCK_PROCESS_CPUTIME_ID, &tspec);
|
|
|
|
|
uint64_t sec = tspec.tv_sec;
|
|
|
|
|
uint64_t nsec = tspec.tv_nsec;
|
|
|
|
|
return ((sec * BILLION) + nsec);
|
|
|
|
|
} else {
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void print_time_difference(long int start_time, long int end_time) {
|
|
|
|
|
if (end_time > 0) {
|
|
|
|
|
LOG(DEBUG, "Time to copy: %u nanoseconds\n", end_time - start_time);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Store all data required for threading the saving to file functionality
|
|
|
|
|
struct ThreadSaveData {
|
|
|
|
|
struct device_data *device_data;
|
|
|
|
|
const char *filename;
|
|
|
|
|
const char *pFramebuffer;
|
|
|
|
|
VkSubresourceLayout srLayout;
|
|
|
|
|
VkFence fence;
|
|
|
|
|
uint32_t const width;
|
|
|
|
|
uint32_t const height;
|
2025-01-29 21:24:23 -08:00
|
|
|
uint32_t const numChannels;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
/* Write the copied image to a PNG file */
|
|
|
|
|
void *writePNG(void *data) {
|
|
|
|
|
struct ThreadSaveData *threadData = (struct ThreadSaveData*)data;
|
|
|
|
|
FILE *file;
|
2024-10-19 22:21:10 -07:00
|
|
|
size_t length = sizeof(char[LARGE_BUFFER_SIZE+STANDARD_BUFFER_SIZE]);
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
const char *tmpStr = ".tmp";
|
|
|
|
|
char *filename = (char *)malloc(length);
|
|
|
|
|
char *tmpFilename = (char *)malloc(length + 4); // Allow for ".tmp"
|
|
|
|
|
VkResult res;
|
2024-10-23 18:47:11 -07:00
|
|
|
png_byte *row_pointer;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
png_infop info;
|
2024-10-22 16:01:01 -07:00
|
|
|
png_struct* png;
|
2024-10-23 18:47:11 -07:00
|
|
|
uint64_t rowPitch = threadData->srLayout.rowPitch;
|
|
|
|
|
uint64_t start_time, end_time;
|
|
|
|
|
const int RGB_NUM_CHANNELS = 3;
|
2025-01-29 21:24:23 -08:00
|
|
|
const int RGBA_NUM_CHANNELS = 4;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
int localHeight = threadData->height;
|
|
|
|
|
int localWidth = threadData->width;
|
2025-01-29 21:24:23 -08:00
|
|
|
int numChannels = threadData->numChannels;
|
2024-10-23 18:47:11 -07:00
|
|
|
int matrixSize = localHeight * rowPitch;
|
|
|
|
|
bool checks_failed = true;
|
2024-10-22 16:01:01 -07:00
|
|
|
memcpy(filename, threadData->filename, length);
|
|
|
|
|
memcpy(tmpFilename, threadData->filename, length);
|
|
|
|
|
strcat(tmpFilename, tmpStr);
|
|
|
|
|
file = fopen(tmpFilename, "wb"); //create file for output
|
2024-10-23 18:47:11 -07:00
|
|
|
if (!file) {
|
2024-10-22 16:01:01 -07:00
|
|
|
LOG(ERROR, "Failed to open output file, '%s', error(%d): %s\n", tmpFilename, errno, strerror(errno));
|
2024-10-23 18:47:11 -07:00
|
|
|
goto cleanup;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
}
|
2024-10-23 18:47:11 -07:00
|
|
|
// TODO: Look into runtime version mismatch issue with some VK workloads
|
|
|
|
|
png = png_create_write_struct(PNG_LIBPNG_VER_STRING, NULL, NULL, NULL); //create structure for write PNG_LIBPNG_VER_STRING
|
|
|
|
|
if (!png) {
|
|
|
|
|
LOG(ERROR, "Create write struct failed. VER_STRING=%s\n", PNG_LIBPNG_VER_STRING);
|
|
|
|
|
goto cleanup;
|
|
|
|
|
}
|
|
|
|
|
info = png_create_info_struct(png);
|
|
|
|
|
if (!info) {
|
|
|
|
|
LOG(ERROR, "Create info struct failed\n");
|
|
|
|
|
goto cleanup;
|
|
|
|
|
}
|
|
|
|
|
if (setjmp(png_jmpbuf(png))) {
|
|
|
|
|
LOG(ERROR, "setjmp() failed\n");
|
|
|
|
|
goto cleanup;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
}
|
2024-10-23 18:47:11 -07:00
|
|
|
threadData->device_data->vtable.WaitForFences(threadData->device_data->device, 1, &threadData->fence, VK_TRUE, UINT64_MAX);
|
|
|
|
|
threadData->pFramebuffer += threadData->srLayout.offset;
|
|
|
|
|
start_time = get_time();
|
|
|
|
|
row_pointer = (png_byte *)malloc(sizeof(png_byte) * matrixSize);
|
|
|
|
|
memcpy(row_pointer, threadData->pFramebuffer, matrixSize);
|
2025-01-29 21:24:23 -08:00
|
|
|
/* Ensure alpha bits are set to 'opaque' if image is of RGBA format */
|
|
|
|
|
if (numChannels == RGBA_NUM_CHANNELS) {
|
|
|
|
|
for (int i = 3; i < matrixSize; i += RGBA_NUM_CHANNELS) {
|
|
|
|
|
row_pointer[i] = 0xFF;
|
|
|
|
|
}
|
|
|
|
|
}
|
2024-10-23 18:47:11 -07:00
|
|
|
end_time = get_time();
|
|
|
|
|
print_time_difference(start_time, end_time);
|
|
|
|
|
// We've created all local copies of data,
|
|
|
|
|
// so let's signal main thread to continue
|
|
|
|
|
pthread_cond_signal(&ptCondition);
|
|
|
|
|
png_init_io(png, file); // Initialize file output
|
|
|
|
|
png_set_IHDR( // Set image properties
|
|
|
|
|
png, // Pointer to png_struct
|
|
|
|
|
info, // Pointer to info_struct
|
|
|
|
|
localWidth, // Image width
|
|
|
|
|
localHeight, // Image height
|
|
|
|
|
8, // Color depth
|
2025-01-29 21:24:23 -08:00
|
|
|
numChannels == RGB_NUM_CHANNELS ? PNG_COLOR_TYPE_RGB : PNG_COLOR_TYPE_RGBA,
|
2024-10-23 18:47:11 -07:00
|
|
|
PNG_INTERLACE_NONE,
|
|
|
|
|
PNG_COMPRESSION_TYPE_DEFAULT,
|
|
|
|
|
PNG_FILTER_TYPE_DEFAULT
|
|
|
|
|
);
|
|
|
|
|
png_set_compression_level(png, 1); // Z_BEST_SPEED=1
|
|
|
|
|
png_set_compression_strategy(png, 2); // Z_HUFFMAN_ONLY=2
|
|
|
|
|
png_set_filter(png, PNG_FILTER_TYPE_BASE, PNG_FILTER_SUB);
|
|
|
|
|
png_set_compression_mem_level(png, 9);
|
|
|
|
|
png_set_compression_buffer_size(png, 65536);
|
|
|
|
|
png_write_info(png, info); // Write png image information to file
|
|
|
|
|
for (int y = 0; y < matrixSize; y+=rowPitch) {
|
|
|
|
|
png_write_row(png, &row_pointer[y]);
|
|
|
|
|
}
|
|
|
|
|
png_write_end(png, NULL); // End image writing
|
|
|
|
|
free(row_pointer);
|
|
|
|
|
|
|
|
|
|
// Rename file, indicating completion, client should be
|
|
|
|
|
// checking for the final file exists.
|
|
|
|
|
if (rename(tmpFilename, filename) != 0 )
|
|
|
|
|
LOG(ERROR, "Could not rename from '%s' to '%s'\n", tmpFilename, filename);
|
|
|
|
|
else
|
|
|
|
|
LOG(INFO, "Successfully renamed from '%s' to '%s'\n", tmpFilename, filename);
|
|
|
|
|
checks_failed = false;
|
|
|
|
|
cleanup:
|
|
|
|
|
if (checks_failed)
|
|
|
|
|
pthread_cond_signal(&ptCondition);
|
|
|
|
|
if (info)
|
|
|
|
|
png_destroy_write_struct(&png, &info);
|
2024-10-22 16:01:01 -07:00
|
|
|
if (file)
|
|
|
|
|
fclose(file);
|
2024-10-23 18:47:11 -07:00
|
|
|
if (filename)
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
free(filename);
|
2024-10-23 18:47:11 -07:00
|
|
|
if (tmpFilename)
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
free(tmpFilename);
|
|
|
|
|
return nullptr;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Write an image to file. Upon encountering issues, do not impact the
|
|
|
|
|
Present operation, */
|
|
|
|
|
static bool write_image(
|
|
|
|
|
const char* filename,
|
|
|
|
|
VkImage image,
|
|
|
|
|
struct device_data* device_data,
|
|
|
|
|
struct instance_data* instance_data,
|
2025-02-10 22:11:59 -08:00
|
|
|
struct queue_data* queue_data,
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
struct swapchain_data* swapchain_data)
|
|
|
|
|
{
|
|
|
|
|
VkDevice device = device_data->device;
|
|
|
|
|
VkPhysicalDevice physical_device = device_data->physical_device;
|
|
|
|
|
VkInstance instance = instance_data->instance;
|
|
|
|
|
|
|
|
|
|
uint32_t const width = swapchain_data->imageExtent.width;
|
|
|
|
|
uint32_t const height = swapchain_data->imageExtent.height;
|
|
|
|
|
VkFormat const format = swapchain_data->format;
|
|
|
|
|
|
vulkan/screenshot-layer: Add region command option
Adds an optional region selection, based off percentages of the
starting/ending of an image's X & Y values.
This is intended as a performance enhancement tradeoff for smaller
images to be created.
With a smaller image size, the screenshotting layer will change the
region boundaries on the GPU side, which will decrease the amount of
time it takes to copy the image over to CPU-accessible memory.
Using vkcube as an example, the original image size is 500x500:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 123530 nanoseconds
Then, by cropping the area to a 100x100 image, we get the following:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Using region: startX = 40% (200), startY = 40% (200), endX = 60% (300), endY = 60% (300)
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 12679 nanoseconds
For this example, this is a ~90% time reduction improvement!
Overall, this option reduces the copy time to a point where it can
become negligible, relative to the frame time of the application.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Reviewed-by: Felix DeGrood felix.j.degrood@intel.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32016>
2024-11-06 12:00:02 -08:00
|
|
|
uint32_t newWidth = width;
|
|
|
|
|
uint32_t newHeight = height;
|
|
|
|
|
uint32_t regionStartX = 0;
|
|
|
|
|
uint32_t regionStartY = 0;
|
|
|
|
|
uint32_t regionEndX = width;
|
|
|
|
|
uint32_t regionEndY = height;
|
|
|
|
|
if (instance_data->region_enabled) {
|
|
|
|
|
regionStartX = int(instance_data->params.region.startX * width);
|
|
|
|
|
regionStartY = int(instance_data->params.region.startY * height);
|
|
|
|
|
regionEndX = int(instance_data->params.region.endX * width);
|
|
|
|
|
regionEndY = int(instance_data->params.region.endY * height);
|
|
|
|
|
newWidth = regionEndX - regionStartX;
|
|
|
|
|
newHeight = regionEndY - regionStartY;
|
|
|
|
|
LOG(DEBUG, "Using region: startX = %.0f% (%d), startY = %.0f% (%d), endX = %.0f% (%d), endY = %.0f% (%d)\n",
|
|
|
|
|
instance_data->params.region.startX*100, regionStartX,
|
|
|
|
|
instance_data->params.region.startY*100, regionStartY,
|
|
|
|
|
instance_data->params.region.endX*100, regionEndX,
|
|
|
|
|
instance_data->params.region.endY*100, regionEndY);
|
|
|
|
|
}
|
|
|
|
|
|
2025-02-10 22:11:59 -08:00
|
|
|
VkQueue queue = getQueueForScreenshot(device_data, instance_data);
|
|
|
|
|
if (!queue) {
|
|
|
|
|
LOG(ERROR, "Unable to find a valid graphics-enabled queue\n");
|
|
|
|
|
return false;
|
|
|
|
|
}
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
|
|
|
|
|
VkResult err;
|
2025-01-29 21:24:23 -08:00
|
|
|
/* Attempt to set destination format to RGB to make writing to file much faster.
|
|
|
|
|
If not available, try to fall back to RGBA. If both fail, abort the screenshot */
|
|
|
|
|
VkFormat supported_formats[] = {VK_FORMAT_R8G8B8_UNORM, VK_FORMAT_R8G8B8A8_UNORM, VK_FORMAT_UNDEFINED};
|
|
|
|
|
uint32_t supported_formats_count = sizeof(supported_formats) / sizeof(VkFormat);
|
|
|
|
|
VkFormat destination_format;
|
|
|
|
|
uint32_t numChannels = 0;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
/* If origin and destination formats are the same, no need to convert */
|
|
|
|
|
bool copyOnly = false;
|
|
|
|
|
bool needs_2_steps = false;
|
2025-01-29 21:24:23 -08:00
|
|
|
bool blt_linear, blt_optimal;
|
|
|
|
|
VkFormatProperties device_format_properties;
|
|
|
|
|
|
|
|
|
|
for (uint32_t i = 0; i < supported_formats_count; i++) {
|
|
|
|
|
destination_format = supported_formats[i];
|
|
|
|
|
instance_data->pd_vtable.GetPhysicalDeviceFormatProperties(physical_device,
|
|
|
|
|
destination_format,
|
|
|
|
|
&device_format_properties);
|
|
|
|
|
if(destination_format == VK_FORMAT_UNDEFINED) {
|
|
|
|
|
LOG(ERROR, "Could not use the supported surface formats!\n");
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
return false;
|
|
|
|
|
}
|
2025-01-29 21:24:23 -08:00
|
|
|
if (destination_format == format && not instance_data->region_enabled) {
|
|
|
|
|
copyOnly = true;
|
|
|
|
|
LOG(DEBUG, "Only copying since the src/dest surface formats are the same.\n");
|
|
|
|
|
break;
|
|
|
|
|
} else {
|
|
|
|
|
blt_linear = device_format_properties.linearTilingFeatures & VK_FORMAT_FEATURE_BLIT_DST_BIT ? true : false;
|
|
|
|
|
blt_optimal = device_format_properties.optimalTilingFeatures & VK_FORMAT_FEATURE_BLIT_DST_BIT ? true : false;
|
|
|
|
|
if (!blt_linear && !blt_optimal) {
|
|
|
|
|
LOG(DEBUG, "Can't blit to linear nor optimal with surface format '%s'\n", vk_Format_to_str(supported_formats[i]));
|
|
|
|
|
} else if (blt_linear) {
|
|
|
|
|
break;
|
|
|
|
|
} else if (blt_optimal) {
|
|
|
|
|
// Can't blit to linear target, but can blit to optimal
|
|
|
|
|
needs_2_steps = true;
|
|
|
|
|
LOG(DEBUG, "Needs 2 steps\n");
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
LOG(DEBUG, "Using surface format '%s' for copy.\n", vk_Format_to_str(destination_format));
|
|
|
|
|
|
|
|
|
|
switch (destination_format)
|
|
|
|
|
{
|
|
|
|
|
case VK_FORMAT_R8G8B8_UNORM:
|
|
|
|
|
numChannels = 3;
|
|
|
|
|
break;
|
|
|
|
|
case VK_FORMAT_R8G8B8A8_UNORM:
|
|
|
|
|
numChannels = 4;
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
LOG(ERROR, "Unsupported format, aborting screenshot!\n");
|
|
|
|
|
break;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
WriteFileCleanupData data = {};
|
|
|
|
|
data.dev_data = device_data;
|
|
|
|
|
|
|
|
|
|
VkImageCreateInfo img_create_info2 = {
|
|
|
|
|
VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
|
|
|
|
|
NULL,
|
|
|
|
|
0,
|
|
|
|
|
VK_IMAGE_TYPE_2D,
|
|
|
|
|
destination_format,
|
vulkan/screenshot-layer: Add region command option
Adds an optional region selection, based off percentages of the
starting/ending of an image's X & Y values.
This is intended as a performance enhancement tradeoff for smaller
images to be created.
With a smaller image size, the screenshotting layer will change the
region boundaries on the GPU side, which will decrease the amount of
time it takes to copy the image over to CPU-accessible memory.
Using vkcube as an example, the original image size is 500x500:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 123530 nanoseconds
Then, by cropping the area to a 100x100 image, we get the following:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Using region: startX = 40% (200), startY = 40% (200), endX = 60% (300), endY = 60% (300)
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 12679 nanoseconds
For this example, this is a ~90% time reduction improvement!
Overall, this option reduces the copy time to a point where it can
become negligible, relative to the frame time of the application.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Reviewed-by: Felix DeGrood felix.j.degrood@intel.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32016>
2024-11-06 12:00:02 -08:00
|
|
|
{newWidth, newHeight, 1},
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
1,
|
|
|
|
|
1,
|
|
|
|
|
VK_SAMPLE_COUNT_1_BIT,
|
|
|
|
|
VK_IMAGE_TILING_LINEAR,
|
|
|
|
|
VK_IMAGE_USAGE_TRANSFER_DST_BIT,
|
|
|
|
|
VK_SHARING_MODE_EXCLUSIVE,
|
|
|
|
|
0,
|
|
|
|
|
NULL,
|
|
|
|
|
VK_IMAGE_LAYOUT_UNDEFINED,
|
|
|
|
|
};
|
|
|
|
|
VkImageCreateInfo img_create_info3 = img_create_info2;
|
|
|
|
|
|
|
|
|
|
if (needs_2_steps) {
|
|
|
|
|
img_create_info2.tiling = VK_IMAGE_TILING_OPTIMAL;
|
|
|
|
|
img_create_info2.usage = VK_IMAGE_USAGE_TRANSFER_SRC_BIT | VK_IMAGE_USAGE_TRANSFER_DST_BIT;
|
|
|
|
|
}
|
|
|
|
|
VkMemoryAllocateInfo mem_alloc_info = {
|
|
|
|
|
VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
|
|
|
|
|
NULL,
|
|
|
|
|
0,
|
|
|
|
|
0
|
|
|
|
|
};
|
|
|
|
|
VkMemoryRequirements mem_requirements;
|
|
|
|
|
VkPhysicalDeviceMemoryProperties mem_properties;
|
|
|
|
|
|
|
|
|
|
VK_CHECK(device_data->vtable.CreateImage(device, &img_create_info2, NULL, &data.image2));
|
|
|
|
|
device_data->vtable.GetImageMemoryRequirements(device, data.image2, &mem_requirements);
|
|
|
|
|
mem_alloc_info.allocationSize = mem_requirements.size;
|
|
|
|
|
instance_data->pd_vtable.GetPhysicalDeviceMemoryProperties(physical_device, &mem_properties);
|
|
|
|
|
if(!get_mem_type_from_properties(&mem_properties,
|
|
|
|
|
mem_requirements.memoryTypeBits,
|
|
|
|
|
needs_2_steps ? VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT : VK_MEMORY_PROPERTY_HOST_CACHED_BIT,
|
|
|
|
|
&mem_alloc_info.memoryTypeIndex)) {
|
|
|
|
|
LOG(ERROR, "Unable to get memory type from the intermediate/final image properties.\n");
|
|
|
|
|
return false;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
VK_CHECK(device_data->vtable.AllocateMemory(device, &mem_alloc_info, NULL, &data.mem2));
|
|
|
|
|
VK_CHECK(device_data->vtable.BindImageMemory(device, data.image2, data.mem2, 0));
|
|
|
|
|
|
|
|
|
|
if (needs_2_steps) {
|
|
|
|
|
VK_CHECK(device_data->vtable.CreateImage(device, &img_create_info3, NULL, &data.image3));
|
|
|
|
|
device_data->vtable.GetImageMemoryRequirements(device, data.image3, &mem_requirements);
|
|
|
|
|
mem_alloc_info.allocationSize = mem_requirements.size;
|
|
|
|
|
instance_data->pd_vtable.GetPhysicalDeviceMemoryProperties(physical_device, &mem_properties);
|
|
|
|
|
|
|
|
|
|
if(!get_mem_type_from_properties(&mem_properties,
|
|
|
|
|
mem_requirements.memoryTypeBits,
|
|
|
|
|
VK_MEMORY_PROPERTY_HOST_CACHED_BIT,
|
|
|
|
|
&mem_alloc_info.memoryTypeIndex)) {
|
|
|
|
|
LOG(ERROR, "Unable to get memory type from the temporary image properties.\n");
|
|
|
|
|
return false;
|
|
|
|
|
}
|
|
|
|
|
VK_CHECK(device_data->vtable.AllocateMemory(device, &mem_alloc_info, NULL, &data.mem3));
|
|
|
|
|
VK_CHECK(device_data->vtable.BindImageMemory(device, data.image3, data.mem3, 0));
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Setup command pool */
|
|
|
|
|
VkCommandPoolCreateInfo cmd_pool_info = {};
|
|
|
|
|
cmd_pool_info.sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO;
|
|
|
|
|
cmd_pool_info.pNext = NULL;
|
2025-02-10 22:11:59 -08:00
|
|
|
cmd_pool_info.queueFamilyIndex = queue_data->familyIndex;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
cmd_pool_info.flags = 0;
|
|
|
|
|
|
|
|
|
|
VK_CHECK(device_data->vtable.CreateCommandPool(device, &cmd_pool_info, NULL, &data.commandPool));
|
|
|
|
|
|
|
|
|
|
/* Set up command buffer */
|
|
|
|
|
const VkCommandBufferAllocateInfo cmd_buf_alloc_info = {VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO, NULL,
|
|
|
|
|
data.commandPool, VK_COMMAND_BUFFER_LEVEL_PRIMARY, 1};
|
|
|
|
|
VK_CHECK(device_data->vtable.AllocateCommandBuffers(device, &cmd_buf_alloc_info, &data.commandBuffer));
|
|
|
|
|
|
|
|
|
|
if (device_data->set_device_loader_data) {
|
|
|
|
|
VK_CHECK(device_data->set_device_loader_data(device, (void *)data.commandBuffer));
|
|
|
|
|
} else {
|
|
|
|
|
*((const void **)data.commandBuffer) = *(void **)device;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
const VkCommandBufferBeginInfo cmd_buf_begin_info = {
|
|
|
|
|
VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,
|
|
|
|
|
NULL,
|
|
|
|
|
VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,
|
|
|
|
|
};
|
|
|
|
|
VK_CHECK(device_data->vtable.BeginCommandBuffer(data.commandBuffer, &cmd_buf_begin_info));
|
|
|
|
|
|
|
|
|
|
// This barrier is used to transition from/to present Layout
|
|
|
|
|
VkImageMemoryBarrier presentMemoryBarrier = {VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
|
|
|
|
|
NULL,
|
|
|
|
|
VK_ACCESS_MEMORY_WRITE_BIT,
|
|
|
|
|
VK_ACCESS_TRANSFER_READ_BIT,
|
|
|
|
|
VK_IMAGE_LAYOUT_PRESENT_SRC_KHR,
|
|
|
|
|
VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
|
|
|
|
|
VK_QUEUE_FAMILY_IGNORED,
|
|
|
|
|
VK_QUEUE_FAMILY_IGNORED,
|
|
|
|
|
image,
|
|
|
|
|
{VK_IMAGE_ASPECT_COLOR_BIT, 0, 1, 0, 1}};
|
|
|
|
|
|
|
|
|
|
// This barrier is used to transition from a newly-created layout to a blt
|
|
|
|
|
// or copy destination layout.
|
|
|
|
|
VkImageMemoryBarrier destMemoryBarrier = {VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
|
|
|
|
|
NULL,
|
|
|
|
|
0,
|
|
|
|
|
VK_ACCESS_TRANSFER_WRITE_BIT,
|
|
|
|
|
VK_IMAGE_LAYOUT_UNDEFINED,
|
|
|
|
|
VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
|
|
|
|
|
VK_QUEUE_FAMILY_IGNORED,
|
|
|
|
|
VK_QUEUE_FAMILY_IGNORED,
|
|
|
|
|
data.image2,
|
|
|
|
|
{VK_IMAGE_ASPECT_COLOR_BIT, 0, 1, 0, 1}};
|
|
|
|
|
|
|
|
|
|
// This barrier is used to transition a dest layout to general layout.
|
|
|
|
|
VkImageMemoryBarrier generalMemoryBarrier = {VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
|
|
|
|
|
NULL,
|
|
|
|
|
VK_ACCESS_TRANSFER_WRITE_BIT,
|
|
|
|
|
VK_ACCESS_MEMORY_READ_BIT,
|
|
|
|
|
VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
|
|
|
|
|
VK_IMAGE_LAYOUT_GENERAL,
|
|
|
|
|
VK_QUEUE_FAMILY_IGNORED,
|
|
|
|
|
VK_QUEUE_FAMILY_IGNORED,
|
|
|
|
|
data.image2,
|
|
|
|
|
{VK_IMAGE_ASPECT_COLOR_BIT, 0, 1, 0, 1}};
|
|
|
|
|
|
|
|
|
|
VkPipelineStageFlags srcStages = VK_PIPELINE_STAGE_TRANSFER_BIT;
|
|
|
|
|
VkPipelineStageFlags dstStages = VK_PIPELINE_STAGE_TRANSFER_BIT;
|
|
|
|
|
|
|
|
|
|
device_data->vtable.CmdPipelineBarrier(data.commandBuffer, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT,
|
|
|
|
|
dstStages, 0, 0, NULL, 0, NULL, 1, &presentMemoryBarrier);
|
|
|
|
|
|
|
|
|
|
device_data->vtable.CmdPipelineBarrier(data.commandBuffer, srcStages, dstStages, 0, 0, NULL, 0, NULL, 1, &destMemoryBarrier);
|
|
|
|
|
|
|
|
|
|
const VkImageCopy img_copy = {
|
|
|
|
|
{VK_IMAGE_ASPECT_COLOR_BIT, 0, 0, 1},
|
|
|
|
|
{0, 0, 0},
|
|
|
|
|
{VK_IMAGE_ASPECT_COLOR_BIT, 0, 0, 1},
|
|
|
|
|
{0, 0, 0},
|
vulkan/screenshot-layer: Add region command option
Adds an optional region selection, based off percentages of the
starting/ending of an image's X & Y values.
This is intended as a performance enhancement tradeoff for smaller
images to be created.
With a smaller image size, the screenshotting layer will change the
region boundaries on the GPU side, which will decrease the amount of
time it takes to copy the image over to CPU-accessible memory.
Using vkcube as an example, the original image size is 500x500:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 123530 nanoseconds
Then, by cropping the area to a 100x100 image, we get the following:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Using region: startX = 40% (200), startY = 40% (200), endX = 60% (300), endY = 60% (300)
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 12679 nanoseconds
For this example, this is a ~90% time reduction improvement!
Overall, this option reduces the copy time to a point where it can
become negligible, relative to the frame time of the application.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Reviewed-by: Felix DeGrood felix.j.degrood@intel.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32016>
2024-11-06 12:00:02 -08:00
|
|
|
{newWidth, newHeight, 1}
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
if (copyOnly) {
|
|
|
|
|
device_data->vtable.CmdCopyImage(data.commandBuffer, image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, data.image2,
|
|
|
|
|
VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &img_copy);
|
|
|
|
|
} else {
|
|
|
|
|
VkImageBlit imageBlitRegion = {};
|
|
|
|
|
imageBlitRegion.srcSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
|
|
|
|
|
imageBlitRegion.srcSubresource.baseArrayLayer = 0;
|
|
|
|
|
imageBlitRegion.srcSubresource.layerCount = 1;
|
|
|
|
|
imageBlitRegion.srcSubresource.mipLevel = 0;
|
vulkan/screenshot-layer: Add region command option
Adds an optional region selection, based off percentages of the
starting/ending of an image's X & Y values.
This is intended as a performance enhancement tradeoff for smaller
images to be created.
With a smaller image size, the screenshotting layer will change the
region boundaries on the GPU side, which will decrease the amount of
time it takes to copy the image over to CPU-accessible memory.
Using vkcube as an example, the original image size is 500x500:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 123530 nanoseconds
Then, by cropping the area to a 100x100 image, we get the following:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Using region: startX = 40% (200), startY = 40% (200), endX = 60% (300), endY = 60% (300)
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 12679 nanoseconds
For this example, this is a ~90% time reduction improvement!
Overall, this option reduces the copy time to a point where it can
become negligible, relative to the frame time of the application.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Reviewed-by: Felix DeGrood felix.j.degrood@intel.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32016>
2024-11-06 12:00:02 -08:00
|
|
|
imageBlitRegion.srcOffsets[0].x = regionStartX;
|
|
|
|
|
imageBlitRegion.srcOffsets[0].y = regionStartY;
|
2025-02-10 22:11:59 -08:00
|
|
|
imageBlitRegion.srcOffsets[0].z = 0;
|
vulkan/screenshot-layer: Add region command option
Adds an optional region selection, based off percentages of the
starting/ending of an image's X & Y values.
This is intended as a performance enhancement tradeoff for smaller
images to be created.
With a smaller image size, the screenshotting layer will change the
region boundaries on the GPU side, which will decrease the amount of
time it takes to copy the image over to CPU-accessible memory.
Using vkcube as an example, the original image size is 500x500:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 123530 nanoseconds
Then, by cropping the area to a 100x100 image, we get the following:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Using region: startX = 40% (200), startY = 40% (200), endX = 60% (300), endY = 60% (300)
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 12679 nanoseconds
For this example, this is a ~90% time reduction improvement!
Overall, this option reduces the copy time to a point where it can
become negligible, relative to the frame time of the application.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Reviewed-by: Felix DeGrood felix.j.degrood@intel.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32016>
2024-11-06 12:00:02 -08:00
|
|
|
imageBlitRegion.srcOffsets[1].x = regionEndX;
|
|
|
|
|
imageBlitRegion.srcOffsets[1].y = regionEndY;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
imageBlitRegion.srcOffsets[1].z = 1;
|
|
|
|
|
imageBlitRegion.dstSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
|
|
|
|
|
imageBlitRegion.dstSubresource.baseArrayLayer = 0;
|
|
|
|
|
imageBlitRegion.dstSubresource.layerCount = 1;
|
|
|
|
|
imageBlitRegion.dstSubresource.mipLevel = 0;
|
vulkan/screenshot-layer: Add region command option
Adds an optional region selection, based off percentages of the
starting/ending of an image's X & Y values.
This is intended as a performance enhancement tradeoff for smaller
images to be created.
With a smaller image size, the screenshotting layer will change the
region boundaries on the GPU side, which will decrease the amount of
time it takes to copy the image over to CPU-accessible memory.
Using vkcube as an example, the original image size is 500x500:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 123530 nanoseconds
Then, by cropping the area to a 100x100 image, we get the following:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Using region: startX = 40% (200), startY = 40% (200), endX = 60% (300), endY = 60% (300)
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 12679 nanoseconds
For this example, this is a ~90% time reduction improvement!
Overall, this option reduces the copy time to a point where it can
become negligible, relative to the frame time of the application.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Reviewed-by: Felix DeGrood felix.j.degrood@intel.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32016>
2024-11-06 12:00:02 -08:00
|
|
|
imageBlitRegion.dstOffsets[1].x = newWidth;
|
|
|
|
|
imageBlitRegion.dstOffsets[1].y = newHeight;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
imageBlitRegion.dstOffsets[1].z = 1;
|
|
|
|
|
|
|
|
|
|
device_data->vtable.CmdBlitImage(data.commandBuffer, image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, data.image2,
|
|
|
|
|
VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &imageBlitRegion, VK_FILTER_NEAREST);
|
|
|
|
|
if (needs_2_steps) {
|
|
|
|
|
// image 3 needs to be transitioned from its undefined state to a
|
|
|
|
|
// transfer destination.
|
|
|
|
|
destMemoryBarrier.image = data.image3;
|
|
|
|
|
device_data->vtable.CmdPipelineBarrier(data.commandBuffer, srcStages, dstStages, 0, 0, NULL, 0, NULL, 1, &destMemoryBarrier);
|
|
|
|
|
|
|
|
|
|
// Transition image2 so that it can be read for the upcoming copy to
|
|
|
|
|
// image 3.
|
|
|
|
|
destMemoryBarrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;
|
|
|
|
|
destMemoryBarrier.newLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;
|
|
|
|
|
destMemoryBarrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
|
|
|
|
|
destMemoryBarrier.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
|
|
|
|
|
destMemoryBarrier.image = data.image2;
|
|
|
|
|
device_data->vtable.CmdPipelineBarrier(data.commandBuffer, srcStages, dstStages, 0, 0, NULL, 0, NULL, 1,
|
|
|
|
|
&destMemoryBarrier);
|
|
|
|
|
|
|
|
|
|
// This step essentially untiles the image.
|
|
|
|
|
device_data->vtable.CmdCopyImage(data.commandBuffer, data.image2, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, data.image3,
|
|
|
|
|
VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &img_copy);
|
|
|
|
|
generalMemoryBarrier.image = data.image3;
|
|
|
|
|
}
|
2024-10-19 15:28:28 -07:00
|
|
|
}
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
|
|
|
|
|
// The destination needs to be transitioned from the optimal copy format to
|
|
|
|
|
// the format we can read with the CPU.
|
|
|
|
|
device_data->vtable.CmdPipelineBarrier(data.commandBuffer, srcStages, dstStages, 0, 0, NULL, 0, NULL, 1, &generalMemoryBarrier);
|
|
|
|
|
|
|
|
|
|
// Restore the swap chain image layout to what it was before.
|
|
|
|
|
// This may not be strictly needed, but it is generally good to restore
|
|
|
|
|
// things to original state.
|
|
|
|
|
presentMemoryBarrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;
|
|
|
|
|
presentMemoryBarrier.newLayout = VK_IMAGE_LAYOUT_PRESENT_SRC_KHR;
|
|
|
|
|
presentMemoryBarrier.srcAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
|
|
|
|
|
presentMemoryBarrier.dstAccessMask = 0;
|
|
|
|
|
device_data->vtable.CmdPipelineBarrier(data.commandBuffer, srcStages, dstStages, 0, 0, NULL, 0, NULL, 1,
|
|
|
|
|
&presentMemoryBarrier);
|
|
|
|
|
VK_CHECK(device_data->vtable.EndCommandBuffer(data.commandBuffer));
|
|
|
|
|
|
|
|
|
|
VkSubmitInfo submitInfo;
|
|
|
|
|
submitInfo.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO;
|
|
|
|
|
submitInfo.pNext = NULL;
|
|
|
|
|
submitInfo.waitSemaphoreCount = semaphoreWaitBeforePresentCount;
|
|
|
|
|
submitInfo.pWaitSemaphores = pSemaphoreWaitBeforePresent;
|
|
|
|
|
submitInfo.pWaitDstStageMask = &dstStageWaitBeforeSubmission;
|
|
|
|
|
submitInfo.commandBufferCount = 1;
|
|
|
|
|
submitInfo.pCommandBuffers = &data.commandBuffer;
|
|
|
|
|
submitInfo.signalSemaphoreCount = 1;
|
|
|
|
|
submitInfo.pSignalSemaphores = &semaphoreWaitAfterSubmission;
|
|
|
|
|
VK_CHECK(device_data->vtable.QueueSubmit(queue, 1, &submitInfo, copyDone));
|
|
|
|
|
|
|
|
|
|
// Map the final image so that the CPU can read it.
|
|
|
|
|
const VkImageSubresource img_subresource = {VK_IMAGE_ASPECT_COLOR_BIT, 0, 0};
|
|
|
|
|
VkSubresourceLayout srLayout;
|
|
|
|
|
const char *pFramebuffer;
|
|
|
|
|
if (!needs_2_steps) {
|
|
|
|
|
device_data->vtable.GetImageSubresourceLayout(device, data.image2, &img_subresource, &srLayout);
|
|
|
|
|
VK_CHECK(device_data->vtable.MapMemory(device, data.mem2, 0, VK_WHOLE_SIZE, 0, (void **)&pFramebuffer));
|
|
|
|
|
data.mem2mapped = true;
|
|
|
|
|
} else {
|
|
|
|
|
device_data->vtable.GetImageSubresourceLayout(device, data.image3, &img_subresource, &srLayout);
|
|
|
|
|
VK_CHECK(device_data->vtable.MapMemory(device, data.mem3, 0, VK_WHOLE_SIZE, 0, (void **)&pFramebuffer));
|
|
|
|
|
data.mem3mapped = true;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Thread off I/O operations
|
|
|
|
|
pthread_t ioThread;
|
|
|
|
|
pthread_mutex_lock(&ptLock); // Grab lock, we need to wait until thread has copied values of pointers
|
2025-01-29 21:24:23 -08:00
|
|
|
struct ThreadSaveData threadData = {device_data, filename, pFramebuffer, srLayout, copyDone, newWidth, newHeight, numChannels};
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
|
|
|
|
|
// Write the data to a PNG file.
|
|
|
|
|
pthread_create(&ioThread, NULL, writePNG, (void *)&threadData);
|
|
|
|
|
pthread_detach(ioThread); // Reclaim resources once thread terminates
|
|
|
|
|
pthread_cond_wait(&ptCondition, &ptLock);
|
|
|
|
|
pthread_mutex_unlock(&ptLock);
|
|
|
|
|
|
|
|
|
|
return true;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static VkResult screenshot_QueuePresentKHR(
|
|
|
|
|
VkQueue queue,
|
|
|
|
|
const VkPresentInfoKHR* pPresentInfo)
|
|
|
|
|
{
|
|
|
|
|
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;
|
|
|
|
|
|
|
|
|
|
VkPresentInfoKHR present_info = *pPresentInfo;
|
|
|
|
|
|
|
|
|
|
static uint32_t frame_counter = 0;
|
|
|
|
|
|
|
|
|
|
VkResult result = VK_SUCCESS;
|
|
|
|
|
loader_platform_thread_lock_mutex(&globalLock);
|
2025-02-10 22:11:59 -08:00
|
|
|
VkSemaphoreCreateInfo semaphoreInfo = {};
|
|
|
|
|
VkFenceCreateInfo fenceInfo = {};
|
|
|
|
|
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
if (pPresentInfo && pPresentInfo->swapchainCount > 0) {
|
|
|
|
|
VkSwapchainKHR swapchain = pPresentInfo->pSwapchains[0];
|
|
|
|
|
|
|
|
|
|
struct swapchain_data *swapchain_data = FIND(struct swapchain_data, swapchain);
|
|
|
|
|
|
|
|
|
|
/* Run initial setup with client */
|
vulkan/screenshot-layer: Add region command option
Adds an optional region selection, based off percentages of the
starting/ending of an image's X & Y values.
This is intended as a performance enhancement tradeoff for smaller
images to be created.
With a smaller image size, the screenshotting layer will change the
region boundaries on the GPU side, which will decrease the amount of
time it takes to copy the image over to CPU-accessible memory.
Using vkcube as an example, the original image size is 500x500:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 123530 nanoseconds
Then, by cropping the area to a 100x100 image, we get the following:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Using region: startX = 40% (200), startY = 40% (200), endX = 60% (300), endY = 60% (300)
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 12679 nanoseconds
For this example, this is a ~90% time reduction improvement!
Overall, this option reduces the copy time to a point where it can
become negligible, relative to the frame time of the application.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Reviewed-by: Felix DeGrood felix.j.degrood@intel.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32016>
2024-11-06 12:00:02 -08:00
|
|
|
if (instance_data->params.enabled[SCREENSHOT_PARAM_ENABLED_comms] && instance_data->socket_fd < 0) {
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
int ret = os_socket_listen_abstract(instance_data->params.control, 1);
|
|
|
|
|
if (ret >= 0) {
|
|
|
|
|
os_socket_block(ret, false);
|
|
|
|
|
instance_data->socket_fd = ret;
|
|
|
|
|
}
|
|
|
|
|
if (instance_data->socket_fd >= 0)
|
|
|
|
|
LOG(INFO, "socket set! Waiting for client input...\n");
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (instance_data->socket_fd >= 0) {
|
vulkan/screenshot-layer: Add region command option
Adds an optional region selection, based off percentages of the
starting/ending of an image's X & Y values.
This is intended as a performance enhancement tradeoff for smaller
images to be created.
With a smaller image size, the screenshotting layer will change the
region boundaries on the GPU side, which will decrease the amount of
time it takes to copy the image over to CPU-accessible memory.
Using vkcube as an example, the original image size is 500x500:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 123530 nanoseconds
Then, by cropping the area to a 100x100 image, we get the following:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Using region: startX = 40% (200), startY = 40% (200), endX = 60% (300), endY = 60% (300)
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 12679 nanoseconds
For this example, this is a ~90% time reduction improvement!
Overall, this option reduces the copy time to a point where it can
become negligible, relative to the frame time of the application.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Reviewed-by: Felix DeGrood felix.j.degrood@intel.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32016>
2024-11-06 12:00:02 -08:00
|
|
|
/* Check client commands first */
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
control_client_check(device_data);
|
|
|
|
|
process_control_socket(instance_data);
|
|
|
|
|
} else if (instance_data->params.frames) {
|
vulkan/screenshot-layer: Add region command option
Adds an optional region selection, based off percentages of the
starting/ending of an image's X & Y values.
This is intended as a performance enhancement tradeoff for smaller
images to be created.
With a smaller image size, the screenshotting layer will change the
region boundaries on the GPU side, which will decrease the amount of
time it takes to copy the image over to CPU-accessible memory.
Using vkcube as an example, the original image size is 500x500:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 123530 nanoseconds
Then, by cropping the area to a 100x100 image, we get the following:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Using region: startX = 40% (200), startY = 40% (200), endX = 60% (300), endY = 60% (300)
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 12679 nanoseconds
For this example, this is a ~90% time reduction improvement!
Overall, this option reduces the copy time to a point where it can
become negligible, relative to the frame time of the application.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Reviewed-by: Felix DeGrood felix.j.degrood@intel.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32016>
2024-11-06 12:00:02 -08:00
|
|
|
/* Else check parameters from env variables */
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
if (instance_data->params.frames->size > 0) {
|
|
|
|
|
struct frame_list *list = instance_data->params.frames;
|
|
|
|
|
struct frame_node *prev = nullptr;
|
|
|
|
|
for (struct frame_node *node = list->head; node!=nullptr; prev = node, node = node->next) {
|
|
|
|
|
if (frame_counter < node->frame_num){
|
|
|
|
|
break;
|
|
|
|
|
} else if (frame_counter == node->frame_num) {
|
|
|
|
|
instance_data->screenshot_enabled = true;
|
|
|
|
|
remove_node(list, prev, node);
|
|
|
|
|
break;
|
|
|
|
|
} else {
|
|
|
|
|
LOG(ERROR, "mesa-screenshot: Somehow encountered a higher number "
|
|
|
|
|
"than what exists in the frame list. Won't capture frame!\n");
|
|
|
|
|
destroy_frame_list(list);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
vulkan/screenshot-layer: Add region command option
Adds an optional region selection, based off percentages of the
starting/ending of an image's X & Y values.
This is intended as a performance enhancement tradeoff for smaller
images to be created.
With a smaller image size, the screenshotting layer will change the
region boundaries on the GPU side, which will decrease the amount of
time it takes to copy the image over to CPU-accessible memory.
Using vkcube as an example, the original image size is 500x500:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 123530 nanoseconds
Then, by cropping the area to a 100x100 image, we get the following:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Using region: startX = 40% (200), startY = 40% (200), endX = 60% (300), endY = 60% (300)
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 12679 nanoseconds
For this example, this is a ~90% time reduction improvement!
Overall, this option reduces the copy time to a point where it can
become negligible, relative to the frame time of the application.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Reviewed-by: Felix DeGrood felix.j.degrood@intel.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32016>
2024-11-06 12:00:02 -08:00
|
|
|
} else if (instance_data->params.frames->all_frames) {
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
instance_data->screenshot_enabled = true;
|
|
|
|
|
}
|
vulkan/screenshot-layer: Add region command option
Adds an optional region selection, based off percentages of the
starting/ending of an image's X & Y values.
This is intended as a performance enhancement tradeoff for smaller
images to be created.
With a smaller image size, the screenshotting layer will change the
region boundaries on the GPU side, which will decrease the amount of
time it takes to copy the image over to CPU-accessible memory.
Using vkcube as an example, the original image size is 500x500:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 123530 nanoseconds
Then, by cropping the area to a 100x100 image, we get the following:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Using region: startX = 40% (200), startY = 40% (200), endX = 60% (300), endY = 60% (300)
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 12679 nanoseconds
For this example, this is a ~90% time reduction improvement!
Overall, this option reduces the copy time to a point where it can
become negligible, relative to the frame time of the application.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Reviewed-by: Felix DeGrood felix.j.degrood@intel.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32016>
2024-11-06 12:00:02 -08:00
|
|
|
if (instance_data->params.region.useImageRegion) {
|
|
|
|
|
instance_data->region_enabled = true;
|
|
|
|
|
}
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (instance_data->screenshot_enabled) {
|
|
|
|
|
LOG(DEBUG, "Screenshot Authorized!\n");
|
|
|
|
|
uint32_t SUFFIX_SIZE = 4; // strlen('.png') == 4;
|
|
|
|
|
uint32_t path_size_used = 0;
|
|
|
|
|
const char *SUFFIX = ".png";
|
|
|
|
|
const char *TEMP_DIR = "/tmp/";
|
2024-10-19 22:21:10 -07:00
|
|
|
char full_path[LARGE_BUFFER_SIZE+STANDARD_BUFFER_SIZE] = "";
|
|
|
|
|
char filename[STANDARD_BUFFER_SIZE] = "";
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
char frame_counter_str[11];
|
|
|
|
|
bool rename_file = true;
|
|
|
|
|
itoa(frame_counter, frame_counter_str);
|
|
|
|
|
|
|
|
|
|
/* Check if we have an output directory given from the env options */
|
|
|
|
|
if (instance_data->params.output_dir &&
|
|
|
|
|
strlen(instance_data->params.output_dir) > 0) {
|
|
|
|
|
strcat(full_path, instance_data->params.output_dir);
|
|
|
|
|
} else {
|
|
|
|
|
memcpy(full_path, TEMP_DIR, strlen(TEMP_DIR));
|
|
|
|
|
}
|
|
|
|
|
path_size_used += strlen(full_path);
|
|
|
|
|
/* Check if we have a filename from the client */
|
|
|
|
|
if (instance_data->filename && strlen(instance_data->filename) > SUFFIX_SIZE) {
|
|
|
|
|
/* Confirm that filename is of form '<name>.png' */
|
|
|
|
|
uint32_t name_len = strlen(instance_data->filename);
|
|
|
|
|
const char *suffix_ptr = &instance_data->filename[name_len - SUFFIX_SIZE];
|
|
|
|
|
if (!strcmp(suffix_ptr, SUFFIX)) {
|
|
|
|
|
rename_file = false;
|
|
|
|
|
strcpy(filename, instance_data->filename);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
if (rename_file) {
|
|
|
|
|
strcat(filename, frame_counter_str);
|
|
|
|
|
strcat(filename, SUFFIX);
|
|
|
|
|
}
|
|
|
|
|
path_size_used += strlen(filename);
|
2024-10-19 22:21:10 -07:00
|
|
|
if(path_size_used <= LARGE_BUFFER_SIZE+STANDARD_BUFFER_SIZE) {
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
strcat(full_path, filename);
|
|
|
|
|
pSemaphoreWaitBeforePresent = pPresentInfo->pWaitSemaphores;
|
|
|
|
|
semaphoreWaitBeforePresentCount = pPresentInfo->waitSemaphoreCount;
|
|
|
|
|
semaphoreInfo.sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO;
|
|
|
|
|
device_data->vtable.CreateSemaphore(device_data->device, &semaphoreInfo, nullptr, &semaphoreWaitAfterSubmission);
|
|
|
|
|
fenceInfo.sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO;
|
|
|
|
|
device_data->vtable.CreateFence(device_data->device, &fenceInfo, nullptr, ©Done);
|
|
|
|
|
if(write_image(full_path,
|
|
|
|
|
swapchain_data->image,
|
|
|
|
|
device_data,
|
|
|
|
|
instance_data,
|
2025-02-10 22:11:59 -08:00
|
|
|
queue_data,
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
swapchain_data)) {
|
|
|
|
|
present_info.pWaitSemaphores = &semaphoreWaitAfterSubmission; // Make semaphore here
|
|
|
|
|
present_info.waitSemaphoreCount = 1;
|
|
|
|
|
}
|
|
|
|
|
} else {
|
2024-10-19 22:21:10 -07:00
|
|
|
LOG(DEBUG, "Cancelling screenshot due to excessive filepath size (max %u characters)\n", LARGE_BUFFER_SIZE);
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
frame_counter++;
|
|
|
|
|
instance_data->screenshot_enabled = false;
|
vulkan/screenshot-layer: Add region command option
Adds an optional region selection, based off percentages of the
starting/ending of an image's X & Y values.
This is intended as a performance enhancement tradeoff for smaller
images to be created.
With a smaller image size, the screenshotting layer will change the
region boundaries on the GPU side, which will decrease the amount of
time it takes to copy the image over to CPU-accessible memory.
Using vkcube as an example, the original image size is 500x500:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 123530 nanoseconds
Then, by cropping the area to a 100x100 image, we get the following:
mesa-screenshot: DEBUG: Screenshot Authorized!
mesa-screenshot: DEBUG: Using region: startX = 40% (200), startY = 40% (200), endX = 60% (300), endY = 60% (300)
mesa-screenshot: DEBUG: Needs 2 steps
mesa-screenshot: DEBUG: Time to copy: 12679 nanoseconds
For this example, this is a ~90% time reduction improvement!
Overall, this option reduces the copy time to a point where it can
become negligible, relative to the frame time of the application.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Reviewed-by: Felix DeGrood felix.j.degrood@intel.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32016>
2024-11-06 12:00:02 -08:00
|
|
|
instance_data->region_enabled = false;
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
loader_platform_thread_unlock_mutex(&globalLock);
|
|
|
|
|
VkResult chain_result = queue_data->device->vtable.QueuePresentKHR(queue, &present_info);
|
|
|
|
|
if (pPresentInfo->pResults)
|
|
|
|
|
pPresentInfo->pResults[0] = chain_result;
|
|
|
|
|
if (chain_result != VK_SUCCESS && result == VK_SUCCESS)
|
|
|
|
|
result = chain_result;
|
|
|
|
|
|
|
|
|
|
if (semaphoreWaitAfterSubmission != VK_NULL_HANDLE) {
|
|
|
|
|
device_data->vtable.DestroySemaphore(device_data->device, semaphoreWaitAfterSubmission, nullptr);
|
|
|
|
|
semaphoreWaitAfterSubmission = VK_NULL_HANDLE;
|
|
|
|
|
}
|
|
|
|
|
if (copyDone != VK_NULL_HANDLE) {
|
|
|
|
|
device_data->vtable.DestroyFence(device_data->device, copyDone, nullptr);
|
|
|
|
|
copyDone = VK_NULL_HANDLE;
|
|
|
|
|
}
|
|
|
|
|
return result;
|
|
|
|
|
}
|
|
|
|
|
|
2024-10-19 15:28:28 -07:00
|
|
|
static VkResult screenshot_AcquireNextImageKHR(
|
|
|
|
|
VkDevice device,
|
|
|
|
|
VkSwapchainKHR swapchain,
|
|
|
|
|
uint64_t timeout,
|
|
|
|
|
VkSemaphore semaphore,
|
|
|
|
|
VkFence fence,
|
|
|
|
|
uint32_t* pImageIndex)
|
|
|
|
|
{
|
|
|
|
|
struct swapchain_data *swapchain_data =
|
|
|
|
|
FIND(struct swapchain_data, swapchain);
|
|
|
|
|
struct device_data *device_data = swapchain_data->device;
|
|
|
|
|
|
|
|
|
|
VkResult result = device_data->vtable.AcquireNextImageKHR(device, swapchain, timeout,
|
|
|
|
|
semaphore, fence, pImageIndex);
|
|
|
|
|
loader_platform_thread_lock_mutex(&globalLock);
|
|
|
|
|
|
|
|
|
|
if (result == VK_SUCCESS) {
|
|
|
|
|
// Use the index given by AcquireNextImageKHR() to obtain the image we intend to copy.
|
|
|
|
|
if(pImageIndex){
|
|
|
|
|
select_image_from_map(swapchain_data, *pImageIndex);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
loader_platform_thread_unlock_mutex(&globalLock);
|
|
|
|
|
return result;
|
|
|
|
|
}
|
|
|
|
|
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
static VkResult screenshot_CreateDevice(
|
|
|
|
|
VkPhysicalDevice physicalDevice,
|
|
|
|
|
const VkDeviceCreateInfo* pCreateInfo,
|
|
|
|
|
const VkAllocationCallbacks* pAllocator,
|
|
|
|
|
VkDevice* pDevice)
|
|
|
|
|
{
|
|
|
|
|
struct instance_data *instance_data =
|
|
|
|
|
FIND(struct instance_data, physicalDevice);
|
|
|
|
|
VkLayerDeviceCreateInfo *chain_info =
|
|
|
|
|
get_device_chain_info(pCreateInfo, VK_LAYER_LINK_INFO);
|
|
|
|
|
assert(chain_info->u.pLayerInfo);
|
|
|
|
|
PFN_vkGetInstanceProcAddr fpGetInstanceProcAddr = chain_info->u.pLayerInfo->pfnNextGetInstanceProcAddr;
|
|
|
|
|
PFN_vkGetDeviceProcAddr fpGetDeviceProcAddr = chain_info->u.pLayerInfo->pfnNextGetDeviceProcAddr;
|
|
|
|
|
PFN_vkCreateDevice fpCreateDevice = (PFN_vkCreateDevice)fpGetInstanceProcAddr(NULL, "vkCreateDevice");
|
|
|
|
|
if (fpCreateDevice == NULL) {
|
|
|
|
|
return VK_ERROR_INITIALIZATION_FAILED;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Advance the link info for the next element on the chain
|
|
|
|
|
chain_info->u.pLayerInfo = chain_info->u.pLayerInfo->pNext;
|
|
|
|
|
|
|
|
|
|
VkDeviceCreateInfo create_info = *pCreateInfo;
|
|
|
|
|
|
|
|
|
|
VkResult result = fpCreateDevice(physicalDevice, &create_info, pAllocator, pDevice);
|
|
|
|
|
if (result != VK_SUCCESS) return result;
|
|
|
|
|
|
|
|
|
|
struct device_data *device_data = new_device_data(*pDevice, instance_data);
|
|
|
|
|
device_data->physical_device = physicalDevice;
|
|
|
|
|
vk_device_dispatch_table_load(&device_data->vtable,
|
|
|
|
|
fpGetDeviceProcAddr, *pDevice);
|
|
|
|
|
|
|
|
|
|
instance_data->pd_vtable.GetPhysicalDeviceProperties(device_data->physical_device,
|
|
|
|
|
&device_data->properties);
|
|
|
|
|
|
|
|
|
|
VkLayerDeviceCreateInfo *load_data_info =
|
|
|
|
|
get_device_chain_info(pCreateInfo, VK_LOADER_DATA_CALLBACK);
|
|
|
|
|
|
|
|
|
|
device_data->set_device_loader_data = load_data_info->u.pfnSetDeviceLoaderData;
|
|
|
|
|
return result;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void screenshot_DestroyDevice(
|
|
|
|
|
VkDevice device,
|
|
|
|
|
const VkAllocationCallbacks* pAllocator)
|
|
|
|
|
{
|
|
|
|
|
struct device_data *device_data = FIND(struct device_data, device);
|
|
|
|
|
device_data->vtable.DestroyDevice(device, pAllocator);
|
|
|
|
|
destroy_device_data(device_data);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static VkResult screenshot_CreateInstance(
|
|
|
|
|
const VkInstanceCreateInfo* pCreateInfo,
|
|
|
|
|
const VkAllocationCallbacks* pAllocator,
|
|
|
|
|
VkInstance* pInstance)
|
|
|
|
|
{
|
|
|
|
|
VkLayerInstanceCreateInfo *chain_info =
|
|
|
|
|
get_instance_chain_info(pCreateInfo, VK_LAYER_LINK_INFO);
|
|
|
|
|
|
|
|
|
|
assert(chain_info->u.pLayerInfo);
|
|
|
|
|
PFN_vkGetInstanceProcAddr fpGetInstanceProcAddr =
|
|
|
|
|
chain_info->u.pLayerInfo->pfnNextGetInstanceProcAddr;
|
|
|
|
|
PFN_vkCreateInstance fpCreateInstance =
|
|
|
|
|
(PFN_vkCreateInstance)fpGetInstanceProcAddr(NULL, "vkCreateInstance");
|
|
|
|
|
if (fpCreateInstance == NULL) {
|
|
|
|
|
return VK_ERROR_INITIALIZATION_FAILED;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Advance the link info for the next element on the chain
|
|
|
|
|
chain_info->u.pLayerInfo = chain_info->u.pLayerInfo->pNext;
|
|
|
|
|
|
|
|
|
|
VkResult result = fpCreateInstance(pCreateInfo, pAllocator, pInstance);
|
|
|
|
|
if (result != VK_SUCCESS) return result;
|
|
|
|
|
|
|
|
|
|
struct instance_data *instance_data = new_instance_data(*pInstance);
|
|
|
|
|
vk_instance_dispatch_table_load(&instance_data->vtable,
|
|
|
|
|
fpGetInstanceProcAddr,
|
|
|
|
|
instance_data->instance);
|
|
|
|
|
vk_physical_device_dispatch_table_load(&instance_data->pd_vtable,
|
|
|
|
|
fpGetInstanceProcAddr,
|
|
|
|
|
instance_data->instance);
|
|
|
|
|
instance_data_map_physical_devices(instance_data, true);
|
|
|
|
|
|
|
|
|
|
parse_screenshot_env(&instance_data->params, getenv("VK_LAYER_MESA_SCREENSHOT_CONFIG"));
|
|
|
|
|
|
|
|
|
|
if (!globalLockInitialized) {
|
|
|
|
|
loader_platform_thread_create_mutex(&globalLock);
|
|
|
|
|
globalLockInitialized = 1;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return result;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void screenshot_DestroyInstance(
|
|
|
|
|
VkInstance instance,
|
|
|
|
|
const VkAllocationCallbacks* pAllocator)
|
|
|
|
|
{
|
|
|
|
|
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);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static const struct {
|
|
|
|
|
const char *name;
|
|
|
|
|
void *ptr;
|
|
|
|
|
} name_to_funcptr_map[] = {
|
|
|
|
|
{ "vkGetInstanceProcAddr", (void *) vkGetInstanceProcAddr },
|
|
|
|
|
{ "vkGetDeviceProcAddr", (void *) vkGetDeviceProcAddr },
|
|
|
|
|
#define ADD_HOOK(fn) { "vk" # fn, (void *) screenshot_ ## fn }
|
|
|
|
|
#define ADD_ALIAS_HOOK(alias, fn) { "vk" # alias, (void *) screenshot_ ## fn }
|
|
|
|
|
ADD_HOOK(CreateSwapchainKHR),
|
|
|
|
|
ADD_HOOK(GetSwapchainImagesKHR),
|
|
|
|
|
ADD_HOOK(DestroySwapchainKHR),
|
|
|
|
|
ADD_HOOK(QueuePresentKHR),
|
2024-10-19 15:28:28 -07:00
|
|
|
ADD_HOOK(AcquireNextImageKHR),
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
|
|
|
|
|
ADD_HOOK(CreateDevice),
|
2025-02-10 22:11:59 -08:00
|
|
|
ADD_HOOK(GetDeviceQueue),
|
|
|
|
|
ADD_HOOK(GetDeviceQueue2),
|
vulkan/screenshot-layer: Add Vulkan screenshot layer
This change adds a Vulkan screenshot layer that allows users to take
screenshots from a Vulkan application, but has an emphasis on
performance, decreasing the performance impact on the application
involved. This allows for automated setups to use this layer to take
screenshots for navigating various in-application menus.
This layer works by hooking into various common Vulkan setup functions, until
it enters the vkQueuePresentKHR function, and from there it copies the current
frame's image from the swapchain as an RGB image to host-cached memory, where
we will receive the information as a framebuffer pointer. From there, we copy
the framebuffer contents to a thread that will detach from the main process
so it can write the image to a PNG file without holding back the main thread.
This layer was created from using the existing overlay layer as a template,
then adding portions of LunarG's VulkanTools screenshot layer:
https://github.com/LunarG/VulkanTools/blob/main/layersvt/screenshot.cpp
More specifically, there were usages of functions, along with modifications of
various functions from screenshot.cpp in the VulkanTools project, used in
screenshot.cpp.
There are some sections of the screenshotting functionality that remain
unmodified from the original screenshot.cpp file in VulkanTools, including the
global locking structures and the writeFile() function, which takes care of
obtaining the images from the swapchain. There were various areas in which
modifications were made, including how images are written to a file (using PNG
instead of PPM, introducing threading, added fences/semaphores, etc), along
with many smaller changes.
v2: Fix segfault upon application exit
v3: Fix filename issue with concatenation, along with some leftover
memory handling that wasn't cleaned up.
v4: Fix some error handling and nits
v5: Fix output directory handling
Reviewed-by: Ivan Briano <ivan.briano@intel.com
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30527>
2024-05-29 23:39:45 -07:00
|
|
|
ADD_HOOK(DestroyDevice),
|
|
|
|
|
|
|
|
|
|
ADD_HOOK(CreateInstance),
|
|
|
|
|
ADD_HOOK(DestroyInstance),
|
|
|
|
|
#undef ADD_HOOK
|
|
|
|
|
#undef ADD_ALIAS_HOOK
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
static void *find_ptr(const char *name)
|
|
|
|
|
{
|
|
|
|
|
for (uint32_t i = 0; i < ARRAY_SIZE(name_to_funcptr_map); i++) {
|
|
|
|
|
if (strcmp(name, name_to_funcptr_map[i].name) == 0)
|
|
|
|
|
return name_to_funcptr_map[i].ptr;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
PUBLIC VKAPI_ATTR PFN_vkVoidFunction VKAPI_CALL vkGetDeviceProcAddr(VkDevice dev,
|
|
|
|
|
const char *funcName)
|
|
|
|
|
{
|
|
|
|
|
void *ptr = find_ptr(funcName);
|
|
|
|
|
if (ptr) return reinterpret_cast<PFN_vkVoidFunction>(ptr);
|
|
|
|
|
|
|
|
|
|
if (dev == NULL) return NULL;
|
|
|
|
|
|
|
|
|
|
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);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
PUBLIC VKAPI_ATTR PFN_vkVoidFunction VKAPI_CALL vkGetInstanceProcAddr(VkInstance instance,
|
|
|
|
|
const char *funcName)
|
|
|
|
|
{
|
|
|
|
|
void *ptr = find_ptr(funcName);
|
|
|
|
|
if (ptr) return reinterpret_cast<PFN_vkVoidFunction>(ptr);
|
|
|
|
|
|
|
|
|
|
if (instance == NULL) return NULL;
|
|
|
|
|
|
|
|
|
|
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);
|
|
|
|
|
}
|