phd-defense/slides/20_foundations/main.tex

\section{Foundations}

\begin{frame}{Bird's eye view of a CPU}
    \centering
    \includegraphics[height=0.94\textheight]{cpu_big_picture.svg}
\end{frame}

\begin{frame}{Possible bottlenecks}
    \begin{columns}
        \begin{column}{0.37\textwidth}
            \begin{center}
                \includegraphics[width=\textwidth]{cpu_big_picture_truncate.svg}
            \end{center}
        \end{column}
        \hfill
        \begin{column}{0.62\textwidth}
            \begin{tightitemize}{0pt}
                \begin{itemize}
                    \item \alert{Frontend:} \uops{} not issued fast enough
                        \bigskip

                    \item \alert{Backend:} saturated execution units
                        \bigskip

                    \item \alert{Dependencies:} computation is stalled waiting
                        for previous results
                \end{itemize}
            \end{tightitemize}
        \end{column}
    \end{columns}
\end{frame}

%\begin{frame}{Dependencies and the ROB}
%    \begin{columns}
%        \begin{column}{0.35\textwidth}
%            \begin{center}
%                \includegraphics[width=\textwidth]{cpu_frontend.svg}
%            \end{center}
%        \end{column}
%        \hfill
%        \begin{column}{0.64\textwidth}
%            \begin{tightitemize}{0pt}
%                \begin{itemize}
%                    \item Dependencies can stall execution
%                    \item Maybe instructions further down can be executed right now?
%                \end{itemize}
%                \begin{center}
%                    \textbf{\alert{$\to$ Out-of-Order CPUs}}
%                \end{center}
%                \begin{itemize}
%                    \item ROB: circular buffer of \uops{}
%                    \item First possible instruction is issued
%                \end{itemize}
%            \end{tightitemize}
%        \end{column}
%    \end{columns}
%\end{frame}

%\begin{frame}{How do we get insights from this complex system?}
%    \textbf{Hardware counters}
%    \begin{itemize}
%        \item Built-in hardware, counters gathered at runtime
%        \item Very accurate
%        \item Available data varies from model to model
%        \item May not even be available at all
%    \end{itemize}
%
%    \textbf{Simulation?}
%    \begin{itemize}
%        \item A modern CPU is \alert{$\sim$\,100e9 transistors}: very complex
%            models!
%        \item Very expensive, even for manufacturers for design validation
%        \item CPU design is industrial secret $\leadsto$ not available anyway
%        \item \ldots{}\ie{} not feasible.
%    \end{itemize}
%\end{frame}

\begin{frame}[fragile]{What do we analyze?}
\begin{columns}
\column{\dimexpr\paperwidth-20pt}
    \begin{minipage}[c]{0.60\textwidth}
    Pieces of code referred as \alert{``microkernels''}:

    \begin{itemize}
        \item body of an (assumed) infinite loop;
        \item in steady-state;
        \item L1-resident (memory model is out of scope);
        \item straight-line code (branches assumed not taken).
    \end{itemize}
    \end{minipage}
    \hfill\begin{minipage}[c]{0.35\textwidth}
    \begin{lstlisting}[language={[x86masm]Assembler}, numbers=none]
loop:
  movsd (%rcx, %rax), %xmm0
  mulsd %xmm1, %xmm0
  addsd (%rdx, %rax), %xmm0
  movsd %xmm0, (%rdx, %rax)
  addq $8, %rax
  cmpq $0x2260, %rax
  jne loop\end{lstlisting}
    \end{minipage}
    \vspace{2em}
    \begin{center}
    Reasonable hypotheses for the category of codes worth optimizing this way!
    \end{center}
\end{columns}
\end{frame}

\begin{frame}{Code analyzers}
    \begin{itemize}
        \item Predict performance of a microkernel
        \item Features microarchitectural models
        \item Most often static analyzers
        \item Predict at least the \alert{\emph{reverse-throughput}} of a kernel (cycles per iteration)
        \item May derive further useful metrics, \eg{} bottlenecks, by
            inspecting their model at will
    \end{itemize}
\end{frame}

\begin{frame}{Existing code analyzers}
    \begin{minipage}[t]{0.55\textwidth}
        \begin{block}{Behavioural}
            \medskip
        \begin{itemize}
            \item \alert{\iaca{}}: Intel, proprietary. Intel CPUs only.
            \item \alert{\llvmmca{}}: \texttt{llvm} project, FOSS.
            \item \alert{\uica{}}, \alert{\uopsinfo{}}: academia. Intel CPUs
                only.
        \end{itemize}
        \end{block}
    \end{minipage}
    \hfill
    \begin{minipage}[t]{0.40\textwidth}
        \begin{block}{ML-based}
            \medskip
        \begin{itemize}
            \item \alert{\ithemal{}}: academia.
        \end{itemize}
        \end{block}
    \end{minipage} \\

    \vfill

    Behavioural tools are (to some extent) based on \alert{manually-made}
    models!\\
    \pause{}
    \bigskip{}
    \begin{center}
        \textbf{\alert{Ambition:}} \alert{automated} model generation.
    \end{center}
\end{frame}

\begin{frame}{When I started my PhD\ldots}
    \centering
    %\includegraphics[height=0.9\textheight]{patate_placeholder.jpg}
    \only<1>{\includegraphics[height=0.9\textheight]{sota_potato.svg}}%
    \only<2>{\includegraphics[height=0.9\textheight]{sota_potato_eval.svg}}\\
\end{frame}