[geeks] Dual Core Rules: your bugs will run twice as fast

[Francois Dion]

On 2/14/07, Sandwich Maker <adh at an.bradford.ma.us> wrote:
> " From: Patrick Giagnocavo <patrick at zill.net>
> "
> " On Feb 13, 2007, at 1:58 PM, Joshua Boyd wrote:
> " >>
> " >
> " > I haven't been impressed with the speed.  The few things were it would
> " > really matter don't seem to be as well optimized.  For instance,
> " > mplayer
> " > and ffmpeg don't really seem to be any faster on a 1.8ghz C2D in 64 bit
> " > mode compared to a 1.7ghz P-M, and I suspect it is because of less hand
> " > assembly tuning.  I might appreciate the compile speed, but I can't use
> " > 64bit mode for compiling (unless I set up 32bit cross compilers, a step
> " > that I just haven't had the time for).
> " >
> "
> " I wonder what would happen if you tried the Sun Studio Compiler for
> " Linux.  Might it generate better 64bit code?
>
> it certainly should, according to sun - at least on the sparc
> platform.  arguable whether they'd be better than gcc on x64.

In my experience at work, in real applications, under Solaris x64, sun
studio makes code that runs faster between 10% and > 300%. On the same
hardware (multi boot), the slowest is Visual Studio .NET on XP64, then
gcc on Ubuntu 64, then gcc on Solaris x64 and finally,  sun studio on
Solaris x64. I suspect that the tcp stack in solaris does make for a
good bit of the differences I'm seeing between ubuntu 64 and solaris
x64 with gcc, but there is no doubt that sun studio improves this
further by a good bit. YMMV.

We've used Visual Studio for as long as I've been here (going on 10
years), gcc equally long, but only started using sun studio in 2003 or
so. We just now switched to Sun Studio Express for the IDE (netbeans 6
based), but still compile with 11 for production. I think August 07 is
when Sun Studio 12 will go RR or maybe GA.


Francois