Add installation instructions

README.md

@@ -88,45 +88,27 @@ Generated by [tools/generate-scalability-results.sh](https://github.com/Sleipnir
 
 Ipopt uses MUMPS by default because it has free licensing. Commercial linear solvers may be much faster.
 
-### How we improved performance
-
-#### Make more decisions at compile time
-
-During problem setup, equality and inequality constraints are encoded as different types, so the appropriate setup behavior can be selected at compile time via operator overloads.
+See [benchmark details](https://github.com/SleipnirGroup/Sleipnir/?tab=readme-ov-file#benchmark-details) for more.
 
-#### Reuse autodiff computation results that are still valid (aka caching)
-
-The autodiff library automatically records the linearity of every node in the computational graph. Linear functions have constant first derivatives, and quadratic functions have constant second derivatives. The constant derivatives are computed in the initialization phase and reused for all solver iterations. Only nonlinear parts of the computational graph are recomputed during each solver iteration.
-
-For quadratic problems, we compute the Lagrangian Hessian and constraint Jacobians once with no problem structure hints from the user.
+## Install
 
-#### Use a performant linear algebra library with fast sparse solvers
-
-[Eigen](https://gitlab.com/libeigen/eigen/) provides these. It also has no required dependencies, which makes cross compilation much easier.
-
-#### Use a pool allocator for autodiff expression nodes
-
-This promotes fast allocation/deallocation and good memory locality.
+### C++ library
 
-We could mitigate the solver's high last-level-cache miss rate (~42% on the machine above) further by breaking apart the expression nodes into fields that are commonly iterated together. We used to use a tape, which gave computational graph updates linear access patterns, but tapes are monotonic buffers with no way to reclaim storage.
+See the [build instructions](https://github.com/SleipnirGroup/Sleipnir/?tab=readme-ov-file#c-library-1).
 
-### Running the benchmarks
+### Python library
 
-Benchmark projects are in the [benchmarks folder](https://github.com/SleipnirGroup/Sleipnir/tree/main/benchmarks). To compile and run them, run the following in the repository root:
 ```bash
-# Install CasADi and [matplotlib, numpy, scipy] pip packages first
-cmake -B build -S .
-cmake --build build
-./tools/generate-scalability-results.sh
+pip install sleipnirgroup-jormungandr
 ```
 
-See the contents of `./tools/generate-scalability-results.sh` for how to run specific benchmarks.
-
 ## Examples
 
 See the [examples](https://github.com/SleipnirGroup/Sleipnir/tree/main/examples) and [optimization unit tests](https://github.com/SleipnirGroup/Sleipnir/tree/main/test/optimization).
 
-## Dependencies
+## Build
+
+### Dependencies
 
 * C++20 compiler
   * On Windows, install [Visual Studio Community 2022](https://visualstudio.microsoft.com/vs/community/) and select the C++ programming language during installation
@@ -147,21 +129,17 @@ See the [examples](https://github.com/SleipnirGroup/Sleipnir/tree/main/examples)
 
 Library dependencies which aren't installed locally will be automatically downloaded and built by CMake.
 
-If [CasADi](https://github.com/casadi/casadi) is installed locally, the benchmark executables will be built.
+The benchmark executables require [CasADi](https://github.com/casadi/casadi) to be installed locally.
 
-## Build instructions
+### C++ library
 
 On Windows, open a [Developer PowerShell](https://learn.microsoft.com/en-us/visualstudio/ide/reference/command-prompt-powershell?view=vs-2022). On Linux or macOS, open a Bash shell.
 
-Clone the repository.
 ```bash
+# Clone the repository
 git clone git@github.com:SleipnirGroup/Sleipnir
 cd Sleipnir
-```
-
-### C++ library
 
-```bash
 # Configure; automatically downloads library dependencies
 cmake -B build -S .
 
@@ -198,7 +176,13 @@ The following build types can be specified via `-DCMAKE_BUILD_TYPE` during CMake
 
 ### Python library
 
+On Windows, open a [Developer PowerShell](https://learn.microsoft.com/en-us/visualstudio/ide/reference/command-prompt-powershell?view=vs-2022). On Linux or macOS, open a Bash shell.
+
 ```bash
+# Clone the repository
+git clone git@github.com:SleipnirGroup/Sleipnir
+cd Sleipnir
+
 # Setup
 pip install --user build
 
@@ -218,6 +202,42 @@ Passing the `--enable-diagnostics` flag to the test executable enables solver di
 
 Some test problems generate CSV files containing their solutions. These can be plotted with [tools/plot_test_problem_solutions.py](https://github.com/SleipnirGroup/Sleipnir/blob/main/tools/plot_test_problem_solutions.py).
 
+## Benchmark details
+
+### Running the benchmarks
+
+Benchmark projects are in the [benchmarks folder](https://github.com/SleipnirGroup/Sleipnir/tree/main/benchmarks). To compile and run them, run the following in the repository root:
+```bash
+# Install CasADi and [matplotlib, numpy, scipy] pip packages first
+cmake -B build -S .
+cmake --build build
+./tools/generate-scalability-results.sh
+```
+
+See the contents of `./tools/generate-scalability-results.sh` for how to run specific benchmarks.
+
+### How we improved performance
+
+#### Make more decisions at compile time
+
+During problem setup, equality and inequality constraints are encoded as different types, so the appropriate setup behavior can be selected at compile time via operator overloads.
+
+#### Reuse autodiff computation results that are still valid (aka caching)
+
+The autodiff library automatically records the linearity of every node in the computational graph. Linear functions have constant first derivatives, and quadratic functions have constant second derivatives. The constant derivatives are computed in the initialization phase and reused for all solver iterations. Only nonlinear parts of the computational graph are recomputed during each solver iteration.
+
+For quadratic problems, we compute the Lagrangian Hessian and constraint Jacobians once with no problem structure hints from the user.
+
+#### Use a performant linear algebra library with fast sparse solvers
+
+[Eigen](https://gitlab.com/libeigen/eigen/) provides these. It also has no required dependencies, which makes cross compilation much easier.
+
+#### Use a pool allocator for autodiff expression nodes
+
+This promotes fast allocation/deallocation and good memory locality.
+
+We could mitigate the solver's high last-level-cache miss rate (~42% on the machine above) further by breaking apart the expression nodes into fields that are commonly iterated together. We used to use a tape, which gave computational graph updates linear access patterns, but tapes are monotonic buffers with no way to reclaim storage.
+
 ## Branding
 
 ### Logo