Optional TikZ Figures - Githubissues

Go through the text and, where possible, replace images with TikZ figures.

Chapter 01

[ ] Figure:theory->CUDA->memory-management->memory-spaces
- Try to use ChatGPT to generate some base, e.g., "Generate a TikZ figure describing CUDA memory structuring, it should contain bla bla bla"
- Consider paying for ChatGPT if it is able to accept images as prompts
[ ] Figure:theory->ICMPP->LUP->CMPP->sum-in-element-computation-dependance
- Can use ChatGPT to generate some base and then expand on it

[ ] Create TikZ figures to aid some explanations - list here:
- [ ] chapter02 -> Implementation -> Decomposition Project -> Implemented Algorithms -> Decomposers -> ICMxPP -> Processing by Sections: Explanation of quality of elements.
- Image showing that the quality of elements in the lower-right corner of the matrix depends on the quality of other elements.
- [ ] \paragraph{Parallel Crout's Method With Partial Pivoting (PCM\textit{x}PP)} -> after the algorithm explanation: Figure showing the dependency of elements described in the text.

[ ] Figure:comparing-decomposers-and-solvers->decomposition-project-benchmarks->solvers-benchmark->comparison-of-execution-times-on-all-matrices->visualization-of-substitutions
- Try to use ChatGPT to generate some base, e.g., "Generate a TikZ figure that visualizes forward substituion".
[ ] Create TikZ figures to aid some explanations - list here:
- [ ] To aid the description of how ICMxPP decomposes Cejka dense matrices - near Figure:comparing-decomposers-and-solvers->decomposition-project-benchmarks->decomposers-benchmark->comparison-of-execution-times-on-subset-of-matrices->matrix-with-metrics->Cejka558
- [ ] TikZ figures explaining how a thread block is divided into warps - near Figure:comparing-decomposers-and-solvers->decomposition-project-benchmarks->decomposers-benchmark->comparison-of-execution-times-on-subset-of-matrices->ICMxPP-double-precision
- [ ] TikZ figures explaining how 8-by-8 blocks of threads stacked on top of each other access the same block of elements multiple times to load the same elements, while a 32-by-32 block of threads does not need to do this as it loads the elements only once - near \begin{quote}