Skip to content

[libc][NFC] Add sollya script to compute worst case range reduction. #104803

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Aug 19, 2024
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
80 changes: 80 additions & 0 deletions libc/utils/mathtools/worst_case.sollya
Original file line number Diff line number Diff line change
@@ -0,0 +1,80 @@
// Implement WorstCase functions to compute the worst case for x mod C, with
// the exponent of x ranges from emin to emax, and precision of x is p.
// Adapted to Sollya from the Maple function in
// J-M. Muller, "Elementary Functions", 3rd ed, Section 11.3.2.
//
// Some examples:
//
// 1) Worst case for trig range reduction fast passes:
//
// Single precision
// > WorstCase(24, -6, 32, pi/32, 128);
// numbermin : 10741887
// expmin : 7
// Worst case: 0x1.47d0fep30
// numberofdigits : -32.888
//
// Double precision
// > WorstCase(53, -8, 32, pi/128, 256);
// numbermin : 6411027962775774
// expmin : -53
// Worst case: 0x1.6c6cbc45dc8dep-1
// numberofdigits : -66.4867
//
// 2) Worst case for exponential function range reduction:
//
// Single precision
// > WorstCase(24, 1, 8, log(2), 128);
// numbermin : 2907270
// expmin : -22
// Worst case: 0x1.62e43p-1
// numberofdigits : -28.9678
//
// Double precision
// > WorstCase(53, 0, 11, log(2), 128);
// numbermin : 7804143460206699
// expmin : -51
// Worst case: 0x1.bb9d3beb8c86bp1
// numberofdigits : -57.4931
//
verbosity=0;
procedure WorstCase(p, emin, emax, C, ndigits) {
epsilonmin = 12345.0;
Digits = ndigits;

powerofBoverC = 2^(emin - p) / C;
for e from emin - p + 1 to emax - p + 1 do {
powerofBoverC = 2 * powerofBoverC;
a = floor(powerofBoverC);
Plast = a;
r = round(1/(powerofBoverC - a), ndigits, RN);
a = floor(r);
Qlast = 1;
Q = a;
P = Plast * a + 1;
while (Q < 2^p - 1) do {
r = round(1/(r - a), ndigits, RN);
a = floor(r);
NewQ = Q * a + Qlast;
NewP = P * a + Plast;
Qlast = Q;
Plast = P;
Q = NewQ;
P = NewP;
};
epsilon = C * abs(Plast - Qlast * powerofBoverC);
if (epsilon < epsilonmin) then {
epsilonmin = epsilon;
numbermin = Qlast;
expmin = e;
};
};
display=decimal!;
print("numbermin : ", numbermin);
print("expmin : ", expmin);
display=hexadecimal!;
print("Worst case: ", numbermin * 2^expmin);
display=decimal!;
ell = round(log2(epsilonmin), ndigits, RN);
print("numberofdigits : ", ell);
};
Loading