By: Patrick Chase (patrickjchase.delete@this.gmail.com), December 9, 2014 1:06 pm
Room: Moderated Discussions
Maynard Handley (name99.delete@this.name99.org) on December 9, 2014 11:41 am wrote:
> Of course, per my previous post, the obvious rejoinder to this
> would be: "get a better language and a better codebase".
> Now that is flip, but it carries a core of truth. IF your goal is to keep on using an
> ancient codebase indefinitely, then you're, more or less by definition, in the business
> of wanting computers that look exactly like what the ancient codebase targeted, only
> faster. Point is, new codebases come along as well, and THEY are willing to adapt
> themselves to new hardware.
>
> IBM seems to be doing OK selling mainframes, but I don't see the rest of us assuming that
> we need to write our code using mainframe paradigms. At some point a Google says "to hell
> with this, we're starting from scratch and can organize things completely differently"...
Being careful as to what I say here as I work (but definitely don't speak) for Google these days...
I think the difference is that the "micro revolution" was a technology-driven sea change that fundamentally made mainframes less competitive. VLSI enabled CMOS microprocessors that rivaled the (traditionally) ECL-based mainframes for performance at vastly lower cost.
While you're right that the specific issue Gabriele raised can be mitigated by a different language choice, that isn't generally true. There have been no fundamental technological breakthroughs that make parallelism easier/cheaper to achieve than it was, say, a decade ago. No magical compilers, no breakthrough methodologies or languages, and therefore no substantive mitigation of Amdahl's law.
While there is certainly pressure from the semiconductor process side that limits the pace of improvement of serial performance, I don't see anything that would actually shift the equilibrium/optimum in favor of smaller cores.
I therefore think 'anon' had it about right: We'll continue to muddle along by improving per-core performance where we can and increasing parallelism where we must. It's not a terribly sexy vision, but sometimes reality works that way.
> Of course, per my previous post, the obvious rejoinder to this
> would be: "get a better language and a better codebase".
> Now that is flip, but it carries a core of truth. IF your goal is to keep on using an
> ancient codebase indefinitely, then you're, more or less by definition, in the business
> of wanting computers that look exactly like what the ancient codebase targeted, only
> faster. Point is, new codebases come along as well, and THEY are willing to adapt
> themselves to new hardware.
>
> IBM seems to be doing OK selling mainframes, but I don't see the rest of us assuming that
> we need to write our code using mainframe paradigms. At some point a Google says "to hell
> with this, we're starting from scratch and can organize things completely differently"...
Being careful as to what I say here as I work (but definitely don't speak) for Google these days...
I think the difference is that the "micro revolution" was a technology-driven sea change that fundamentally made mainframes less competitive. VLSI enabled CMOS microprocessors that rivaled the (traditionally) ECL-based mainframes for performance at vastly lower cost.
While you're right that the specific issue Gabriele raised can be mitigated by a different language choice, that isn't generally true. There have been no fundamental technological breakthroughs that make parallelism easier/cheaper to achieve than it was, say, a decade ago. No magical compilers, no breakthrough methodologies or languages, and therefore no substantive mitigation of Amdahl's law.
While there is certainly pressure from the semiconductor process side that limits the pace of improvement of serial performance, I don't see anything that would actually shift the equilibrium/optimum in favor of smaller cores.
I therefore think 'anon' had it about right: We'll continue to muddle along by improving per-core performance where we can and increasing parallelism where we must. It's not a terribly sexy vision, but sometimes reality works that way.
