chore: add draft for outline and updated benchmark ingest script

benchmark/benchmark-iter.py

@@ -9,6 +9,9 @@ import sys
 filename_pattern = re.compile(
     r"(?P<job>[0-9]+)\.(?P<type>cpp-bench.)?(?P<partition>[^.]+).bench.txt"
 )
+query_pattern = re.compile(
+    r"^(?P<query>Q[0-9].[0-9]) +\| +(?P<run>[0-9]) +\| +(?P<duration>[0-9]+\.[0-9]+) +(?P<unit>\w+)$"
+)
 
 
 def init_db(db: sqlite3.Connection):
@@ -87,6 +90,19 @@ def main(ingest: pathlib.Path, db: sqlite3.Connection):
             (bench_id,) = bench_id
         benches[bench] = bench_id
 
+    # create or get cpp compiler
+    cpp_compiler = db.execute(
+        "select id from compilers where name = 'cpp-ref'"
+    ).fetchone()
+    if cpp_compiler is None:
+        (cpp_compiler,) = db.execute(
+            "insert into compilers (name) values (?) returning id",
+            ["cpp-ref"],
+        ).fetchone()
+        print(f"Fetched [c] {cpp_compiler}")
+        db.commit()
+        compilers["cpp-ref"] = cpp_compiler
+
     for filename in os.listdir(ingest):
         match = filename_pattern.search(filename)
 
@@ -103,6 +119,22 @@ def main(ingest: pathlib.Path, db: sqlite3.Connection):
             if match.get("type") is not None:
                 bench = benches["benchmark-cpp"]
 
+            # set compiler based on bench
+            compiler_name = "adaptive"
+            if match.get("type") is not None:
+                compiler_name = "cpp-ref"
+
+            compiler = compilers.get(compiler_name)
+
+            if compiler is None:
+                (compiler,) = db.execute(
+                    "insert into compilers (name) values (?) returning id",
+                    [compiler_name],
+                ).fetchone()
+                print(f"Fetched [c] {compiler}")
+                db.commit()
+                compilers[compiler_name] = compiler
+
             # figure out partition
             partition_name = match["partition"]
 
@@ -121,14 +153,51 @@ def main(ingest: pathlib.Path, db: sqlite3.Connection):
                 print(f"Fetched [p] {partition}")
                 partitions[partition_name] = partition
 
-            # TODO: parse file iter by iter
-
             print(f"Handling file for {partition} on {bench}")
 
             fd = open(ingest / filename)
 
             content = fd.read()
-            print(content.split("||")[2])
+
+            for line in content.splitlines():
+                match = query_pattern.search(line)
+                if match is not None:
+                    match = match.groupdict()
+                    print(f"Handling line {line}")
+
+                    query_name = match["query"].replace(".", "-").lower()
+                    print(f"Handling query {query_name}")
+
+                    # fetch or insert query
+                    query = queries.get(query_name)
+                    if query is None:
+                        (query,) = db.execute(
+                            "insert into queries (name) values (?) returning id",
+                            [query_name],
+                        ).fetchone()
+                        print(f"Fetched [q] {query}")
+                        db.commit()
+                        queries[query_name] = query
+
+                    try:
+                        if match["run"] != "0":
+                            db.execute(
+                                "insert into measurements (compiler, query, partition, bench, duration, unit, iter_count) values (?, ?, ?, ?, ?, ?, ?)",
+                                [
+                                    compiler,
+                                    query,
+                                    partition,
+                                    bench,
+                                    float(match["duration"]),
+                                    match["unit"],
+                                    1,
+                                ],
+                            )
+                            print("inserted value")
+                        else:
+                            print("discard warmup")
+                    except Exception as e:
+                        print(f"Failed to insert row: {e}")
 
             fd.close()
             db.commit()

presentation.tex

-\documentclass{beamer}
+\documentclass[draft]{beamer}
 \usepackage[T1]{fontenc}
-\usepackage[german]{babel}
+\usepackage[german,english]{babel}
 \usepackage{tikz,pgfplots,verbatim}
+\usepackage{listings}
+\definecolor{GrayCodeBlock}{RGB}{64, 160, 43}
+\definecolor{BlackText}{RGB}{76, 79, 105}
+\definecolor{RedTypename}{RGB}{210, 15, 57}
+\definecolor{GreenString}{RGB}{96,172,57}
+\definecolor{PurpleKeyword}{RGB}{136, 57, 239}
+\definecolor{GrayComment}{RGB}{64, 160, 43}
+\definecolor{GoldDocumentation}{RGB}{180,165,45}
+\lstdefinelanguage{rust}
+{
+    columns=fullflexible,
+    keepspaces=true,
+    frame=single,
+    framesep=0pt,
+    framerule=0pt,
+    framexleftmargin=4pt,
+    framexrightmargin=4pt,
+    framextopmargin=5pt,
+    framexbottommargin=3pt,
+    xleftmargin=4pt,
+    xrightmargin=4pt,
+    backgroundcolor=\color{GrayCodeBlock},
+    basicstyle=\ttfamily\color{BlackText},
+    keywords={
+        true,false,
+        unsafe,async,await,move,
+        use,pub,crate,super,self,mod,
+        struct,enum,fn,const,static,let,mut,ref,type,impl,dyn,trait,where,as,
+        break,continue,if,else,while,for,loop,match,return,yield,in
+    },
+    keywordstyle=\color{PurpleKeyword},
+    ndkeywords={
+        bool,u8,u16,u32,u64,u128,i8,i16,i32,i64,i128,char,str,
+        Self,Option,Some,None,Result,Ok,Err,String,Box,Vec,Rc,Arc,Cell,RefCell,HashMap,BTreeMap,
+        macro_rules
+    },
+    ndkeywordstyle=\color{RedTypename},
+    comment=[l][\color{GrayComment}\slshape]{//},
+    morecomment=[s][\color{GrayComment}\slshape]{/*}{*/},
+    morecomment=[l][\color{GoldDocumentation}\slshape]{///},
+    morecomment=[s][\color{GoldDocumentation}\slshape]{/*!}{*/},
+    morecomment=[l][\color{GoldDocumentation}\slshape]{//!},
+    morecomment=[s][\color{RedTypename}]{\#![}{]},
+    morecomment=[s][\color{RedTypename}]{\#[}{]},
+    stringstyle=\color{GreenString},
+    string=[b]"
+}
+
+\newcommand{\code}[1]{\lstinline[language=rust]{#1}}
 
 \usetheme[sectionpage=none]{metropolis}           % Use metropolis theme
-\title{PPDS Präsentation}
+\title{PPDS Presentation}
 \date{\today}
 \author{Elias Engel and Joshua Kobschätzki}
-\institute{PPDS Sommersemester 2024}
-\section{Übersicht}
+\institute{PPDS SoSe 2024}
+\section{Overview}
 \begin{document}
 \maketitle
-\section{Lernergebnisse}
+\section{Architecture}
+\begin{frame}[fragile]
+	\note{2 Min}
+	\frametitle{\secname: AdaptiveCompiler}
+	\begin{itemize}
+		\item Compiling \code{struct PlanNode} $\to$ \code{Box<dyn ONC>}
+		      \begin{itemize}
+			      \item[$\to$] Translates PlanNode into IR
+			            \pause
+			      \item[$\to$] optimized in passes
+			            \begin{enumerate}
+				            \item Predicate
+				            \item Unify
+				                  \pause
+				            \item Pushdown Projection
+				                  \pause
+				            \item Projection Fuse
+			            \end{enumerate}
+			            \pause
+			      \item[$\to$] translated into operators, who may opportunistically use SIMD
+			            \pause
+			      \item[$\to$] (optional) SIMD operators use an intermediate ONC layer, \verb|ChunkedONC|
+		      \end{itemize}
+	\end{itemize}
+\end{frame}
+\section{Benchmark results}
+\begin{frame}
+	\note{2 Min}
+	\frametitle{\secname}
+	TODO
+\end{frame}
+\section{Learning Outcomes}
 \begin{frame}[fragile]{\secname}
 	\begin{itemize}
 		\item {
-		      \verb|std::simd|, High-level SIMD ist suboptimal
+		      \verb|std::simd|, High-level SIMD is suboptimal
 
 		      \begin{itemize}
-			      \item Kosten der Abstraktion hoch
-			      \item Optimierung von skalaren Teilen oder SIMD per intrinsics ist effektiver
-			      \item Mikroarchitekturen haben sehr unterschiedliche Charakteristiken
-			            \note{Implied here, high level SIMD can't accurately reflect these and is a compromise}
+			      \item Abstraction cost was too high
+			      \item Optimising of scalar code via SIMD intrinsics is significantly more effective
+			      \item Does not adapt well across microarchitectures
 		      \end{itemize}
 		      }
-		\item ONC ist nicht gut vektorisierbar
+		\item ONC interface is not well suitable for vectorization
 		      \begin{itemize}
-			      \item Design des Interfaces/ der Datenstrukturen elementar
+			      \item Vectorization requires certain data structure layouts (columnar)
+			      \item Vectorization of ONC has a fixed overhead (unpacking)
 		      \end{itemize}
 	\end{itemize}
 \end{frame}
-\section{Herausforderungen}
-\begin{frame}
-\end{frame}
 
 \begin{frame}[standout]
-	Danke für ihre Aufmerksamkeit\\
-	Raum für Fragen
+	Thank you for your Attention
 \end{frame}
 
 \section*{Backup Slides}