(feat) changed indexing instrategy for traces and added less paralel …

…mode for profiling
TalusBio · Jul 22, 2024 · ec2f3c5 · ec2f3c5
1 parent c9b8993
commit ec2f3c5
Show file tree

Hide file tree

Showing 14 changed files with 694 additions and 245 deletions.
diff --git a/README.md b/README.md
@@ -91,6 +91,19 @@ IONMESH_PROFILE_NUM_WINDOWS
     # windows.
 ```
 
+### Profiling
+
+Mostly notes to self but someone else might benefit ...
+
+```
+# Making a cool flamegraph of a run
+sudo RUST_LOG=debug CARGO_PROFILE_RELEASE_DEBUG=true IONMESH_PROFILE_NUM_WINDOWS=2 \
+    cargo flamegraph \
+    --output flamegraph_secondversion.svg \
+    --features less_parallel \
+    -- --config ./benchmark/default_ionmesh_config.toml benchmark/${MYFAVORITEFILE}.d -o tmp/
+```
+
 ## Roadmap
 
 1. Use aggregation metrics to re-score sage search.

diff --git a/src/aggregation/aggregators.rs b/src/aggregation/aggregators.rs
@@ -181,33 +181,35 @@ fn parallel_aggregate_clusters<
 
     let mut cluster_vecs = out2.into_iter().flatten().collect::<Vec<_>>();
 
-    let unclustered_elems: Vec<usize> = cluster_labels
-        .iter()
-        .enumerate()
-        .filter(|(_, x)| match x {
-            ClusterLabel::Unassigned => true,
-            ClusterLabel::Noise => keep_unclustered,
-            _ => false,
-        })
-        .map(|(i, _elem)| i)
-        .collect();
+    if keep_unclustered {
+        let unclustered_elems: Vec<usize> = cluster_labels
+            .iter()
+            .enumerate()
+            .filter(|(_, x)| match x {
+                ClusterLabel::Unassigned => true, // Should there be any unassigned?
+                ClusterLabel::Noise => true,
+                ClusterLabel::Cluster(_) => false,
+            })
+            .map(|(i, _elem)| i)
+            .collect();
 
-    // if unclustered_elems.len() > 0 {
-    //     log::debug!("Total Orig elems: {}", cluster_labels.len());
-    //     log::debug!("Unclustered elems: {}", unclustered_elems.len());
-    //     log::debug!("Clustered elems: {}", cluster_vecs.len());
-    // }
+        // if unclustered_elems.len() > 0 {
+        //     log::debug!("Total Orig elems: {}", cluster_labels.len());
+        //     log::debug!("Unclustered elems: {}", unclustered_elems.len());
+        //     log::debug!("Clustered elems: {}", cluster_vecs.len());
+        // }
 
-    let unclustered_elems = unclustered_elems
-        .iter()
-        .map(|i| {
-            let mut oe = def_aggregator();
-            oe.add(&elements.get_aggregable_at_index(*i));
-            oe
-        })
-        .collect::<Vec<_>>();
+        let unclustered_elems = unclustered_elems
+            .iter()
+            .map(|i| {
+                let mut oe = def_aggregator();
+                oe.add(&elements.get_aggregable_at_index(*i));
+                oe
+            })
+            .collect::<Vec<_>>();
 
-    cluster_vecs.extend(unclustered_elems);
+        cluster_vecs.extend(unclustered_elems);
+    }
 
     timer.stop(true);
     cluster_vecs

diff --git a/src/aggregation/chromatograms.rs b/src/aggregation/chromatograms.rs
@@ -254,7 +254,12 @@ impl<
         let mut mag_a = T::default();
         let mut mag_b = T::default();
         for i in 0..NBINS {
-            let other_index = i + other_vs_self_offset as usize;
+            let other_index = i as i32 + other_vs_self_offset;
+            if other_index < 0 {
+                continue;
+            }
+
+            let other_index = other_index as usize;
             if other_index >= other.chromatogram.len() {
                 continue;
             }

diff --git a/src/aggregation/dbscan/dbscan.rs b/src/aggregation/dbscan/dbscan.rs
@@ -20,7 +20,7 @@ fn reassign_centroid<
     const N: usize,
     T: Send + Clone + Copy,
     C: NDPointConverter<R, N>,
-    I: QueriableIndexedPoints<'a, N> + std::marker::Sync,
+    I: QueriableIndexedPoints<N> + std::marker::Sync,
     G: Sync + Send + ClusterAggregator<T, R>,
     R: Send,
     RE: Send + Sync + AsAggregableAtIndex<T> + ?Sized,
@@ -167,7 +167,7 @@ pub fn dbscan_aggregate<
         + Sync
         + std::fmt::Debug
         + ?Sized,
-    IND: QueriableIndexedPoints<'a, N> + std::marker::Sync + Send + std::fmt::Debug,
+    IND: QueriableIndexedPoints<N> + std::marker::Sync + Send + std::fmt::Debug,
     NAI: AsNDPointsAtIndex<N> + std::marker::Sync + Send,
     T: HasIntensity + Send + Clone + Copy + Sync,
     D: Send + Sync,