SVN: z3c.pt/trunk/ Refactored compiler classes; reimplemented disk-cache module.

[Hanno Schlichting]

Hi Malthe.

Refactored compiler classes; reimplemented disk-cache module.

Is there any specific reason why you switched to a pickle based cache
instead of a byte-code based one? This made the cache twelve times
slower than before :( At least using the binary format (protocol 2) made
it only six times slower, but still something I would like to know a
reason for.

As a side note: Something causes considerable startup overhead now.
Instead of being three times faster in rendering the most simple 'Hello
World' benchmark, we are now slower with a 0.6x factor :(

Your performance addict,
Hanno


--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "z3c.pt" group.
To post to this group, send email to ***@googlegroups.com
To unsubscribe from this group, send email to z3c_pt+***@googlegroups.com
For more options, visit this group at http://groups.google.com/group/z3c_pt?hl=en
-~----------~----~----~----~------~----~------~--~---

[Malthe Borch]

Hey Hanno, –––

Right, so the main feature of the new disk cache is that it actually
correctly represents the entire state of the template class, something
that was not the case previously. It can probably be optimized and we
can also revert to the previous strategy, but it's important that
we've got something that behaves correctly now.

The challenge was to keep all of the compiled stuff in a single file;
using pickle was a shortcut; we can certainly experiment with other
strategies.

As for Hello World, then yes, I've noticed this and I wonder what
causes it; unfortunately, my refactor was so substantial that I could
not incrementally check the benchmark.

Let's meet in #repoze.

\malthe

Post by Hanno Schlichting
Hi Malthe.

Refactored compiler classes; reimplemented disk-cache module.

Is there any specific reason why you switched to a pickle based cache
instead of a byte-code based one? This made the cache twelve times
slower than before :( At least using the binary format (protocol 2) made
it only six times slower, but still something I would like to know a
reason for.
As a side note: Something causes considerable startup overhead now.
Instead of being three times faster in rendering the most simple 'Hello
World' benchmark, we are now slower with a 0.6x factor :(
Your performance addict,
Hanno

--
--=====================--
mail: ***@gmail.com
homepage: zeitmaschine.dk
--=====================--

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "z3c.pt" group.
To post to this group, send email to ***@googlegroups.com
To unsubscribe from this group, send email to z3c_pt+***@googlegroups.com
For more options, visit this group at http://groups.google.com/group/z3c_pt?hl=en
-