By: Richard Cownie (tich.delete@this.pobox.com), November 24, 2010 10:44 am
Room: Moderated Discussions
ajensen (@.) on 11/24/10 wrote:
---------------------------
>I'm sure if you can dig up the most difficult workload in the world with low ILP,
>random branches, cache misses everywhere ...
I think you just described databases and web servers
and browsers ...
>then IA64 would absolutely
>suck at it no mater how you do the uarch. And OOOE x86 or >RISC would excel at it. Big deal.
Yes, applications with those characteristics *are* a big
deal. A huge deal. Which is part of the reason why
IA64 finds itself confined to a small segment of the
total market for computing devices - while x86 is
increasingly dominant in "computers", and ARM is sneaking
up in other kinds of devices (and an honorable mention
for MIPS).
The other parts of the reason are poor execution and
politics - customers like the rapid progress that you
tend to get with architectures available from multiple
competing suppliers.
---------------------------
>I'm sure if you can dig up the most difficult workload in the world with low ILP,
>random branches, cache misses everywhere ...
I think you just described databases and web servers
and browsers ...
>then IA64 would absolutely
>suck at it no mater how you do the uarch. And OOOE x86 or >RISC would excel at it. Big deal.
Yes, applications with those characteristics *are* a big
deal. A huge deal. Which is part of the reason why
IA64 finds itself confined to a small segment of the
total market for computing devices - while x86 is
increasingly dominant in "computers", and ARM is sneaking
up in other kinds of devices (and an honorable mention
for MIPS).
The other parts of the reason are poor execution and
politics - customers like the rapid progress that you
tend to get with architectures available from multiple
competing suppliers.