GPU backend

Adds GPU backend and "CUDA" feature to all the tests. As of MR creation, there were no forks in tested IRs so only vanilla non-parallel (single block, single thread) codegen has been tested.

hercules_cg/src/fork_tree.rs

+use std::collections::{HashMap, HashSet};
+
+use crate::*;
+
+/* Construct a map from each fork node F to all forks satisfying:
+ * a) domination by F
+ * b) no domination by F's join
+ * c) no domination by any other fork that's also dominated by F, where we don't count self-domination
+ * Note that the fork_tree also includes the non-fork start node, as unique root node.
+ */
+pub fn fork_tree(function: &Function, fork_join_nesting: &HashMap<NodeID, Vec<NodeID>>) -> HashMap<NodeID, HashSet<NodeID>> {
+    let mut fork_tree = HashMap::new();
+    for (control, forks) in fork_join_nesting {
+        if function.nodes[control.idx()].is_fork() {
+            fork_tree.entry(*control).or_insert_with(HashSet::new);
+            let nesting_fork = forks.get(1).copied().unwrap_or(NodeID::new(0));
+            fork_tree.entry(nesting_fork).or_insert_with(HashSet::new).insert(*control);
+        }
+    }
+    fork_tree
+}
+
+/*
+ * Construct a map from fork node to all control nodes (including itself) satisfying:
+ * a) domination by F
+ * b) no domination by F's join
+ * c) no domination by any other fork that's also dominated by F, where we do count self-domination
+ * Here too we include the non-fork start node, as key for all controls outside any fork.
+ */
+pub fn fork_control_map(fork_join_nesting: &HashMap<NodeID, Vec<NodeID>>) -> HashMap<NodeID, HashSet<NodeID>> {
+    let mut fork_control_map = HashMap::new();
+    for (control, forks) in fork_join_nesting {
+        let fork = forks.first().copied().unwrap_or(NodeID::new(0));
+        fork_control_map.entry(fork).or_insert_with(HashSet::new).insert(*control);
+    }
+    fork_control_map
+}