fosdem2024_bigint: start adding multiplication section

[libreriscv.git] / conferences / fosdem2024 / fosdem2024_bigint / fosdem2024_bigint.tex

diff --git a/conferences/fosdem2024/fosdem2024_bigint/fosdem2024_bigint.tex b/conferences/fosdem2024/fosdem2024_bigint/fosdem2024_bigint.tex
index a484c63e1b1bc23b97bdf300128d72df5f347385..79d1fd380d761a72788e3369490c8a7435c84813 100644 (file)
--- a/conferences/fosdem2024/fosdem2024_bigint/fosdem2024_bigint.tex
+++ b/conferences/fosdem2024/fosdem2024_bigint/fosdem2024_bigint.tex
@@ -5,6 +5,11 @@
 \usetheme{default}
 \usepackage[english]{babel}
 \usepackage{tikz}
+\usepackage{minted}
+\usemintedstyle{monokai}
+\definecolor{codebg}{rgb}{0.1,0.09,0.08}
+\newminted[codeenv]{python3}{escapeinside=@@,fontsize=\small,bgcolor=codebg}
+\newmintinline[codeinline]{python3}{escapeinside=@@,fontsize=\small,bgcolor=codebg}
 
 \title[Fast Big-Integer Arithmetic on SVP64 ...]{
     Fast Big-Integer Arithmetic on SVP64 at up to 256-bits/cycle and beyond
@@ -31,18 +36,86 @@
         \pause \\
         \medskip
         Simple Example:
-        \begin{semiverbatim}
+        \begin{codeenv}
 setvl 0, 0, 3, 0, 1, 1 # makes stuff run 3 times
 sv.add *r3, *r15, r12  # adds 3 times
-\pause
+@\pause@
 # expands to:
-add r3, r15, r12  \only<+(1)->{# no * means r12 doesn't increment}
-add r4, r16, r12  \only<+(1)->{# * means r3 and r15 increment}
+add r3, r15, r12  # no * means r12 doesn't increment
+add r4, r16, r12  # * means r3 and r15 increment
 add r5, r17, r12
-        \end{semiverbatim}
+        \end{codeenv}
     \end{itemize}
 \end{frame}
 
+\begin{frame}[fragile]
+    \frametitle{Big-Integer Addition on SVP64}
+    How can we use SVP64 to add 256-bit integers?
+    \pause
+    \begin{codeenv}
+setvl 0, 0, 4, 0, 1, 1 # makes stuff run 4 times
+addic r0, r0, 0  # clear CA (carry flag)
+sv.adde *r4, *r4, *r8  # carry-propagating add
+@\pause@
+# expands to:
+addic r0, r0, 0  # clear CA (carry flag)
+adde r4, r4, r8
+adde r5, r5, r9
+adde r6, r6, r10
+adde r7, r7, r11
+    \end{codeenv}
+\end{frame}
+
+\begin{frame}
+    \frametitle{Big-Integer Addition on an example CPU}
+    Disclaimer:
+    SVP64 is designed for everything from tiny to big and fast CPUs, this example only shows a hypothetical big and fast CPU design
+\end{frame}
+
+\begin{frame}
+    \frametitle{Big-Integer Addition on an example CPU}
+    \input{bigint-add-pipe.dia-tex}
+\end{frame}
+
+\begin{frame}[fragile]
+    \frametitle{Big-Integer Multiply on SVP64}
+    How can we use SVP64 to Multiply a 64-bit by a 256-bit integer?
+    \pause
+    \begin{itemize}
+        \item new instruction: \codeinline{maddedu RT, RA, RB, RC}
+        \pause
+        \item $64 \times 64 + 64 \rightarrow 128$-bit Multiply-Add
+        \pause
+        \item Semantics as used in this presentation (somewhat simplified):
+        \begin{codeenv}
+result = (RA * RB) + RC
+RT = LSB_HALF(result)
+RC = MSB_HALF(result)
+        \end{codeenv}
+    \end{itemize}
+\end{frame}
+
+\begin{frame}[fragile]
+    \frametitle{Big-Integer Multiply on SVP64}
+    How can we use SVP64 to Multiply a 64-bit by a 256-bit integer?
+    \pause
+    \begin{codeenv}
+# 64-bit input in r3
+# 256-bit input in r20-23
+# 320-bit output in r4-8
+setvl 0, 0, 4, 0, 1, 1 # makes stuff run 4 times
+li r8, 0  # clear carry register
+sv.maddedu *r4, r3, *r20, r8  # carrying multiply
+@\pause@
+# expands to:
+li r8, 0
+maddedu r4, r3, r20, r8
+maddedu r5, r3, r21, r8
+maddedu r6, r3, r22, r8
+maddedu r7, r3, r23, r8
+    \end{codeenv}
+\end{frame}
+
 \begin{frame}
     \input{test.dia-tex}
 \end{frame}