 |
|  |
 |
|
Ari Rahikkala
|
 |
Unreported loss of streaming in atomic read operation
Oct 1999 time: 07:29
|
|
quote: Originally posted by Asher
IBM and Apple have been heavily optimizing the GCC compiler for the G5 for a long time now.
Intel optimizes its own Intel compiler rather than GCC.
So Apple saw this, thought it'd make a brilliant marketing move -- we can compile both using the same compiler, because 99.9% of people won't know that GCC sucks for Pentium 4 code. But just to be safe, they disabled SSE2 (which is a huge performance booster, akin to the G5's AltiVec) because they claimed it reduced performance (bullshit).
p://www.spec.org/cpu2000/results/res2003q3/cpu2000-20030616-02265.html">1205 |
Just by the way, there's an effort to add support for autovectorization in gcc with SSA for Trees (Tree-SSA is expected to be merged in in 3.5). This means that gcc will be able to use vector extension optimizations even if they haven't been explicitly coded in. icc already does this. Intel's engineers probably know their own stuff better than outsiders, so if they aren't going to help the gcc folks, I don't think gcc will be able to reach the speeds of icc on Intel's stuff. Not to mention that compiler development is rather slowish - SSE2 wasn't even supported in gcc before 3.3 (which isn't considered stable enough for general use by most distros yet), and one can expect to wait at least a year for 3.5... perhaps even longer for autovectorization. Still, the point is, gcc is going to produce faster code than it does now...
|
|
|  |
 |
|
mrmitchell
|
|
I was under the impression that it would indeed be ****ing fast, but it would also be ****ing expensive. Now, off to the Research!
|
|
|  |
All times are GMT. The time now is 05:29. Apolyton Time is 00:29. |
top of page
|
| archivepost |
|
Forum Rules:
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
|
HTML code is ON
vB code is ON
Smilies are ON
[IMG] code is ON
|
|
|
|
|
|