Skip to content

Commit

Permalink
docs: Make flame graph docs reflect icicle default
Browse files Browse the repository at this point in the history
Signed-off-by: Matt Wozniski <[email protected]>
  • Loading branch information
tjmcewan authored and godlygeek committed Sep 25, 2023
1 parent c016ce5 commit 05e8d36
Showing 1 changed file with 68 additions and 50 deletions.
118 changes: 68 additions & 50 deletions docs/flamegraph.rst
Original file line number Diff line number Diff line change
Expand Up @@ -5,45 +5,78 @@ The flame graph reporter generates an HTML file containing a flame graph
representation of the allocations contributing to the tracked process's peak
memory usage.

Flame graphs are a way to visualize how your program is spending its time.
A few important things about flame graphs:
Flame graphs are a way to visualize where your program is spending its
memory. A few important things about flame graphs:

- The flame graph displays the superposition of all stack traces that
lead to memory allocations at a given time (normally the time when the
total amount of allocated memory was highest).
led to all memory allocations active at a given time (normally the
time when the total amount of allocated memory was highest).

- A stack trace is represented as a column of boxes, where each box
represents one function call in that call stack.

- The y-axis shows the stack depth. One special row represents the root,
the next row represents the top level function calls that eventually
led to memory being allocated, the next row represents things directly
called by each of those top level calls, and so on. The furthest box
from the root in any given vertical slice of the flame graph is the
function that actually allocated the memory, and all of the boxes
between it and the root tell you the full call stack that led to that
allocation.

- The x-axis **does not show the passage of time**, so the left-to-right
ordering has no special meaning. Two functions called by the same
caller will appear side by side in the flame graph, but can appear in
either order.

- The width of each box represents how much memory was allocated by that
function call or its children. Wider boxes led to more memory being
allocated than narrower ones, in proportion to their width.

- A flame graph can't tell you how many times a function was called,
only how much memory that function allocated.

- A stack trace is represented as a column of boxes, where each box
represents a function call.
Flames versus Icicles
---------------------

- The x-axis **does not show the passage of time**, so the left-to-right
ordering has no special meaning: every level just shows the collection
of functions that were called by the nodes immediately under them.
In what's traditionally called a flame graph, each function in a call
stack is shown directly above its caller, with the root at the bottom.
This is called a "flame graph" because the wide base with narrowing
columns above it looks sort of like a burning log with flames leaping
into the air above it.

By default, Memray instead generates what's sometimes called an "icicle
graph", which instead has the root at the top. In an icicle graph, each
function is below its caller, and there is a wide ceiling that thinner
columns descend from, like icicles hanging from a roof. We default to
the icicle representation because it plays more nicely with the
browser's scroll bar: the most information-dense portion of an icicle
graph is at the top of the screen, and the further down the page you
scroll, the sparser the graphed data becomes, as shallower call stacks
drop off.

You can switch between the flame graph orientation and the icicle graph
orientation with this toggle button:

- The y-axis shows the stack depth, ordered from root at the bottom to
leaf at the top. The top box shows the function that made a memory
allocation and everything beneath that is its ancestry. The function
beneath a function is its parent.
.. image:: _static/images/icicle_flame_toggle.png
:align: center

- The width of each function box represents how much memory was
allocated by that function or its children. Functions with wider boxes
allocated more bytes of memory than those with narrower boxes, in
proportion to their width.
Whichever of these modes you choose, the data shown in the table is the
same, just mirrored vertically.

Interpreting flame graphs
-------------------------

Flame graphs can be interpreted as follows:
Memray's default (icicle mode) flame graphs can be interpreted as
follows:

- The nodes at the bottom of the flame graph represent functions that
- The nodes at the bottom of the graph represent functions that
allocated memory.

- For quickly identifying the functions that allocated more memory
directly, look for large plateaus along the bottom edge, as these show
a single stack trace was responsible for a large chunk of the total
memory of the snapshot that the flame graph represents.
memory of the snapshot that the graph represents.

- Reading flame graphs from the bottom up shows ancestry relationships.
Every function is called by its parent, which is shown directly above
Expand All @@ -60,15 +93,19 @@ Flame graphs can be interpreted as follows:
mean more memory was allocated by the given node, so those are the
most important to understand first.

- Major forks in the flame graph (when a node splits into several ones in
the next level) can be useful to study: these nodes can indicate
a logical grouping of code, where a function processes work in stages,
each with its own function. It can also be caused by a conditional
statement, which chooses which function to call.
- Major forks in the graph (when a node splits into several ones in the
next level) can be useful to study: these nodes can indicate a logical
grouping of code, where a function processes work in stages, each with
its own function. It can also be caused by a conditional statement,
which chooses which function to call.

- If the application is multi-threaded, the stacks of all the threads
that contribute to the memory peak will appear commingled in the flame
graph by default.
that contribute to the memory peak will appear commingled in the graph
by default.

And of course, if you switch from the "icicle" view to the "flame" view,
the root jumps to the bottom of the page, and call stacks grow upwards
from it instead of downwards.

Simple example
--------------
Expand All @@ -90,7 +127,8 @@ Simple example
a(100000)
This code allocates memory from the system allocator in just 2 places:
``c()``, and ``d()``. This is how the flame graph looks:
``c()``, and ``d()``. This is how a flame graph (with the root at the
bottom) looks:

.. image:: _static/images/simple_example.png

Expand Down Expand Up @@ -138,7 +176,7 @@ A more complete example
a(100000)
This code allocates memory from the system allocator in 5 places:
``e()``, ``d()``, ``g()``, ``i()`` and ``missing()``. The associated
``d()``, ``e()``, ``g()``, ``i()`` and ``missing()``. The associated
flame graph looks like this:

.. image:: _static/images/complex_example.png
Expand All @@ -163,7 +201,7 @@ and ``h()``. Understanding why the code does this may be a major clue to
its logical organization. This may be the result of a conditional (if
conditional, call ``b()``, else call ``h()``) or a logical grouping of
stages (where ``a()`` is processed in two parts: ``b()`` and ``h()``).
In our case we know is the second case, as ``a()`` is creating a list
In our case we know it's the second case, as ``a()`` is creating a list
with the result of ``b()`` and ``h()``.

If you look carefully you can notice that ``missing()`` allocates
Expand Down Expand Up @@ -196,26 +234,6 @@ checkbox:
Note that allocations in these frames will still be accounted for
in parent frames, even if they're hidden.

Flames versus Icicles
---------------------

The flame graphs explained above show each function above its caller,
with the root at the bottom. This is what's traditionally called
a "flame graph", because the wide base with narrowing columns above it
looks sort of like a burning log with flames leaping into the air above
it. Memray also supports what's sometimes called an "icicle graph",
which has the root at the top. In an icicle graph, each function is
below its caller, and there is a wide ceiling that thinner columns
descend from, like icicles hanging from a roof. Whichever of these modes
you choose, the data shown in the table is the same, just mirrored
vertically.

You can switch between showing a flame graph and an icicle graph with
this toggle button:

.. image:: _static/images/icicle_flame_toggle.png
:align: center

.. _memory-leaks-view:

Memory Leaks View
Expand Down

0 comments on commit 05e8d36

Please sign in to comment.