landmarks 1.4 · OCaml Package

All about exporting profiling results

A callgraph is a tree where each node is an instance of a Landmark.landmark representing the entering and exiting of instrumented code in the execution path.

type id = int

Identifies nodes

type kind =

| Normal
(*
Usual landmarks
*)
| Root
(*
The special node that started with the profiling.
*)
| Counter
(*
Counters (see Landmark.counter)
*)
| Sampler
(*
Samplers (set Landmark.sampler)
*)

The kind of node.

type node = {

id : id;
(*
Unique identifier.
*)
kind : kind;
landmark_id : string;
(*
The node is an instance of this landmark.
*)
name : string;
(*
Name of the landmark (see Landmark.register).
*)
location : string;
(*
Location of the landmark (see Landmark.register).
*)
calls : int;
(*
Number of time this node was entered.
*)
time : float;
(*
Time (in cycles) spent between enter and exit.
*)
children : id list;
(*
The list of instances of landmarks that was entered while the node was opened.
*)
sys_time : float;
(*
Time (using Sys.time) spent between enter and exit.
*)
allocated_bytes : float;
(*
Gc.allocated_bytes between enter and exit.
*)
distrib : floatarray;
(*
For samplers only. The list of collected samples.
*)

}

The type exported view of a node.

Callgraph

type graph = {

nodes : node array;
label : string;
root : id;

}

The type of callgraphs.

val nodes : graph -> node list

Returns all nodes of a graph.

val root : graph -> node

Returns the root of a graph.

val children : graph -> node -> node list

Returns the children of node(children graph node is equivalent to List.map (node_of_id graph) node.children

val subgraph : graph -> node -> graph

Change the root of a graph

val label : graph -> node -> string

Returns a fully qualified name if it is needed.

val graph_of_nodes : ?label:string -> ?root:id -> node list -> graph

Build a graph from a list of nodes.

Traversal

val path_dfs : 
  (bool -> node list -> node -> unit) ->
  (node list -> node -> unit) ->
  graph ->
  unit

path_dfs f g graph traverses the graph in the depth-first fashion starting from the root. At each step we call f visited path v or g path v where v is the visited node and path is the path from the root that led us to that node. The function g is called when the visited node v belongs to path; it indicates a loop (and the traversal does not continue with the children of g). The function f is called when v does not belong to path. The flag visited is true when the vertex has already been visited.

val dfs : 
  (node list -> node -> bool) ->
  (node list -> node -> unit) ->
  graph ->
  unit

A specialization of path_dfs that does not need to read the visited flag. The returned values of the first function tells whether or not the traversal should continue visiting the children of the current node.

Utility functions

val depth : graph -> node -> int

Returns the depth to the root of the node (it is better to partially apply the function, if you need to call multiple times on the same graph).

val shallow_ancestor : graph -> node -> node

Returns the oldest ancestor of a node that is not the root (if it exists) or the root if it does not exist.

val intensity : ?proj:(node -> float) -> graph -> node -> float

Returns an arbitrary number between 0.0 and 1.0.

val total_number_of_calls : graph -> int

Compute the sum of all calls field.

Simplification / Merge / Quotienting.

val prune : graph -> graph

Remove the unaccessible nodes from a graph.

val aggregate_landmarks : graph -> graph

aggregate_landmarks g computes the quotient by the relation "being an instance of the same landmark".

Output

val output : ?threshold:float -> out_channel -> graph -> unit

Pretty printed output a call graph on an out_channel.

val output_json : out_channel -> graph -> unit

Output a JSON representation of a call graph on an out_channel.

package landmarks

Callgraph

Traversal

Utility functions

Simplification / Merge / Quotienting.

Output