At last something is drawn !!! That's so cool !! Vulkan is sooo hard that seing this sweet triangle is somehow a victory :-D I like the way you structured your engine. Good job Andrew, quite inspiring !! I hope that object loading will come soon, why not using glTF format for a change ? ;-)
Thanks! Yeah, it's very satisfying. Model loading will be coming soon, I have a few things to look at first. glTF sounds interesting! I've seen it popping up a bit recently, thanks for the suggestion!
03:46 and 14:10 The "inputChunk" parameter should be passed by 'const&', to avoid copying 24 bytes. 07:44 The "frame" variable is created as a COPY at each iteration loop. Use 'const&' to avoid unnecessarily copying 96 bytes (32 bytes * 3 frames)."
just a sidenote here. I know this is only for demonstration process, and debug work, but measuring time with floating points lead to precision errors over time. In actual production uint64 timers should be used for storing timestamps (for measuring delta it pretty much doesn't matter). Grabbing that in this context is easily done via glfwGetTimerValue(), and to get time in seconds: double InSeconds = double(Timer)/10000000; Again: not an error, but just to reinforce good habits. I'm writing this, cause I had the priviledge of having to change entire time measuring codebase for our game at work, because of similar "storing time in FPs" issue.
27:04 There's a C++11 feature called "member initializer list". It's better. Because it creates an instance of the class with those values already set (in one go, at creation time), instead of first creating it then running each line from the constructor where those member variables are individually set. No offence, but you write code like you come from C# or Java... Like: "return nullptr;" when the return type is a struct (and not a POINTER). 😆 But I really appreciate the tutorial series. I'm sorry if it doesn't show with all these comments. I'm really grateful!
I don't really like that glfw is initialized inside of "App" and terminated inside of "Engine". It's not like this is wrong but I just feel like initializing and terminating should be the responsibility of a single class. Still great video. This series is probably the best resource to get started with vulkan
Hi! Maybe anyone knows what should I do if I see "The procedure entry point vkCmdEndRendering could not be located in the dynamic link library" message? regsvr32 didn't work, sfc /scannow showed nothing either 🥺
Hi, your tutorials are awesome, I am following and trying to use them with Qt Qml, but I realize that I need to render to texture, can you show me how can I do it? I would like to abstract the output but I do not understand how to do it
Thanks Andre, I'm glad you're enjoying the tutorials! Yeah I know what you're saying, QT sort of deals with views rather than a global renderer. I might look at that at a future point, thanks for reaching out!
After this step my window says it's locked at 60fps, do you know why? I'm using linux, so I think it's because of the lack of the mailbox support by my nvidia drivers, the processor would be stuck waiting for the images in the swapchain to be presented... But if that's the case, do' you know how to unlock the FPS? The monitor I'm running the program is 165hz, so one would think the loop would run at an equivalent fps
Ah, nevermind, I forced the present mode to be mailbox and got the fps improvement I was looking for. Even tho it's not "supported" by my 3080ti on linux, it seems to be working fine. I guess I'll figure if there's a problem once the image starts moving, idk
If you're running into an error that say's "CommandBufferBeginInfo does not have function VkGetHeaderVersion". Remove the BeginInfo object from commandBuffer.end() The only input commandBuffer.end() takes is a dispatcher. I was dealing with that for way too long- and it's my fault too, it's done correctly in the video
Your implementation of an fps counter is broken, it says "58" (on a 60Hz monitor) with the FIFO present mode. It should say 60. 😤 Well, ideally, it should say in milliseconds (ms). Because FPS is a poor metric. But a lot of people are more comfortable with FPS (which is just "1000.0 / timeInMs"). There's an article called "An FPS counter" on OpenGL-Tutorial.
Seeing your title I went "Wtf is C++ 12? That's not a thing.".
05:46 Member variables in "engine.h" are initialized with "nullptr", except they're not pointers. A simple "{}" is enough.
At last something is drawn !!! That's so cool !! Vulkan is sooo hard that seing this sweet triangle is somehow a victory :-D I like the way you structured your engine. Good job Andrew, quite inspiring !! I hope that object loading will come soon, why not using glTF format for a change ? ;-)
Thanks! Yeah, it's very satisfying. Model loading will be coming soon, I have a few things to look at first. glTF sounds interesting! I've seen it popping up a bit recently, thanks for the suggestion!
03:46 and 14:10 The "inputChunk" parameter should be passed by 'const&', to avoid copying 24 bytes.
07:44 The "frame" variable is created as a COPY at each iteration loop. Use 'const&' to avoid unnecessarily copying 96 bytes (32 bytes * 3 frames)."
just a sidenote here. I know this is only for demonstration process, and debug work, but measuring time with floating points lead to precision errors over time. In actual production uint64 timers should be used for storing timestamps (for measuring delta it pretty much doesn't matter). Grabbing that in this context is easily done via glfwGetTimerValue(), and to get time in seconds: double InSeconds = double(Timer)/10000000;
Again: not an error, but just to reinforce good habits. I'm writing this, cause I had the priviledge of having to change entire time measuring codebase for our game at work, because of similar "storing time in FPs" issue.
Awesome, thanks, that’s really good to know!
27:04 There's a C++11 feature called "member initializer list". It's better. Because it creates an instance of the class with those values already set (in one go, at creation time), instead of first creating it then running each line from the constructor where those member variables are individually set.
No offence, but you write code like you come from C# or Java... Like: "return nullptr;" when the return type is a struct (and not a POINTER). 😆 But I really appreciate the tutorial series. I'm sorry if it doesn't show with all these comments. I'm really grateful!
I don't really like that glfw is initialized inside of "App" and terminated inside of "Engine". It's not like this is wrong but I just feel like initializing and terminating should be the responsibility of a single class.
Still great video. This series is probably the best resource to get started with vulkan
Good point! I'll have to fix that.
Hi! Maybe anyone knows what should I do if I see "The procedure entry point vkCmdEndRendering could not be located in the dynamic link library" message? regsvr32 didn't work, sfc /scannow showed nothing either 🥺
okay I updated graphic drivers and faced that: Exception thrown at 0x0000000000000000 Access violation (it refers to vkCmdEndRendering function)
I forgot to mention that problem had solved. I should have used commandBuffer.end() instead of commandBuffer.endRenderinf()
Hi, your tutorials are awesome, I am following and trying to use them with Qt Qml, but I realize that I need to render to texture, can you show me how can I do it? I would like to abstract the output but I do not understand how to do it
Thanks Andre, I'm glad you're enjoying the tutorials! Yeah I know what you're saying, QT sort of deals with views rather than a global renderer. I might look at that at a future point, thanks for reaching out!
If you use c++200 fps routine causes an error you have to go back to c++17
After this step my window says it's locked at 60fps, do you know why? I'm using linux, so I think it's because of the lack of the mailbox support by my nvidia drivers, the processor would be stuck waiting for the images in the swapchain to be presented... But if that's the case, do' you know how to unlock the FPS? The monitor I'm running the program is 165hz, so one would think the loop would run at an equivalent fps
Ah, nevermind, I forced the present mode to be mailbox and got the fps improvement I was looking for. Even tho it's not "supported" by my 3080ti on linux, it seems to be working fine. I guess I'll figure if there's a problem once the image starts moving, idk
for some reason the triangle is not showing up I almost copy pasted everything from the tutorial but cannot find where the problem.
never mind it's working I missed something in the shader.vert
nicee
If you're running into an error that say's "CommandBufferBeginInfo does not have function VkGetHeaderVersion". Remove the BeginInfo object from commandBuffer.end()
The only input commandBuffer.end() takes is a dispatcher.
I was dealing with that for way too long- and it's my fault too, it's done correctly in the video
Your implementation of an fps counter is broken, it says "58" (on a 60Hz monitor) with the FIFO present mode.
It should say 60. 😤
Well, ideally, it should say in milliseconds (ms). Because FPS is a poor metric. But a lot of people are more comfortable with FPS (which is just "1000.0 / timeInMs").
There's an article called "An FPS counter" on OpenGL-Tutorial.