When a frame is dropped, the previous non-dropped frame is considered to be repeated, but the strip chart would compute new cached frame data for it, causing performance to be unusably slow at high frame rate data gathering even when most frames are skipped
This makes performance remotely usable in that situation
Menus would only update properly if there was a strip chart open that would cause the PStatView to be updated
Calling `check_update()` in `new_collector()` doesn't work properly because the views haven't been updated there yet; calling `do_update()` there does work around it but I think the better fix is just to make sure we always update the view so that the `check_update()` (which is also called by `idle()`) works properly
It changes the following things:
* Changes the counts in the PStatFrameData from uint16 to uint32, since I was hitting the value limit in some heavy frames with Python profiling enabled - no good reason for this limitation, so this allows removing it later
* Adds a T_expire_thread message, which I can use later to fix#450
These features are not used on the client side, but will be used on master. Adding these changes here now makes it possible to use a master client with a 1.10.13 version of the server (can be useful if you can't compile Panda on the host).
Similar to what we're doing on the flame graph now
Also, tweak tooltip message for collectors that overlap frame boundary if it's partially on a dropped frame, so that it isn't misleading
- Properly display and stack bars overlapping frame boundaries
- Ignore "App:Show code:General" collector (already disabled on master, but to support 1.10 clients) - it's an "inverted" collector that doesn't nest properly so it messes up the stacking
- Better handling / display of dropped frames
- Better handling of clock skew due to `PStatClient::resume_after_pause()` (but please don't use it, because it's not possible to fully handle it properly)
- Vertical scrolling (Windows only)
Upon reading data from datagram, the bars would animate from zero width, which looks odd when opening an existing session.
Also, fix the scale area not updating properly in GTK version.
Currently, the client doesn't send nested start/stop pairs, but there is no good reason not to handle this case in the server, and in some cases it is useful to send them (I will soon make changes on master to take advantage of this).
Client version 3.1 indicates that the client may send nested start/stop pairs.
It can be very useful to know how often a collector was invoked in a frame, not just how long it took. This adds a number to the upper-right corner showing exactly that (but not aggregated, just for leaf collectors).
Remove the use of set and list, which are allocator-heavy and insertion was a bottleneck. Since each sample occurs only once on the linked list, we can more efficiently roll our own linked list with next and prev pointers, so no allocation needed. Instead of the set, we can just store a per-collector flag.
* Windows stay open after client disconnects, for further inspection
* Ability to save session results to a file, and reopening those files
* Ability to save the current graph window layout for new sessions
* Ability to change colors (by right-clicking on bar)
* SI prefixes for Hz units (kHz, MHz, etc.)
* Ability to export session to Chrome Tracing JSON format
* "Close All Graphs" menu option
* Graphs now properly show data when opened while Pause is on
* Some fixes for weird graph window minimize behavior on Windows
- New powerful scrolling Timeline view for seeing all time events across all threads
- Redo flame graph to use stack-based nesting rather than the standard collector nesting
- Rewrite flame graph drawing to not use labels
- Status bar appears in main window showing top-level level collectors; double-clicking them brings up their chart and right-clicking them shows their children
- Context menus are added when right-clicking labels and charts
- Tooltips now appear when mouse hovers over collector area in a chart
- Strip chart windows now automatically determine the appropriate scale better
- Graph menus redone to allow opening flame chart anywhere as well as strip chart
- Instead of just ms everywhere, also use s / us / ns where appropriate
- Don't disable smoothing right away on mouse down on strip chart, only after dragging
- Windows: The MDI child windows are quite ugly and overlap with the status bar, so instead they are now top-level windows, but some code is added to make them spawn inside and move with the parent window, and minimize to its corner. I can back this out if people prefer the old behavior despite the ugly decoration
- Windows: Label text shows ellipsis when cut off
- Windows: Graph windows no longer have icons
- Windows: Graph windows no longer spawn perfectly on top of each other, rather cascading
- GTK: Render at high resolution when GDK_SCALE is not 1
- GTK: Graph windows are forced to be floating in tiling WMs
- GTK: Flame chart window no longer has useless dividing bar
- GTK: Use more efficient cairo surface types
* New "Flame Graph" chart for seeing all collectors in a frame, much easier to read than piano roll
* Update controls, fonts, background color to more modern visual style on Windows
* Proper support for high DPI monitors (with correct scaling)
* Add tooltips for collector labels showing full name and averaged value
* Colors of collectors are now converted to sRGB transfer encoding
* Major performance improvement to piano roll view on Windows
* Movering mouse over labels now highlights the corresponding area in chart
* Label hover effect changed to darkening effect instead of border
* Reimplement graph as static common control on Windows
* Check boxes are now clickable by their label on Windows
* Graph windows have minimum sizes on Windows
Useful for uniquely identifying the process in a situation where multiple clients on the same host connect with the same server.
New version bump is not necessary, as old servers should just ignore the extra field in the datagram.
Instead of wrapping the whole subdirectory's CMakeLists in
a gigantic if block, we use if(NOT HAVE_FOO) + return()
This is tidier since it keeps the indentation generally consistent
across all CMakeLists files.
This changes the target_link_libraries() declarations so they only
go directly to a component library if it's in the same metalib.
If it's outside of the metalib, the correct thing to do is link against
the metalib. CMake takes care of the transitive linking for us, so
there isn't actually any big change here.
rdb
Sam Edwards