Skip to content

Commit

Permalink
David's Review changes
Browse files Browse the repository at this point in the history
- Mentioned Chef-FP in relevant places
- Explained Chef-FP in the beginning, linked to paper
- changed FPEE to full form in all places (except in code elements)
  • Loading branch information
QuillPusher committed Nov 5, 2023
1 parent 3c47e28 commit 4b10c27
Showing 1 changed file with 19 additions and 18 deletions.
37 changes: 19 additions & 18 deletions docs/userDocs/source/user/FloatingPointErrorEstimation.rst
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
FP Error Estimation
*******************
Floating Point Error Estimation using CHEF-FP
*********************************************

============
Introduction
Expand Down Expand Up @@ -33,11 +33,17 @@ gradient computation for large and complex codebases, and is deeply
integrated with the compiler, allowing automatic generation of error
estimation code.

Where does the FPEE logic reside?
=================================
Where does the CHEF-FP logic reside?
====================================

**CHEF-FP** (the tool created for Floating Point Error Estimation) is a
flexible, scalable, and easy-to-use source-code transformation tool based on
Automatic Differentiation (AD) for analyzing approximation errors in HPC
applications.

The main logic for Floating Point Error Estimation resides in the following
files:
For more details, please view `Fast and Automatic Floating Point Error Analysis with CHEF-FP`_.

The main logic for CHEF-FP resides in the following files:

- `include/clad/Differentiator/ErrorEstimator.h`_

Expand Down Expand Up @@ -78,8 +84,8 @@ This model will return a formula that is represented using a Clang
expression.This Clang expression can, in turn, be written into the derivative
code that is generated using Clad.

What else is FPEE framework capable of?
=======================================
What else is CHEF-FP capable of?
================================

Sensitivity Analysis
--------------------
Expand All @@ -105,7 +111,7 @@ How do I create my own Custom model?
====================================

Custom Models may be one of the main reasons that new users may be interested
in adapting the FPEE code to their specific use cases.
in adapting the CHEF-FP code to their specific use cases.

Top define a custom model using Clad:

Expand Down Expand Up @@ -135,15 +141,15 @@ Further Reading

For more technical details, please view:

- `Fast and Automatic Floating Point Error Analysis`_ - (published paper)
- `Fast and Automatic Floating Point Error Analysis with CHEF-FP`_ - (published paper)

- `How to Estimate Floating Point Errors Using AD`_ - (tutorial)

- `CHEF-FP Examples Repo`_ - (includes benchmarks)

- `Estimating Floating-Point Errors Using Automatic Differentiation`_ - (presentation, slides and video)

- `FP-EE Proposal`_ - (PDF, slightly outdated, useful for background information)
- `Floating-Point Error Estimation Proposal`_ - (PDF, slightly outdated, useful for background information)


Appendix - Notable Classes
Expand All @@ -159,11 +165,6 @@ error information exchange between Clad and the Error Estimation module
(CHEF-FP). This class is responsible for a lot of the housekeeping tasks as
well.

**CHEF-FP** is a flexible, scalable, and easy-to-use source-code
transformation tool based on Automatic Differentiation (AD) for analyzing
approximation errors in HPC applications. For more details, please view `Fast
and Automatic Floating Point Error Analysis`_.

clad::ErrorEstimator::EmitFinalErrorStmts
-----------------------------------------

Expand All @@ -177,7 +178,7 @@ forward block.

.. _demos/ErrorEstimation: https://github.com/vgvassilev/clad/tree/master/demos/ErrorEstimation

.. _Fast and Automatic Floating Point Error Analysis: https://arxiv.org/pdf/2304.06441.pdf
.. _Fast and Automatic Floating Point Error Analysis with CHEF-FP: https://arxiv.org/pdf/2304.06441.pdf

.. _CustomModel: https://github.com/vgvassilev/clad/blob/master/demos/ErrorEstimation/CustomModel/README.md

Expand All @@ -187,7 +188,7 @@ forward block.

.. _Estimating Floating-Point Errors Using Automatic Differentiation: https://compiler-research.org/presentations/#FPErrorEstADSIAMUQ2022

.. _FP-EE Proposal: https://compiler-research.org/assets/docs/Garima_Singh_Proposal_2020.pdf
.. _Floating-Point Error Estimation Proposal: https://compiler-research.org/assets/docs/Garima_Singh_Proposal_2020.pdf

.. _CHEF-FP Examples Repo: https://github.com/grimmmyshini/chef-fp-examples

Expand Down

0 comments on commit 4b10c27

Please sign in to comment.