slides_software_dev_3.lyx

#LyX 2.3 created this file. For more info see http://www.lyx.org/
\lyxformat 544
\begin_document
\begin_header
\save_transient_properties true
\origin unavailable
\textclass foils
\use_default_options false
\maintain_unincluded_children false
\language english
\language_package default
\inputencoding auto
\fontencoding global
\font_roman "default" "default"
\font_sans "tgheros" "default"
\font_typewriter "beramono" "default"
\font_math "auto" "auto"
\font_default_family default
\use_non_tex_fonts false
\font_sc false
\font_osf false
\font_sf_scale 100 100
\font_tt_scale 100 100
\use_microtype false
\use_dash_ligatures true
\graphics default
\default_output_format default
\output_sync 0
\bibtex_command default
\index_command default
\paperfontsize 20
\spacing single
\use_hyperref false
\papersize a4paper
\use_geometry true
\use_package amsmath 1
\use_package amssymb 1
\use_package cancel 1
\use_package esint 1
\use_package mathdots 1
\use_package mathtools 1
\use_package mhchem 1
\use_package stackrel 1
\use_package stmaryrd 1
\use_package undertilde 1
\cite_engine basic
\cite_engine_type default
\biblio_style plain
\use_bibtopic false
\use_indices false
\paperorientation landscape
\suppress_date false
\justification true
\use_refstyle 1
\use_minted 0
\index Index
\shortcut idx
\color #008000
\end_index
\leftmargin 1.5cm
\topmargin 2cm
\rightmargin 1.5cm
\bottommargin 3cm
\secnumdepth 1
\tocdepth 1
\paragraph_separation skip
\defskip smallskip
\is_math_indent 0
\math_numbering_side default
\quotes_style english
\dynamic_quotes 0
\papercolumns 1
\papersides 1
\paperpagestyle default
\tracking_changes false
\output_changes false
\html_math_output 0
\html_css_as_file 0
\html_be_strict false
\end_header

\begin_body

\begin_layout My Logo
\begin_inset space ~
\end_inset


\end_layout

\begin_layout Title
\begin_inset space ~
\end_inset


\begin_inset Newline newline
\end_inset


\begin_inset space ~
\end_inset


\begin_inset Newline newline
\end_inset

More thoughts on scientific code development
\end_layout

\begin_layout Author
Christopher Albert
\end_layout

\begin_layout Foilhead
Lessons learned since last time
\end_layout

\begin_layout Itemize
Three mostly equivalent realizations
\end_layout

\begin_deeper
\begin_layout Itemize
A) External routines, modules as structs
\end_layout

\begin_layout Itemize
B) Module routines, types as structs
\end_layout

\begin_layout Itemize
C) Additionally put subroutines as type procedures
\end_layout

\end_deeper
\begin_layout Itemize
Main requirements
\end_layout

\begin_deeper
\begin_layout Itemize
Easy to write, understand and document
\end_layout

\begin_layout Itemize
Easy error identification, testing and debugging
\end_layout

\begin_layout Itemize
High (cache) performance and parallelizability
\end_layout

\end_deeper
\begin_layout Itemize
The question is not mainly 
\emph on
which
\emph default
, but 
\emph on
how
\emph default
 to use
\end_layout

\begin_deeper
\begin_layout Itemize
Modular: Split to managable pieces
\end_layout

\begin_layout Itemize
Explicit: Know what data a routine manipulates
\end_layout

\end_deeper
\begin_layout Itemize
Misunderstanding between 
\emph on
features for OO
\emph default
 and 
\emph on
OO design
\end_layout

\begin_layout Standard
\begin_inset Separator plain
\end_inset


\end_layout

\begin_layout Foilhead
Claims
\end_layout

\begin_layout Enumerate

\emph on
The limiting design cases are data oriented and object oriented
\end_layout

\begin_deeper
\begin_layout Enumerate
DO design: 
\emph on
Structures
\emph default
 are a set of 
\emph on
data
\emph default
 that are operated on by 
\emph on
implied routines
\end_layout

\begin_layout Enumerate
OO design: 
\emph on
Objects
\emph default
 are a set of 
\emph on
routines
\emph default
 that operate on 
\emph on
implied data
\end_layout

\end_deeper
\begin_layout Enumerate

\emph on
Using 
\begin_inset Quotes eld
\end_inset

OO
\begin_inset Quotes erd
\end_inset

 features of F03+ can also mean doing DO
\emph default

\begin_inset Newline newline
\end_inset

Usually: put routines that 
\begin_inset Quotes eld
\end_inset

belong
\begin_inset Quotes erd
\end_inset

 mainly to a type inside this type
\end_layout

\begin_layout Enumerate

\emph on
Actually it's good that 
\begin_inset Quotes eld
\end_inset

real
\begin_inset Quotes erd
\end_inset

 OO isn't used most of the time in physics codes
\begin_inset Newline newline
\end_inset


\emph default
...
 for the reasons mentioned in the main requirements.
 
\end_layout

\begin_layout Standard
\begin_inset CommandInset href
LatexCommand href
target "http://blog.cleancoder.com/uncle-bob/2019/06/16/ObjectsAndDataStructures.html"
literal "false"

\end_inset


\end_layout

\begin_layout Standard
\begin_inset Separator plain
\end_inset


\end_layout

\begin_layout Foilhead
Anti-pattern 1: God Struct
\end_layout

\begin_layout Standard
In style A this would look like
\end_layout

\begin_layout Standard
\begin_inset listings
lstparams "basicstyle={\footnotesize},frame=single"
inline false
status open

\begin_layout Plain Layout

module god_module
\end_layout

\begin_layout Plain Layout

  real :: thing1
\end_layout

\begin_layout Plain Layout

  real :: thing2
\end_layout

\begin_layout Plain Layout

  real :: thing3
\end_layout

\begin_layout Plain Layout

  real :: thing4
\end_layout

\begin_layout Plain Layout

  real :: thing5
\end_layout

\begin_layout Plain Layout

  ...
\end_layout

\begin_layout Plain Layout

end
\end_layout

\begin_layout Plain Layout

\end_layout

\begin_layout Plain Layout

subroutine do_something
\end_layout

\begin_layout Plain Layout

  use god_module
\end_layout

\begin_layout Plain Layout

  thing1 = thing4
\end_layout

\begin_layout Plain Layout

end
\end_layout

\end_inset


\end_layout

\begin_layout Standard
In style A the problem can immediately mitigated by 
\end_layout

\begin_layout Standard
\begin_inset listings
lstparams "basicstyle={\footnotesize},frame=single"
inline false
status open

\begin_layout Plain Layout

use god_module, only: thing1, thing4
\end_layout

\end_inset


\end_layout

\begin_layout Standard
This allows to use well-defined 
\begin_inset Quotes eld
\end_inset

parts
\begin_inset Quotes erd
\end_inset

 of a structure.
\end_layout

\begin_layout Standard
\begin_inset Separator plain
\end_inset


\end_layout

\begin_layout Foilhead
Anti-pattern 1: God Struct
\end_layout

\begin_layout Standard
In style B/C this would look like
\end_layout

\begin_layout Standard
\begin_inset listings
lstparams "basicstyle={\footnotesize},frame=single"
inline false
status open

\begin_layout Plain Layout

module god_module
\end_layout

\begin_layout Plain Layout

  type god_type
\end_layout

\begin_layout Plain Layout

    real :: thing1
\end_layout

\begin_layout Plain Layout

    real :: thing2
\end_layout

\begin_layout Plain Layout

    real :: thing3
\end_layout

\begin_layout Plain Layout

    real :: thing4
\end_layout

\begin_layout Plain Layout

    real :: thing5
\end_layout

\begin_layout Plain Layout

    ...
\end_layout

\begin_layout Plain Layout

    
\end_layout

\begin_layout Plain Layout

    contains
\end_layout

\begin_layout Plain Layout

    
\end_layout

\begin_layout Plain Layout

    subroutine do_something
\end_layout

\begin_layout Plain Layout

      thing1 = thing2
\end_layout

\begin_layout Plain Layout

    end
\end_layout

\begin_layout Plain Layout

end
\end_layout

\begin_layout Plain Layout

\end_layout

\end_inset


\end_layout

\begin_layout Standard
In 
\emph on
both
\emph default
 styles B and C we have 
\emph on
no possibility
\emph default
 to specify 
\emph on
only
\emph default
 the parts of 
\emph on
god_type
\emph default
 relevant to our routine.
 The only solution is to split 
\emph on
god_type
\emph default
.
\begin_inset Separator plain
\end_inset


\end_layout

\begin_layout Foilhead
Splitting the God Type
\end_layout

\begin_layout Standard
TODO
\end_layout

\begin_layout Standard
\begin_inset Separator plain
\end_inset


\end_layout

\begin_layout Foilhead
Row- versus column-oriented storage
\end_layout

\begin_layout Standard
Let's say we act on 
\begin_inset Formula $N$
\end_inset

 data at once.
 There are two ways to do this: 
\end_layout

\begin_layout Enumerate

\emph on
Array of structs (AoS) 
\emph default
or 
\emph on
row-oriented
\end_layout

\begin_layout Enumerate

\emph on
Struct of arrays (SoA) 
\emph default
or 
\emph on
column-oriented 
\end_layout

\begin_layout Standard
What is better depends on the access pattern with the goal to be serial,
 not random! Rule of thumb:
\end_layout

\begin_layout Itemize
Good: SoA if each array is processed separately or for array-kind of operations
 (SIMD)
\end_layout

\begin_layout Itemize
Bad: SoA with large 
\begin_inset Formula $N$
\end_inset

 and access to multiple members without SIMD operations
\end_layout

\begin_layout Itemize
Good: AoS with small structs that group together 
\emph on
only
\emph default
 the data that is actually accessed in the right order.
\end_layout

\begin_layout Itemize
Bad: AoS with god structs and few members accessed (danger in OO design)
\end_layout

\begin_layout Standard
\begin_inset Separator plain
\end_inset


\end_layout

\begin_layout Foilhead
Row- versus column-oriented storage
\end_layout

\begin_layout Standard
In style A, SOA would look like
\end_layout

\begin_layout Standard
\begin_inset listings
lstparams "basicstyle={\footnotesize},frame=single"
inline false
status open

\begin_layout Plain Layout

module gods_module
\end_layout

\begin_layout Plain Layout

  real :: thing1(N)
\end_layout

\begin_layout Plain Layout

  real :: thing2(N)
\end_layout

\begin_layout Plain Layout

  ...
\end_layout

\begin_layout Plain Layout

end
\end_layout

\begin_layout Plain Layout

\end_layout

\begin_layout Plain Layout

subroutine do_something
\end_layout

\begin_layout Plain Layout

  use gods_module, only: thing1, thing2
\end_layout

\begin_layout Plain Layout

  thing1(:) = thing2(:)
\end_layout

\begin_layout Plain Layout

end
\end_layout

\end_inset


\end_layout

\begin_layout Standard
In more complex cases one would need a for loop inside every 
\emph on
do_something
\emph default
 kind of routine.
 Style A can put types inside modules to get AOS:
\end_layout

\begin_layout Standard
\begin_inset listings
lstparams "basicstyle={\footnotesize},frame=single"
inline false
status open

\begin_layout Plain Layout

module gods_module
\end_layout

\begin_layout Plain Layout

  type(stuff12) :: stuff12(N)
\end_layout

\begin_layout Plain Layout

  type(stuff12) :: stuff12(N)
\end_layout

\begin_layout Plain Layout

end
\end_layout

\end_inset


\end_layout

\begin_layout Standard
\begin_inset Separator plain
\end_inset


\end_layout

\begin_layout Foilhead
Style A
\end_layout

\begin_layout Standard
Pro:
\end_layout

\begin_layout Standard
Con:
\end_layout

\begin_layout Itemize
Subroutines in global namespace are hard to find from reading (no 
\begin_inset Quotes eld
\end_inset

use only
\begin_inset Quotes erd
\end_inset

)
\end_layout

\end_body
\end_document