Comments on: Danny’s .NET 2.0 talk: Part Two http://blog.excastle.com/2004/09/11/dannys-net-20-talk-part-two/ Life, .NET, and Cats Wed, 07 Jan 2009 03:07:49 +0000 http://wordpress.org/?v=2.7 hourly 1 By: Per Larsen http://blog.excastle.com/2004/09/11/dannys-net-20-talk-part-two/comment-page-1/#comment-4223 Per Larsen Mon, 13 Sep 2004 08:17:00 +0000 http://blog.excastle.com/2004/09/11/dannys-net-20-talk-part-two/#comment-4223 Danny, <br> <br>"If profile based optimizations and relinking were so simple, why isn't everyone doing it?" <br> <br>I guess it hasn't been important enough to warrant the investment until now. <br> <br>"...Rewriting those to take a second input to alter the codegen patterns is far from a trivial exercise." <br> <br>Okay, I obviously don't have intimate knowledge of the Delphi compiler's inner workings, so I admit that "without too much effort" may be an exaggeration :) <br> Danny,

"If profile based optimizations and relinking were so simple, why isn’t everyone doing it?"

I guess it hasn’t been important enough to warrant the investment until now.

"…Rewriting those to take a second input to alter the codegen patterns is far from a trivial exercise."

Okay, I obviously don’t have intimate knowledge of the Delphi compiler’s inner workings, so I admit that "without too much effort" may be an exaggeration :)

]]> By: Danny Thorpe http://blog.excastle.com/2004/09/11/dannys-net-20-talk-part-two/comment-page-1/#comment-4222 Danny Thorpe Sun, 12 Sep 2004 16:34:00 +0000 http://blog.excastle.com/2004/09/11/dannys-net-20-talk-part-two/#comment-4222 Per, <br> If profile based optimizations and relinking were so simple, why isn't everyone doing it? <br> Building a compiler to emit code and a linker to sew it together is one thing. Rewriting those to take a second input to alter the codegen patterns is far from a trivial exercise. Please leave the armchair coaching to sports programs. Per,

If profile based optimizations and relinking were so simple, why isn’t everyone doing it?

Building a compiler to emit code and a linker to sew it together is one thing. Rewriting those to take a second input to alter the codegen patterns is far from a trivial exercise. Please leave the armchair coaching to sports programs.

]]> By: Per Larsen http://blog.excastle.com/2004/09/11/dannys-net-20-talk-part-two/comment-page-1/#comment-4221 Per Larsen Sun, 12 Sep 2004 07:44:00 +0000 http://blog.excastle.com/2004/09/11/dannys-net-20-talk-part-two/#comment-4221 "MS won't license it out, so Delphi is out of luck (and so is C#)" <br> <br>This feature is obviously target at native code (as opposed to .NET code) - otherwise, it would be part of the JIT-compiler and not specific to C++. <br> <br>Until Borland announces a Delphi with native 64-bit support, this feature is not really relevant for them/us. However, when/if Borland does decide to do a native 64-bit version of Delphi in the future, this feature sounds simple enough in principle that they should be able to mimic it without too much effort. "MS won’t license it out, so Delphi is out of luck (and so is C#)"

This feature is obviously target at native code (as opposed to .NET code) - otherwise, it would be part of the JIT-compiler and not specific to C++.

Until Borland announces a Delphi with native 64-bit support, this feature is not really relevant for them/us. However, when/if Borland does decide to do a native 64-bit version of Delphi in the future, this feature sounds simple enough in principle that they should be able to mimic it without too much effort.

]]>