Do the Math [LWN.net]

Do the Math

Posted Nov 16, 2006 23:03 UTC (Thu) by GreyWizard (subscriber, #1026)
In reply to: Resisting the binary blob by landley
Parent article: Resisting the binary blob

An 8-bit processor can address only 256 bytes of memory with a single word, yet the maximum practical memory was 16 kilobytes. A 16-bit processor can address only 16 kilobytes of memory with a single word, yet the maximum practical memory was one megabyte. A 32-bit processor can address only 4 gigabytes of memory with a single word but you insist that a rapid transition to 64-bit machines will happen when 4 gigabytes "falls off the low end of the market." Are you unaware of PAE? Have you not noticed that people have been deploying 32-bit server systems with up to 16 gigabytes for quite some time with good results?

Do you dispute that desktop and laptop systems could do the same or do you admit that your deadline is at least three years too early? Either way, this might be a clue about the credibility of the rest of the document.

(Log in to post comments)

Do the Math

Posted Nov 18, 2006 6:41 UTC (Sat) by roelofs (subscriber, #2599) [Link]

A 16-bit processor can address only 16 kilobytes of memory with a single word

Er...2^16 == 64K. ;-)

Greg

s/16 kilobytes/64 kilobyes/

Posted Dec 3, 2006 20:50 UTC (Sun) by GreyWizard (subscriber, #1026) [Link]

Yes, 16K was a typo. Well spotted.

Do the Math

Posted Nov 19, 2006 2:21 UTC (Sun) by landley (guest, #6789) [Link]

Your first question (about why the 8 bit machines could access more than
256 bytes) was answered in footnote 8. Your second question (about PAE)
was answered in footnote 13. As for question #3, "Will the laptop slow
things down" is the point of the bit at the end of Section 1, first
paragraph under the title "Is 2008 a hard deadline?", the sentence
starting with "At the time, we doubted our own findings..."

I take it you skimmed a bit?

Doing some skimming yourself?

Posted Dec 3, 2006 20:50 UTC (Sun) by GreyWizard (subscriber, #1026) [Link]

Actually, I read the entire paper, but forgot that you mentioned PAE without credibly addressing the implications. Contrary to the claims in your paper, there is no need to rewrite user space software to get performance benefits in most cases. The kernel manages the memory through virtual memory hardware. (This would not be enough for applications that require more than 3GB in a single process, but if your claim is that this is the important limit then your deadline is too late rather than too early -- and still wrong.)

On the other hand, you apparently failed to read the relatively short comment you responded to. The first sentence about 8-bit computers did not end in a question mark because it was not a question, so there is no need for a footnote to answer it. The rest of your response is difficult to decipher. Since you seem to regard yourself as too cool for punctuation, exactly which sentence in my previous message are you counting as "question #3"? That you once doubted your findings is no excuse for being obviously wrong, even according to your own othewise dubious logic.