Stories
Slash Boxes
Comments

SoylentNews is people

What's Worked in Computer Science

posted by cmn32480 on Monday November 30 2015, @08:19AM   Printer-friendly
from the thought-provoking dept.

Phoenix666 writes:

In 1999, Butler Lampson gave a talk about the past and future of "computer systems research" [PDF]. Here are his opinions from 1999 on "what worked".

Yes

Virtual memory
Address spaces
Packet nets
Objects / subtypes
RDB and SQL
Transactions
Bitmaps and GUIs
Web
Algorithms

Maybe

Parallelism
RISC
Garbage collection
Reuse

No

Capabilities
Fancy type systems
Functional programming
Formal methods
Software engineering
RPC
Distributed computing
Security

Basically everything that was a Yes in 1999 is still important today.

The article is a current snapshot on those issues. Do you agree?

Original Submission

 
This discussion has been archived. No new comments can be posted.
What's Worked in Computer Science | Log In/Create an Account | Top | 125 comments | Search Discussion
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.

  • (Score: 2) by TheRaven on Monday November 30 2015, @06:23PM

    by TheRaven (270) on Monday November 30 2015, @06:23PM (#269827) Journal

    You would be correct, if not for one minor point: power consumption. The decoder is one bit of the CPU that you have to have powered whenever you're executing any instructions. Xeons avoid this slightly by caching micro-op traces in hot loops, but generally you're taking the power consumption hit from a complex decoder all of the time. The only way for CISC to recover this is by having a sufficiently dense encoding that you can get away with a smaller L1 and maintain the same L1 hit rate as a RISC chip.

    This is particularly important if you remember that we hit the end of Dennard Scaling a few generations ago. The number of transistors per die is still going up, but the number that you can power (for a given power budget) is barely moving. This means that the best use of transistors is to provide functionality that gives a big speedup when in use, but can be powered down when not. This means that the current best power/performance comes from SoCs with lots of specialised (RISC) accelerator cores.

    --
    sudo mod me up
    Starting Score:    1  point
    Karma-Bonus Modifier   +1  
    Total Score:   2