[OldRocketGuy]

Intel has created some C++ programming tools to make simple parallelizing code on their multi-core architecture. Tools like this are needed to get the promised performance out of multi-core architectures and will become truly critical as the number of cores increase. The problem is that Intel is charging $$$ for these tools.

One of the great advantages of Xcode, is that it comes with OS X at no additional cost. While the cost of the Intel tools may not be a factor for large software vendors, it is a barrier to entry for much of the Open Source developers, the scientific and engineering communities and others who do their own modeling and students learning to do parallel programming.

It makes no sense for Intel to place barriers in the way of developers or potential developers which make their CPU far more effective. The attempt to turn compiler tools into a profit center only hurts the demand for their own chips in the long run.

What is Apple doing to provide compiler technology to take advantage of multi-core architectures, specifically the Intel architecture, now that Apple is only making computers based on this technology? Have they made a deal with Intel, so that the new tools can be provided as part of Xcode?

[Camelot]

This is nothing new.

For years the Intel compilers have excelled over other compilers in generating optimized code for Intel processors. There's even been talk that the Intel compilers deliberately produce sub-standard code for AMD processors. In any case any benchmark you see posted about Intel processors are going to be based on the Intel compiler.

They're not the only ones though. For many years now the best compiler you can get for the PowerPC has been an IBM compiler - for which IBM charge similar fees.

GCC, the popular open source compiler which XCode is built on produces decent code output, but not the fastest on any processor. That's one of the reasons why knowing the compiler used in any benchmark is essential to understanding the results.

[Thinine]

GCC should be adding support for a similar feature sometime soon. However, there's the issue of whether or not to support Intel's version or making their own.

[OldRocketGuy]

GCC should be adding support for a similar feature sometime soon. However, there's the issue of whether or not to support Intel's version or making their own.

That is exactly what I was hoping for.

[Gametes]

I find it surprising that Apple doesn't (or do they?) devote the manpower to making GCC - at least for chips /languages run on Macs - the best compiler available. I mean, this is the same tool they use in-house, right? So it stands to reason that they would protect it.

[Catfish_Man]

Originally Posted by Gametes

I find it surprising that Apple doesn't (or do they?) devote the manpower to making GCC - at least for chips /languages run on Macs - the best compiler available. I mean, this is the same tool they use in-house, right? So it stands to reason that they would protect it.

They have quite a few people working on it.

Examples of Apple additions:
ObjC2
Autovectorization (in collaboration with IBM and others)
ObjC++
@synchronized
@try/@catch
ObjC access modifiers
Lots of LLVM-GCC enhancements recently (they hired the lead dev of LLVM)
-fno-nil-receivers
-fobjc-direct-dispatch