Real hardware plots

src/appendix.tex

@@ -3,69 +3,105 @@
 
 \subsection{Simulation Results}
 
-\begin{table}[!ht]
+ TODO !! nochmal aktualisieren!
-\centering
+
-\input{tables/vadd_3GHz}
+\begin{table}[H]
-\caption{Runtime of the VADD benchmark in $\unit{\pico\second}$ on the generic ARM system.}
+\csvreader[
+        head to column names,
+        centered tabularray = {
+            hlines,
+            vlines,
+            hline{3} = {2}{-}{},
+            column{1,2} = {c},
+            column{3} = {r},
+            column{4} = {r},
+            row{1,2} = {c},
+            cell{1}{1} = {r=2}{},
+            cell{1}{2} = {r=2}{},
+            cell{1}{3} = {c=2}{},
+            cell{1}{5} = {c=2}{},
+            cell{3}{1} = {r=4}{},
+            cell{7}{1} = {r=4}{},
+            cell{11}{1} = {r=4}{},
+            cell{15}{1} = {r=4}{},
+            cell{19}{1} = {r=4}{},
+        },
+        table head = {
+            Workload & Level & Runtime [$\unit{\pico\second}$] & \\
+            & & Non-PIM & PIM \\
+         }
+    ]{tables/simulations_3GHz.csv}{}{
+        \csvexpval\workload & \csvexpval\level & \csvexpval\hbm & \csvexpval\pim
+    }
+\caption{Runtimes of the microbenchmarks on the generic ARM-based system for non-\acs*{pim} and \acs*{pim} mode.}
 \end{table}
 
-\begin{table}[!ht]
+\begin{table}[H]
-\centering
+\csvreader[
-\input{tables/vmul_3GHz}
+        head to column names,
-\caption{Runtime of the VMUL benchmark in $\unit{\pico\second}$ on the generic ARM system.}
+        centered tabularray = {
+            hlines,
+            vlines,
+            hline{3} = {2}{-}{},
+            column{1,2} = {c},
+            column{3} = {r},
+            column{4} = {r},
+            row{1,2} = {c},
+            cell{1}{1} = {r=2}{},
+            cell{1}{2} = {r=2}{},
+            cell{1}{3} = {c=2}{},
+            cell{1}{5} = {c=2}{},
+            cell{3}{1} = {r=4}{},
+            cell{7}{1} = {r=4}{},
+            cell{11}{1} = {r=4}{},
+            cell{15}{1} = {r=4}{},
+            cell{19}{1} = {r=4}{},
+        },
+        table head = {
+            Workload & Level & Runtime [$\unit{\pico\second}$] & \\
+            & & Non-PIM & PIM \\
+         }
+    ]{tables/simulations_100GHz.csv}{}{
+        \csvexpval\workload & \csvexpval\level & \csvexpval\hbm & \csvexpval\pim
+    }
+\caption{Runtimes of the microbenchmarks on the infinite compute system for non-\acs*{pim} and \acs*{pim} mode.}
 \end{table}
 
-\begin{table}[!ht]
+\begin{table}[H]
-\centering
+\csvreader[
-\input{tables/haxpy_3GHz}
+        head to column names,
-\caption{Runtime of the \ac{haxpy} benchmark in $\unit{\pico\second}$ on the generic ARM system.}
+        centered tabularray = {
-\end{table}
+            hlines,
-
+            vlines,
-\begin{table}[!ht]
+            hline{3} = {2}{-}{},
-\centering
+            column{1,2} = {c},
-\input{tables/gemv_3GHz}
+            column{3} = {r},
-\caption{Runtime of the \ac{gemv} benchmark in $\unit{\pico\second}$ on the generic ARM system.}
+            column{4} = {r},
-\end{table}
+            row{1,2} = {c},
-
+            cell{1}{1} = {r=2}{},
-\begin{table}[!ht]
+            cell{1}{2} = {r=2}{},
-\centering
+            cell{1}{3} = {c=2}{},
-\input{tables/gemv_layers_3GHz}
+            cell{1}{5} = {c=2}{},
-\caption{Runtime of the \ac{dnn} benchmark in $\unit{\pico\second}$ on the generic ARM system.}
+            cell{3}{1} = {r=4}{},
-\end{table}
+            cell{7}{1} = {r=4}{},
-
+            cell{11}{1} = {r=4}{},
-\begin{table}[!ht]
+            cell{15}{1} = {r=4}{},
-\centering
+            cell{19}{1} = {r=4}{},
-\input{tables/vadd_100GHz}
+        },
-\caption{Runtime of the VADD benchmark in $\unit{\pico\second}$ on the infinite compute system.}
+        table head = {
-\end{table}
+            Workload & Level & Runtime [$\unit{\pico\second}$] & \\
-
+            & & Vega & Tesla \\
-\begin{table}[!ht]
+         }
-\centering
+    ]{tables/torch.csv}{}{
-\input{tables/vmul_100GHz}
+        \csvexpval\workload & \csvexpval\level & \csvexpval\vega & \csvexpval\tesla
-\caption{Runtime of the VMUL benchmark in $\unit{\pico\second}$ on the infinite compute system.}
+    }
-\end{table}
+\caption{Runtimes of the microbenchmarks on the different \acs*{gpu} platforms.}
-
-\begin{table}[!ht]
-\centering
-\input{tables/haxpy_100GHz}
-\caption{Runtime of the \ac{haxpy} benchmark in $\unit{\pico\second}$ on the infinite compute system.}
-\end{table}
-
-\begin{table}[!ht]
-\centering
-\input{tables/gemv_100GHz}
-\caption{Runtime of the \ac{gemv} benchmark in $\unit{\pico\second}$ on the infinite compute system.}
-\end{table}
-
-\begin{table}[!ht]
-\centering
-\input{tables/gemv_layers_100GHz}
-\caption{Runtime of the \ac{dnn} benchmark in $\unit{\pico\second}$ on the infinite compute system.}
 \end{table}
 
+\newpage
 \subsection{Microkernels}
 
-\begin{listing}[!ht]
+\begin{listing}[H]
 \begin{verbatim}
 MOV GRF_A #0, BANK
 MOV GRF_A #1, BANK
@@ -97,7 +133,7 @@ EXIT
 	\label{lst:vadd_bench}
 \end{listing}
 
-\begin{listing}[!ht]
+\begin{listing}[H]
 \begin{verbatim}
 MOV GRF_A #0, BANK
 MOV GRF_A #1, BANK
@@ -129,7 +165,7 @@ EXIT
 	\label{lst:vmul_bench}
 \end{listing}
 
-\begin{listing}[!ht]
+\begin{listing}[H]
 \begin{verbatim}
 MOV SRF_M #0, BANK
 MOV GRF_A #0, BANK
@@ -162,7 +198,7 @@ EXIT
 	\label{lst:haxpy_bench}
 \end{listing}
 
-\begin{listing}[!ht]
+\begin{listing}[H]
 \begin{verbatim}
 MOV GRF_A #0, BANK
 MOV GRF_A #1, BANK
@@ -188,28 +224,33 @@ EXIT
 	\label{lst:gemv_bench}
 \end{listing}
 
+\newpage
 \subsection{Source Code}
 
-\begin{listing}[!ht]
+\begin{listing}[H]
-\begin{minted}{rust}
+\begin{minted}[linenos]{rust}
 pub fn execute<const X16R: usize, const R: usize, const X16C: usize>(
     matrix: &Matrix<X16R, X16C>,
     input_vector: &Vector<X16C>,
     output_partial_sum_vector: &mut SVector<F16x16, R>,
     dummy: &impl PimOperand,
 ) {
+    // Load input vector into GRF-A registers
     for chunk in input_vector.0.iter() {
         chunk.execute_read();
     }
 
+    // Execute the MAC instructions without memory barriers
     for sub_matrix in matrix.0.iter() {
         for column_block in sub_matrix.fixed_rows::<1>(0).iter() {
             column_block.execute_read_async();
         }
     }
 
+    // Verify all memory accesses have finished
     barrier::dsb(barrier::SY);
 
+    // Copy the partial sums into the bank
     for chunk in output_partial_sum_vector
         .fixed_rows_with_step_mut::<X16R>(0, 16)
         .iter_mut()
@@ -217,6 +258,7 @@ pub fn execute<const X16R: usize, const R: usize, const X16C: usize>(
         chunk.execute_write();
     }
 
+    // Execute the EXIT instruction
     dummy.execute_read();
 }
 \end{minted}

src/chapters/results.tex

@@ -99,6 +99,7 @@ The levels X1-X4 denote the increasing dimensions, with each successive level do
 \begin{table}
 \centering
 \begin{tblr}{
+  column{1} = {c},
   cell{2}{2} = {r},
   cell{3}{2} = {r},
   cell{4}{2} = {r},
@@ -168,6 +169,7 @@ Again, several different dimensions of the benchmark inputs are used, whose matr
 \begin{table}
 \centering
 \begin{tblr}{
+  column{1} = {c},
   cell{2}{2} = {r},
   cell{3}{2} = {r},
   cell{4}{2} = {r},
@@ -259,7 +261,7 @@ Since the Samsung \ac{fpga} platform can be assumed to be a highly optimized acc
 \begin{figure}
     \centering
     \includegraphics[width=0.8\linewidth]{plots/samsung}
-    \caption{Relative performance of the \ac{gemv} and ADD microbenchmark for different batch sizes on the hardware implementation of Samsung \cite{lee2021}.}
+    \caption[Relative performance of the \ac{gemv} and ADD microbenchmark for different batch sizes on the hardware implementation of Samsung.]{Relative performance of the \ac{gemv} and ADD microbenchmark for different batch sizes on the hardware implementation of Samsung \cite{lee2021}.}
     \label{fig:samsung_speedup}
 \end{figure}
 
@@ -269,10 +271,29 @@ The \ac{gemv} microbenchmark on the other hand shows a more drastic speedup with
 Although the dimensions used by Samsung are different from the simulations of this thesis, the highest achieved speedup of $\qty{6.1}{\times}$ is well within the reach of the real hardware implementation.
 
 \subsubsection{Comparison to Real Hardware}
+TODO: check all ranges
 
 In addition to the comparison of Samsung's real hardware implementation, the same benchmarks of the performed simulations are run on a [...] with HBM2 [...].
 As this system is using a generic \aca{hbm} \ac{dram} and not \aca{fimdram}, the measurements are only intended to serve as a vague estimation of the runtimes in a non-\ac{pim} case.
 
+\begin{figure}
+    \centering
+    \resizebox{\linewidth}{!}{%
+    \input{plots/runtimes_vector}
+    }
+    \caption{}
+    \label{fig:runtimes_vector}
+\end{figure}
+
+\begin{figure}
+    \centering
+    % \resizebox{\linewidth}{!}{%
+    \input{plots/runtimes_matrix}
+    % }
+    \caption{}
+    \label{fig:runtimes_matrix}
+\end{figure}
+
 % \subsubsection{Initialization Overhead}
 % conversion der operanden im verhältnis zur laufzeit abschätzen
 

src/index.tex

@@ -28,12 +28,15 @@
 \usepackage{makecell}
 \usepackage{minted}
 \usepackage{lscape}
+\usepackage{float}
+\usepackage[l3]{csvsimple}
 
 % Configurations
 \usetikzlibrary{matrix}
 \usetikzlibrary{automata}
 \usetikzlibrary{fit}
 \usetikzlibrary{positioning}
+\usepgfplotslibrary[groupplots]
 
 \addbibresource{doc.bib}
 
@@ -82,13 +85,13 @@
 \setcounter{page}{1}
 
 % Chapters
-\include{chapters/introduction}
+% \include{chapters/introduction}
-\include{chapters/dram}
+% \include{chapters/dram}
-\include{chapters/pim}
+% \include{chapters/pim}
-\include{chapters/vp}
+% \include{chapters/vp}
-\include{chapters/implementation}
+% \include{chapters/implementation}
 \include{chapters/results}
-\include{chapters/conclusion}
+% \include{chapters/conclusion}
 
 % Appendix
 \appendix

src/plots/matrix_infinite.tex

@@ -1,19 +1,16 @@
 \begin{tikzpicture}
-    \pgfplotstableread[col sep=comma]{plots/tables/gemv_100GHz.csv}\gemv
+    \pgfplotstableread[col sep=comma]{plots/speedup_tables/matrix_100GHz.csv}\csv
-    \pgfplotstableread[col sep=comma]{plots/tables/gemv_layers_100GHz.csv}\gemvlayers
     \begin{axis}[
             width=0.9\textwidth,
             ybar=1pt,
             bar width = 15pt,
             ymin=0.1,
-            ymax=100,
+            ymax=10,
-            ymode=log,
-            log origin=infty,
             ymajorgrids,
             ylabel={Relative Performance},
             tick pos=left,
             xtick=data,
-            xticklabels from table={\gemv}{level},
+            xticklabels from table={\csv}{level},
             enlarge x limits=0.25,
             legend style={
                 at={(current bounding box.south-|current axis.south)},
@@ -23,10 +20,10 @@
                 /tikz/every even column/.append style={column sep=0.5cm}
             },
         ]
-        \addplot[fill=_blue!90] table [x expr=\coordindex, y={speedup}]{\gemv};
+        \addplot[fill=_blue!90] table [x expr=\coordindex, y={gemv}]{\csv};
         \addlegendentry{GEMV}
     
-        \addplot[fill=_orange!90] table [x expr=\coordindex, y={speedup}]{\gemvlayers};
+        \addplot[fill=_orange!90] table [x expr=\coordindex, y={dnn}]{\csv};
         \addlegendentry{DNN}
     \end{axis}
 \end{tikzpicture}

src/plots/matrix_normal.tex

@@ -1,20 +1,16 @@
 \begin{tikzpicture}
-    \pgfplotstableread[col sep=comma]{plots/tables/gemv_3GHz.csv}\gemv
+    \pgfplotstableread[col sep=comma]{plots/speedup_tables/matrix_3GHz.csv}\csv
-    \pgfplotstableread[col sep=comma]{plots/tables/gemv_layers_3GHz.csv}\gemvlayers
     \begin{axis}[
             width=0.9\textwidth,
             ybar=1pt,
             bar width = 15pt,
             ymin=0.1,
-            ymax=100,
+            ymax=75,
-            ymode=log,
-            log origin=infty,
-            % minor y tick num = 5,
             ymajorgrids,
             ylabel={Relative Performance},
             tick pos=left,
             xtick=data,
-            xticklabels from table={\gemv}{level},
+            xticklabels from table={\csv}{level},
             enlarge x limits=0.25,
             legend style={
                 at={(current bounding box.south-|current axis.south)},
@@ -24,10 +20,10 @@
                 /tikz/every even column/.append style={column sep=0.5cm}
             },
         ]
-        \addplot[fill=_blue!90] table [x expr=\coordindex, y={speedup}]{\gemv};
+        \addplot[fill=_blue!90] table [x expr=\coordindex, y={gemv}]{\csv};
         \addlegendentry{GEMV}
     
-        \addplot[fill=_orange!90] table [x expr=\coordindex, y={speedup}]{\gemvlayers};
+        \addplot[fill=_orange!90] table [x expr=\coordindex, y={dnn}]{\csv};
         \addlegendentry{DNN}
     \end{axis}
 \end{tikzpicture}

src/plots/runtime_tables/hbm_100GHz.csv

@@ -0,0 +1,5 @@
+level,vadd,vmul,haxpy,gemv,dnn
+X1,11768899990,9758441990,16772264000,7916400980,176584310
+X2,23071196990,18975123990,33462372990,15800020980,1860990350
+X3,46873992980,37109877990,65993469990,62587326980,7063630630
+X4,91264808000,74066441980,134134323970,127514526980,27344749530

src/plots/runtime_tables/hbm_3GHz.csv

@@ -0,0 +1,5 @@
+level,vadd,vmul,haxpy,gemv,dnn
+X1,21334729581,20112009858,36613117899,72834815610,5504078079
+X2,43268334354,39990439863,73515923487,151235040573,21478024143
+X3,85485583179,80576580096,146061622170,614362203486,85912073262
+X4,166942414809,163020260223,294729236073,1228045754304,322188095061

src/plots/runtime_tables/pim_100GHz.csv

@@ -0,0 +1,5 @@
+level,vadd,vmul,haxpy,gemv,dnn
+X1,911446480,911416480,954454480,951904860,536177760
+X2,1822806480,1822776480,1908822480,1814530860,738329760
+X3,3645526480,3645496480,3817558480,6990944860,1547139760
+X4,7290966480,7290936480,7635030480,13892610860,4782339760

src/plots/runtime_tables/pim_3GHz.csv

@@ -0,0 +1,5 @@
+level,vadd,vmul,haxpy,gemv,dnn
+X1,1475510346,1475512344,1543044078,1377734886,933823908
+X2,2950962084,2950964082,3085992252,2601142920,1220409702
+X3,5901852240,5901848244,6171893928,9942655392,2367353610
+X4,11803639878,11803641876,12343693950,19731271644,6955629408

src/plots/runtime_tables/tesla.csv

@@ -0,0 +1,5 @@
+level,vadd,vmul,haxpy,gemv,dnn
+X1,69572650,67408281,69791189,750246152,231093065
+X2,123217536,103994272,123543145,648714601,431703456
+X3,207693503,182162140,207947543,2454455479,877622611
+X4,378089165,350280326,377434890,4968984949,2175751385

src/plots/runtime_tables/vega.csv

@@ -0,0 +1,5 @@
+level,vadd,vmul,haxpy,gemv,dnn
+X1,69572650,67408281,69791189,750246152,231093065
+X2,123217536,103994272,123543145,648714601,431703456
+X3,207693503,182162140,207947543,2454455479,877622611
+X4,378089165,350280326,377434890,4968984949,2175751385

src/plots/runtimes_matrix.tex

@@ -0,0 +1,74 @@
+\begin{tikzpicture}
+\pgfplotstableread[col sep=comma]{plots/runtime_tables/hbm_3GHz.csv}\hbmarm
+\pgfplotstableread[col sep=comma]{plots/runtime_tables/pim_3GHz.csv}\hbmpim
+\pgfplotstableread[col sep=comma]{plots/runtime_tables/hbm_100GHz.csv}\hbminf
+\pgfplotstableread[col sep=comma]{plots/runtime_tables/pim_100GHz.csv}\piminf
+\pgfplotstableread[col sep=comma]{plots/runtime_tables/vega.csv}\vega
+\pgfplotstableread[col sep=comma]{plots/runtime_tables/tesla.csv}\tesla
+\begin{groupplot}[
+    group style={
+        group size=2 by 1,
+        horizontal sep=0pt,
+        vertical sep=0pt,
+        xticklabels at=edge bottom,
+        yticklabels at=edge left,
+        xlabels at=edge bottom,
+        ylabels at=edge left,
+    },
+    height=8cm,
+    width=0.45\linewidth,
+    ybar=1pt,
+    axis line style={draw=none},
+    tick style={draw=none},
+    ymin=0.1,
+    ymax=1e13,
+    ymode=log,
+    % ymax=25,
+    ymajorgrids,
+    ylabel={Runtime [$\unit{\pico\second}$]},
+    tick pos=left,
+    xtick=data,
+    xticklabels from table={\vega}{level},
+    legend style={
+        at={(1.0,-0.15)},
+        anchor=north,
+        legend columns=-1,
+        draw=none,
+        /tikz/every even column/.append style={column sep=0.5cm}
+    },
+    enlarge x limits=0.2,
+]
+    \nextgroupplot[
+        xlabel=GEMV,
+        bar width=3pt,
+        % width=3cm,
+    ]
+    
+    \addplot[fill=_blue!90]     table [x expr=\coordindex, y={gemv}]{\hbmarm};
+    \addlegendentry{Non-PIM ARM}
+    \addplot[fill=_orange!90]   table [x expr=\coordindex, y={gemv}]{\hbmpim};
+    \addlegendentry{PIM ARM}
+    \addplot[fill=_yellow!90]   table [x expr=\coordindex, y={gemv}]{\hbminf};
+    \addlegendentry{Non-PIM Inf}
+    \addplot[fill=_green!90]    table [x expr=\coordindex, y={gemv}]{\piminf};
+    \addlegendentry{PIM Inf}
+    \addplot[fill=_darkblue!90] table [x expr=\coordindex, y={gemv}]{\vega};
+    \addlegendentry{Vega}
+    \addplot[fill=violet!90]    table [x expr=\coordindex, y={gemv}]{\tesla};
+    \addlegendentry{Tesla}
+    
+    \nextgroupplot[
+        xlabel=DNN,
+        bar width=3pt,
+        % width=3cm,
+    ]
+
+    \addplot[fill=_blue!90]     table [x expr=\coordindex, y={dnn}]{\hbmarm};
+    \addplot[fill=_orange!90]   table [x expr=\coordindex, y={dnn}]{\hbmpim};
+    \addplot[fill=_yellow!90]   table [x expr=\coordindex, y={dnn}]{\hbminf};
+    \addplot[fill=_green!90]    table [x expr=\coordindex, y={dnn}]{\piminf};
+    \addplot[fill=_darkblue!90] table [x expr=\coordindex, y={dnn}]{\vega};
+    \addplot[fill=violet!90]    table [x expr=\coordindex, y={dnn}]{\tesla};
+    
+\end{groupplot}
+\end{tikzpicture}

src/plots/runtimes_vector.tex

@@ -0,0 +1,87 @@
+\begin{tikzpicture}
+\pgfplotstableread[col sep=comma]{plots/runtime_tables/hbm_3GHz.csv}\hbmarm
+\pgfplotstableread[col sep=comma]{plots/runtime_tables/pim_3GHz.csv}\hbmpim
+\pgfplotstableread[col sep=comma]{plots/runtime_tables/hbm_100GHz.csv}\hbminf
+\pgfplotstableread[col sep=comma]{plots/runtime_tables/pim_100GHz.csv}\piminf
+\pgfplotstableread[col sep=comma]{plots/runtime_tables/vega.csv}\vega
+\pgfplotstableread[col sep=comma]{plots/runtime_tables/tesla.csv}\tesla
+\begin{groupplot}[
+    group style={
+        group size=3 by 1,
+        horizontal sep=0pt,
+        vertical sep=0pt,
+        xticklabels at=edge bottom,
+        yticklabels at=edge left,
+        xlabels at=edge bottom,
+        ylabels at=edge left,
+    },
+    height=8cm,
+    width=0.45\linewidth,
+    ybar=1pt,
+    axis line style={draw=none},
+    tick style={draw=none},
+    ymin=0.1,
+    ymax=1e12,
+    ymode=log,
+    % ymax=25,
+    ymajorgrids,
+    ylabel={Runtime [$\unit{\pico\second}$]},
+    tick pos=left,
+    xtick=data,
+    xticklabels from table={\vega}{level},
+    legend style={
+        at={(0.5,-0.15)},
+        anchor=north,
+        legend columns=-1,
+        draw=none,
+        /tikz/every even column/.append style={column sep=0.5cm}
+    },
+    enlarge x limits=0.2,
+]
+    \nextgroupplot[
+        xlabel=VADD,
+        bar width=3pt,
+        % width=3cm,
+    ]
+    
+    \addplot[fill=_blue!90]     table [x expr=\coordindex, y={vadd}]{\hbmarm};
+    \addplot[fill=_orange!90]   table [x expr=\coordindex, y={vadd}]{\hbmpim};
+    \addplot[fill=_yellow!90]   table [x expr=\coordindex, y={vadd}]{\hbminf};
+    \addplot[fill=_green!90]    table [x expr=\coordindex, y={vadd}]{\piminf};
+    \addplot[fill=_darkblue!90] table [x expr=\coordindex, y={vadd}]{\vega};
+    \addplot[fill=violet!90]    table [x expr=\coordindex, y={vadd}]{\tesla};
+    
+    \nextgroupplot[
+        xlabel=VMUL,
+        bar width=3pt,
+        % width=3cm,
+    ]
+
+    \addplot[fill=_blue!90]     table [x expr=\coordindex, y={vmul}]{\hbmarm};
+    \addlegendentry{Non-PIM ARM}
+    \addplot[fill=_orange!90]   table [x expr=\coordindex, y={vmul}]{\hbmpim};
+    \addlegendentry{PIM ARM}
+    \addplot[fill=_yellow!90]   table [x expr=\coordindex, y={vmul}]{\hbminf};
+    \addlegendentry{Non-PIM Inf}
+    \addplot[fill=_green!90]    table [x expr=\coordindex, y={vmul}]{\piminf};
+    \addlegendentry{PIM Inf}
+    \addplot[fill=_darkblue!90] table [x expr=\coordindex, y={vmul}]{\vega};
+    \addlegendentry{Vega}
+    \addplot[fill=violet!90]    table [x expr=\coordindex, y={vmul}]{\tesla};
+    \addlegendentry{Tesla}
+    
+    \nextgroupplot[
+        xlabel=HAXPY,
+        bar width=3pt,
+        % width=3cm,
+    ]
+    
+    \addplot[fill=_blue!90]     table [x expr=\coordindex, y={haxpy}]{\hbmarm};
+    \addplot[fill=_orange!90]   table [x expr=\coordindex, y={haxpy}]{\hbmpim};
+    \addplot[fill=_yellow!90]   table [x expr=\coordindex, y={haxpy}]{\hbminf};
+    \addplot[fill=_green!90]    table [x expr=\coordindex, y={haxpy}]{\piminf};
+    \addplot[fill=_darkblue!90] table [x expr=\coordindex, y={haxpy}]{\vega};
+    \addplot[fill=violet!90]    table [x expr=\coordindex, y={haxpy}]{\tesla};
+
+\end{groupplot}
+\end{tikzpicture}

src/plots/speedup_tables/matrix_100GHz.csv

@@ -0,0 +1,5 @@
+level,gemv,dnn
+X1,8.316378361593825,0.3293391169376365
+X2,8.707496426927674,2.520540889480061
+X3,8.952627753954278,4.565606038073768
+X4,9.178586247394538,5.717860064798073

src/plots/speedup_tables/matrix_3GHz.csv

@@ -0,0 +1,5 @@
+level,gemv,dnn
+X1,52.86562483836241,5.894128466670185
+X2,58.14176507187079,17.599027693570402
+X3,61.79055586904123,36.290342473171975
+X4,62.23855088820042,46.32048031346181

src/plots/speedup_tables/vector_100GHz.csv

@@ -0,0 +1,5 @@
+level,vadd,vmul,haxpy
+X1,12.912332482758615,10.706896577073085,17.57261802574388
+X2,12.656964545133722,10.410011429377231,17.530374532261376
+X3,12.857948841452387,10.179649930700249,17.28682620992881
+X4,12.517518527941442,10.158700762676236,17.568276160961705

src/plots/speedup_tables/vector_3GHz.csv

@@ -0,0 +1,5 @@
+level,vadd,vmul,haxpy
+X1,14.459220593631812,13.63052633194372,23.727849658355645
+X2,14.66245011706494,13.551652528382078,23.822458866951166
+X3,14.484534634672588,13.652770583844921,23.665607976080565
+X4,14.143299569834605,13.81101374775393,23.87690729103017

src/plots/tables/gemv_100GHz.csv

@@ -1,5 +0,0 @@
-workload,level,frequency,speedup
-gemv,X1,100GHz,0.2108059965502083
-gemv,X2,100GHz,0.40509080127411157
-gemv,X3,100GHz,0.8462958338758609
-gemv,X4,100GHz,4.7274497979448125

src/plots/tables/gemv_3GHz.csv

@@ -1,5 +0,0 @@
-workload,level,frequency,speedup
-gemv,X1,3GHz,3.468782996825547
-gemv,X2,3GHz,6.723879985176877
-gemv,X3,3GHz,12.744110856471028
-gemv,X4,3GHz,23.645526777997713

src/plots/tables/gemv_layers_100GHz.csv

@@ -1,5 +0,0 @@
-workload,level,frequency,speedup
-gemv_layers,X1,100GHz,0.17890250001597863
-gemv_layers,X2,100GHz,0.6097840333112959
-gemv_layers,X3,100GHz,3.9637284525723304
-gemv_layers,X4,100GHz,6.088778065749799

src/plots/tables/gemv_layers_3GHz.csv

@@ -1,5 +0,0 @@
-workload,level,frequency,speedup
-gemv_layers,X1,3GHz,2.992752194063702
-gemv_layers,X2,3GHz,11.246371082010572
-gemv_layers,X3,3GHz,34.94598413478715
-gemv_layers,X4,3GHz,72.33604077371677

src/plots/tables/haxpy_100GHz.csv

@@ -1,5 +0,0 @@
-workload,level,frequency,speedup
-haxpy,X1,100GHz,2.0481358611246403
-haxpy,X2,100GHz,2.3234133539462776
-haxpy,X3,100GHz,2.272582592673281
-haxpy,X4,100GHz,2.3895030032424387

src/plots/tables/haxpy_3GHz.csv

@@ -1,5 +0,0 @@
-workload,level,frequency,speedup
-haxpy,X1,3GHz,19.816741597088416
-haxpy,X2,3GHz,25.395400082633245
-haxpy,X3,3GHz,28.676005064893953
-haxpy,X4,3GHz,31.783592582828017

src/plots/tables/vadd_100GHz.csv

@@ -1,5 +0,0 @@
-workload,level,frequency,speedup
-vadd,X1,100GHz,2.398047786583271
-vadd,X2,100GHz,1.823243660465808
-vadd,X3,100GHz,1.562017010059411
-vadd,X4,100GHz,1.7888939829610704

src/plots/tables/vadd_3GHz.csv

@@ -1,5 +0,0 @@
-workload,level,frequency,speedup
-vadd,X1,3GHz,12.766775777414075
-vadd,X2,3GHz,14.19338061465721
-vadd,X3,3GHz,15.313227057302887
-vadd,X4,3GHz,16.430379164365913

src/plots/tables/vmul_100GHz.csv

@@ -1,5 +0,0 @@
-workload,level,frequency,speedup
-vmul,X1,100GHz,2.4019901321627604
-vmul,X2,100GHz,2.2189241847884267
-vmul,X3,100GHz,1.86705278821741
-vmul,X4,100GHz,1.7484391189395834

src/plots/tables/vmul_3GHz.csv

@@ -1,5 +0,0 @@
-workload,level,frequency,speedup
-vmul,X1,3GHz,14.157521157521158
-vmul,X2,3GHz,15.915413533834586
-vmul,X3,3GHz,16.959713823354058
-vmul,X4,3GHz,18.215465292791755

src/plots/vector_infinite.tex

@@ -1,20 +1,16 @@
 \begin{tikzpicture}
-    \pgfplotstableread[col sep=comma]{plots/tables/vadd_100GHz.csv}\vadd
+    \pgfplotstableread[col sep=comma]{plots/speedup_tables/vector_100GHz.csv}\csv
-    \pgfplotstableread[col sep=comma]{plots/tables/vmul_100GHz.csv}\vmul
-    \pgfplotstableread[col sep=comma]{plots/tables/haxpy_100GHz.csv}\haxpy
     \begin{axis}[
             width=0.8\textwidth,
             ybar=1pt,
             bar width = 15pt,
             ymin=0,
-            ymax=5,
+            ymax=25,
-            % ymode=log,
-            % log origin=infty,
             ymajorgrids,
             ylabel={Relative Performance},
             tick pos=left,
             xtick=data,
-            xticklabels from table={\vadd}{level},
+            xticklabels from table={\csv}{level},
             enlarge x limits=0.25,
             legend style={
                 at={(current bounding box.south-|current axis.south)},
@@ -24,13 +20,13 @@
                 /tikz/every even column/.append style={column sep=0.5cm}
             },
         ]
-        \addplot[fill=_blue!90] table [x expr=\coordindex, y={speedup}]{\vadd};
+        \addplot[fill=_blue!90] table [x expr=\coordindex, y={vadd}]{\csv};
         \addlegendentry{VADD}
     
-        \addplot[fill=_orange!90] table [x expr=\coordindex, y={speedup}]{\vmul};
+        \addplot[fill=_orange!90] table [x expr=\coordindex, y={vmul}]{\csv};
         \addlegendentry{VMUL}
     
-        \addplot[fill=_yellow!90] table [x expr=\coordindex, y={speedup}]{\haxpy};
+        \addplot[fill=_yellow!90] table [x expr=\coordindex, y={haxpy}]{\csv};
         \addlegendentry{HAXPY}
     \end{axis}
 \end{tikzpicture}

src/plots/vector_normal.tex

@@ -1,21 +1,16 @@
 \begin{tikzpicture}
-    \pgfplotstableread[col sep=comma]{plots/tables/vadd_3GHz.csv}\vadd
+    \pgfplotstableread[col sep=comma]{plots/speedup_tables/vector_3GHz.csv}\csv
-    \pgfplotstableread[col sep=comma]{plots/tables/vmul_3GHz.csv}\vmul
-    \pgfplotstableread[col sep=comma]{plots/tables/haxpy_3GHz.csv}\haxpy
     \begin{axis}[
             width=0.8\textwidth,
             ybar=1pt,
             bar width = 15pt,
             ymin=0,
-            ymax=35,
+            ymax=25,
-            % ymode=log,
-            % log origin=infty,
-            % minor y tick num = 5,
             ymajorgrids,
             ylabel={Relative Performance},
             tick pos=left,
             xtick=data,
-            xticklabels from table={\vadd}{level},
+            xticklabels from table={\csv}{level},
             enlarge x limits=0.25,
             legend style={
                 at={(current bounding box.south-|current axis.south)},
@@ -25,13 +20,13 @@
                 /tikz/every even column/.append style={column sep=0.5cm}
             },
         ]
-        \addplot[fill=_blue!90] table [x expr=\coordindex, y={speedup}]{\vadd};
+        \addplot[fill=_blue!90] table [x expr=\coordindex, y={vadd}]{\csv};
         \addlegendentry{VADD}
     
-        \addplot[fill=_orange!90] table [x expr=\coordindex, y={speedup}]{\vmul};
+        \addplot[fill=_orange!90] table [x expr=\coordindex, y={vmul}]{\csv};
         \addlegendentry{VMUL}
     
-        \addplot[fill=_yellow!90] table [x expr=\coordindex, y={speedup}]{\haxpy};
+        \addplot[fill=_yellow!90] table [x expr=\coordindex, y={haxpy}]{\csv};
         \addlegendentry{HAXPY}
     \end{axis}
 \end{tikzpicture}

src/tables/gemv_100GHz.tex

@@ -1,22 +0,0 @@
-
-\begin{tblr}{
-  hlines,
-  vlines,
-  cell{2}{2} = {r},
-  cell{2}{3} = {r},
-  cell{3}{2} = {r},
-  cell{3}{3} = {r},
-  cell{4}{2} = {r},
-  cell{4}{3} = {r},
-  cell{5}{2} = {r},
-  cell{5}{3} = {r},
-  hline{2} = {-}{solid,black},
-  hline{2} = {2}{-}{solid,black},
-}
-Level & Non-\ac{pim} & \ac{pim} \\
-X1&96465760&457604440 \\
-X2&192178090&474407440 \\
-X3&430015980&508115440 \\
-X4&2720535980&575476440 \\
-
-\end{tblr}

src/tables/gemv_3GHz.tex

@@ -1,22 +0,0 @@
-
-\begin{tblr}{
-  hlines,
-  vlines,
-  cell{2}{2} = {r},
-  cell{2}{3} = {r},
-  cell{3}{2} = {r},
-  cell{3}{3} = {r},
-  cell{4}{2} = {r},
-  cell{4}{3} = {r},
-  cell{5}{2} = {r},
-  cell{5}{3} = {r},
-  hline{2} = {-}{solid,black},
-  hline{2} = {2}{-}{solid,black},
-}
-Level & Non-\ac{pim} & \ac{pim} \\
-X1&2734886043&788428116 \\
-X2&5462015184&812330856 \\
-X3&10958710653&859903902 \\
-X4&22594486896&955550160 \\
-
-\end{tblr}

src/tables/gemv_layers_100GHz.tex

@@ -1,22 +0,0 @@
-
-\begin{tblr}{
-  hlines,
-  vlines,
-  cell{2}{2} = {r},
-  cell{2}{3} = {r},
-  cell{3}{2} = {r},
-  cell{3}{3} = {r},
-  cell{4}{2} = {r},
-  cell{4}{3} = {r},
-  cell{5}{2} = {r},
-  cell{5}{3} = {r},
-  hline{2} = {-}{solid,black},
-  hline{2} = {2}{-}{solid,black},
-}
-Level & Non-\ac{pim} & \ac{pim} \\
-X1&8957090&50066880 \\
-X2&33647200&55178880 \\
-X3&299035090&75442880 \\
-X4&951182090&156218880 \\
-
-\end{tblr}

src/tables/gemv_layers_3GHz.tex

@@ -1,22 +0,0 @@
-
-\begin{tblr}{
-  hlines,
-  vlines,
-  cell{2}{2} = {r},
-  cell{2}{3} = {r},
-  cell{3}{2} = {r},
-  cell{3}{3} = {r},
-  cell{4}{2} = {r},
-  cell{4}{3} = {r},
-  cell{5}{2} = {r},
-  cell{5}{3} = {r},
-  hline{2} = {-}{solid,black},
-  hline{2} = {2}{-}{solid,black},
-}
-Level & Non-\ac{pim} & \ac{pim} \\
-X1&262765971&87800778 \\
-X2&1070441487&95181057 \\
-X3&4332005991&123962913 \\
-X4&17236314765&238281147 \\
-
-\end{tblr}

src/tables/haxpy_100GHz.tex

@@ -1,22 +0,0 @@
-
-\begin{tblr}{
-  hlines,
-  vlines,
-  cell{2}{2} = {r},
-  cell{2}{3} = {r},
-  cell{3}{2} = {r},
-  cell{3}{3} = {r},
-  cell{4}{2} = {r},
-  cell{4}{3} = {r},
-  cell{5}{2} = {r},
-  cell{5}{3} = {r},
-  hline{2} = {-}{solid,black},
-  hline{2} = {2}{-}{solid,black},
-}
-Level & Non-\ac{pim} & \ac{pim} \\
-X1&2374720&1088170 \\
-X2&3677220&1604170 \\
-X3&5875080&2708170 \\
-X4&11640050&4703250 \\
-
-\end{tblr}

src/tables/haxpy_3GHz.tex

@@ -1,22 +0,0 @@
-
-\begin{tblr}{
-  hlines,
-  vlines,
-  cell{2}{2} = {r},
-  cell{2}{3} = {r},
-  cell{3}{2} = {r},
-  cell{3}{3} = {r},
-  cell{4}{2} = {r},
-  cell{4}{3} = {r},
-  cell{5}{2} = {r},
-  cell{5}{3} = {r},
-  hline{2} = {-}{solid,black},
-  hline{2} = {2}{-}{solid,black},
-}
-Level & Non-\ac{pim} & \ac{pim} \\
-X1&30413223&1555443 \\
-X2&59499108&2417913 \\
-X3&117506376&4207788 \\
-X4&234943821&7578414 \\
-
-\end{tblr}

src/tables/simulations_100GHz.csv

@@ -0,0 +1,21 @@
+workload,level,hbm,pim
+VADD,X1,11768899990,455723240
+VADD,X2,23071196990,911403240
+VADD,X3,46873992980,1822763240
+VADD,X4,91264808000,3645483240
+VMUL,X1,9758441990,455708240
+VMUL,X2,18975123990,911388240
+VMUL,X3,37109877990,1822748240
+VMUL,X4,74066441980,3645468240
+HAXPY,X1,16772264000,477227240
+HAXPY,X2,33462372990,954411240
+HAXPY,X3,65993469990,1908779240
+HAXPY,X4,134134323970,3817515240
+GEMV,X1,7916400980,475952430
+GEMV,X2,15800020980,907265430
+GEMV,X3,62587326980,3495472430
+GEMV,X4,127514526980,6946305430
+DNN,X1,176584310,268088880
+DNN,X2,1860990350,369164880
+DNN,X3,7063630630,773569880
+DNN,X4,27344749530,2391169880

src/tables/simulations_3GHz.csv

@@ -0,0 +1,21 @@
+workload,level,hbm,pim
+VADD,X1,21334729581,737755173
+VADD,X2,43268334354,1475481042
+VADD,X3,85485583179,2950926120
+VADD,X4,166942414809,5901819939
+VMUL,X1,20112009858,737756172
+VMUL,X2,39990439863,1475482041
+VMUL,X3,80576580096,2950924122
+VMUL,X4,163020260223,5901820938
+HAXPY,X1,36613117899,771522039
+HAXPY,X2,73515923487,1542996126
+HAXPY,X3,146061622170,3085946964
+HAXPY,X4,294729236073,6171846975
+GEMV,X1,72834815610,688867443
+GEMV,X2,151235040573,1300571460
+GEMV,X3,614362203486,4971327696
+GEMV,X4,1228045754304,9865635822
+DNN,X1,5504078079,466911954
+DNN,X2,21478024143,610204851
+DNN,X3,85912073262,1183676805
+DNN,X4,322188095061,3477814704

src/tables/torch.csv

@@ -0,0 +1,21 @@
+workload,level,vega,tesla
+VADD,X1,69572650,TODO
+VADD,X2,123217536,TODO
+VADD,X3,207693503,TODO
+VADD,X4,378089165,TODO
+VMUL,X1,67408281,TODO
+VMUL,X2,103994272,TODO
+VMUL,X3,182162140,TODO
+VMUL,X4,350280326,TODO
+HAXPY,X1,69791189,TODO
+HAXPY,X2,123543145,TODO
+HAXPY,X3,207947543,TODO
+HAXPY,X4,377434890,TODO
+GEMV,X1,750246152,TODO
+GEMV,X2,648714601,TODO
+GEMV,X3,2454455479,TODO
+GEMV,X4,4968984949,TODO
+DNN,X1,231093065,TODO
+DNN,X2,431703456,TODO
+DNN,X3,877622611,TODO
+DNN,X4,2175751385,TODO

src/tables/vadd_100GHz.tex

@@ -1,22 +0,0 @@
-
-\begin{tblr}{
-  hlines,
-  vlines,
-  cell{2}{2} = {r},
-  cell{2}{3} = {r},
-  cell{3}{2} = {r},
-  cell{3}{3} = {r},
-  cell{4}{2} = {r},
-  cell{4}{3} = {r},
-  cell{5}{2} = {r},
-  cell{5}{3} = {r},
-  hline{2} = {-}{solid,black},
-  hline{2} = {2}{-}{solid,black},
-}
-Level & Non-\ac{pim} & \ac{pim} \\
-X1&2024360&844170 \\
-X2&2574730&1412170 \\
-X3&3899060&2496170 \\
-X4&8037930&4493240 \\
-
-\end{tblr}

src/tables/vadd_3GHz.tex

@@ -1,22 +0,0 @@
-
-\begin{tblr}{
-  hlines,
-  vlines,
-  cell{2}{2} = {r},
-  cell{2}{3} = {r},
-  cell{3}{2} = {r},
-  cell{3}{3} = {r},
-  cell{4}{2} = {r},
-  cell{4}{3} = {r},
-  cell{5}{2} = {r},
-  cell{5}{3} = {r},
-  hline{2} = {-}{solid,black},
-  hline{2} = {2}{-}{solid,black},
-}
-Level & Non-\ac{pim} & \ac{pim} \\
-X1&15585399&1220778 \\
-X2&29988981&2112885 \\
-X3&59177430&3864465 \\
-X4&118902645&7236756 \\
-
-\end{tblr}

src/tables/vmul_100GHz.tex

@@ -1,22 +0,0 @@
-
-\begin{tblr}{
-  hlines,
-  vlines,
-  cell{2}{2} = {r},
-  cell{2}{3} = {r},
-  cell{3}{2} = {r},
-  cell{3}{3} = {r},
-  cell{4}{2} = {r},
-  cell{4}{3} = {r},
-  cell{5}{2} = {r},
-  cell{5}{3} = {r},
-  hline{2} = {-}{solid,black},
-  hline{2} = {2}{-}{solid,black},
-}
-Level & Non-\ac{pim} & \ac{pim} \\
-X1&2030090&845170 \\
-X2&3109090&1401170 \\
-X3&4654880&2493170 \\
-X4&7829930&4478240 \\
-
-\end{tblr}

src/tables/vmul_3GHz.tex

@@ -1,22 +0,0 @@
-
-\begin{tblr}{
-  hlines,
-  vlines,
-  cell{2}{2} = {r},
-  cell{2}{3} = {r},
-  cell{3}{2} = {r},
-  cell{3}{3} = {r},
-  cell{4}{2} = {r},
-  cell{4}{3} = {r},
-  cell{5}{2} = {r},
-  cell{5}{3} = {r},
-  hline{2} = {-}{solid,black},
-  hline{2} = {2}{-}{solid,black},
-}
-Level & Non-\ac{pim} & \ac{pim} \\
-X1&17269047&1219779 \\
-X2&33834132&2125872 \\
-X3&66308292&3909753 \\
-X4&131863338&7239087 \\
-
-\end{tblr}