C / C++ / MFC

29-Mar-16 20:52

I have a problem when i have being composed a compiler program for my processor.  A system operation( windows, linux ...), how can it run on another processor AMD or Intel? While AMD processors and Intel processors have instructions set different. Can someone answer my question?

Jochen Arndt29-Mar-16 21:33

29-Mar-16 21:33

Every compiler has options to specify the target processor type and version (the supported commands). If such options are not used, the compilers will generate code that can run on all Intel and AMD CPUs limited to a minimum processor generation depending on the compiler and version.

For GCC see i386 and x86-64 Options - Using the GNU Compiler Collection (GCC)[^].
For Visual C/C++ there is only one option: /arch (x86)[^].

Ngoc Linh Nguyen30-Mar-16 3:14

30-Mar-16 3:14

Thank you very much! Jochen Arndt,

Regarding GPU, I know that, graphic module in a program can be run by CPU. Also, I know that GPU's instruction set of ATI (or Nvidia) and CPU's instruction set (Intel, AMD) are different. Is it that, insider a operation system has be exiting a dynamic compiler or a normal compiler compile a program to all machine architecture targets?

Newest Ubuntu 64 bits have about 1GB size. If it was compiled to specify one machine architecture, Its size would be smaller ? And how many about size?

Regards!

Jochen Arndt30-Mar-16 4:15

30-Mar-16 4:15

GPU's are complete different. They are accessed by operation system dependant drivers for normal operation. While it is possible to perform calculations on modern GPU's, these use their own compilers (e.g. The LLVM Compiler Infrastructure Project[^]).

Intel and AMD have basically the same instruction set. The same applies to other x86 based CPUs (VIA, Cyrix). There are only a very few manufacturer specific instructions. Compilers will not use these instructions by default and might even not provide options to use them. But even then compiling for a specific manufacturer would result only in marginal decreased code size. See also Differences between AMD64 and Intel 64 at x86-64 - Wikipedia, the free encyclopedia[^]

The code size can be slightly decreased (together with slightly increased performance) when generating code for newer processor generations. This is mainly achieved by enabling Streaming SIMD Extensions - Wikipedia, the free encyclopedia[^] for older processors and Advanced Vector Extensions - Wikipedia, the free encyclopedia[^] and the FMA instruction set - Wikipedia, the free encyclopedia[^] for newer ones.

Operating system distributions should run on all systems matching the minimum requirements. Therefore, they are usually build to support also older CPU's. However, they (and also applications) may detect the CPU and use modern instructions when supported. While this increases the code size it will also increase performance.

I can't tell you how much the size decreases when compiling a Linux kernel for a more recent CPU. If you really want to know it, you have to do it.

But reducing the code size should be not necessary nowadays where systems have a lot of storage (RAM and drives).

Ngoc Linh Nguyen30-Mar-16 5:36

30-Mar-16 5:36

Hi Jochen Arndt,

I am trying to develop a processor ( you can think it is a IP ). It will be a IP module beside ARM core and the other IPs into a SoC ( example Snapdragon 820 etc... ). My processor can't running a program which were compiled for ARM architecture. And it only can run one program at a time. It was used to boost up the speed of a program when user need. The ISA for my processor is difference ARM's ISA. I want to make a driver/compiler for my IP, which can compile machine code of a program for the ARM core to machine code for my processor. Have you ever seem a driver/compiler like that? And It has ability to do?

Thank for your responds!

Jochen Arndt30-Mar-16 5:53

30-Mar-16 5:53

I have not seen such a compiler (which does not mean they do not exist). But the behaviour of a "driver" is similar to an emulator or the virtual execution system for intermediate languages like CIL used with .Net languages.

Ngoc Linh Nguyen30-Mar-16 5:58

30-Mar-16 5:58

Thank you!

Best Regards!

Ngoc Linh Nguyen31-Mar-16 0:39

31-Mar-16 0:39

Hi Jochen Arndt,

Actually, I want to invite you for my project( processor that i am trying to make compiler/driver for it ). I will send project's material for you so that you clearly understand.

Are you ok ?

Regards!

Jochen Arndt31-Mar-16 0:51

31-Mar-16 0:51

Thank you for your invitation. But I have not enough free time to be engaged in such a project.

problem with compiler

Ngoc Linh Nguyen29-Mar-16 20:52

29-Mar-16 20:52

I have a problem when i have being composed a compiler program for my processor.  A system operation( windows, linux ...), how can it run on another processor AMD or Intel? While AMD processors and Intel processors have instructions set different. Can someone answer my question?

InitDialog in splitter

Member 1228273828-Mar-16 10:07

28-Mar-16 10:07

With MFCm I am trying to create an OpenGL control on a split form the uses the splitter class to create the panes. My problem is I don't know where to put the OnInitDialog() function or how to wire it up. I have a main form CMainFrame and 2 child forms CFormRight and CFormLeft but don't see any place for the OnInitDialog function. The OnCreateClient function in MainForm calls m_splitter.CreateView but not OnInitDialog.

Any Suggestions?

John C

Richard MacCutchan28-Mar-16 21:46

Richard MacCutchan

28-Mar-16 21:46

What does OnInitDialog have to do with this, where is the dialog in the above application?

Victor Nijegorodov29-Mar-16 2:20

Victor Nijegorodov

29-Mar-16 2:20

You seem to need the OnInitialUpdate overrides for your left and right CView derived classes.

Member 1228273829-Mar-16 3:04

29-Mar-16 3:04

Victor,

Thanks so much but I still don't know what to do. my CForm right and CFormLeft (where I want to put the OpenGL window) inherit from CFormView. Where do put the over ride? I have CMainForm : public CFrameWnd, and COpenGL: public CWnd. The OnCreateClient() function in CmainForm creates the splitter

BOOL CMainFrame::OnCreateClient(LPCREATESTRUCT lpcs, CCreateContext* pContext)
{
	// TODO: Add your specialized code here and/or call the base class

	// create splitter window
	if (!m_wndSplitter.CreateStatic(this, 1, 2)){
		return FALSE;
	}

	if (!m_wndSplitter.CreateView(0, 0, RUNTIME_CLASS(CFormLeft), CSize(300, 100), pContext) ||
		!m_wndSplitter.CreateView(0, 1, RUNTIME_CLASS(CFormRight), CSize(100, 100), pContext))
	{
		m_wndSplitter.DestroyWindow();
		return FALSE;
	}

Thanks.
John

Richard MacCutchan29-Mar-16 4:23

Richard MacCutchan

29-Mar-16 4:23

See my previous response. Where is the dialog?

Member 1228273829-Mar-16 7:12

29-Mar-16 7:12

Richard,

The function that I am trying to get to execute is COpenGLControl::oglInitialize() which was called in OnInitDialog in the example template I am following. the oglInitialize function, listed below should setup the OpenGL context. But my case is somewhat different as you can see from my most recent post and I don't know where to place the call to oglInitalize()

void COpenGLControl::oglInitialize()
{
	// Initial Setup:
	//
	static PIXELFORMATDESCRIPTOR pfd =
	{
		sizeof(PIXELFORMATDESCRIPTOR),
		1,
		PFD_DRAW_TO_WINDOW | PFD_SUPPORT_OPENGL | PFD_DOUBLEBUFFER,
		PFD_TYPE_RGBA,
		32,    // bit depth
		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
		16,    // z-buffer depth
		0, 0, 0, 0, 0, 0, 0,
	};

	// Get device context only once.
	hdc = GetDC()->m_hDC;

	// Pixel format.
	m_nPixelFormat = ChoosePixelFormat(hdc, &pfd);
	SetPixelFormat(hdc, m_nPixelFormat, &pfd);

	// Create the OpenGL Rendering Context.
	hrc = wglCreateContext(hdc);
	wglMakeCurrent(hdc, hrc);

	// Basic Setup:
	//
	// Set color to use when clearing the background.
	glClearColor(0.0f, 0.0f, 0.0f, 1.0f);
	glClearDepth(1.0f);

	// Turn on backface culling
	glFrontFace(GL_CCW);
	glCullFace(GL_BACK);

	// Turn on depth testing
	glEnable(GL_DEPTH_TEST);
	glDepthFunc(GL_LEQUAL);

	// Send draw request
	OnDraw(NULL);
}

In that example template I have three classes CMFCOpenGLDlg : public CDialogEx COpenGLControl : public CWnd and CMFCOpenGLApp : public CWinApp in this example the CMFCOpenGLDlg::OnInitDialog() function calls COpenGLControl::oglCreate() and COpenGLControl::OnCreate() calls COpenGLControl::oglInitialize() but I just don't see how to wire this up in my splitt form example where the panes are created from the splitter class.

Thanks,

John

...

// Create OpenGL Control window
m_oglWindow.oglCreate(rect, this);

Richard MacCutchan29-Mar-16 7:19

Richard MacCutchan

29-Mar-16 7:19

OK, but you are not calling this from a dialog, but from a CView, so you probably need to call the initializer from the CView::OnInitialUpdate[^] function.

Member 1228273829-Mar-16 10:41

29-Mar-16 10:41

Richard,

Thanks for bearing with me. I looked in MDN and found this comment about OnInitialUpdate

Called by the framework after the view is first attached to the document, but before the view is initially displayed.

So I assume that OnInitalUpdate is called automaticaly and I don't need to do anything to invoke it but where would I put the override code in CFormRight : public CFormView or COpenGLControl : public CWnd ?

Thanks again,

John

David Crow29-Mar-16 14:49

David Crow

29-Mar-16 14:49

Member 12282738 wrote:
CFormRight

Yes.

Victor Nijegorodov30-Mar-16 5:09

Victor Nijegorodov

30-Mar-16 5:09

Member 12282738 wrote:
So I assume that OnInitalUpdate is called automaticaly and I don't need to do anything to invoke it but where would I put the override code in CFormRight : public CFormView or COpenGLControl : public CWnd ?

Yes, it is called automatically. Note that if you do not override the OnInitalUpdate in your derived class then the base class implementation will be called.

And note, that OnInitalUpdate is a method of a CView, not a CDialog. So - override it in CFormRight class.