Update Install instructions

CERN · May 24, 2023 · 6885857 · 6885857
1 parent 78185b0
commit 6885857
Show file tree

Hide file tree

Showing 2 changed files with 58 additions and 23 deletions.
diff --git a/Frontispiece/MATLAB_installation.md b/Frontispiece/MATLAB_installation.md
@@ -11,7 +11,7 @@ Installation Instructions for MATLAB
    - [Requirements](#requirements-1)
    - [Simple Instructions](#simple-instructions-1)
    - [Step by Step Instructions](#step-by-step-instructions1)
-   
+- [Advanced](#advanced)
    *****
 
 ## Windows
@@ -20,24 +20,24 @@ Installation Instructions for MATLAB
 
 1. MATLAB
 2. Visual Studio (Community or Profesional)
-3. A CUDA capable GPU from NVIDIA with [compute capability greater or equal to 3.0](https://en.wikipedia.org/wiki/CUDA#GPUs_supported)
+3. A CUDA capable GPU from NVIDIA with [compute capability greater or equal to 3.5](https://en.wikipedia.org/wiki/CUDA#GPUs_supported)
 4. CUDA Toolkit (9.2 or newer)
 
 Tested on
 
 | Software        | Version           | 
 | ------------- |:-------------:|
 |**Windows**| 7, 8, 10.|
-|**MATLAB**| 2014b 2016b 2017a 2018b|
-|**CUDA**| 9.2 10|
-|**Visual Studio**| 2010 2013 2015|
+|**MATLAB**|Any MATLAB >2016b|
+|**CUDA**|Any CUDA 9.2>|
+|**Visual Studio**| 2010 2013 2015 2019 2022|
 
 
 
 ### Simple Instructions
 
-1. Install MATLAB, Visual Studio and CUDA
-2. Rename either `mex_CUDA_win64_MVS2013.xml` (Visual Studio 2013 or older) or `mex_CUDA_win64_MVS2015.xml`(Visual Studio 2015 or newer) to `mex_CUDA_win64.xml`
+1. Install MATLAB, Visual Studio and CUDA (Remember to install C++ when isntalling Visual Studio!)
+2. Rename the XML file corresponding to the Visual Studio you have. e.g. `mex_CUDA_win64_MVS2015.xml`(Visual Studio 2015/2017) to `mex_CUDA_win64.xml`
 3. Run `Compile.m`
 
 A succesfull installation should be able to execute the script at `TIGRE/MATLAB/Demos/d03_generateData.m` without errors.
@@ -129,9 +129,9 @@ Tested on
 
 | Software       | Version    | 
 | ------------- |:-------------:|
-| **Ubuntu**|  16.04 17.10| 
-| **MATLAB**|  2017a 2018b| 
-| **CUDA**| 9.2 10.0| 
+| **Ubuntu**|  Any ubuntu 16.04>| 
+| **MATLAB**|  Any MATLAB 2016b>| 
+| **CUDA**| Any Cuda 0.2>| 
 | **gcc**|  6.4.0 7.2.0| 
 
 ### Simple Instructions
@@ -141,7 +141,7 @@ Tested on
 
 A succesfull installation should be able to execute the script at `TIGRE/MATLAB/Demos/d03_generateData.m` without errors.
 
-### Step by Step Instructions:<sup>1</sup>
+### Step by Step Instructions:
 
 1. Install MATLAB
 
@@ -181,4 +181,13 @@ If none of this works, please contact the authors at [[email protected]](m
 
 ****
 
-<sup>1</sup> Testing by the TIGRE team in Linux is limited, thus the step by step instructions are less detailed than expected. Please do [contact us](mailto:[email protected]) if you are having trouble or would like to contribute to the instructions. 
+## Advanced
+
+If you are doing reconstruction of large datasets, and you want to use swap memory, you will need to deactivate TIGREs pinned memory feature at compile time. This will allow you to use swap memory, but it will make the operators in TIGRE slower, as pinned memory is used for simultaneous memory and compute. 
+
+
+You can do this by calling the `Compile.m` file from the MATLAB command line as `Compile --no_pinned_memory`.
+
+
+
+
diff --git a/Frontispiece/python_installation.md b/Frontispiece/python_installation.md
@@ -1,22 +1,35 @@
 Installation Instructions for Python
 ======
 
+## Table of contents
+
+- [Windows](#windows)
+   - [Requirements](#requirements)
+   - [Simple Instructions](#simple-instructions)
+   - [Step by Step Instructions](#step-by-step-instructions)
+- [Linux](#linux)
+   - [Requirements](#requirements-1)
+   - [Simple Instructions](#simple-instructions-1)
+   - [Step by Step Instructions](#step-by-step-instructions1)
+- [Optional Code Style Enforcement](#Optional-Code-Style-Enforcement)
+- [Advanced](#advanced)
+   *****
 ## Windows
 
 ### Requirements:
 
-1. Python 3
+1. Python 3.7-3.10
 2. MVS
-3. A CUDA capable GPU from NVIDIA with [compute capability greater or equal to 3.0](https://en.wikipedia.org/wiki/CUDA#GPUs_supported)
+3. A CUDA capable GPU from NVIDIA with [compute capability greater or equal to 3.5](https://en.wikipedia.org/wiki/CUDA#GPUs_supported)
 4. CUDA Toolkit
 
 Tested on
 
 | Software        | Version           | 
 | ------------- |:-------------:|
 |**Windows**| 10 |
-|**Python**| 3.7 3.8 |
-|**CUDA**| 10.1 |
+|**Python**| 3.7 3.8 3.9 3.10|
+|**CUDA**| 9.2>|
 |**MSVC**| 19.24 |
 
 ### Simple Instructions
@@ -63,19 +76,19 @@ A succesfull installation should be able to execute the script at `TIGRE/Python/
 
 ### Requirements:
 
-1. Python 2/Python 3
+1. Python 3
 2. gcc
-3. A CUDA capable GPU from NVIDIA with [compute capability greater or equal to 3.0](https://en.wikipedia.org/wiki/CUDA#GPUs_supported)
+3. A CUDA capable GPU from NVIDIA with [compute capability greater or equal to 3.5](https://en.wikipedia.org/wiki/CUDA#GPUs_supported)
 4. CUDA Toolkit
 
 
 Tested on
 
 | Software        | Version           | 
 | ------------- |:-------------:|
-|**Ubuntu**| 16.04 17.10|
-|**Python**| 2.7 3.7 |
-|**CUDA**| 8.0 9.2 10.1 10.2|
+|**Ubuntu**| 16.04>|
+|**Python**| 3.7-3.10|
+|**CUDA**| 9.2>|
 |**gcc**|  7.6.0|
 
 ### Simple Instructions
@@ -90,14 +103,17 @@ A succesfull installation should be able to execute the script at `TIGRE/Python/
 
 For Ubuntu
 
-1. Install python and pip (you can use 2 or 3, example show for 2)
+1. Install python and pip
+
+	Recommended to do it via [Anaconda3](https://docs.anaconda.com/free/anaconda/install/linux/)
 
 	```
 	sudo apt update
 	sudo apt upgrade
-	sudo apt install python2.7 python-pip
+	sudo apt install python3.10 python-pip
 	```
 
+
 2. Install CUDA
 
    Installing CUDA in linux (specially one with a GUI) can be a challenge. Please follow [NVIDIAs instructions](https://developer.download.nvidia.com/compute/cuda/10.0/Prod/docs/sidebar/CUDA_Installation_Guide_Linux.pdf) carefully.\
@@ -185,3 +201,13 @@ check blanket noqa.......................................................Passed
 ```
 
 **NOTE:** pre-commit may also be manually invoked against *all* files (staged and unstaged) using the `pre-commit run --all-files`. However, some changes made to Python's TIGRE codebase by `black` have been manually reverted for readability reasons and should not be committed in their blackened state.
+
+## Advanced 
+
+If you are doing reconstruction of large datasets, and you want to use swap memory, you will need to deactivate TIGREs pinned memory feature at compile time. This will allow you to use swap memory, but it will make the operators in TIGRE slower, as pinned memory is used for simultaneous memory and compute. 
+
+You can do this by calling the `stup.py` with the flag `--no_pinned_memory`. 
+
+
+
+