[pypy-dev] PyPy 1.1 beta release

Carl Friedrich Bolz cfbolz at gmx.de
Tue Apr 21 09:50:24 CEST 2009


Niko Matsakis wrote:
> On Apr 20, 2009, at 8:36 PM, Jacob Hallén wrote:
> 
>> I think this formulation is a bit confusing. It is not our speed 
>> that is 0.8-2x CPython, it is our performance relative to CPython
>> that is between 0.8 and 2, with 0.8 meaning that we are faster than
>> CPython on those benchmarks, and 2 meaning that we need twice the
>> time to run the benchmark.
> 
> Maybe I am a bit confused, but I don't see a difference between those
>  two things?
> 
> i.e., if the speed is 0.8x CPython, to me that means that it runs in
>  80% of CPython's time (i.e., faster), whereas 2x CPython would be 
> twice as much time.
> 
> In any case, I agree that the second formulation is phrased more 
> clearly, just curious if my understanding of 0.8x is flawed.
> 

Hi Niko,

FWIW, that's the reasoning we had when we wrote the thing. However, we 
keep having this discussion every time we write something about 
performance, so I guess it's not as clear as you and me think :-).

Carl Friedrich



More information about the Pypy-dev mailing list