Opened 18 years ago
Last modified 22 months ago
#1074 assigned enhancement
Identify and fix performance bottlenecks in our app_server
| Reported by: | axeld | Owned by: | jua |
|---|---|---|---|
| Priority: | normal | Milestone: | R1.1 |
| Component: | Servers/app_server | Version: | |
| Keywords: | Cc: | kaoutsis | |
| Blocked By: | Blocking: | ||
| Platform: | All |
Description
The current app_server implementation is mostly not optimized yet. As this is actually noticeable, and we aren't that close to the speed of BeOS R5 yet, this should be changed, though :-)
Attachments (1)
Change History (41)
comment:1 by , 18 years ago
| Cc: | added |
|---|
comment:2 by , 18 years ago
comment:3 by , 18 years ago
| Cc: | added; removed |
|---|
Well, you'd have to use a profiler and compile with profiling options. Stephan has already done some profiling. I think you need to add this to the Jamfiles for the app_server (Ingo will know better):
{
SubDirCcFlags -pg ;
SubDirC++Flags -pg ;
}
I also found a short intro to this topic. I've rarely used profilers, though, and I don't know if you will also need to set linker flags in the Jamfile. You could just try and see if it works. I suggest that you ask on the main mailing list if you need more (=better ;) help.
comment:4 by , 18 years ago
It will also require us to measure the impact of context switches, and to find ways to reduce their cost (ie. by trying to make sure that drawing threads always use their full quantum if there is something left to draw).
follow-up: 9 comment:5 by , 18 years ago
The SubDirCcFlags/SubDirC++Flags should work. You can also add the compiler flags in your build/jam/UserBuildConfig file, which is the recommended way, since you don't need to meddle with any Jamfiles that are under version control. To add the "-pg" flag in src/servers/app and all subdirectories use the following:
AppendToConfigVar CCFLAGS : HAIKU_TOP src servers app : -pg : global ; AppendToConfigVar C++FLAGS : HAIKU_TOP src servers app : -pg : global ;
If you want to add it only for specific directories, replace the "global" by "local" and add respective lines for the directories you're interested in.
comment:6 by , 18 years ago
Does the profiling switches -pg working for sure in gcc 2.95.3-beos-060710 ? If i enable the switches, it seems that i can not produce gmon.out: $ gprof -v GNU gprof 2.15 $ gprof MyCompiledAndLinkedWithProf gmon.out: No such file or directory
follow-up: 8 comment:7 by , 18 years ago
Did you run the app before you started the profiler? I.e.: $ MyCompiledAndLinkedWithProf $ gprof MyCompiledAndLinkedWithProf
follow-up: 12 comment:9 by , 18 years ago
follow-up: 11 comment:10 by , 18 years ago
You're misusing gprof :-) You're supposed to run the executable, and quit it (killing it won't help), then it writes out a file called something like profile* - and *that* file is the one you have to use gprof on. Because of these semantics, you can more or less only use profiling in the app_server test environment. Or write your own profiling code (instead of using the one provided by i386-mcount.o).
comment:11 by , 18 years ago
Replying to axeld:
You're misusing gprof :-)
this is a nice opportunity to clear this issue: a) i follow this guide http://www.network-theory.co.uk/docs/gccintro/gccintro_80.html that Waldemar kindly provided. Doesn't working on hrev5 as described, (did i miss something?) but i believe on linux it would (i will test it tomorrow to be sure). b) I just found this: taken from our old newsletter issue 20: [...] Daniel Reinhold writes: [...] There is a GNU profiler that is included with many POSIX systems called 'gprof'. It is included with BeOS R5 as well (or at least with the developer tools mentioned above), but frankly, it doesn't work. However, Be did provide their own profiler called "profile". [...] Could you please clear the situation: Does gprof working on hrev5, should we use be's profile, or we may use Daniel's Ezprof (it is working fine), and we may extend to fit our needs?
You're supposed to run the executable, and quit it (killing it won't help),
Yes, i already do that, ...
then it writes out a file called something like profile* -
Yes, this file is generated...
and *that* file is the one you have to use gprof on.
How?
Because of these semantics, you can more or less only use profiling in the app_server test environment. Or write your own profiling code (instead of using the one provided by i386-mcount.o).
That is a very important information to go on. Thanks!
follow-up: 13 comment:12 by , 18 years ago
Replying to kaoutsis:
Replying to bonefish: Very elegant. Now something to finish that work for linking: in my hrev5 & gcc 2.95.3-beos-060710 need to link with /boot/develop/lib/x86/i386-mcount.o to resolve the profiling symbols.
Just use the LinkAgainst rule to add it:
LinkAgainst app_server : /boot/develop/lib/x86/i386-mcount.o ;
Note, that the dependency is old, so the first time you might need to remove the generated executable or otherwise trigger re-linking.
comment:14 by , 18 years ago
gprof is working perfectly on linux, as described in the Waldemar's short info (but unfortunately not on hrev5). Also gprof is found in binutils package.
comment:15 by , 18 years ago
Sorry, my bad. Under BeOS, you can just use "bprof" instead of "gprof"; usage is the same.
follow-ups: 24 35 comment:16 by , 18 years ago
Hi,
if someone manages to get gprof/bprof working properly on R5 (for use in the test environment), that would be great. Don't forget to tell me how... :-)
I did manage to get some useful info from ezprof, but I fixed all issues to the point that the info retrieved from ezprof is now pretty much inconclusive.
I have thought very much about the performance issues, and have come to the conclusion, that the problem could very well be in the way "update sessions" work. Currently, the app_server maintains repainting on the *window* level of things. There are two update sessions (per window): a current and a pending update session. The client window paints views in the current update session which touch the dirty region. The problem here is, that the mechanism is probably too "coarse": The update session cannot grow after the window has been informed that repainting needs to be done. So additional dirty regions are placed in the next (pending) update session. When the window finally gets to repainting a particular view after having painted a few views, large portions of it might already be dirty in the next update session. So it has to be painted again. If the update sessions worked on the *view* level, instead of the window level (as seems to be the case in R5, since there is one update message per view), the dirty region (for each view) might grow for a much longer time on the server side (while the client window is busy painting other views), and the window might have to paint the view only once.
I have yet to think of a way, how the update stuff can be moved to the view level, without wasting too much resources for all those regions. But I think this is defenitely one very important reason for the current "slowness" of the app_server. The drawing speed itself is likely not the problem for the majority of the views you see during typical use.
HTH, Stephan
comment:17 by , 18 years ago
I also made some research about ezprof, since i am interesting to make it working in haiku also. The main problem about ezprof is well... speed. From my point of view, if ezprof can speed up, is very useful. To give an example of slowness: i take the collatz.c from http://www.network-theory.co.uk/docs/gccintro/gccintro_80.html for examing with ezprof and the results were the same as gprof (in linux), with a difference: ezprof take 3 hours! to complete. The "profile" the most speedy profiling tool (is a tool that comes with bprof) give the same results to the same machine in 20 seconds. ps: i guess collatz.c is the test for profilers:-)
bye,
Vasilis
comment:18 by , 18 years ago
"profile" is using a very different concept, though: it just periodically checks where the executable is and counts the functions. That's very cheap, but also very inexact. Also, profilers that work via instrumenting like ezprof make function calls much more expensive, and thus, may distort the results.
comment:19 by , 15 years ago
| Cc: | removed |
|---|---|
| Owner: | changed from to |
| Version: | R1/pre-alpha1 |
comment:20 by , 8 years ago
| Owner: | changed from to |
|---|---|
| Status: | new → assigned |
comment:21 by , 6 years ago
Is it worth leaving this open? We can open new tickets for specific still-existing issues.
follow-up: 37 comment:23 by , 4 years ago
Well, the conclusions in https://dev.haiku-os.org/ticket/1074#comment:16 should be turned into a new ticket if still applicable. Other than that, yes, this seems quite open-ended and not really useful.
comment:24 by , 4 years ago
Replying to stippi:
I have yet to think of a way, how the update stuff can be moved to the view level, without wasting too much resources for all those regions.
It is not possible because all views in a window share same thread and it can't be changed without breaking API behavior. Instead of this, dirty region collection can be improved.
comment:25 by , 4 years ago
I run test with small window and placeholder view that call BView::Invalidate at end of BView::Draw. profile results:
profiling results for thread "w:2284:TestApp" (2287):
tick interval: 1000 us
total ticks: 27072 (27072000 us)
unknown ticks: 0 (0 us, 0.00%)
dropped ticks: 0 (0 us, 0.00%)
hits unknown image
------------------------------------------------------------------------------
7154 0 1 kernel_x86_64
7336 0 3078 /boot/system/servers/app_server
2742 0 3080 /boot/system/lib/libbe.so
56 0 3100 /boot/system/lib/libstdc++.so.6.0.25
9784 0 3101 /boot/system/lib/libroot.so
hits in us in % image function
------------------------------------------------------------------------------
6080 6080000 22.46 3101 system_time
3199 3199000 11.82 3078 Window::BeginUpdate(BPrivate::PortLink&)
2435 2435000 8.99 1 ConditionVariable::_Notify(bool, int)
1642 1642000 6.07 1 ConditionVariableEntry::Wait(unsigned int, long)
737 737000 2.72 1 x86_64_syscall_entry
688 688000 2.54 1 _mutex_lock
681 681000 2.52 3101 __mutex_lock
570 570000 2.11 3101 __mutex_unlock
489 489000 1.81 3080 agg::math_stroke<agg::pod_bvector<agg::point_base<double>, 6u> >::calc_miter(agg::pod_bvector<agg::point_base<double>, 6u>&, agg::vertex_dist const&, agg::vertex_dist const&, agg::vertex_dist const&, double, double, double, double, agg::line_join_e, double)
458 458000 1.69 1 _mutex_unlock
431 431000 1.59 3101 BPrivate::hoardHeap::findAvailableSuperblock(int, BPrivate::block*&, BPrivate::processHeap*)
422 422000 1.56 3078 void agg::render_scanlines<agg::rasterizer_scanline_aa<agg::rasterizer_sl_clip<agg::ras_conv_int> >, agg::scanline_p8, agg::renderer_scanline_aa_solid<agg::renderer_region<PixelFormat> > >(agg::rasterizer_scanline_aa<agg::rasterizer_sl_clip<agg::ras_conv_int> >&, agg::scanline_p8&, agg::renderer_scanline_aa_solid<agg::renderer_region<PixelFormat> >&)
352 352000 1.30 3080 agg::vcgen_stroke::vertex(double*, double*)
317 317000 1.17 3101 BPrivate::hoardHeap::freeBlock(BPrivate::block*&, BPrivate::superblock*&, int, BPrivate::processHeap*)
270 270000 1.00 3078 agg::conv_adaptor_vcgen<agg::path_base<agg::vertex_block_storage<double, 8u, 256u> >, agg::vcgen_stroke, agg::null_markers>::vertex(double*, double*)
...
system_time seems to be quite slow.
comment:26 by , 4 years ago
system_time uses uint128_t on x86_64: https://xref.landonf.org/source/xref/haiku/src/system/libroot/os/arch/x86_64/system_time.cpp#27
Possibly this can be optimized like the bootloader one was? But that still seems odd.
follow-up: 28 comment:27 by , 4 years ago
Can you use profile -f to get the full call stacks? This should then show what is calling system_time so often, as well as other things that a profile without -f often misses.
comment:28 by , 4 years ago
Replying to waddlesplash:
Can you use
profile -fto get the full call stacks? This should then show what is calling system_time so often, as well as other things that a profile without -f often misses.
profiling results for thread "w:1250:TestApp" (1253):
tick interval: 1000 us
total ticks: 43459 (43459000 us)
unknown ticks: 0 (0 us, 0.00%)
dropped ticks: 0 (0 us, 0.00%)
samples/tick: 10.4
hits unknown image
------------------------------------------------------------------------------
8832 0 1 kernel_x86_64
43459 0 3078 /boot/system/servers/app_server
16303 0 3080 /boot/system/lib/libbe.so
2350 0 3100 /boot/system/lib/libstdc++.so.6.0.25
43459 0 3101 /boot/system/lib/libroot.so
60161 0 51 commpage
hits in us in % image function
------------------------------------------------------------------------------
43459 43459000 100.00 3078 MessageLooper::_message_thread(void*)
43459 43459000 100.00 3101 thread_entry
43244 43244000 99.51 3078 ServerWindow::_MessageLooper()
35525 35525000 81.74 3078 ServerWindow::_DispatchMessage(int, BPrivate::LinkReceiver&)
19138 19138000 44.04 3078 Window::BeginUpdate(BPrivate::PortLink&)
13539 13539000 31.15 3078 ServerWindow::_DispatchViewMessage(int, BPrivate::LinkReceiver&)
11922 11922000 27.43 3078 ServerWindow::_DispatchViewDrawingMessage(int, BPrivate::LinkReceiver&)
11447 11447000 26.34 3101 system_time
10172 10172000 23.41 3078 BRect Painter::_StrokePath<agg::path_base<agg::vertex_block_storage<double, 8u, 256u> > >(agg::path_base<agg::vertex_block_storage<double, 8u, 256u> >&, cap_mode) const
8949 8949000 20.59 3078 BRect Painter::_RasterizePath<agg::conv_stroke<agg::path_base<agg::vertex_block_storage<double, 8u, 256u> >, agg::null_markers> >(agg::conv_stroke<agg::path_base<agg::vertex_block_storage<double, 8u, 256u> >, agg::null_markers>&) const
8825 8825000 20.31 1 x86_64_syscall_entry
6373 6373000 14.66 3080 BPrivate::LinkReceiver::GetNextMessage(int&, long)
5800 5800000 13.35 3078 agg::conv_adaptor_vcgen<agg::path_base<agg::vertex_block_storage<double, 8u, 256u> >, agg::vcgen_stroke, agg::null_markers>::vertex(double*, double*)
5779 5779000 13.30 3078 DrawingEngine::StrokeLine(BPoint const&, BPoint const&)
5638 5638000 12.97 3078 Painter::StrokeLine(BPoint, BPoint)
5077 5077000 11.68 3078 DrawingEngine::StrokeRect(BRect)
4931 4931000 11.35 3078 Painter::StrokeRect(BRect const&) const
4197 4197000 9.66 3080 BPrivate::LinkReceiver::ReadFromPort(long)
4140 4140000 9.53 3101 _kern_port_buffer_size_etc
3724 3724000 8.57 1 _user_port_buffer_size_etc
3712 3712000 8.54 1 _get_port_message_info_etc
3395 3395000 7.81 3080 agg::vcgen_stroke::vertex(double*, double*)
2740 2740000 6.30 3101 malloc
2523 2523000 5.81 1 ConditionVariableEntry::Wait(unsigned int, long)
2191 2191000 5.04 3101 _kern_write_port_etc
2171 2171000 5.00 3100 operator new(unsigned long)
2089 2089000 4.81 3101 BPrivate::threadHeap::malloc(unsigned long)
2087 2087000 4.80 3101 _kern_read_port_etc
1997 1997000 4.60 1 _user_write_port_etc
1931 1931000 4.44 1 writev_port_etc
1867 1867000 4.30 3101 free
1829 1829000 4.21 1 _user_read_port_etc
1792 1792000 4.12 1 read_port_etc
1771 1771000 4.08 1 ConditionVariable::_Notify(bool, int)
1703 1703000 3.92 3080 agg::vcgen_stroke::add_vertex(double, double, unsigned int)
1623 1623000 3.73 3078 Window::_SendUpdateMessage()
1584 1584000 3.64 3080 agg::vertex_sequence<agg::vertex_dist, 6u>::add(agg::vertex_dist const&)
1390 1390000 3.20 3080 BRegion::IntersectWith(BRegion const*)
1318 1318000 3.03 1 _mutex_unlock
1309 1309000 3.01 3078 void agg::rasterizer_sl_clip<agg::ras_conv_int>::line_to<agg::rasterizer_cells_aa<agg::cell_aa> >(agg::rasterizer_cells_aa<agg::cell_aa>&, int, int)
1294 1294000 2.98 3078 ServerWindow::SendMessageToClient(BMessage const*, int) const
1219 1219000 2.80 3101 BPrivate::hoardHeap::findAvailableSuperblock(int, BPrivate::block*&, BPrivate::processHeap*)
1193 1193000 2.75 3080 agg::math_stroke<agg::pod_bvector<agg::point_base<double>, 6u> >::calc_miter(agg::pod_bvector<agg::point_base<double>, 6u>&, agg::vertex_dist const&, agg::vertex_dist const&, agg::vertex_dist const&, double, double, double, double, agg::line_join_e, double)
1177 1177000 2.71 3078 agg::rasterizer_scanline_aa<agg::rasterizer_sl_clip<agg::ras_conv_int> >::add_vertex(double, double, unsigned int)
1131 1131000 2.60 3101 BPrivate::processHeap::free(void*)
995 995000 2.29 3080 BPrivate::LinkReceiver::HasMessages() const
965 965000 2.22 1 get_locked_port(int)
962 962000 2.21 3078 void agg::render_scanlines<agg::rasterizer_scanline_aa<agg::rasterizer_sl_clip<agg::ras_conv_int> >, agg::scanline_p8, agg::renderer_scanline_aa_solid<agg::renderer_region<PixelFormat> > >(agg::rasterizer_scanline_aa<agg::rasterizer_sl_clip<agg::ras_conv_int> >&, agg::scanline_p8&, agg::renderer_scanline_aa_solid<agg::renderer_region<PixelFormat> >&)
947 947000 2.18 3078 agg::rasterizer_cells_aa<agg::cell_aa>::line(int, int, int, int)
942 942000 2.17 3101 _kern_port_count
918 918000 2.11 3080 agg::math_stroke<agg::pod_bvector<agg::point_base<double>, 6u> >::calc_cap(agg::pod_bvector<agg::point_base<double>, 6u>&, agg::vertex_dist const&, agg::vertex_dist const&, double)
895 895000 2.06 3101 __mutex_lock
828 828000 1.91 3080 agg::pod_bvector<agg::point_base<double>, 6u>::allocate_block(unsigned int)
755 755000 1.74 1 _mutex_lock
755 755000 1.74 3080 BRegion::_SetSize(int)
749 749000 1.72 3101 __mutex_unlock
714 714000 1.64 3078 Window::EndUpdate()
640 640000 1.47 3078 Window::InvalidateView(View*, BRegion&)
535 535000 1.23 3101 BPrivate::hoardHeap::freeBlock(BPrivate::block*&, BPrivate::superblock*&, int, BPrivate::processHeap*)
501 501000 1.15 3080 agg::math_stroke<agg::pod_bvector<agg::point_base<double>, 6u> >::calc_join(agg::pod_bvector<agg::point_base<double>, 6u>&, agg::vertex_dist const&, agg::vertex_dist const&, agg::vertex_dist const&, double, double)
496 496000 1.14 3080 BRegion::_AdoptRegionData(BRegion&)
462 462000 1.06 3078 Window::VisibleContentRegion()
452 452000 1.04 3078 ServerWindow::_UpdateCurrentDrawingRegion()
435 435000 1.00 3078 HWInterface::InvalidateRegion(BRegion&)
...
comment:29 by , 4 years ago
system_time with int128 as it currently exists:
0000000000000010 <system_time>: 10: 0f 31 rdtsc 12: 48 c1 e2 20 shl $0x20,%rdx 16: 48 09 d0 or %rdx,%rax 19: 48 f7 25 00 00 00 00 mulq 0x0(%rip) # 20 <system_time+0x10> 20: 48 89 d0 mov %rdx,%rax 23: c3 retq
system_time with the logic as used in the bootloader involving uint64:
0000000000000010 <system_time>: 10: 0f 31 rdtsc 12: 48 8b 0d 00 00 00 00 mov 0x0(%rip),%rcx # 19 <system_time+0x9> 19: 48 c1 e2 20 shl $0x20,%rdx 1d: 48 09 d0 or %rdx,%rax 20: 89 c2 mov %eax,%edx 22: 48 c1 e8 20 shr $0x20,%rax 26: 48 0f af d1 imul %rcx,%rdx 2a: 48 0f af c1 imul %rcx,%rax 2e: 48 c1 ea 20 shr $0x20,%rdx 32: 48 01 d0 add %rdx,%rax 35: c3 retq
So the uint128 version is actually fewer instructions, so unless imul is somehow significantly faster than mulq, the current version is faster.
follow-up: 32 comment:31 by , 4 years ago
Do you mean you enabled PROFILE_TIMING in app_server? Most of the system_time calls are behind that. It still seems rather strange.
comment:32 by , 4 years ago
Replying to waddlesplash:
Do you mean you enabled PROFILE_TIMING in app_server? Most of the system_time calls are behind that. It still seems rather strange.
No, this is regular app_server, not test_app_server and no additional debug options were used.
comment:33 by , 4 years ago
Well, I'm not sure where all these calls are coming from then: https://xref.landonf.org/source/search?q=system_time&defs=&refs=&path=servers+app+-tests&hist=&type=&project=haiku
Note most of those are behind disabled #ifs.
comment:34 by , 4 years ago
I benchmarked both system_time versions and libroot one seems a bit faster:
system_time_libroot: 133254624 calls/sec system_time_boot: 130536885 calls/sec
Assembly is the same as above.
comment:35 by , 4 years ago
Replying to stippi:
If the update sessions worked on the *view* level, instead of the window level (as seems to be the case in R5, since there is one update message per view)
BeOS seems to do the same thing as Haiku, updates are managed on per-window, not per-view level:
fd04213c 80005d0c TestView::Draw(BRect) + 00000028 fd042188 ec223731 BWindow::do_draw_views(void) + 0000026d fd0421e0 ec2256b6 BWindow::DispatchMessage(BMessage *, BHandler *) + 000005a6 fd042240 ec2244be BWindow::task_looper(void) + 00000362 fd04228c ec16ce96 BLooper::_task0_(void *) + 00000036 fd0422a0 ec08558d thread_start + 00000039
comment:36 by , 4 years ago
I managed to run BeOS in VMware player without major speed drop. Now speed is comparable to Haiku, thread switching score (#15509):
BeOS: 25812.424108 Haiku: 28583.646038
BeOS app_server feels faster than Haiku.
comment:37 by , 4 years ago
Replying to pulkomandy:
Well, the conclusions in https://dev.haiku-os.org/ticket/1074#comment:16 should be turned into a new ticket if still applicable. Other than that, yes, this seems quite open-ended and not really useful.
Done in #16356. However, I did't notice there has been a lengthy discussion after your comment.
comment:38 by , 4 years ago
| Milestone: | R1 → R1.1 |
|---|
comment:39 by , 22 months ago
DrawingEngine::FillRect seems inefficient and call malloc/free on each call.
by , 22 months ago
| Attachment: | callgrind2.png added |
|---|
comment:40 by , 22 months ago
It seems there are significant overhead in heap memory allocator when creating and destroying windows:
profiling results for summary "all" (1):
tick interval: 1000 us
total ticks: 140988 (140988000 us)
unknown ticks: 0 (0 us, 0.00%)
dropped ticks: 0 (0 us, 0.00%)
hits unknown image
------------------------------------------------------------------------------
852 0 5 /Haiku/data/packages/haiku2/generated.x86_64.native/tests/libbe_test/x86_64/apps/lib/libtestappserver.so
34 0 10 /boot/system/lib/libfreetype.so.6.18.3
2 0 3 /boot/system/runtime_loader
8 0 8 /Haiku/data/packages/haiku2/generated.x86_64.native/tests/libbe_test/x86_64/apps/lib/libhwinterfaceimpl.so
63 0 21 /boot/system/lib/libbe.so
134642 0 12 /boot/system/lib/libroot.so
17 0 11 /boot/system/lib/libstdc++.so.6.0.29
4914 0 6 /Haiku/data/packages/haiku2/generated.x86_64.native/tests/libbe_test/x86_64/apps/lib/libbe_test.so
1 0 27 /boot/system/lib/libcrypto.so.1.1
21 0 22 /boot/home/Tests/VideoStreams/build.x86_64/install/lib/libvideostreams.so
4 0 7 /Haiku/data/packages/haiku2/generated.x86_64.native/tests/libbe_test/x86_64/apps/lib/libhwinterface.so
hits in us in % image function
------------------------------------------------------------------------------
65970 65970000 46.79 12 BPrivate::hoardSbrk(long)
64626 64626000 45.84 12 BPrivate::hoardUnsbrk(void*, long)
1790 1790000 1.27 6 BRegion::Support::miRegionOp(BRegion*, BRegion const*, BRegion const*, int (*)(BRegion*, clipping_rect*, clipping_rect*, clipping_rect*, clipping_rect*, int, int), int (*)(BRegion*, clipping_rect*, clipping_rect*, int, int), int (*)(BRegion*, clipping_rect*, clipping_rect*, int, int))
1074 1074000 0.76 6 BRegion::Support::miUnionNonO(BRegion*, clipping_rect*, clipping_rect*, int, int)
648 648000 0.46 12 __mutex_lock
605 605000 0.43 6 BRegion::Support::miCoalesce(BRegion*, int, int)
465 465000 0.33 12 _kern_mutex_lock
446 446000 0.32 6 BRegion::Support::miSetExtents(BRegion*)
414 414000 0.29 12 __mutex_unlock
359 359000 0.25 12 _kern_mutex_unlock
259 259000 0.18 6 BRegion::Support::miIntersectO(BRegion*, clipping_rect*, clipping_rect*, clipping_rect*, clipping_rect*, int, int)
225 225000 0.16 6 BRegion::Support::miSubtractO(BRegion*, clipping_rect*, clipping_rect*, clipping_rect*, clipping_rect*, int, int)
208 208000 0.15 12 BPrivate::hoardHeap::findAvailableSuperblock(int, BPrivate::block*&, BPrivate::processHeap*)
163 163000 0.12 12 BPrivate::threadHeap::malloc(unsigned long)
154 154000 0.11 12 memcpy
149 149000 0.11 12 free
130 130000 0.09 12 _kern_write_port_etc
111 111000 0.08 12 BPrivate::processHeap::free(void*)
107 107000 0.08 12 malloc
i find interesting this topic. Can you think a way, that i could help?