README.md

sr-non-linear-bounds

Scripts to reproduce the numerical experiments of the paper "Bounds on Non-linear Errors for Variance Computation with Stochastic Rounding".

Before running the scripts, ensure that you have installed Verificarlo v0.8.0, Python 3, and matplotlib on your computer.

Textbook algorithm

To reproduce textbook algorithm experiments use the following commands:

# To generate figure 3 with 1 - lambda = 0.9 (left)

sr-non-linear-bounds$ cd textbook

# The generated plot is a pdf file named textbook-n.pdf
sr-non-linear-bounds/textbook$ verificarlo-c -O2 --function=textbook_sr ./textbook.c -o textbook -lm
sr-non-linear-bounds/textbook$ ./run-and-plot_over_n.py

# To generate figure 3 with n= 10^6 (right)

sr-non-linear-bounds$ cd textbook_proba

# The generated plot is a pdf file named textbook-p.pdf
sr-non-linear-bounds/textbook_proba$ gcc ./textbook.c -o textbook -lm
sr-non-linear-bounds/textbook_proba$ ./run-and-plot_over_n.py

textbook against two-pass

To reproduce the experiments of textbook algorithm against two-pass algorithm use the following commands:

sr-non-linear-bounds$ cd textbook_vs_two-pass

# To generate figure 4 (left)
# The generated plot is a pdf file named text-vs-tp.pdf
sr-non-linear-bounds/textbook_vs_two-pass$ verificarlo-c -O2 --function=text_vs_tp_sr ./text_vs_tp_sr.c -o text_vs_tp_sr -lm
sr-non-linear-bounds/textbook_vs_two-pass$ text-tp.py

# To generate figure 4 (right)
# First, modify the random_array function in the text_vs_tp_sr.c file to generate random values between 1024 and 1025.
# The generated plot is a pdf file named text-vs-tp.pdf
sr-non-linear-bounds/textbook_vs_two-pass$ verificarlo-c -O2 --function=text_vs_tp_sr ./text_vs_tp_sr.c -o text_vs_tp_sr -lm
sr-non-linear-bounds/textbook_vs_two-pass$ text-tp.py