DragonFly BSD
DragonFly users List (threaded) for 2005-02
[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]

Re: Dragonfly and Hyperthreading....

From: EM1897@xxxxxxx
Date: Tue, 22 Feb 2005 09:51:00 -0500

In a message dated 2/22/2005 1:09:38 AM Eastern Standard Time, Matthew Dillon <dillon@xxxxxxxxxxxxxxxxxxxx> writes:

>    I agree with all your comments and will add a few more negative effects
>    from having a long pipeline... and that is that pipeline stall and miss
>    conditions are seriously aggravated when you have a longer pipeline.
>    Intel underestimated the effect branch prediction misses, register
>    collisions, and main memory access delays had on their pipeline.
>    Perfectly predictable, hand-optimized code can run very, very fast on an
>    Intel cpu, but since most code is not perfectly predictable and is
>    definitely not hand-optimized, they wound up hitting these situations
>    more often then they liked.
>    Also, the latches separating each pipeline stage (even using the
>    trick of alternating the clock phase for each stage) impose a minimum
>    of two gate delays plus clock slop plus wire routing delay and this
>    puts a limit on how little logic you can have in each stage and
>    still be effective.  Having fewer, larger pipeline stages can actually
>    wind up being faster in some cases, especially when you are shoving
>    data out to a slower unit (like main memory) which can absorb additional
>    time slop.  So, e.g. if you have a pipeline stage which can output 
>    either to a register or to a memory buffer it could very well be that
>    the logic going to the memory buffer can exceed the nominal stage
>    time without creating a problem.

Wow. And I thought I was a geek :-)

You know its funny, we talk about intel like they are
a bunch of bumbling twits, but the truth of the matter
is that without them we'd all be shoveling gravel or
delivering packages or something. Perhaps a moment to 
ponder the days of our dual-floppy wordstar systems, 
or reviewing some 6502 assembly code would give us
a fresh perspective?

And while AMD is getting all the press on the Opteron,
Intel is beating their brains out in the flash memory
market, they've completely taken over the notebook 
market with the P-M, and AMD posted a loss last quarter. 
They know what they're doing. Its not all about who
has the fastest processor.

[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]