Implement spadd_build_pattern

This commit is contained in:
Andreas Longva 2020-12-04 14:13:07 +01:00
parent 7c68950614
commit 4420237ede
5 changed files with 120 additions and 4 deletions

View File

View File

@ -3,7 +3,7 @@ use crate::ops::{Transpose};
use nalgebra::{Scalar, DMatrixSlice, ClosedAdd, ClosedMul, DMatrixSliceMut}; use nalgebra::{Scalar, DMatrixSlice, ClosedAdd, ClosedMul, DMatrixSliceMut};
use num_traits::{Zero, One}; use num_traits::{Zero, One};
/// Sparse-dense matrix-matrix multiplication `C = beta * C + alpha * trans(A) * trans(B)`. /// Sparse-dense matrix-matrix multiplication `C <- beta * C + alpha * trans(A) * trans(B)`.
pub fn spmm_csr_dense<'a, T>(c: impl Into<DMatrixSliceMut<'a, T>>, pub fn spmm_csr_dense<'a, T>(c: impl Into<DMatrixSliceMut<'a, T>>,
beta: T, beta: T,
alpha: T, alpha: T,

View File

@ -32,6 +32,8 @@ macro_rules! assert_compatible_spmm_dims {
mod coo; mod coo;
mod csr; mod csr;
mod pattern;
pub use coo::*; pub use coo::*;
pub use csr::*; pub use csr::*;
pub use pattern::*;

View File

@ -0,0 +1,77 @@
use crate::pattern::SparsityPattern;
use std::mem::swap;
use std::iter;
/// Sparse matrix addition pattern construction, `C <- A + B`.
///
/// Builds the pattern for `C`, which is able to hold the result of the sum `A + B`.
/// The patterns are assumed to have the same major and minor dimensions. In other words,
/// both patterns `A` and `B` must both stem from the same kind of compressed matrix:
/// CSR or CSC.
/// TODO: Explain that output pattern is only used to avoid allocations
pub fn spadd_build_pattern(pattern: &mut SparsityPattern,
a: &SparsityPattern,
b: &SparsityPattern)
{
// TODO: Proper error messages
assert_eq!(a.major_dim(), b.major_dim());
assert_eq!(a.minor_dim(), b.minor_dim());
let input_pattern = pattern;
let mut temp_pattern = SparsityPattern::new(a.major_dim(), b.minor_dim());
swap(input_pattern, &mut temp_pattern);
let (mut offsets, mut indices) = temp_pattern.disassemble();
offsets.clear();
offsets.reserve(a.major_dim() + 1);
indices.clear();
offsets.push(0);
for lane_idx in 0 .. a.major_dim() {
let lane_a = a.lane(lane_idx);
let lane_b = b.lane(lane_idx);
indices.extend(iterate_intersection(lane_a, lane_b));
offsets.push(indices.len());
}
// TODO: Consider circumventing format checks? (requires unsafe, should benchmark first)
let mut new_pattern = SparsityPattern::try_from_offsets_and_indices(
a.major_dim(), a.minor_dim(), offsets, indices)
.expect("Pattern must be valid by definition");
swap(input_pattern, &mut new_pattern);
}
/// Iterate over the intersection of the two sets represented by sorted slices
/// (with unique elements)
fn iterate_intersection<'a>(mut sorted_a: &'a [usize],
mut sorted_b: &'a [usize]) -> impl Iterator<Item=usize> + 'a {
// TODO: Can use a kind of simultaneous exponential search to speed things up here
iter::from_fn(move || {
if let (Some(a_item), Some(b_item)) = (sorted_a.first(), sorted_b.first()) {
let item = if a_item < b_item {
sorted_a = &sorted_a[1 ..];
a_item
} else if b_item < a_item {
sorted_b = &sorted_b[1 ..];
b_item
} else {
// Both lists contain the same element, advance both slices to avoid
// duplicate entries in the result
sorted_a = &sorted_a[1 ..];
sorted_b = &sorted_b[1 ..];
a_item
};
Some(*item)
} else if let Some(a_item) = sorted_a.first() {
sorted_a = &sorted_a[1..];
Some(*a_item)
} else if let Some(b_item) = sorted_b.first() {
sorted_b = &sorted_b[1..];
Some(*b_item)
} else {
None
}
})
}

View File

@ -1,8 +1,9 @@
use nalgebra_sparse::coo::CooMatrix; use nalgebra_sparse::coo::CooMatrix;
use nalgebra_sparse::ops::serial::{spmv_coo, spmm_csr_dense}; use nalgebra_sparse::ops::serial::{spmv_coo, spmm_csr_dense, spadd_build_pattern};
use nalgebra_sparse::ops::{Transpose}; use nalgebra_sparse::ops::{Transpose};
use nalgebra_sparse::csr::CsrMatrix; use nalgebra_sparse::csr::CsrMatrix;
use nalgebra_sparse::proptest::csr; use nalgebra_sparse::proptest::{csr, sparsity_pattern};
use nalgebra_sparse::pattern::SparsityPattern;
use nalgebra::{DVector, DMatrix, Scalar, DMatrixSliceMut, DMatrixSlice}; use nalgebra::{DVector, DMatrix, Scalar, DMatrixSliceMut, DMatrixSlice};
use nalgebra::proptest::matrix; use nalgebra::proptest::matrix;
@ -10,6 +11,7 @@ use nalgebra::proptest::matrix;
use proptest::prelude::*; use proptest::prelude::*;
use std::panic::catch_unwind; use std::panic::catch_unwind;
use std::sync::Arc;
#[test] #[test]
fn spmv_coo_agrees_with_dense_gemv() { fn spmv_coo_agrees_with_dense_gemv() {
@ -99,6 +101,19 @@ fn trans_strategy() -> impl Strategy<Value=Transpose> + Clone {
proptest::bool::ANY.prop_map(Transpose) proptest::bool::ANY.prop_map(Transpose)
} }
fn pattern_strategy() -> impl Strategy<Value=SparsityPattern> {
sparsity_pattern(0 ..= 6usize, 0..= 6usize, 40)
}
/// Constructs pairs (a, b) where a and b have the same dimensions
fn spadd_build_pattern_strategy() -> impl Strategy<Value=(SparsityPattern, SparsityPattern)> {
pattern_strategy()
.prop_flat_map(|a| {
let b = sparsity_pattern(Just(a.major_dim()), Just(a.minor_dim()), 40);
(Just(a), b)
})
}
/// Helper function to help us call dense GEMM with our transposition parameters /// Helper function to help us call dense GEMM with our transposition parameters
fn dense_gemm<'a>(c: impl Into<DMatrixSliceMut<'a, i32>>, fn dense_gemm<'a>(c: impl Into<DMatrixSliceMut<'a, i32>>,
beta: i32, beta: i32,
@ -167,4 +182,26 @@ proptest! {
"The SPMM kernel executed successfully despite mismatch dimensions"); "The SPMM kernel executed successfully despite mismatch dimensions");
} }
#[test]
fn spadd_build_pattern_test((c, (a, b)) in (pattern_strategy(), spadd_build_pattern_strategy()))
{
// (a, b) are dimensionally compatible patterns, whereas c is an *arbitrary* pattern
let mut pattern_result = c.clone();
spadd_build_pattern(&mut pattern_result, &a, &b);
// To verify the pattern, we construct CSR matrices with positive integer entries
// corresponding to a and b, and convert them to dense matrices.
// The sum of these dense matrices will then have non-zeros in exactly the same locations
// as the result of "adding" the sparsity patterns
let a_csr = CsrMatrix::try_from_pattern_and_values(Arc::new(a.clone()), vec![1; a.nnz()])
.unwrap();
let a_dense = DMatrix::from(&a_csr);
let b_csr = CsrMatrix::try_from_pattern_and_values(Arc::new(b.clone()), vec![1; b.nnz()])
.unwrap();
let b_dense = DMatrix::from(&b_csr);
let c_dense = a_dense + b_dense;
let c_csr = CsrMatrix::from(&c_dense);
prop_assert_eq!(&pattern_result, &*c_csr.pattern());
}
} }