Replies to post #4932 on Advanced Micro Devices Inc (AMD)

[Haddock]

Without any help, this would almost assuredly be true.

In general isn't the point that you can't help software vendors port to one architecture without helping them with the other. After all, the OSs and compilers are the same ones. Unless you are suggesting help in exchange for promises not to port to the competitor architecture, which I don't think the software companies would be likely to fall for/accept.

Are you saying that you know such agreements are in the offing?

[Petz]

One thing I'm sure Intel helped "port" to Itanium is the compilers and OS's used to run specINT and specFP. According to this link, http://www.spec.org/osg/cpu2000/results/res2001q2/cpu2000-20010522-00660.html
Windows Advanced Server Limited Edition has been available for Itanium for two years, but SPEC has never published a specFP result for Itanium 2 running under any versino of Windows.

Itanium was able to get world-beating specFP results two years ago running Windows Advanced Server. For Itanium 1, there are 13 specFP results published, 10 of them were running Windows.

For Itanium 2, there are even more results -- 16 -- but NONE of them running Windows.

The only Itanium 2 result for specINT running Windows is the Bull NovaScale 4040, and there is no corresponding specFP result!

Windows might not be as ubiquitous amongst scientific workstation users as it is everywhere else, but I smell an incompatibility in the Windows/Itanium 2/SPEC combination.

Perhaps the real problem is that the only way to get decent spec results is to write the code in assembly language and have the compiler recognize and insert it when it sees the SPEC code. Intel's army of programmers hasn't finished the job yet.

I'd sure like to see the SPEC scores obtained by running the same compiler and compiler switches used for the Itanium 1 SPEC results in 2001 -- but running the code on an Itanium 2. Must be more embarrassing than to leave zero results at spec.org.

Petz