mirror/tracy
1
0
Fork 0
mirror of https://github.com/wolfpld/tracy.git synced 2024-09-20 05:42:18 +00:00
Code Issues Packages Projects Releases Wiki Activity
6,766 Commits 2 Branches 28 Tags 40 MiB
4347cc58b1
Commit Graph

1189 Commits

Author SHA1 Message Date
Bartosz Taudul
46d2977694
prev_state is long. 2022-02-14 22:34:20 +01:00
Bartosz Taudul
9f2ffb05ac
Check if memory can be allocated in a thread. 
Consider running the following code with operator new and delete overloaded to
track allocations with call stacks:

std::thread( []({ thread_local std::string str; });

Each call stack requires a memory allocation to be performed by the profiler,
to make the stack available at a later time. When the thread is created, the
TLS block is initialized and the std::string buffer can be allocated. To track
this allocation, rpmalloc has to be initialized. This initialization also
happens within the TLS block.

Now, when the thread exits, the heap managed by rpmalloc may be released first
during the TLS block destruction (and if the destruction is performed in
reverse creation order, then it *will* be destroyed first, as rpmalloc was
initialized only after the std::string initialization, to track the allocation
performed within). The next thing to happen is destruction of std::string and
release of the memory block it contains.

The release is tracked by the profiler, and as mentioned earlier, to save the
call stack for later use, a memory allocation is needed. But the allocator is
no longer available in this thread, because rpmalloc was released just before!

As a solution to this issue, profiler will detect whether the allocator is
still available and will ignore the call stack, if it's not. The other
solution is to disable the rpmalloc thread cleanup, which may potentially
cause leak-like behavior, in case a large number of threads is spawned and
destroyed.

Note that this is not a water-tight solution. Other functions will still want
to allocate memory for call stacks, but it is rather unlikely that such calls
would be performed during TLS block destruction. It is also possible that the
event queue will run out of allocated space for events at this very moment,
and in such a case the allocator will also fail.
 2022-02-14 17:55:46 +01:00
Bartosz Taudul
793e6b0e29
Add rpmalloc thread state accessor. 2022-02-14 17:53:27 +01:00
Bartosz Taudul
20680fa6a3
Keep track of rpmalloc thread shutdown state. 2022-02-14 17:52:33 +01:00
Balazs Kovacsics
aa44cedcd2 Fix indentation 2022-02-11 17:14:16 +01:00
Balazs Kovacsics
a0a56eaf9e Move TRACY_UWP define to own header 2022-02-10 20:26:45 +01:00
Balazs Kovacsics
7169f5f30a Fix compiling TracyClient.cpp on UWP with defined TRACY_ENABLE 2022-02-09 22:14:36 +01:00
Bartosz Taudul
f058ad01fc
Don't process ring buffers when not connected in on-demand mode. 2022-01-31 20:53:10 +01:00
hulakdar
e98f7e5019 Fix potentially unitialized value warning 2022-01-26 21:08:54 +02:00
hulakdar
9fda60820a Silence padding warnings, as this is intended behaviour 2022-01-26 21:08:54 +02:00
Bartosz Taudul
4731bb9ec1
Cast GetProcAddress() ptr to const void*. 2022-01-05 20:03:20 +01:00
Bartosz Taudul
621c68352b
Call RtlWalkFrameChain directly from inlined function. 2022-01-01 17:33:39 +01:00
Bartosz Taudul
e368266837
Add custom demangler interface. 2021-12-22 18:17:48 +01:00
Bartosz Taudul
2f0d4a7707
Merge pull request #294 from Lectem/symbols_post_init 
Call SymLoadModuleEx for modules loaded after init (Fixes #293)
 2021-12-22 16:36:08 +01:00
Bartosz Taudul
29dc2fadf9
Do not bail-out on software sampling failure. 2021-12-21 22:09:40 +01:00
Bartosz Taudul
f1fa352a13
/sys/kernel/debug/tracing/... files are accessible on android. 2021-12-21 21:10:17 +01:00
Bartosz Taudul
efae847690
Replace getline(). 2021-12-21 20:28:18 +01:00
Bartosz Taudul
7f10265e34
Keep ring buffer size unsigned. 2021-12-21 20:24:52 +01:00
Bartosz Taudul
81ae7c06b8
Android libstdc++ is really stupid, it seems. 2021-12-21 19:56:32 +01:00
Bartosz Taudul
5741bcfd32
Make ring buffer size adjustable. 
If call stack capture is enabled for context switch data, the 64KB buffer is
too small to work without overruns. However, if the default buffer size is
increased, then the maximum locked memory limit is hit.

This change keeps the small buffer size for all the buffers that may be used
without escalated privileges. The context switch buffer is bigger, but it does
not need to obey the limits, as the application is running as root, if it is
to be used.
 2021-12-21 15:48:40 +01:00
Bartosz Taudul
d88bf2c7a8
Separate message type for context switch callstack samples. 2021-12-21 14:05:07 +01:00
Bartosz Taudul
0ba2c0a86e
Collect context switch callstack on Linux. 2021-12-21 13:59:16 +01:00
Bartosz Taudul
0c041795b6
Move Linux calltrace processing to a separate function. 2021-12-21 13:58:43 +01:00
Bartosz Taudul
4a89a30556
Implement checking if any TID is within current process. 2021-12-21 13:52:52 +01:00
Bartosz Taudul
b8d7824a4f
Increase ring buffer size. 2021-12-21 13:27:08 +01:00
Lectem
5d0466b729 Rename GetModuleName and call it before sym* use 2021-12-12 14:32:24 +01:00
Lectem
17855cbac5 Call SymLoadModuleEx for modules loaded after init 
This fixes issue #293. Symbols are not loaded if the module is loaded dynamically after the SymInitialize call.
This may stall the symbol resolver thread a bit the first time a module is loaded.
 2021-12-08 21:30:06 +01:00
Bartosz Taudul
14cdf3d43e
Use perf_event_open for context switch tracking. 2021-12-05 12:02:28 +01:00
Bartosz Taudul
0a3da9efdb
Optional CPU idx storage in RingBuffer. 2021-12-05 03:03:14 +01:00
Bartosz Taudul
6a0d233613
Proper mmap failure detection. 2021-12-05 01:32:33 +01:00
Bartosz Taudul
22284d57e7
Don't identify samples if there will be no context switches. 2021-12-05 01:32:30 +01:00
Nathan Day
bdafd0c7b6 Added IsConnected C API Equivalent 2021-12-04 16:55:41 +00:00
Bartosz Taudul
20f8c8b8a4
Add IdentifySamples flag to the protocol. 
On Windows there is no way to distinguish callstack data coming from random
sampling and from context switches. Each callstack timestamp has to be matched
against the context switch data in order to decide its origin. This is
obviously non-trivial.

On some other platforms, the origin information may be available right away,
in which case the process of matching against the context switch data, which
possibly includes postponing callstacks for processing in the future, may be
completely omitted.
 2021-12-04 14:50:25 +01:00
Bartosz Taudul
2008c88983
Silence compiler warnings. 2021-12-04 13:41:55 +01:00
Bartosz Taudul
74e2a63516
Don't require time to be available for samples. 2021-12-02 01:29:22 +01:00
Bartosz Taudul
051b1ba8f5
Display sampling disabled message on debug log channel. 2021-12-02 00:49:01 +01:00
slembcke
87a727dd7f Add fibers to TracyC.h 2021-11-26 16:49:39 -06:00
Bartosz Taudul
d41447b0c9
Pass full module path to LoadLibraryEx(). 2021-11-25 23:27:20 +01:00
Bartosz Taudul
a27d1902ae
Store paths to kernel modules. 2021-11-25 23:27:05 +01:00
Bartosz Taudul
8428ede59d
Retrieve kernel symbols on Windows. 2021-11-25 23:12:41 +01:00
Bartosz Taudul
4bde9022e1
Properly acknowledge unavailable symbol code. 2021-11-25 22:44:56 +01:00
Bartosz Taudul
e8fb2abb2a
Dispatch kernel code retrieval. 2021-11-25 22:27:35 +01:00
Bartosz Taudul
01db142144
Rpmalloc is initialized here. 2021-11-25 22:23:28 +01:00
Bartosz Taudul
1ed174c1b6
Expose GetModuleName() outside TracyCallstack.cpp. 2021-11-25 21:56:13 +01:00
Bartosz Taudul
2bd1e5bca7
Merge pull request #283 from expenses/c-gpu-zones 
Implement a bare-bones C API for graphics profiling
 2021-11-24 12:50:20 +01:00
Ashley Ruglys
2db109ab98
Add serial functions 2021-11-24 11:56:43 +01:00
Ashley Ruglys
2995db4b5e
Remove the 'active' argument 2021-11-24 11:45:37 +01:00
Bartosz Taudul
cad65ab52f
Handle magic dbghelp line numbers. 
https://github.com/Microsoft/microsoft-pdb/issues/16
https://docs.microsoft.com/en-us/archive/blogs/jmstall/line-hidden-and-0xfeefee-sequence-points
 2021-11-20 17:09:20 +01:00
Bartosz Taudul
bbbd78883c
Allow crash signal selection by user. 2021-11-18 20:20:10 +01:00
Ashley Ruglys
bf1896f6ba
memset the thread to 0 when writing the GpuZoneEnd as this seems to prevent an assert failure 2021-11-17 18:14:57 +01:00
Ashley Ruglys
cc45865239
Small formatting changes 2021-11-17 11:31:02 +01:00
Ashley Ruglys
91b4058dc4
Implement bare-bones functionality for gpu contexts in the C api 2021-11-17 11:27:52 +01:00
Bartosz Taudul
590a4afb89
Full dequeue is needed to handle symbol queries. 2021-11-14 19:24:05 +01:00
Bartosz Taudul
a5f5dcc00f
Remove capacity safeties. 2021-11-14 18:52:57 +01:00
Bartosz Taudul
5306034893
Remove unused methods. 2021-11-14 18:52:27 +01:00
Bartosz Taudul
9055a31156
Decrease number of SPSCQueue items to 8K. 
This matches the query limit set on the server.
 2021-11-14 18:51:11 +01:00
Bartosz Taudul
a203eb5aef
Adapt to SPSCQueue interface. 2021-11-14 18:50:59 +01:00
Bartosz Taudul
93f83fc50c
Remove allocator silliness. 2021-11-14 18:42:07 +01:00
Bartosz Taudul
fe04cf0b35
Move SPSCQueue to tracy namespace. 2021-11-14 18:31:58 +01:00
Bartosz Taudul
20e3ec92b9
Replace readerwriterqueue with SPSCQueue. 2021-11-14 18:27:25 +01:00
Bartosz Taudul
97c519811c
Capture context switch call stacks on Windows. 2021-11-12 12:22:29 +01:00
Bartosz Taudul
987f70a9d4
Signal initialization failure through exit code. 2021-11-11 12:43:59 +01:00
Bartosz Taudul
765561de81
Serialize continuous frame events. 2021-11-05 01:13:24 +01:00
Nathan Day
4fece233d9 Added unused parameters when TRACY_HAS_CALLSTACK is undefined 2021-11-04 15:24:58 +00:00
Bartosz Taudul
887826a28a
Remove dead code. 2021-11-03 18:58:53 +01:00
Bartosz Taudul
fceb065329
Fiber name transport. 2021-11-03 18:57:30 +01:00
Bartosz Taudul
229d547e91
Add fiber support on client side. 2021-11-02 00:11:09 +01:00
Bartosz Taudul
8b15cd049c
Skip comparing known to be matching memory. 2021-10-30 01:26:31 +02:00
Bartosz Taudul
1ecdb8aa2a
Report device name on Android. 2021-10-26 18:09:56 +02:00
Bartosz Taudul
bd108220cb
Drain symbols queue if on-demand and not connected. 2021-10-23 01:54:38 +02:00
Bartosz Taudul
3d13ec7018
Pump symbols queue during profiler termination. 2021-10-23 01:47:04 +02:00
Bartosz Taudul
fe9c63b8d2
Signal symbol resolution thread destruction. 2021-10-23 01:46:00 +02:00
Bartosz Taudul
02ce9b7d8b
Extract symbol queue item handling. 2021-10-23 01:35:01 +02:00
Bartosz Taudul
2b9265cc40
Move symbol resolution to a separate thread. 2021-10-22 22:25:09 +02:00
Bartosz Taudul
b4cfb1f2d5
Add SPSC symbol queue. 2021-10-22 22:22:27 +02:00
Bartosz Taudul
f2b0aadedc
Return external names, instead of sending them. 
Note that the old function is still used by code, so this commit won't
compile.
 2021-10-22 22:20:30 +02:00
Bartosz Taudul
af5da6f53c
Move inline string copy helpers to a separate header. 2021-10-22 21:53:46 +02:00
Bartosz Taudul
75f77a068f
Use rpmalloc in readerwriterqueue. 2021-10-22 20:39:10 +02:00
Bartosz Taudul
60b36df077
Cosmetics. 2021-10-21 01:49:20 +02:00
Bartosz Taudul
6c53c36bfe
Remove server query quota. 2021-10-21 01:48:52 +02:00
Bartosz Taudul
93b6fd72c3
Keep lock until we're done with the data. 2021-10-21 01:31:03 +02:00
Bartosz Taudul
4348c5428d
Returned strings may be invalidated, so copy them. 2021-10-21 01:30:17 +02:00
Bartosz Taudul
ff54317a87
Move EnsureReadable() and co. to top of source file. 2021-10-21 00:06:14 +02:00
Bartosz Taudul
6d490ffd28
Revert "Use SPSC queue for frame images." 
This reverts commit 02e76faff7.

Let's not limit frame image reporting to just one thread.
 2021-10-20 23:18:03 +02:00
Bartosz Taudul
ce2a7d6741
Offset field is not used. 2021-10-20 22:57:23 +02:00
Bartosz Taudul
02e76faff7
Use SPSC queue for frame images. 2021-10-20 22:57:23 +02:00
Bartosz Taudul
ff3382391e
Remove BlockingReaderWriterQueue and semephores. 2021-10-20 22:57:22 +02:00
Bartosz Taudul
3d21c4e7ea
Add amalgamated readerwriterqueue 1.0.5. 2021-10-20 22:45:30 +02:00
Bartosz Taudul
a1511a39bd
Delta-encode code information pointers. 2021-10-16 16:29:06 +02:00
Bartosz Taudul
5ccf369919
Implement callstack serialization. 2021-10-10 16:14:17 +02:00
Bartosz Taudul
a421083e58
Crash reports may be serialized. 2021-10-10 15:58:24 +02:00
Bartosz Taudul
082255f940
Messages may now use serialized queue. 2021-10-10 15:54:21 +02:00
Bartosz Taudul
20074e64f4
Fix processing of serialized zone color, value and validation. 2021-10-10 14:43:36 +02:00
Bartosz Taudul
909f5916b1
Use possibly-synchronous macros for C API zones. 2021-10-09 15:51:29 +02:00
Bartosz Taudul
2cea5bc5b2
Add C API async-or-sync queue macros. 2021-10-09 15:50:30 +02:00
Bartosz Taudul
5835011d0f
Allow serialized zone validation. 2021-10-09 15:50:09 +02:00
Bartosz Taudul
7527f33d6c
Handle serialized zones. 2021-10-09 15:31:45 +02:00
Bartosz Taudul
2eeade89ba
Add macro for handling serial dequeue thread context checks. 2021-10-09 15:30:58 +02:00
Bartosz Taudul
2d5d4293a9
Move thread context check to a separate function. 2021-10-09 15:16:36 +02:00
Bartosz Taudul
13acec38f7
Use the possibly-synchronous macros in C++ zones. 2021-10-09 14:58:33 +02:00