Docstoc

award-Colwell

Document Sample
award-Colwell Powered By Docstoc
					  What Computer
Architects Really
        Do
                          “Or should.”
     Bob Colwell
      E-M Talk ISCA ‘05
But First, I‟d Like to Thank
 Ron Hoelzeman (Pitt), Doug Jensen, Dan
  Siewiorek (CMU), George Cox, Kevin Kahn
 Paul Rodman, Dave Papworth, Rich Lethin,
  Josh Fisher & amazing Multiflow team
 Incredible P6 team, esp. Randy Steck, Glenn
  Hinton, Mike Fetterman, Andy Glew, Dave
  Papworth, Gurbir Singh
 My parents, Ellen, Kelly, Ken, Kristen
 Joe Malingowski
 Yale Patt, Wen-mei Hwu, Guri Sohi, Tom
  Conte, Computer Arch community            2
                                            2
Is there a trend here…




                         3
                         3
A Misconception About What
Computer Architects Do
EE Times May 23, 2005
“Is the day of the architect over?”
“Microprocessor architects managed to re-
  create almost the whole history of the
  mainframe computer industry…they used
  all the tricks, from microprogramming
  and stripped-down pipelines with load-
  store architectures to speculative
  execution and branch prediction. Best of
 all, hardly anyone was unkind enough to
 comment that all this ground had been
 covered already, just at a lesser level
 of integration.”
                                           4
                                           4
What would we have done…
Had we been born 300 years ago
  Same IQ‟s but no computers, no electronics
  Power source = waterwheels and oxen
  Same brains as today but different challenges
Likewise with computer pioneers
  Issue isn‟t “why were they so innovative & why
    aren‟t we”
  They did what we do: whatever is necessary
      Those things change over time

                                                   5
                                                   5
Intellectual Giant Theory
Intellectual giants did walk the earth in
  the ‟60‟s
  – Eckert, Mauchly, von N, Conway, Cocke, Brooks,
    Flynn, Tomasulo…
  – we should honor pioneering contributions
  – but today‟s designers are not leeches living off
    that legacy


Intellectual giants did walk the earth in
  the ‟60‟s. They still do.
                                                  6
                                                  6
Design today: more complex

  system                 1960‟s complexity
                     – Poor tools, interaction of
                       electronics, packaging & ISA



           CPU
                         Today‟s complexity
                     – Today‟s complexity from speed,
                       hyper-aggressive uArch‟s, power
                       limits, SW compatibility, number of
                       usage models, CPU-Mesmerization



    CPU-Mesmerization. Root cause? Profits.
                                                       7
                                                        7
64-bit extensions to IA32


                    CPU architect




IA32+64-bit exts         IA32



                                    8
                                    8
Multicore

        $100B industry




                         9
                         9
Design today: scarier
           much harder than design-
 No-recalls
 for-minimal-field-service
  – Pioneers designed for 1,000 users
      Design errors? Charge „em for service calls
  – Today we design for 1,000,000,000 users
      Design errors? Pray…




                                                    10
                                                    10
What architects really do
   Insidious error: thinking architects design
    instruction sets & uArch mechanisms
   We have, and do, but that misses the point
     – Architects start out as generals, moonlight as Special
       Forces
     – Range freely, identify needs, apply appropriate force
     – Ensure that biggest risks are attacked first
     – Make sure project goals are clear & focussed
     – Seek odd viewing angles to drive out problems
     – Supply judgment calls where data is lacking
          And judgment as to when data should be collected
                                                                11
                                                                11
Recent arch history
  – For past two decades an architect‟s point of
    highest leverage has been microarchitecture
       Re-use what works
          – Pipelining, caches, shared buses, superscalar
       Invent where necessary
          – Microdataflow/OOO, trace caches, speculative branch
            predictors, cache coherency
       With some major ISA work on RISCs
  – But this is changing. Right now.

CPU architects must evolve into system designers

                                                                  12
                                                                  12
System Designers todo list
“Whatever needs doing” has become…
  1. Products, not CPUs
  2. Power-constrained system design
  3. Multicore (gotta pay the bills, too)
  4. Reliable systems from unreliable
     components


  No longer “what I‟d like to sell you” but
    designing what buyers want
                                              13
                                              13
Products Not CPUs
$$$

                             2004




                                                Cell phones
                                                 Ray tracing
                                    iPOD     Portable computing


                                                         Time

            “PC era”                       Ubiq. comp.
      “Speed at any price”          “What’s in it for me?” -buyer 14
                                                                  14
Power-constrained design

   “fast as possible at max power” will yield to
    “fast enough, no faster”
     – Lesson from the embedded space

   Thermal variability vs. guaranteed real-time
     – Throw in wireless links for good measure

   Battery life, not just cooling cost
   Global warming, energy crisis looms
     – It ain‟t just cars and oil prices
     – Be synchronized to public taste or lose


                                                  15
                                                  15
Multicore
 “HereI come, ready or not…”
 We can build „em. Can we…
  – Compile to them?
  – Feed them? (bandwidth)
  – Cool them? (power)
  – Write apps for them?
 Clear   and present challenge
  – There are pots of gold associated w/ this

                                                16
                                                16
Reliable Systems from
Unreliable Components
 N-mod  redundancy too expensive
 Transient errors, manufacturing
  defects, design errors
  – Must survive them all
  – Solution can‟t drive power up
  – Still want guaranteed performance for real-time
       Shekhar Borkar says we have at
 Intel‟s
 most 10 years to figure this out
                                                  17
                                                      17
Adjurations
 Computer     revolution is only getting
  started
 Role of architects is changing
  – If we don‟t do it who will?
 Your   grandchildren will thank you
  – And wonder if they‟re as smart as you were 




                                                   18
                                                   18
backups




          19
          19

				
DOCUMENT INFO