Hacker News Reader

Comment by mbrubeck | Hacker News Reader

"2009Q3 uses up to 930% less memory than the 2009Q2 release."

Does that math make sense in some way I'm not seeing?

From their benchmarks rietveld: $ ./perf.py -r -b rietveld --args "-j always," --track_memory ../q2/python ../q3/python Mem max: 575116.000 -> 55952.000: 927.87% smaller

This 930% basically means that 2009Q3 can use as little as ~1/10 of the memory of 2009Q2.

"Uses 1/10 the memory" would have been much easier to parse.

For a human, maybe. We're programmers.

It's unconventional, but 9.3x less memory can be seen in some sense as 1/9.3=10.7 percent as much memory. Not defending it, just trying to interpret :P

You can't have 9.3x less memory, either. You can have (1/9.3)x as much memory, or they could just say 10% as much or 90% less.

I agree that it is technically unsound, but the meaning of "Nx less of something" is still relatively unambiguous when N>1.

Yes, it’s stupid (ambiguous, imprecise) terminology, but pretty common usage when describing computer benchmark results.

Actually, it's 10.3x less memory (they've subtracted 1), so it's using 9.7% as much memory as the previous release. I would have said that 90.3% less than before, but maybe that's just me

I never checked but I assumed it meant that the memory savings of Q3 compared to the savings of Q2 increased 930%

There's no "savings" vs. CPython: Memory usage is still 2-3x that of Python 2.6.1.

(Edit: Or did you mean that the 2009Q2->Q3 savings was 930% greater than the Q1->Q2 savings?)

Right, they're comparing against themselves.

Reply on news.ycombinator.com