? (0xe2.0x9a.0x9b@gmail.com) on 7/21/10 wrote:
---------------------------
>anon (anon@anon.com) on 7/21/10 wrote:
>---------------------------
>>But you're wrong about this as an example of how pipelining requires speculation
>>because CPUs disallow such sequences without cache and pipeline draining instructions.
>
>There are two options of how to design a pipelined version of a CPU:
>
>1. Maintain full compatibility between the pipelined and non-pipelined versions,
>even for all the unlikely corner-cases such as intense self-modifying code. The
>two CPUs give mathematically identical results in all possible cases.
>
>2. Do not maintain full compatibility. One specimen of this approach are the transitions
>from 8086 to 286 to 386 to 486 etc.
>
>I was mainly talking about case (1).

snip

>>Exceptions are another one, but again it could have imprecise exceptions and not require speculative execution.
>
>Well, but you cannot go back to 1970 and fix 8086 so that it knows about the concept of imprecise exceptions.
>

snip

>>On the contrary, I think your assertion that pipelining requires speculative execution is wrong.
>
>I am not saying that. I am saying this: Pipelining, the way it is implemented in
>current CPUs (e.g: Core2), involves certain assumptions

I'm sorry, but you're just modifying your argument as you go along now. You can't tell me you were talking specifically about x86 and constraints of market realities in CPU design. Because we can all read what you posted to Linus.

"NOTE that this theoretical OoO CPU I am mentioning here
is *not* doing any *speculation* (except maybe those
which also present in the in-order CPU). Once you have
speculation, there by necessity exist some cases in which
the CPU's prediction engine will make a wrong decision.

Even mere pipelining entails a small dose of speculative
execution !!!

I will repeat it one more time, because it seems to me
that you are completely missing this: Even mere
pipelining entails a small dose of speculation !!!"

You are clearliy talking about theoretical designs, and you were saying that pipelining requires speculation.

snip

>>Caches have nothing to do with speculative execution that you were talking about.
>
>What are you saying? That if a contemporary CPU (Core2 or whatever) decides to
>allocate a cache-line for data at address 0x1230, it is not making any speculations
>about future uses of that piece of data?
>
>(Note: I am *not* against caches)
>

Yes there is a speculative nature about caches, of course it is nothing to do with speculative execution or anything in the CPU pipeline which is what this discussion was supposed to be about (out of order versus in order, pipelining, speculative execution etc). I'll quote you again

"Even mere pipelining entails a small dose of speculative
execution !!!"

I'm not saying you don't know the difference, but I'm saying that the argument about execution engines is only muddied by you bringing up caches.