Release v1.0.5

README.md

@@ -11,6 +11,7 @@ Stable versions:
 
 | Version  | Date        | Document                                                                                                       |
 |----------|:-----------:|:--------------------------------------------------------------------------------------------------------------:|
+| 1.0.5    | 2023-Apr-06 | [Link](https://arxiv.org/pdf/2011.15028v6.pdf)                                                                 |
 | 1.0.4    | 2023-Feb-15 | [Link](https://arxiv.org/pdf/2011.15028v5.pdf)                                                                 |
 | 1.0.3    | 2022-Mar-31 | [Link](https://arxiv.org/pdf/2011.15028v4.pdf)                                                                 |
 | 1.0.2    | 2021-Apr-13 | [Link](https://arxiv.org/pdf/2011.15028v3.pdf)                                                                 |

ldbc.cls

@@ -232,7 +232,7 @@
         \begin{center}
         \Large
         The specification was built on the source code available at \\
-        \url{https://github.com/ldbc/ldbc_graphalytics_docs/tree/main}
+        \url{https://github.com/ldbc/ldbc_graphalytics_docs/releases/tag/v1.0.5}
         \end{center}
         }
 

tex/definition.tex

@@ -224,7 +224,7 @@ \section{Algorithms}
 
 The Graphalytics benchmark consists of six algorithms (also known as \emph{kernels}~\cite{DBLP:conf/hipc/BaderM05}) which need to be executed on the different datasets: five algorithms for unweighted graphs and one algorithm for weighted graphs. These algorithms have been selected based on the results of multiple surveys and expert advice from the participants of the LDBC Technical User Community (TUC) meeting.
 
-Each workload of Graphalytics consists of executing a single algorithm on a single dataset. Below, abstract descriptions are provided for the six algorithms; pseudo-code is given in \autoref{chap:algorithms}. Furthermore, a link to the reference implementation is presented in \autoref{sec:instructions:drivers}. However, Graphalytics does not impose any constraint on the implementation of algorithms. Any implementation is allowed, as long as its correctness can be validated by comparing its output to correct reference output (\autoref{sec:definitions_validation}). 
+Each workload of Graphalytics consists of executing a single algorithm on a single dataset. Below, abstract descriptions are provided for the six algorithms; pseudo-code is given in \autoref{chap:algorithms}. Furthermore, a link to the reference implementation is presented in \autoref{sec:instructions:core}. However, Graphalytics does not impose any constraint on the implementation of algorithms. Any implementation is allowed, as long as its correctness can be validated by comparing its output to correct reference output (\autoref{sec:definitions_validation}).
 
 In the following sections, a graph $G$ consists of a set of vertices $V$ and a set of edges $E$. For undirected graphs, each edge is bidirectional, so if $(u,v)\in E$ then $(v,u)\in E$. Each vertex $v$ has a set of outgoing neighbors
 $N_\mathrm{out}(v) = \{u \in V | (v, u) \in E \}$, a set of incoming neighbors
@@ -332,7 +332,7 @@ \subsection{Single-Source Shortest Paths (SSSP)}
 \section{Output Validation}
 \label{sec:definitions_validation}
 
-The output of every execution of an algorithm on a dataset must be validated for the result to be admissible. All algorithms in the Graphalytics benchmark are deterministic and can therefore be validated by comparing to reference output for correctness. The reference output is typically generated by a specifically chosen reference platform, the implementation of which is cross-validated with at least two other platforms up to target scale~L. \futureinversion{2.0}{Define target scale in this chapter [Gabor]} The results are tested by cross-validating multiple platforms and implementations against each others (see \autoref{sec:instructions:drivers}).
+The output of every execution of an algorithm on a dataset must be validated for the result to be admissible. All algorithms in the Graphalytics benchmark are deterministic and can therefore be validated by comparing to reference output for correctness. The reference output is typically generated by a specifically chosen reference platform, the implementation of which is cross-validated with at least two other platforms up to target scale~L. \futureinversion{2.0}{Define target scale in this chapter [Gabor]} The results are tested by cross-validating multiple platforms and implementations against each other.
 
 The validation output presents numbers either as integers or floating-point numbers, depending on the algorithm definition. Note that these numbers are stored in the file system as decimal values in plain text (ASCII).
 For floating-point numbers, a scientific notation with 15~significant digits (e.g., $2.476\,533\,217\,845\,853\mathrm{e-}08$) is used.

tex/version.tex

@@ -14,4 +14,4 @@
   \versionLogEntry{02/09/2016}{0.2.4}{}{First draft}
   \versionLogEntry{16/07/2016}{0.1}{}{First draft}
 }
-\lastVersion{1.0.5-draft}
+\lastVersion{1.0.5}