Jump to content
  • Advertisement
Sign in to follow this  

DX11 Questions about good coding practices for DirectX (C++)

This topic is 2556 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

I've been using SlimDX (DX11) for quite some time now and I am fluent in C# and feel quite comfortable coding for DX11.
However, I figured that my lack of C++ skills is unacceptable, so I set out to learn C++.

So far, everything seems quite straightforward and obvious, though I obviously had some difficulties in the beginning.
To test and further develop my C++ skills, I decided to port my current engine, which I've written in C#/SlimDX to C++.
I'm not really experiencing any trouble with porting and solving problems...

... however I'm quite at a loss as to how to produce "good / clean code".

Online reading materials do a great job at explaining programming logic and how to solve problems, but they seem to give no indication on what would be a good or clean way to do so.
Due to this, I feel rather insecure about my current way of coding C++.

Here's a random class as an example on how I'm currently approaching things:

[source]#pragma once
#include "stdafx.h"

// Forward Declarations
class CDeferredRenderer;

class CGBuffer

ID3D11Texture2D *GDiffuse;
ID3D11Texture2D *GNormal;
ID3D11Texture2D *GDepth;
ID3D11Texture2D *GLight;
ID3D11RenderTargetView *RTDiffuse;
ID3D11RenderTargetView *RTNormal;
ID3D11RenderTargetView *RTDepth;
ID3D11RenderTargetView *RTLight;
ID3D11ShaderResourceView *GDiffuseView;
ID3D11ShaderResourceView *GNormalView;
ID3D11ShaderResourceView *GDepthView;
ID3D11ShaderResourceView *GLightView;

void Init(CDeferredRenderer * Renderer);
void Clear();
void BeginGeometryStage(ID3D11DepthStencilView *DepthBufferView);
void Release();

CDeferredRenderer *Renderer;
ID3D11Device *Device;
ID3D11DeviceContext *Context;

void CreateGBufferTextures();
void CreateGTexture( DXGI_FORMAT Format, ID3D11Texture2D* & Texture, ID3D11RenderTargetView* & RTView, ID3D11ShaderResourceView* & ResView );

[source]#include "StdAfx.h"
#include "GBuffer.h"
#include "DeferredRenderer.h"
#include "Engine.h"
#include "Log.h"
#include "StringParser.h"



void CGBuffer::Init( CDeferredRenderer * Renderer )
this->Renderer = Renderer;
this->Device = Renderer->Device;
this->Context = Renderer->Context;


void CGBuffer::CreateGBufferTextures()
CreateGTexture(DXGI_FORMAT_R8G8B8A8_UNORM, GDiffuse, RTDiffuse, GDiffuseView);
CreateGTexture(DXGI_FORMAT_R16G16_UNORM, GNormal, RTNormal, GNormalView);
CreateGTexture(DXGI_FORMAT_R32_FLOAT, GDepth, RTDepth, GDepthView);
CreateGTexture(DXGI_FORMAT_R16G16B16A16_FLOAT, GLight, RTLight, GLightView);

void CGBuffer::CreateGTexture( DXGI_FORMAT Format, ID3D11Texture2D* & Texture, ID3D11RenderTargetView* & RTView, ID3D11ShaderResourceView* & ResView )
// Create Texture
D3D11_TEXTURE2D_DESC TextureDesc;
ZeroMemory(&TextureDesc, sizeof(D3D11_TEXTURE2D_DESC));

TextureDesc.Width = SCREEN_WIDTH;
TextureDesc.Height = SCREEN_HEIGHT;
TextureDesc.MipLevels = 1;
TextureDesc.ArraySize = 1;
TextureDesc.Format = Format;
TextureDesc.SampleDesc.Count = 1;
TextureDesc.SampleDesc.Quality = 0;
TextureDesc.Usage = D3D11_USAGE_DEFAULT;
TextureDesc.CPUAccessFlags = 0;
TextureDesc.MiscFlags = 0;

HRESULT hr = Device->CreateTexture2D(&TextureDesc, NULL, &Texture);

// Create RenderView
ZeroMemory(&Desc, sizeof(D3D11_RENDER_TARGET_VIEW_DESC));

Desc.Format = Format;
Desc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2D;

hr = Device->CreateRenderTargetView(Texture, &Desc, &RTView);

// Create Shader Resource View
hr = Device->CreateShaderResourceView(Texture, NULL, &ResView);

void CGBuffer::Release()



GDiffuse = NULL;
GNormal = NULL;
GDepth = NULL;
GLight = NULL;

RTDiffuse = NULL;
RTNormal = NULL;
RTDepth = NULL;
RTLight = NULL;

GDiffuseView = NULL;
GNormalView = NULL;
GDepthView = NULL;
GLightView = NULL;

void CGBuffer::Clear()
Context->ClearRenderTargetView(RTDiffuse, D3DXCOLOR(0.0f, 0.0f, 0.0f, 0.0f));
Context->ClearRenderTargetView(RTNormal, D3DXCOLOR(0.5f, 0.5f, 0.0f, 0.0f));
Context->ClearRenderTargetView(RTDepth, D3DXCOLOR(1.0f, 0.0f, 0.0f, 0.0f));
Context->ClearRenderTargetView(RTLight, D3DXCOLOR(0.0f, 0.0f, 0.0f, 0.0f));

void CGBuffer::BeginGeometryStage(ID3D11DepthStencilView *DepthBufferView)
ID3D11RenderTargetView* RenderTargetViews[3] =
Context->OMSetRenderTargets(3, RenderTargetViews, DepthBufferView);

This works fine, but there's a lot of questions popping up in my head:
  • I only include stdafx.h (containing rarely changing libs such as DX libs) in my header files. I then add forward declarations for all other classes used in this one.
    In my .cpp I include all header files of other classes needed.
    It seems like I cannot run into problems this way... is this the correct approach for managing classes in c++?
  • I keep only pointers to all DX specific stuff (Textures, Rendertargetviews etc) which I can easily pass around.
    I do the same thing with my own classes (My Renderer Class keeps only a Pointer to my GBuffer class)
    Is this okay to do?
    Or should I rather define things not as pointers, and then write get methods to pass references to member variables to other classes?
    class CDeferredRenderer
    CGBuffer *GBuffer;
    void CDeferredRenderer::Init( CEngine * Engine, HWND hWnd )
    GBuffer = new CGBuffer();
  • This also means I have to do things like this (although rarely):
    [source]void CGBuffer::CreateGTexture( DXGI_FORMAT Format, ID3D11Texture2D* & Texture, ID3D11RenderTargetView* & RTView, ID3D11ShaderResourceView* & ResView )[/source]
    Is this okay?
  • Should I make all members private and write Getters for them?
    In C# I'm used to do things like:
    [source]public CGBuffer GBuffer { get; private set; }[/source]

    I obviously can't reproduce this in C++ without writing a Get method for each variable I want to have read-only access to.
    Should I really do this, or is it better to focus on functionality rather than wrapping everything with accessor methods?
  • Should I prefer using references over pointers if I'm in a situation where I can use both to achieve the same thing?
    I noticed the DirectX11 api doesn't make much use of references. For example:
    [source]D3DXMatrixIdentity(D3DXMATRIX *pOut);[/source]
    instead of
    [source]D3DXMatrixIdentity(D3DXMATRIX & Out);[/source]
  • I've seen some code where all member variables are preceded with "m" and pointers with "p" resulting in code like:
    [source]ID3D11Texture2D* mpGDiffuse;[/source]
    Is this considered a good practice?


Share this post

Link to post
Share on other sites
In general, it is best to use references if you are passing anything because it allows the compiler to do its job and optimize the code better because a reference cannot change, but a pointer can. Getters and Setters I do not use much because its a pain to write out all that code, when I can simply access the variable directly. For simplicity, I usually directly access the member. If I can type the function GetTexture(), I sure as well can type .Texture At least, thats how I look at it. Getters and Setters in my mind should be used when you want to ensure that other things happen when a variable is set. For example, if you set a texture, maybe you want to delete the old one if there is one, and then replace it with the new one. In that case, you need a setter function. But, sometimes you dont need to do any work, nor will you ever need to. In that case, I just go straight to the variable. Same concept with getters. If you think you might need to modify an internal state for each Get call, then create a function. If you are sure you wont need to do any internal work, save yourself some work and access the variable straight.

To your second question about headers and forward declarations, your method is correct. Put your forward declarations in the header, and include the necessary header in the cpp file.

Share this post

Link to post
Share on other sites
1. Yes, you should generally prefer to handle includes this way whenever possible. Putting includes in header files can cause longer recompilation times when you change an included header file.

2. Generally you only want to use pointers when you have to. In your example of the CDeferredRenderer containing a CGBuffer, there's really no reason to make it a pointer and then dynamically allocate an instance of CGBuffer since you're always going to have one. Instead the C++ way of doing it is to just declare the CGBuffer as a regular member. For passing something to a function you should almost always pass as a reference. If necessary that function/class can convert that reference to a pointer so that it can hold onto it.

3. This is more of an OOP thing and not really C++ specific. If you're used to using properties, you can just stick with writing getters and setters.

4. Yes, if you can use a reference you really probably should. Using references can save you from the many pitfalls of pointers. The reason the DirectX API uses pointers everywhere is because it maintains C compatibility, and C doesn't have references.

5. This also isn't really C++-related, some people do the same in C#. It comes from Microsoft, who used this notation frequently for the Win32 API and MFC library. Some people still like it, some hate it...I would say it's totally your call.

Share this post

Link to post
Share on other sites
Thank you both for your suggestions and explanations, I greatly appreciate it!
This really helps, I was feeling very insecure about these things, but now I have a good idea on how to do things better.


Share this post

Link to post
Share on other sites
Sign in to follow this  

  • Advertisement

Important Information

By using GameDev.net, you agree to our community Guidelines, Terms of Use, and Privacy Policy.

We are the game development community.

Whether you are an indie, hobbyist, AAA developer, or just trying to learn, GameDev.net is the place for you to learn, share, and connect with the games industry. Learn more About Us or sign up!

Sign me up!