Antheus

For past 20 years, CPUs capacity has increased exponentially.
Meanwhile, RAM throughput has increased only linearly. Same for other forms of storage.

Today, given commodity hardware, all algorithms are IO bound. Or, even RAM is no longer capable of providing data fast enough for CPU to consume. Or, given enough optimizations, all algorithms today will end up IO bound.

This trend is unlikely to reverse, since many of these limits are physical, such as distances between RAM, voltages, heat, ... Especially when talking about real-time, where entire state might need to be visited once every 40ms or more.


The situation with large data sets is so bad, that using languages which run 1000-1 million times slower than what native code could do has no impact on final running time (aka, the web and petabyte databases). CPUs are just sitting there, waiting for data to return.


It would be possible to design computing machines which have different physical design, but this immediately makes costs prohibitive.

In real world, storage is not a viable trade off on current hardware.

I feel old because I had an A: and a B: drive. And I approve of every post in this thread.

Luxury. We could only dream of having letters for our drives. Or colons. We did a LOAD "",8 and we were grateful for it.

Floppys are bound to make a comeback.

Put 5 of them in RAID and they can stream a single mp3 file in real time.

I bet half the people who "leave" in the next week come back eventually anyhow.

Relevant: "... devs should avoid forums." WHAAAA?

Quote:

Original post by Nik02 The dereferencing itself is the source of the overhead. It is negligible in most practical cases, though.

Except, in most cases, when dealing with math. Such as large number of vectors, coordinate transforms, graphics... Where nullable doesn't even make sense. A point or coordinate is.

Even if it's not overhead in code, it's effectively a nonsensical situation. A matrix where one column is null doesn't make sense. A null matrix has no use. A null vector has no use. It just adds an unneeded complication.

Quote:

It sure would be sweet if the profiler could just say "Hey dude, this function is 60% slower than it could be!

Halting problem.

This type of optimization problems can be tackled using genetic programming, where existing code is mutated and profiled repeatedly. The only thing it can say is whether there exists a faster solution within given search space, possible under local minima. Obviously, such test would need to run to infinity, or have a well-defined search space that would signal termination.

Quote:

You need to really do something here!" ... but then again I guess this is akin to wishing my compiler would just write my program for me. :-)

It's quite doable. If program inputs are specified as unit tests, then applying the above method (given truly vast computational resources), it would be possible to evolve actual code.

Quote:

P.S. I've been searching the internet for tutorials that can give advice on how to use the various metrics to spot potential problems in a program, but I haven't had much success so far.

Is there a problem? No? Leave it alone.

Does click of a button take 600ms? Use profiler on all code paths starting from button event handler.

Does the renderer run at 7FPS. Use profiler on anything rendering related.

But first you need a measurable problem.

Quote:

Original post by Butabee Regardless, Unreal is not a native MMO engine. It requires heavy modification to use it as such.

And PHP is a crappy language that can only be used to build trivial web sites. It's what HTML+JS offers to user that matters, not how it's generated.

Quote:

You also don't get the source code with the indie version so it's very unlikely anyone could make a MMO with the indie license.

It's impossible to build a 3D AAA-grade MMO as an indie. The upfront cost is simply too high given the competition, and tools like UE or other similar ones cannot grow organically. When working with this type of engine, one cannot start with stick figures, then progress to simply polygons and use metrics to iterate into what works. It's either full-feature from day one or bust. Even big studios fell into this trap, trying to launch with content for 20 levels out of 50, and hoping to catch up.

Meanwhile, Farmville, Habbo and similar started small, saw what floats and what doesn't, killed fast and iterated furiously. Same for other kinds of start ups.

UE and similar are Big Iron. They work since they are standardized, so one can quickly scale to 10*X new hires if needed, since those have been trained externally and can jump in instantly into existing scaffolding and management structures.

Unfortunately, same goes for big names, who do exactly the same, but with $1 billion budget and much more efficiently.

Quote:

Original post by Butabee I would think many indie developers would be dropping bricks right about now.

Indie and other small teams grow organically. They focus on the niche specialties each individual has, and maximize those. The cost of learning a complete vendor locked vertical stack is often detrimental in this setting, since it polarizes the available talent (similar to Java vs. .Net shops and Ruby vs. Python, ...). And this type of fashion statements matter a lot when on shoestring budget.

An indie team also do not have the capacity to produce AAA-grade product such engine implies, and anything less will result in sub-par result.

Teams that secure funding that can afford to not only hire arbitrary talent, but also afford to purchase any cost beneficial technology along with needed oversight and management.

In this regard, this release has no impact. It's probably more of a response to Unreal, which definitely does not have the "they have yet to have a game launched" problem.

Quote:

Original post by SoldierX But now I planning to start a rather advanced game project, and I want it both to be cross-platform (Win+Linux+Mac, that is) and have high performance (stunning 3d graphics, yadda yadda). So, C++ fulfills these requirements. But what alternatives do I have?

UDK.

Quote:

Original post by nullsquared Indeed, I used the following UTF-8 helper library: http://utfcpp.sourceforge.net/.

Did you get the library verified by the lawyer to ensure it is compatible with your project's and company release policy? Did he verify that the implementation does not violate any patents? Did you verify that the author of the library, or any third parties associated with the library are not residents of a country that is banned from doing trade in any of the markets where you wish to ship? Has all of the above been documented and filed? Has the code in the library been verified as original work, or has author signed a "work-for-hire" or other form of transfer of ownership to your company?

Did you submit the library to your build master for integration with the build system? Have the integration and other tests been written?

Did you coordinate the introduction of new third-party library with the project manager for inclusion into the project life-cycle? Has risk analysis with regard to third-party library lifecycle been integrated into your product's life cycle?


All of the above are issues that will need to be cleared in any reasonably-sized company. This 1 minute download just caused 50 hour workload for your company across multiple departments.

Imagine that library happens to contain something that violates some patent (perhaps it uses a linked list), and it gets integrated into your flagship product. Suddenly, you have IBM suing your for hundreds of millions of dollars.

This is why open source matters only in some rare cases, the rest of the code lying around source sites simply doesn't exist. It's too expensive.

As Microsoft once said, Linux is only free if your time is worthless. Most people do not understand that and laugh at Microsoft. And for a startup with no revenue it's true, FOSS is great. But as soon as money is involved, things get very complicated.

The benefit of built-in libraries, or those provided by software vendors is that they are safe from most of the above. Hence, writing something from scratch, despite potential bugs, is considerably better choice in such setting.

CPUs are really good at data parallel (which is a crappy explanation) operations. SIMD and GPUs are all about that on beefier hardware. A Tutorial.

SIMD is just a CPU-friendly way to formulate an algorithm as data parallel.

Many problems cannot be expressed as such. Some are naturally data parallel, some are called embarrassingly parallel. For those that can be, intrinsics are a way of encoding such algorithm in most hardware-friendly way.

Quote:

What exactly is the reason, that x87 code is often faster

- The algorithm is not suitable for data parallelism
- The algorithm isn't applied properly

Code which relies heavily on conditional execution, or one which has highly non-deterministic flow is not suitable for most forms of data parallelism, with exception of certain types of hardware, or abundance of memory. SIMD has neither, GPUs are just barely better.

Quote:

that i just call one function thousands of times

The overhead here is a killer.

The point of SIMD (especially Streaming part) is to call function once, and the intrinsics process million elements in same call. Same for GPU and batching. Minimize number of calls, especially data conversions. Ideally, you compute square root

Think Formula 1. It needs to be brought to race track by truck, it needs to be tuned, ... Then it keeps going 300mph for 3 hours it is on the track. That is SIMD. It takes time to prepare, but then it's bitchin' fast.
Calling function for each element is Formula 1 in New York downtown. 50mph, red light, 50 mph, red light, 50 mph, red light, 50 mph, red light.

Quote:

Original post by Viral_Fury I don't think downloading a game and trying it out once is a "commitment", but that might just be me.

Youtube vs. downloadable trailer. Which wins?

Quote:

Games that are played in a browser don't often have a good reputation. Why not just include a video of gameplay on the site?

They currently lack fidelity for what counts as AAA products. But unless one is actually developing such product, in-browser presents a very low barrier to entry.

There's more than just download. Browser games can be played on any computer with browser, often even on locked down computers. Depending on target demographic, they might be quite a reasonable choice. There's also platform dependency to consider.

And think of viral marketing. People like your game, they click on a link and e-mail to their friends. When those open the e-mail, the game is running inside the mail itself already.

Quote:

will eventually become an MMORPG. Since I'm using DirectX 9.0

Unless you have a million $ budget, a 3D MMO is out of question, simply due to sheer cost of asset creation. 2D however is perfectly managable, and 2D works without a problem in browser.