Scriptionary

So today marks the first week of Windows 7 RTM usage and I have to say that I like it. As I said before, Vista’s gaming performance was sub par but it looks like Windows 7 has picked up the slack.

It simply seems like the same system performs better with Windows 7, it’s definitely not simply a Vista update/Service Pack. And I’m happy to report that Fallout 3 (for the players) works OK on Windows 7 although there is the occasional crash and Alt-Tabbing is out of the question it seems.

Another very good thing thing I’ve noticed is the lack of UAC pop-ups, the only ones I’ve seen are when I try to run downloaded software – which is the same as in Windows XP.

Maybe in the long run some quirks will show up but for now I’m a happy camper indeed.

Windows 7 RTM was released for MSDN subscribers today so I’m taking this chance to upgrade to Windows 7 early on. I’ve been using Vista fulltime a bit over a year now and have been a bit disappointed in the long run with its performance when it comes to gaming and other intense apps.

Although I’ve been disappointed with Vista’s gaming performance, everything else has been smooth sailing. In fact a bunch of features that are apparently introduced in Windows 7 (Start Search), were already available in Vista.

Anyway, I’ll soon find out if the Windows 7 talk is all hype or if there’s some truth to the claims I’ve been hearing.

The upcoming C++ revision, C++0x has been delayed and is now scheduled for release somewhere in the 2010′s, thus effectively becoming C++1x.

Here’s a link to an article published in Dr. Dobbs Journal by C++’s creator Bjarne Stroustrup explaining why (page 3).

I don’t think anyone is really surprised with the delay as the new standard would lift C++ up to a more modern level in order compete with contemporary programming languages, thereby implementing much addendum. The list of new features and STL additions is certainly impressive and when C++0x, er, C++1x is released, we’ll probably all need to brush up on our C++ skills.

I don’t know how many times I’ve ran into people that need help with a piece of C code that should run perfectly fine but for some reason it just doesn’t. Of course, after much going back and forward, you find out that the person you’re trying to help is using Borland’s Turbo C compiler version 2.01, released 20 years ago in 1989..

While the compiler is freely download-able from several websites labeled as an “Antique”, it’s really not a great tool to start programming with on modern systems. If you really (really) want to use Turbo C, buy an old 286 and go at it but don’t use your Intel Core i7. Here’s why. PS, this also applies to Turbo C++

1. The compiler is 20 years old, it won’t support the current C99 standard so many things you might have learned, you now have to unlearn in order to get the program working.

2. Turbo C is an MS-DOS compiler, chances are if you are learning C now, you have never used DOS in your life; there’s no reason to go back, nostalgia doesn’t apply to DOS. Furthermore, Windows does not support DOS programs from the 80′s, so you would have to emulate DOS in order to get your programs working properly.

3. There are many free compilers available that are 100% compatible with today’s hardware and not bound by any limitations. Even if your Turbo C compiler is 32 bit compatible, it’s not a great match for your 64 bit CPU and Operating System.

4. You’re stuck on the command line interface with no way out. Even if you make the best command line program for DOS, nobody besides yourself will ever want to use it. Converting your program to Win32 requires a new compiler that will yell at you for using coding practices from 1989 and break your program.

In other words, if the following code compiles without warning, you know you need another compiler.

#include <stdio.h>

main()
{
    printf("Hello, World!");
}

Been gone for a couple of days but I’m back with some cool stuff from Intel. Intel has posted a whole bunch of stuff from GDC 2009 on their site, you can check it out right here.

Interestingly, my postscript got answered, Intel has posted (a preview) of Tom Forsyth’s talk on Larrabee’s SIMD extensions, the full version should come online in one month according to the site.

As always with Intel (and many others), you have to read through the marketing crud and filter out the core.

I don’t often recommend books, especially when it comes to DirectX/Direct3D. There are many monstrosities out there that should never have seen the light of day.

Once per week or so I browse a local Barnes and Noble hoping to catch a book I haven’t seen before. It rarely happens since their inventory is very stagnant, but still. A month or two ago I accidentally browsed the Graphics/Web section and found “Introduction to 3D Game Programming with DirectX 10″ by Frank D. Luna.

The title, publisher (Wordware) and cover were kind of off-putting so I almost didn’t buy it. But I’m glad I did since, in my case, it has become my number one DX10 reference. It’s compact, concise, always around my desk and well written so give it a shot if you have 30 bucks or so to spare.

Today Intel released it’s new issue of “Intel Visual Adrenaline” featuring a three page interview with Tom Forsyth about Intel’s upcoming Larrabee GPU, which is x86 based and fully programmable.

Click here to read the PDF, scroll down to page eight (8)

Larrabee will support a rasterization pipeline as well as raytracing but Forsyth mentions raytracing to be more of a technical feature than a mainstream implementation. Regardless of this, for graphics programmers this should be good as there will finally be a piece of hardware that actually supports realtime raytracing.

Direct3D as well as OpenGL will be supported in addition to the much anticipated programmable route, either through C++ or pure assembly, which should open up the card for people interested in parallel computing.

Sadly there’s no definitive answer from Forsyth on how many cores Larrabee will actually contain. I guess we’ll have to wait for that a bit longer.

PS, if anyone is at GDC listening to Abrash and Forsyth tomorrow, let me know what you got from it.

I haven’t been active at all in the OpenGL circles recently but apparently OpenGL 3.1 was released yesterday. Check out the specs by clicking here at Khronos.ORG.

I have been extremely skeptical about OpenGL 3.0 since it was more like an OpenGL 2.2 and did not implement any of the promised object oriented features that competing APIs like Direct3D implement. Thankfully, this release of OpenGL 3.1 seems to be better at first glance as it not only deprecates but removes the old functionality which has been in the way for so long.

It looks like Khronos has heard and listened to its outraged end-users and finally created a better specification. At least, upon first glance, that’s what it looks like.

With the release of OpenGL 3.1 also comes a new version of GLSL, version 1.40. Most excitingly, NVIDIA has released OpenGL 3.1 drivers immediately after the release of GL, showing that they’re way ahead in the game already (as opposed to AMD or Intel).

It is amazing to me that no major outlets such as Slashdot have picked up this this story and one has to find out about 3.1 by accident or RSS feeds.