Author: Maciej Fijalkowski <fij...@gmail.com>
Branch: extradoc
Changeset: r3717:029b8e343f97
Date: 2011-06-16 21:48 +0200
http://bitbucket.org/pypy/extradoc/changeset/029b8e343f97/
Log: start describing benchmarks
diff --git a/talk/iwtc11/benchmarks/benchmark.sh
b/talk/iwtc11/benchmarks/benchmark.sh
--- a/talk/iwtc11/benchmarks/benchmark.sh
+++ b/talk/iwtc11/benchmarks/benchmark.sh
@@ -5,7 +5,7 @@
if [ "$1" == "gcc" ]; then
./runner.py -n 5 -c "$*" sqrt/sqrt_double.c
./runner.py -n 5 -c "$*" sqrt/sqrt_long.c
- #./runner.py -n 5 -c "$*" sqrt/sqrt_fix16.c
+ ./runner.py -n 5 -c "$*" sqrt/sqrt_fix16.c
#./runner.py -n 5 -c "$* -lm" convolution/conv3.c 1
#./runner.py -n 5 -c "$* -lm" convolution/conv5.c 1
./runner.py -n 5 -c "$* -lm" convolution/conv3.c 100
@@ -20,7 +20,7 @@
else
$* ./runner.py -n 10 sqrt/sqrt.py main int
$* ./runner.py -n 10 sqrt/sqrt.py main float
- #$* ./runner.py -n 10 sqrt/sqrt.py main Fix16
+ $* ./runner.py -n 10 sqrt/sqrt.py main Fix16
#$* ./runner.py -n 10 convolution/convolution.py conv3 1
#$* ./runner.py -n 10 convolution/convolution.py conv5 1
$* ./runner.py -n 10 convolution/convolution.py conv3 100
diff --git a/talk/iwtc11/benchmarks/new_result.txt
b/talk/iwtc11/benchmarks/new_result.txt
--- a/talk/iwtc11/benchmarks/new_result.txt
+++ b/talk/iwtc11/benchmarks/new_result.txt
@@ -1,63 +1,37 @@
-pypy --jit enable_opts=intbounds:rewrite:virtualize:heap:unroll
-sqrt(int): 1.79892385006 +- 0.00194840037512
-sqrt(float): 0.983013772964 +- 0.00221919586293
-conv3(1e6): 0.766417503357 +- 0.00866699505143
-conv5(1e6): 0.996688437462 +- 0.012036835877
-conv3(1e5): 0.730618429184 +- 0.00375146136701
-conv5(1e5): 1.03531208038 +- 0.0111413026874
-conv3x3(3): 0.069846701622 +- 0.000501920798166
-conv3x3(1000): 0.0522719621658 +- 0.0357056076979
-dilate3x3(1000): 0.38942694664 +- 0.00619291977785
-NoBorderImagePadded: 1.89698078632 +- 0.0208055951105
-NoBorderImagePadded(iter): 0.519681739807 +- 0.0200662890046
-NoBorderImagePadded(range): 0.450081467628 +- 0.00105444417894
-NoBorderImage: 2.13951308727 +- 0.00576674378529
-NoBorderImage(iter): 1.46965010166 +- 0.00394661836239
-NoBorderImage(range): 1.35105161667 +- 0.00249887289286
-sobel(NoBorderImagePadded): 0.45955350399 +- 0.00145458444751
-sobel_uint8(NoBorderImagePadded): 0.498426914215 +- 0.00665320862997
+pypy
+sqrt(int): 1.81218411922 +- 0.0217793211373
+sqrt(float): 0.987752747536 +- 0.010412866907
+sqrt(Fix16): 2.21716473103 +- 0.00636762886973
+conv3(1e6): 0.756063270569 +- 0.0225483799804
+conv5(1e6): 1.07853357792 +- 0.00813703177389
+conv3(1e5): 0.731228137016 +- 0.0161735983974
+conv5(1e5): 1.05882668495 +- 0.0341698164686
+conv3x3(3): 0.0685305118561 +- 0.000278350915493
+conv3x3(1000): 0.0520143270493 +- 0.0342524928471
+dilate3x3(1000): 0.415372800827 +- 0.044525202511
+NoBorderImagePadded: 1.89855155945 +- 0.00727766698755
+NoBorderImagePadded(iter): 0.477786374092 +- 0.00111654321125
+NoBorderImagePadded(range): 0.451609492302 +- 0.00325743719387
+NoBorderImage: 2.22490911484 +- 0.0605514739511
+NoBorderImage(iter): 1.46266727448 +- 0.00415733919764
+NoBorderImage(range): 1.37843291759 +- 0.0630244441411
+sobel(NoBorderImagePadded): 0.463502717018 +- 0.00257025834276
+sobel_uint8(NoBorderImagePadded): 0.52193570137 +- 0.0198069947724
pypy --jit enable_opts=intbounds:rewrite:virtualize:heap
-sqrt(int): 2.27739796638 +- 0.0271040580427
-sqrt(float): 1.364168787 +- 0.0235396053333
-conv3(1e6): 1.72038755417 +- 0.0280206343663
-conv5(1e6): 1.93043384552 +- 0.0302489061093
-conv3(1e5): 1.6559261322 +- 0.0364074757582
-conv5(1e5): 1.85165474415 +- 0.032410582414
-conv3x3(3): 0.107097601891 +- 0.00457118866065
-conv3x3(1000): 0.0721160173416 +- 0.00365968876656
-dilate3x3(1000): 0.43175163269 +- 0.0720869033105
-NoBorderImagePadded: 2.00819942951 +- 0.0260239930765
-NoBorderImagePadded(iter): 1.22523207664 +- 0.026102105011
-NoBorderImagePadded(range): 1.113205266 +- 0.0381177388909
-NoBorderImage: 2.21718068123 +- 0.0503771001922
-NoBorderImage(iter): 1.39955751896 +- 0.0034236237913
-NoBorderImage(range): 1.34794125557 +- 0.0379578329049
-sobel(NoBorderImagePadded): 1.00590751171 +- 0.0175536088063
-sobel_uint8(NoBorderImagePadded): 1.03622698784 +- 0.00533611100064
-
-
-gcc -O2
-sqrt(float): 0.98 +- 0.00707106781187
-sqrt(int): 0.792 +- 0.004472135955
-conv3(1e6): 0.77 +- 0.0141421356237
-conv5(1e6): 1.026 +- 0.00894427191
-conv3(1e5): 0.686 +- 0.00894427191
-conv5(1e5): 0.976 +- 0.00547722557505
-conv3x3(3): 0.282 +- 0.00836660026534
-conv3x3(1000): 0.244 +- 0.00894427191
-dilate3x3(1000): 0.252 +- 0.004472135955
-sobel_magnitude: 0.18 +- 0.0
-
-gcc -O3 -march=native -fno-tree-vectorize
-sqrt(float): 0.978 +- 0.004472135955
-sqrt(int): 0.79 +- 0.0
-conv3(1e6): 0.74 +- 0.00707106781187
-conv5(1e6): 0.768 +- 0.01788854382
-conv3(1e5): 0.576 +- 0.00547722557505
-conv5(1e5): 0.652 +- 0.00836660026534
-conv3x3(3): 0.27 +- 0.0
-conv3x3(1000): 0.242 +- 0.004472135955
-dilate3x3(1000): 0.25 +- 0.0
-sobel_magnitude: 0.196 +- 0.00894427191
+sqrt(int): 2.25307536125 +- 0.00499253786735
+sqrt(float): 1.35373134613 +- 0.0033224112985
+sqrt(Fix16): 3.76334681511 +- 0.0499187749691
+conv3(1e6): 1.68589990139 +- 0.00487515973416
+conv5(1e6): 1.89262311459 +- 0.00528534075689
+conv3(1e5): 1.65629634857 +- 0.0281428596263
+conv5(1e5): 1.87937986851 +- 0.0451342539426
+conv3x3(3): 0.100474405289 +- 0.00342971298052
+conv3x3(1000): 0.0688207149506 +- 0.00054547298268
+dilate3x3(1000): 0.415153169632 +- 0.0341885675395
+NoBorderImagePadded: 1.9624298811 +- 0.0123395556581
+NoBorderImagePadded(iter): 1.1778367281 +- 0.00914555357417
+NoBorderImagePadded(range): 1.13030288219 +- 0.00631341257207
+NoBorderImage: 2.17767724991 +- 0.0252138424223
+NoBorderImage(iter): 1.45121192932 +- 0.0274908528137
diff --git a/talk/iwtc11/benchmarks/runall.sh b/talk/iwtc11/benchmarks/runall.sh
--- a/talk/iwtc11/benchmarks/runall.sh
+++ b/talk/iwtc11/benchmarks/runall.sh
@@ -1,10 +1,10 @@
-#!/bin/sh
+#!/bin/bash
-#./benchmark.sh pypy
-./benchmark.sh pypy --jit enable_opts=intbounds:rewrite:virtualize:heap:unroll
+./benchmark.sh pypy
+#./benchmark.sh pypy --jit enable_opts=intbounds:rewrite:virtualize:heap:unroll
./benchmark.sh pypy --jit enable_opts=intbounds:rewrite:virtualize:heap
#./benchmark.sh gcc
./benchmark.sh gcc -O2
./benchmark.sh gcc -O3 -march=native -fno-tree-vectorize
-#./benchmark.sh python2.7
-
+./benchmark.sh python2.7
+./benchmark.sh python2.6 psyco-wrapper.py
diff --git a/talk/iwtc11/paper.tex b/talk/iwtc11/paper.tex
--- a/talk/iwtc11/paper.tex
+++ b/talk/iwtc11/paper.tex
@@ -889,6 +889,25 @@
under a nice interface without loosing performance.
\end{itemize}
+Benchmarks were run on Intel i7 M620 @2.67GHz with 4M cache and 8G of RAM in
+32bit mode.
+The machine was otherwise unoccupied. We run the following software
+for benchmarks:
+
+\begin{itemize}
+\item PyPy 1.5
+\item CPython 2.7.2
+\item Psyco 1.6 with CPython 2.6.6
+\item GCC 4.4.5 shipped with Ubuntu 11.4
+\end{itemize}
+
+We run GCC both with -O2 optimization and -O3 -march=native, disabling the
+automatic loop vectorization. In all cases, SSE2 instructions were used for
+floating point operations, except Psyco which uses x87 FPU instructions.
+We also run PyPy with loop peeling optimization and without (but otherwise
+identical).
+
+XXX discuss some more, explain why gcc is faster
XXX we need Psyco numbers
\subsection{Numpy}
_______________________________________________
pypy-commit mailing list
pypy-commit@python.org
http://mail.python.org/mailman/listinfo/pypy-commit
