Skip to content

Rounding improvements for log2 and log256 #106

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.

Sign up for GitHub

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
immrsd merged 2 commits into from
Nov 28, 2025
Merged

Rounding improvements for log2 and log256 #106

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
a89ec74
Improve rounding approach for log2 and log256
immrsd Nov 28, 2025
295b607
Merge main
immrsd Nov 28, 2025
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
30 changes: 14 additions & 16 deletions math/core/sources/internal/macros.move
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -252,7 +252,8 @@ public(package) macro fun mul_mod<$Int>($a: $Int, $b: $Int, $modulus: $Int): $In
/// - `$bit_width`: The bit width of the type (8, 16, 32, 64, 128, or 256).
///
/// #### Returns
/// The zero-based position of the most significant bit as a `u8`. Returns `0` if `$value` is 0.
/// The zero-based position of the most significant bit as a `u8`.
/// Returns `0` if `$value` is 0.
public(package) macro fun msb<$Int>($value: $Int, $bit_width: u16): u8 {
common::msb($value as u256, $bit_width)
}
Expand Down Expand Up @@ -291,10 +292,8 @@ public(package) macro fun log2<$Int>(
floor_log
} else if (rounding_mode == rounding::up()) {
floor_log + 1
} else if (log2_should_round_up(value, floor_log)) {
floor_log + 1
} else {
floor_log
round_log2_to_nearest(value, floor_log)
}
}

Expand All @@ -312,7 +311,7 @@ public(package) macro fun log2<$Int>(
/// - `$rounding_mode`: Rounding strategy drawn from `rounding::RoundingMode`.
///
/// #### Returns
/// The base-256 logarithm as a `u16`, rounded according to the specified mode.
/// The base-256 logarithm as a `u8`, rounded according to the specified mode.
/// Returns `0` if `$value` is 0.
public(package) macro fun log256<$Int>(
$value: $Int,
Expand All @@ -323,8 +322,8 @@ public(package) macro fun log256<$Int>(
if (value == 0) {
return 0
};
let floor_log2 = common::msb(value, bit_width) as u16;
let floor_log256 = (floor_log2 / 8) as u8;
let floor_log2 = common::msb(value, bit_width);
let floor_log256 = floor_log2 / 8;

if (rounding_mode == rounding::down()) {
floor_log256
Expand All @@ -333,10 +332,8 @@ public(package) macro fun log256<$Int>(
floor_log256
} else if (rounding_mode == rounding::up()) {
floor_log256 + 1
} else if (log256_should_round_up(value, floor_log256 as u16)) {
floor_log256 + 1
} else {
floor_log256
round_log256_to_nearest(value, floor_log256)
}
}

Expand Down Expand Up @@ -701,11 +698,11 @@ public(package) fun round_division_result(
///
/// #### Returns
/// `true` if the value should round up, `false` otherwise.
public(package) fun log2_should_round_up(value: u256, floor_log: u16): bool {
public(package) fun round_log2_to_nearest(value: u256, floor_log: u16): u16 {
let threshold_exp = 2 * floor_log + 1;
let max_small = std::u128::max_value!() as u256;
let fast_path = threshold_exp < 256 && value <= max_small;
if (fast_path) {
let should_round_up = if (fast_path) {
// Fast path: both value² and exponent fit in u256
let value_squared = value * value;
let threshold = 1 << (threshold_exp as u8);
Expand All @@ -720,7 +717,8 @@ public(package) fun log2_should_round_up(value: u256, floor_log: u16): bool {
u512::from_u256(1 << (threshold_exp as u8))
};
value_squared.ge(&threshold)
}
};
if (should_round_up) { floor_log + 1 } else { floor_log }
}

/// Nearest-integer rounding for log256 without floats.
Expand All @@ -743,12 +741,12 @@ public(package) fun log2_should_round_up(value: u256, floor_log: u16): bool {
///
/// #### Returns
/// `true` if the value should round up, `false` otherwise.
public(package) fun log256_should_round_up(value: u256, floor_log: u16): bool {
public(package) fun round_log256_to_nearest(value: u256, floor_log: u8): u8 {
// For u256 values, floor_log ∈ [0, 31], so `threshold_exp = 8 * floor_log + 4 ≤ 252`
// and the power-of-two threshold fits safely in u256.
let threshold_exp = 8 * floor_log + 4;
let threshold = 1 << (threshold_exp as u8);
value >= threshold
let threshold = 1 << threshold_exp;
if (value >= threshold) { floor_log + 1 } else { floor_log }
Comment thread
immrsd marked this conversation as resolved.
}

/// Apply nearest-integer rounding to log10 without floats.
Expand Down
Loading