Changeset 169
- Timestamp:
- Jan 22, 2008, 2:38:27 PM (18 years ago)
- Author:
- neil.c.c.brown
- Message:
-
Tweaked the slides some more (changed date, changed a few bullet points)
- File:
-
- 1 edited
- docs/trunk/omega-test-slides/omega-test.tex (modified) (6 diffs)
Legend:
- Unmodified
- Added
- Removed
-
docs/trunk/omega-test-slides/omega-test.tex
r168 r169 35 35 \subject{Talks} 36 36 37 \date{ (削除) 24 (削除ここまで)January 2008}37 \date{(追記) 31 (追記ここまで) January 2008} 38 38 39 39 \begin{document} … … 149 149 \item General aim is to eliminate variables (reduce dimensions) 150 150 \item Deal with equalities first (easiest) then inequalities 151 \item Integer division is not reversible (and requires rational numbers), so is avoided. 151 \item Inexact integer division requires rational numbers, so is avoided. 152 \begin{itemize} 153 \item Division is only used to remove common factors in coefficients 154 \end{itemize} 152 155 \end{itemize} 153 156 \end{frame} … … 738 741 \end{lstlisting} 739 742 \begin{itemize} 740 \item For parallel replication (e.g. 0 FOR 10), we introduce a ghost variable $i'$: 0ドル \leq i (削除) \leq i' - 1,ドル $ (削除ここまで)i' \leq 9$743 \item For parallel replication (e.g. 0 FOR 10), we introduce a ghost variable $i'$: 0ドル \leq i (追記) < (追記ここまで)i' \leq 9$ 741 744 \item Effectively checking for any two arbitrary (non-equal) indices 742 745 \item Must check both ways! … … 748 751 749 752 \begin{frame} 750 \frametitle{Flattening (削除) (削除ここまで)Expressions}753 \frametitle{Flattening (追記) occam (追記ここまで)Expressions} 751 754 \begin{itemize} 752 755 \item \lstinline|3*(7-i)| becomes 21ドル + (-3)i$ 753 \item Addition is left alone, subtraction is turned into addition (削除) (削除ここまで)756 \item Addition is left alone, subtraction is turned into addition(追記) ($x - y$ becomes $x + (-y)$) (追記ここまで) 754 757 \item Multiplied expressions are multiplied out 755 758 \item Division and modulo are special \ldots … … 763 766 \begin{itemize} 764 767 \item We can't express anything useful about them 765 \end{itemize} 766 \item They often cancel out anyway: 768 \item Can get false alarms (e.g. \lstinline|a[x*x],a[x+1] := 0,0| treated as $y = x+1$) 769 \end{itemize} 770 \item They may cancel out anyway: 767 771 \begin{lstlisting} 768 772 PAR i = 0 FOR 10 … … 1108 1112 \frametitle{Questions?} 1109 1113 \begin{itemize} 1110 %TODO put references here (the Pugh Omega Test paper, occam 2.1 manual, and maybe the Kessler paper 1111 \item Blah 1114 \item TODO put references here (the Pugh Omega Test paper, occam 2.1 manual, and maybe the Kessler paper 1112 1115 \end{itemize} 1113 1116 \end{frame}
Note:
See TracChangeset
for help on using the changeset viewer.