tooling quality and some random rant

retard re at tard.com.invalid
Mon Feb 14 08:39:25 PST 2011


Mon, 14 Feb 2011 04:44:43 +0200, so wrote:

>> Unfortunately DMC is always out of the question because the performance
>> is 10-20 (years) behind competition, fast compilation won't help it.
> 
> Can you please give a few links on this?

What kind of proof you need then? Just take some existing piece of code 
with high performance requirements and compile it with dmc. You lose.

http://biolpc22.york.ac.uk/wx/wxhatch/wxMSW_Compiler_choice.html
http://permalink.gmane.org/gmane.comp.lang.c++.perfometer/37
http://lists.boost.org/boost-testing/2005/06/1520.php
http://www.digitalmars.com/d/archives/c++/chat/66.html
http://www.drdobbs.com/cpp/184405450

Many of those are already old. GCC 4.6, LLVM 2.9, and ICC 12 are much 
faster, especially on multicore hardware. A quick look at DMC changelog 
doesn't reveal any significant new optimizations durin the past 10 years 
except some Pentium 4 opcodes and fixes on library level.

I rarely see a benchmark where DMC produces fastest code. In addition, 
most open source projects are not compatible with DMC's toolchain out of 
the box. If execution performance of the generated code is your top 
priority, I wouldn't recommend using DigitalMars products.


More information about the Digitalmars-d mailing list