From f8fa402a2a95bcd4ea1bbf5837e95df58a8867c6 Mon Sep 17 00:00:00 2001
From: Ayaka Yorihiro <ayakayorihiro@gmail.com>
Date: Fri, 10 Jan 2025 15:35:30 -0500
Subject: [PATCH] A script to generate a proof of concept flame graph

---
 tools/profiler/poc-lift-flame-graph.py | 100 +++++++++++++++++++++++++
 1 file changed, 100 insertions(+)
 create mode 100644 tools/profiler/poc-lift-flame-graph.py

diff --git a/tools/profiler/poc-lift-flame-graph.py b/tools/profiler/poc-lift-flame-graph.py
new file mode 100644
index 000000000..a2aad8ffe
--- /dev/null
+++ b/tools/profiler/poc-lift-flame-graph.py
@@ -0,0 +1,100 @@
+# a super simple script to show ADL level profiling proof of concept/experimentation
+import sys
+
+class Component:
+    def __init__(self, name, pos_id):
+        self.name = name
+        self.position_id = pos_id
+        self.position = None
+        self.cells = {} # cell_name --> pos_id
+        self.groups = {} # group_name --> pos_id
+
+    def rewrite(self, position_map):
+        # replace position IDs with the filename, linenum pair.
+        self.position = position_map[self.position_id]
+        for cell in self.cells:
+            pos_id = self.cells[cell]
+            self.cells[cell] = position_map[pos_id]
+        for group in self.groups:
+            pos_id = self.groups[group]
+            self.groups[group] = position_map[pos_id]
+
+    def __str__(self):
+        s = f"Component {self.name}:\n"
+        s += f"\tpos: {self.position_id}\n"
+        s += f"\tcells:\n"
+        for cell in self.cells:
+            s += f"\t\t{cell}: {self.cells[cell]}\n"
+        s += f"\tgroups:\n"
+        for group in self.groups:
+            s += f"\t\t{group}: {self.groups[group]}\n"
+        return s
+
+def parse(calyx_file):
+    # a really hacky parser.
+    metadata = False
+    file_map = {}
+    position_map = {}
+    components = {} # name --> Component
+    curr_component = None
+    with open(calyx_file, "r") as r:
+        for line in r:
+            line_strip = line.strip()
+            # start metadata
+            if line_strip.startswith("metadata #{"):
+                metadata = True
+                continue
+            if metadata:
+                if line_strip.startswith("file-"):
+                    line_split = line_strip.split("file-")[1].split(":")
+                    file_id = line_split[0]
+                    filename = line_split[1].strip()
+                    file_map[file_id] = filename
+                if line_strip.startswith("pos-"):
+                    line_split = line_strip.split("pos-")[1].split(":")
+                    position_id = line_split[0]
+                    rest_split = line_split[1].strip().strip("(").strip(")").replace(" ", "").split(",")
+                    file_id = rest_split[0]
+                    line_num = rest_split[1]
+                    position_map[position_id] = (file_map[file_id], line_num) # if the query fails sth went wrong
+            else:
+                # shoddy attempt at parsing an eDSL-generated Calyx file
+                if line_strip.startswith("component"):
+                    if curr_component is not None: # store the previous component
+                        components[curr_component.name] = curr_component
+                    name = line_strip.split(" ")[1].split("<")[0]
+                    pos_id = line_strip.split('"pos"=')[1].split(">")[0]
+                    curr_component = Component(name, pos_id)
+                elif line_strip.startswith("group "):
+                    groupname = line_strip.split("<")[0].split(" ")[-1]
+                    pos_id = line_strip.split('"pos"=')[1].split(">")[0]
+                    if groupname in curr_component.groups:
+                        raise RuntimeError(f"Group {groupname} already recorded in component {curr_component.name}")
+                    curr_component.groups[groupname] = pos_id
+                elif "@pos" in line_strip: # currently the other thing that has positions is cells. probably will change in the future
+                    cellname = line_strip.split("=")[0].strip().split(" ")[-1]
+                    pos_id = line_strip.split("@pos(")[1].split(")")[0]
+                    if cellname in curr_component.cells:
+                        raise RuntimeError(f"Cell {cellname} already recorded in component {curr_component.name}")
+                    curr_component.cells[cellname] = pos_id
+
+    return components, position_map
+
+def main(calyx_file):
+    components, position_map = parse(calyx_file)
+    for component_name in components:
+        component = components[component_name]
+        component.rewrite(position_map)
+        print(component)
+        print()
+
+if __name__ == "__main__":
+    if len(sys.argv) > 1:
+        calyx_file = sys.argv[1]
+        main(calyx_file)
+    else:
+        args_desc = [
+            "CALYX_FILE"
+        ]
+        print(f"Usage: {sys.argv[0]} {' '.join(args_desc)}")
+        sys.exit(-1)