On Fri, Feb 08, 2002 at 01:41:11AM +0100, Myk Melez wrote: > I recently ran some performance tests in which I compared 2.04 with 2.06 > and subsequent developer releases. The difference between 2.04 and 2.06 > is statistically insignificant; 2.06a-c are slightly slower than 2.06; > and 2.06d is much slower.
This is definately a sub-optimal state of affairs. I must admit I'm at a loss as to why that should be. Even looking back at the Changes file, I can't anything that might have caused this slowdown. In fact, when it comes to compiled templates, 2.06c was known to be broken - something which was then fixed in 2.06d: * Applied another patch from Simon to fix a bug with compiled templates which were never being loaded due to a '<' comparison of timestamps rather than '<='. See http://tt2.org/pipermail/templates/2002-January/002361.html Furthermore, running a very simple benchmark of my own, I'm seeing what I would expect to see: that 2.06d is faster because the COMPILE_DIR option now works correctly (I'm deliberately not re-using the same template object each time, to defeat the internal caching): 2.06c: 11 wallclock secs (10.77 usr + 0.12 sys = 10.89 CPU) @ 9.18/s 2.06d: 7 wallclock secs ( 6.98 usr + 0.09 sys = 7.07 CPU) @ 14.14/s Hmm... let's see what happens when we ignore parsing/compiling and just benchmark the template execution: 2.06c: 6 wallclock secs ( 6.16 usr + 0.02 sys = 6.18 CPU) @ 16.18/s 2.06d: 6 wallclock secs ( 6.10 usr + 0.06 sys = 6.16 CPU) @ 16.23/s Nope, I'm still not seeing the slowdown that you. Admittedly my benchmark is rather primitive. Can you elaborate more on what you're doing in your benchmark to see if we can identified where the problem might lie? Thanks A
