derek/memsys24-slides
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Some updating of slides

Browse Source
This commit is contained in:
Derek Christ
2024-09-28 20:41:34 +02:00
parent bf09b0b6f9
commit ca81974866
3 changed files with 253 additions and 85 deletions
Download Patch File Download Diff File Expand all files Collapse all files

243
images/bare_metal.svg
View File

File diff suppressed because one or more lines are too long
Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 69 KiB

After

Width:  |  Height:  |  Size: 74 KiB

BIN
main.pdf
View File

Binary file not shown.

95
main.tex
View File

@@ -223,7 +223,6 @@
} }
\begin{frame}{Software Library} \begin{frame}{Software Library}
Software support library
\begin{columns} \begin{columns}
\begin{column}{0.5\textwidth} \begin{column}{0.5\textwidth}
\begin{itemize} \begin{itemize}
@@ -235,7 +234,6 @@
\item Execution \item Execution
\begin{itemize} \begin{itemize}
\item Generation of RD and WR requests \item Generation of RD and WR requests
\item Insetion of memory barriers
\end{itemize} \end{itemize}
\end{itemize} \end{itemize}
\end{column} \end{column}
@@ -270,7 +268,7 @@
\begin{frame}{Virtual Prototype Platform} \begin{frame}{Virtual Prototype Platform}
\begin{columns} \begin{columns}
\begin{column}{0.5\textwidth} \begin{column}{0.4\textwidth}
\begin{itemize} \begin{itemize}
\item ARM processor model \item ARM processor model
\item Bare-metal kernel \item Bare-metal kernel
@@ -281,7 +279,7 @@
\end{itemize} \end{itemize}
\end{itemize} \end{itemize}
\end{column} \end{column}
\begin{column}{0.5\textwidth} \begin{column}{0.6\textwidth}
\begin{figure} \begin{figure}
\includesvg[width=0.8\textwidth]{images/bare_metal.svg} \includesvg[width=0.8\textwidth]{images/bare_metal.svg}
\end{figure} \end{figure}
@@ -300,11 +298,11 @@
\begin{column}{0.5\textwidth} \begin{column}{0.5\textwidth}
\begin{itemize} \begin{itemize}
\item Vector benchmarks (BLAS level 1) \item Vector benchmarks (BLAS level 1)
\begin{itemize} \begin{itemize}
\item VADD: $z = x + y$ \item VADD: $z = x + y$
\item VMUL: $z = x \cdot y$ \item VMUL: $z = x \cdot y$
\item HAXPY: $z = a \cdot x + y$ \item HAXPY: $z = a \cdot x + y$
\end{itemize} \end{itemize}
\item Vector-Matrix benchmarks (BLAS level 2) \item Vector-Matrix benchmarks (BLAS level 2)
\begin{itemize} \begin{itemize}
@@ -345,27 +343,27 @@
} }
\begin{frame}{System Configuration} \begin{frame}{System Configuration}
\begin{columns}[t] % \begin{columns}[t]
\begin{column}{0.5\textwidth} % \begin{column}{0.5\textwidth}
Two simulated systems: Two simulated systems:
\begin{itemize} \begin{itemize}
\item Generic ARM system \item Generic ARM system
\item Infinite compute system \item Infinite compute system
\begin{itemize} \begin{itemize}
\item Unrealistic high frequency of 100 GHz \item Infinite CPU clock frequency
\item Completely memory bound \item Completely memory bound
\item Lower bound of possible speedup \item Lower bound of possible PIM speedup
\end{itemize} \end{itemize}
\end{itemize} \end{itemize}
\end{column} % \end{column}
\begin{column}{0.5\textwidth} % \begin{column}{0.5\textwidth}
Two real GPUs using HBM2: % Two real GPUs using HBM2:
\begin{itemize} % \begin{itemize}
\item AMD RX Vega 56 % \item AMD RX Vega 56
\item NVIDIA Tesla V100 % \item NVIDIA Tesla V100
\end{itemize} % \end{itemize}
\end{column} % \end{column}
\end{columns} % \end{columns}
\end{frame} \end{frame}
\begin{frame}{Speedups / Generic ARM System} \begin{frame}{Speedups / Generic ARM System}
@@ -398,24 +396,6 @@
\item GPU has no speculative execution \item GPU has no speculative execution
} }
\begin{frame}{Runtimes / Vector Benchmarks}
\begin{figure}
\includesvg[width=0.8\textwidth]{images/runtimes_vector.svg}
\end{figure}
\end{frame}
\note[itemize]{
\item Real GPUs use multiple memory channels
\item Memory barriers
\item Also architectural differences
}
\begin{frame}{Runtimes / Matrix Benchmarks}
\begin{figure}
\includesvg[width=0.8\textwidth]{images/runtimes_matrix.svg}
\end{figure}
\end{frame}
\begin{frame}{Conclusion and Future Work} \begin{frame}{Conclusion and Future Work}
\textbf{Conclusion} \textbf{Conclusion}
\begin{itemize} \begin{itemize}
@@ -500,9 +480,26 @@
\end{figure} \end{figure}
\end{frame} \end{frame}
\note[itemize]{ \note[itemize]{
\item Data layout in program and address mapping must match \item Data layout in program and address mapping must match
} }
\begin{frame}{Runtimes / Vector Benchmarks}
\begin{figure}
\includesvg[width=0.8\textwidth]{images/runtimes_vector.svg}
\end{figure}
\end{frame}
\note[itemize]{
\item Real GPUs use multiple memory channels
\item Memory barriers
\item Also architectural differences
}
\begin{frame}{Runtimes / Matrix Benchmarks}
\begin{figure}
\includesvg[width=0.8\textwidth]{images/runtimes_matrix.svg}
\end{figure}
\end{frame}
\end{document} \end{document}