perf: use passthrough hasher and preallocate capacity for outputs
This commit is contained in:
1 parent
b6ff1f81e6
commit
c02740798b
4 files changed
+52
-22
No files matched your search
+2
-2
@@ -176,7 +176,7 @@ fn test_deps_tuple_large_tuples() {
|
||||
let node_ids = (&handles[0], &handles[1], &handles[2], &handles[3]).to_node_ids();
|
||||
assert_eq!(node_ids.len(), 4);
|
||||
for (i, node_id) in node_ids.iter().enumerate().take(4) {
|
||||
assert_eq!(*node_id, NodeId(i));
|
||||
assert_eq!(*node_id, NodeId(i as u32));
|
||||
}
|
||||
|
||||
// Test 8-tuple
|
||||
@@ -193,7 +193,7 @@ fn test_deps_tuple_large_tuples() {
|
||||
.to_node_ids();
|
||||
assert_eq!(node_ids.len(), 8);
|
||||
for (i, node_id) in node_ids.iter().enumerate().take(8) {
|
||||
assert_eq!(*node_id, NodeId(i));
|
||||
assert_eq!(*node_id, NodeId(i as u32));
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
Reference in new issue
Block a user