From: Francis Russell <francis@unchartedbackwaters.co.uk>
Date: Tue, 11 Jun 2013 12:45:24 +0000 (+0100)
Subject: Add better benchmark information and minor fixes.
X-Git-Url: https://git.unchartedbackwaters.co.uk/w/?a=commitdiff_plain;h=ba406b2f98c88b6f8931c31519594d65241db95d;p=francis%2Fpsl_presentation_20130611.git

Add better benchmark information and minor fixes.
---

diff --git a/presentation.tex b/presentation.tex
index 5e21281..896cd75 100644
--- a/presentation.tex
+++ b/presentation.tex
@@ -92,7 +92,7 @@ dimension.
 
 \small
 ONETEP performs transforms in each dimension so each FFT only operates on 50\%
-zeroes instead of 87.5\% of the naïve strategy.
+zeroes instead of the 87.5\% of the naïve strategy.
 
 \vspace{1em}
 
@@ -165,7 +165,7 @@ then trasform back to compute the values of the midpoints.
 
 \frame{
 
-\frametitle{Performance Results with FFTW\footnote{Core i7-2600, 3.4GHz, 8MB L2
+\frametitle{Performance Results with FFTW\footnote{Intel Core i7-2600, 3.4GHz, 8MB L2
 cache, FFTW 3.3}}
 
 \footnotesize
@@ -181,11 +181,11 @@ problem sizes due to the factorisations it chooses.
 
 \frame{
 
-\frametitle{Performance Results with FFTW\footnote{Core i7-2600, 3.4GHz, 8MB L2
+\frametitle{Performance Results with FFTW\footnote{Intel Core i7-2600, 3.4GHz, 8MB L2
 cache, FFTW 3.3}}
 
 \footnotesize
-We can filter the results to those which FFTW likes best (products of small
+Filtering the results to those which FFTW likes best (products of small
 primes). Specifically, sizes of the form $2^a3^b5^c7^d11^e13^f$ where $e+f<2$.
 
 \centering
@@ -204,7 +204,7 @@ primes). Specifically, sizes of the form $2^a3^b5^c7^d11^e13^f$ where $e+f<2$.
 \item In our standalone benchmarks, our results give around a 35\% reduction in
   execution time over ONETEP's approach for FFTW's preferred sizes.
 
-\item In practise, we found the actual reduction to be a lot less and overall
+\item In practice, we found the actual reduction to be a lot less and overall
 reduction in execution time to usually be less than 5\%.
 
 \item When doing Fourier interpolation, ONETEP spends a lot of time in its
@@ -247,8 +247,8 @@ possible to pass in data in data in the split format.
 
 \frame{
 
-\frametitle{ONETEP Interpolation Routine Timings\footnote{Core i7-2600, 3.4GHz,
-8MB L2 182 cache, FFTW 3.3}}
+\frametitle{ONETEP Interpolation Routine Timings\footnote{ONETEP 3.3.9.5, Intel
+Core i7-2600, 3.4GHz, 8MB L2 cache, FFTW 3.3}}
 
 \centering
 \resizebox{0.65\textwidth}{!}{
@@ -259,8 +259,8 @@ possible to pass in data in data in the split format.
 
 \frame{
 
-\frametitle{ONETEP Total Execution Time Timings\footnote{Core i7-2600, 3.4GHz,
-8MB L2 182 cache, FFTW 3.3}}
+\frametitle{ONETEP Total Execution Time Timings\footnote{ONETEP 3.3.9.5, Intel
+Core i7-2600, 3.4GHz, 8MB L2 cache, FFTW 3.3}}
 
 \centering
 \resizebox{0.65\textwidth}{!}{