add spmm example and change the kernel

This commit is contained in:
Saurabh 2022-02-15 17:38:20 -07:00
parent f637013aa0
commit e7d8a00836
2 changed files with 54 additions and 18 deletions

View File

@ -24,6 +24,7 @@ io = [ "pest", "pest_derive" ]
slow-tests = []
[dependencies]
criterion = { version = "0.3", features = ["html_reports"] }
nalgebra = { version="0.30", path = "../" }
num-traits = { version = "0.2", default-features = false }
proptest = { version = "1.0", optional = true }
@ -31,6 +32,7 @@ matrixcompare-core = { version = "0.1.0", optional = true }
pest = { version = "2", optional = true }
pest_derive = { version = "2", optional = true }
serde = { version = "1.0", default-features = false, features = [ "derive" ], optional = true }
itertools = "0.10"
[dev-dependencies]
itertools = "0.10"
@ -38,6 +40,11 @@ matrixcompare = { version = "0.3.0", features = [ "proptest-support" ] }
nalgebra = { version="0.30", path = "../", features = ["compare"] }
serde_json = "1.0"
[[example]]
name = "spmm"
required-features = ["io"]
path = "example/spmm.rs"
[package.metadata.docs.rs]
# Enable certain features when building docs for docs.rs
features = [ "proptest-support", "compare" ]

View File

@ -1,16 +1,19 @@
use std::collections::HashSet;
use crate::cs::CsMatrix;
use crate::ops::serial::{OperationError, OperationErrorKind};
use crate::ops::Op;
use crate::SparseEntryMut;
use itertools::Itertools;
use nalgebra::{ClosedAdd, ClosedMul, DMatrixSlice, DMatrixSliceMut, Scalar};
use num_traits::{One, Zero};
fn spmm_cs_unexpected_entry() -> OperationError {
OperationError::from_kind_and_message(
OperationErrorKind::InvalidPattern,
String::from("Found unexpected entry that is not present in `c`."),
)
}
//fn spmm_cs_unexpected_entry() -> OperationError {
// OperationError::from_kind_and_message(
// OperationErrorKind::InvalidPattern,
// String::from("Found unexpected entry that is not present in `c`."),
// )
//}
/// Helper functionality for implementing CSR/CSC SPMM.
///
@ -32,28 +35,54 @@ where
{
for i in 0..c.pattern().major_dim() {
let a_lane_i = a.get_lane(i).unwrap();
let some_val = Zero::zero();
let mut scratchpad_values: Vec<T> = vec![some_val; b.pattern().minor_dim()];
let mut scratchpad_indices: HashSet<usize> = HashSet::new();
let mut c_lane_i = c.get_lane_mut(i).unwrap();
for c_ij in c_lane_i.values_mut() {
*c_ij = beta.clone() * c_ij.clone();
}
//let (indices, values) = c_lane_i.indices_and_values_mut();
//indices
// .iter()
// .zip(values.iter())
// .for_each(|(id, val)| scratchpad_values[*id] = beta.clone() * val.clone());
//for (index, c_ij) in c_lane_i.indices_and_values_mut() {
// *c_ij = beta.clone() * c_ij.clone();
//}
for (&k, a_ik) in a_lane_i.minor_indices().iter().zip(a_lane_i.values()) {
let b_lane_k = b.get_lane(k).unwrap();
let (mut c_lane_i_cols, mut c_lane_i_values) = c_lane_i.indices_and_values_mut();
//let (mut c_lane_i_cols, mut c_lane_i_values) = c_lane_i.indices_and_values_mut();
let alpha_aik = alpha.clone() * a_ik.clone();
for (j, b_kj) in b_lane_k.minor_indices().iter().zip(b_lane_k.values()) {
// Determine the location in C to append the value
let (c_local_idx, _) = c_lane_i_cols
.iter()
.enumerate()
.find(|(_, c_col)| *c_col == j)
.ok_or_else(spmm_cs_unexpected_entry)?;
// TODO make a scratchpad and defer the accumulation into C after processing one
// full row of A.
scratchpad_values[*j] += alpha_aik.clone() * b_kj.clone();
scratchpad_indices.insert(*j);
//let (c_local_idx, _) = c_lane_i_cols
// .iter()
// .enumerate()
// .find(|(_, c_col)| *c_col == j)
// .ok_or_else(spmm_cs_unexpected_entry)?;
c_lane_i_values[c_local_idx] += alpha_aik.clone() * b_kj.clone();
c_lane_i_cols = &c_lane_i_cols[c_local_idx..];
c_lane_i_values = &mut c_lane_i_values[c_local_idx..];
//c_lane_i_values[c_local_idx] += alpha_aik.clone() * b_kj.clone();
//c_lane_i_cols = &c_lane_i_cols[c_local_idx..];
//c_lane_i_values = &mut c_lane_i_values[c_local_idx..];
}
}
// sort the indices, and then access the relevant indices (in sorted order) from values
// into C.
let sorted_indices: Vec<usize> =
Itertools::sorted(scratchpad_indices.into_iter()).collect();
c_lane_i
.values_mut()
.iter_mut()
.zip(sorted_indices.into_iter())
.for_each(|(output_ref, index)| {
*output_ref = beta.clone() * output_ref.clone() + scratchpad_values[index].clone()
});
}
Ok(())