python simply not scaleable enough for google?

Terry Reedy tjreedy at udel.edu
Wed Nov 11 21:15:02 CET 2009


Robert P. J. Day wrote:
> http://groups.google.com/group/unladen-swallow/browse_thread/thread/4edbc406f544643e?pli=1
> 
>   thoughts?

Program_cost = human_writing&maintance_cost + running_cost*number_of_runs

Nothing new here. The builtin types and many modules are written in C to 
reduce running cost for frequency used components. The first killer ap 
for Python was scientific computing with early numerical python, where 
people often run one-time combinations of inputs and constantly reused 
linear algebra functions.

Google has an unusually high number of runs for many programs, making 
running_cost minimization important.

At one time, C was not 'scaleable enough' for constantly rerun aps. 
Hotspots were hand rewritten in assembler. Apparently now, with 
processors more complex and compilers much improved, that is not longer 
much the case.

I can imagine a day when code compiled from Python is routinely 
time-competitive with hand-written C.

Terry Jan Reedy




More information about the Python-list mailing list