[Beowulf] NVIDIA buying PGI

[Joe Landman]

On 07/31/2013 03:27 PM, mathog wrote:
> Douglas Eadline wrote:
>
>> It makes sense, PGI accelerator is nice and keeping
>> it out of the hands of competitors makes strategic
>> sense I suppose, I would not be expecting any further support
>> for AMD APU from PGI.
>
> Am I the only one who hates it when companies make such blatantly
> anti-competitive moves

Er ... blatantly anticompetitive?  In what way?  Are there not still 
many good compilers, including good open source compilers, on the market?

It makes sense from a competitive "gird your loins" sort of way, as 
Intel has its compiler happily able to emit Larrabee ... er ... MIC 
executables.  Is that technology anticompetitive in Intel's hands?

Is the potential for less AMD APU support anticompetitive?  There are 
multiple choices in the market for accelerators, AMD APUs are just one 
of them.

> (and the antitrust people look the other way)?  This may (may) be good
> for Nvidia, but I don't see
> it as being at all good for the customers.

How would it be bad for customers?  What choices are being removed or 
denied you that you *would* take?  I am not talking about potential 
choices such as AMD ... if enough people took them seriously, I am 
guessing they would be in far better economic health than they are now.

>
> At least Nvidia isn't being sneaky about it.  Unlike when Intel's
> compilers were needlessly generating
> inferior code for AMD targets, or when Microsoft made early Windows
> throw an error when it ran on DR-DOS

Intel didn't look to see if the CPU ID said AMD.  It applied a simple 
pattern match to determine capability.  One it should not have done, as 
there are infinitely superior mechanisms (say looking at the flags 
entries in /proc/cpuinfo) of determining chip capabilities.  I don't 
know if they are still doing that.  But when they were, a very simple 
binary patch disabled that code pathway.

But I'd argue that it doesn't matter anymore.

AMD has GCC compilers, PG compilers (for the time being), as well as 
Open64, and the PathScale compilers.  Not to mention the NAG compilers.

So I guess I am missing something here.

Its good for NVidia or they wouldn't do it.  I don't see it as 
anti-competitive.  I do see it as shoring up a weak point in their 
offerings.

And AMD had the option of doing this.  As did others.

I'd say it was anti-competitive if Intel bought them, as choice in the 
market place would have been reduced.


-- 
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics, Inc.
email: landman at scalableinformatics.com
web  : http://scalableinformatics.com
        http://scalableinformatics.com/siflash
phone: +1 734 786 8423 x121
fax  : +1 866 888 3112
cell : +1 734 612 4615