Review vmgen branch for merge into trunk

[GoogleCodeExporter]

Purpose of code changes on this branch:
To use vmgen to generate the main interpretation loop in PyEval_EvalFrameEx().

When reviewing my code changes, please focus on:
Readability, understandability, and anything stupid I did.

After the review, I'll merge this branch into:
/trunk

Original issue reported on code.google.com by jyass...@gmail.com on 31 Jan 2009 at 1:00

[GoogleCodeExporter]

I'll add performance comparisons as I run them.

32-bit build, running on an Intel Core2 E5420 @ 2.50GHz.

Report on Linux enki.corp.google.com 2.6.24-21gg10-generic #1 SMP Sat Nov 8 
13:10:36
PST 2008 x86_64                                                                 

2to3:                                                                           

Min: 24.786 -> 23.145: 7.09% faster                                             

Avg: 24.901 -> 23.289: 6.92% faster                                             

Django:                                                                         

Min: 0.271 -> 0.268: 1.24% faster                                               

Avg: 0.277 -> 0.269: 3.12% faster                                               

PyBench:                                                                        

Min: 54232 -> 47880: 13.27% faster                                              

Avg: 55485 -> 48814: 13.67% faster                                              

SlowSpitfire:                                                                   

Min: 0.743 -> 0.684: 8.62% faster                                               

Avg: 0.747 -> 0.688: 8.65% faster                                               

Spitfire:                                                                       

Min: 0.510 -> 0.508: 0.39% faster                                               

Avg: 0.518 -> 0.510: 1.74% faster   

Original comment by jyass...@gmail.com on 31 Jan 2009 at 1:09

[GoogleCodeExporter]

Times are "trunk -> vmgen"

64-bit build, running on an Intel Core2 E5420 @ 2.50GHz.

Report on Linux enki.corp.google.com 2.6.24-21gg10-generic #1 SMP Sat Nov 8 
13:10:36
PST 2008 x86_64

2to3:
Min: 21.913 -> 20.129: 8.86% faster
Avg: 22.057 -> 20.194: 9.23% faster

Django:
Min: 0.256 -> 0.245: 4.49% faster
Avg: 0.263 -> 0.249: 5.98% faster

PyBench:
Min: 43165 -> 35775: 20.66% faster
Avg: 43979 -> 36675: 19.92% faster

SlowSpitfire:
Min: 0.684 -> 0.617: 10.85% faster
Avg: 0.691 -> 0.621: 11.30% faster

Original comment by jyass...@gmail.com on 31 Jan 2009 at 1:53

[GoogleCodeExporter]

64-bit build on an AMD Opteron 8214 HE

2to3:
Min: 30.441 -> 29.048: 4.80% faster
Avg: 30.693 -> 29.107: 5.45% faster

Django:
Min: 0.312 -> 0.320: 2.50% slower
Avg: 0.321 -> 0.322: 0.42% slower

PyBench:
Min: 69089 -> 61642: 12.08% faster
Avg: 70091 -> 62977: 11.30% faster

SlowSpitfire:
Min: 1.309 -> 1.076: 21.67% faster
Avg: 1.333 -> 1.097: 21.54% faster

Original comment by jyass...@gmail.com on 2 Feb 2009 at 9:50

[GoogleCodeExporter]

32-bit build on the AMD Opteron 8214 HE
Btw, all these runs have been with perf.py --rigorous

2to3:
Min: 33.861 -> 32.768: 3.34% faster
Avg: 33.969 -> 33.032: 2.84% faster

Django:
Min: 0.354 -> 0.340: 4.07% faster
Avg: 0.358 -> 0.345: 3.62% faster

PyBench:
Min: 83170 -> 75183: 10.62% faster
Avg: 84751 -> 77114: 9.90% faster

SlowSpitfire:
Min: 1.174 -> 1.046: 12.15% faster
Avg: 1.189 -> 1.061: 12.09% faster

Spitfire:
Min: 0.753 -> 0.744: 1.15% faster
Avg: 0.762 -> 0.753: 1.22% faster

Original comment by jyass...@gmail.com on 2 Feb 2009 at 11:23

[GoogleCodeExporter]

Set review issue status to: Started

Original comment by jyass...@gmail.com on 7 Feb 2009 at 1:37

• Changed state: Started

[GoogleCodeExporter]

Set review issue status to: Fixed

Original comment by jyass...@gmail.com on 18 Feb 2009 at 7:27

• Changed state: Fixed