cessen/psychopath
1
0
Fork 0
Code Issues 1 Pull Requests Projects Releases Wiki Activity

More code tidying of the Sobol sampler.

Browse Source 
In particular, clearer and more concise documentation.
This commit is contained in:
Nathan Vegdahl 2020-05-01 10:42:02 +09:00
parent 45241784fb
commit fd75a72655
5 changed files with 10 additions and 4114 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1024
sub_crates/sobol/direction_numbers/joe-kuo-other-0.1024.txt
View File

File diff suppressed because it is too large Load Diff

1024
sub_crates/sobol/direction_numbers/joe-kuo-other-2.1024.txt
View File

File diff suppressed because it is too large Load Diff

1024
sub_crates/sobol/direction_numbers/joe-kuo-other-3.1024.txt
View File

File diff suppressed because it is too large Load Diff

1024
sub_crates/sobol/direction_numbers/joe-kuo-other-4.1024.txt
View File

File diff suppressed because it is too large Load Diff

28
sub_crates/sobol/src/lib.rs
View File

@ -1,10 +1,7 @@
//! An implementation of the Sobol sequence with Owen scrambling. //! A seedable, Owen-scrambled Sobol sequence.
//! //!
//! This implementation also allows seeding to create multiple statistically //! This implementation is limited to `2^16` samples, and will loop back to
//! independent Sobol sequences, using a technique due to Brent Burley. This //! the start of the sequence after that limit.
//! is useful for any situation where you want to decorrelate sampling.
//!
//! This implementation is SIMD accelerated with SSE 4.1 on x86-64 platforms.
mod wide; mod wide;
use wide::Int4; use wide::Int4;
@ -15,21 +12,16 @@ include!(concat!(env!("OUT_DIR"), "/vectors.inc"));
pub const MAX_DIMENSION_SET: u32 = MAX_DIMENSION / 4; pub const MAX_DIMENSION_SET: u32 = MAX_DIMENSION / 4;
/// Compute four dimensions of a sample from the Owen-scrambled Sobol sequence. /// Compute four dimensions of a single sample in the Sobol sequence.
/// ///
/// `sample_index` specifies which sample in the Sobol sequence to compute /// `sample_index` specifies which sample in the Sobol sequence to compute.
/// the dimensions for. This implementation produces up to 2^16 samples total,
/// and will loop back to the start of the sequence after that.
/// ///
/// `dimension_set` specifies which four dimensions to compute: /// `dimension_set` specifies which four dimensions to compute. `0` yields the
/// * 0 = dimensions 0-3 /// first four dimensions, `1` the second four dimensions, and so on.
/// * 1 = dimensions 4-7
/// * 2 = dimensions 8-11
/// * And so on.
/// ///
/// Passing a different `seed` will produce a completely statistically /// `seed` produces statistically independent Sobol sequences. Passing two
/// independent Sobol sequence, with no stratification with or correlation to /// different seeds will produce two different sequences that are only randomly
/// sequences with other seeds. /// associated, with no stratification or correlation between them.
#[inline] #[inline]
pub fn sample_4d(sample_index: u32, dimension_set: u32, seed: u32) -> [f32; 4] { pub fn sample_4d(sample_index: u32, dimension_set: u32, seed: u32) -> [f32; 4] {
assert!(dimension_set < MAX_DIMENSION_SET); assert!(dimension_set < MAX_DIMENSION_SET);