Annotate bitwise013_1

This commit is contained in:
Matt Mascarenhas 2018-04-09 18:06:36 +01:00
parent c86e33ef32
commit 305275829a
2 changed files with 33 additions and 0 deletions

View File

@ -0,0 +1,31 @@
[video member=pervognsen stream_platform=twitch project=bitwise title="More Code Generation, Part 1" vod_platform=youtube id=uIjd094EKPM annotator=Miblo]
[0:08][Recap and set the stage for the day][:speech]
[0:45][:Run the program and consult the :performance profile with a few words on our :memory allocation]
[5:18][Explain some of the :performance considerations involved in allocating :memory, including TLB (translation lookaside buffer) misses and demand clearing of memory pages][:speech]
[8:02][Recommend Intel VTune[ref
site="Intel® Software"
page="Choose & Download | Intel® System Studio"
url=https://software.intel.com/en-us/system-studio/choose-download] and Windows Performance Toolkit[ref
site="Microsoft Docs"
page="Windows Performance Toolkit"
url=https://docs.microsoft.com/en-us/windows-hardware/test/wpt/]][:research]
[14:44][:Run the program and consult VTune's Advanced Hotspot Analysis, and ReactOS[ref
site=ReactOS
page="ntoskrnl/mm/mmfault.c Source File"
url=https://doxygen.reactos.org/db/d74/mmfault_8c_source.html]][:performance]
[19:25][:Run the program and consult VTune's General Exploration Microarchitecture Analysis, with an explanation of basic sampling vs instrumentation :profiling, and mentions of PerfMon Events[ref
site="PerfMon Events"
url=https://download.01.org/perfmon/index/index.html#pltfrm=skylake.html&pltfrmevent=DTLB_LOAD_MISSES.WALK_ACTIVE] and perf[ref
site="Brendan Gregg"
page="Linux perf Examples"
url=http://brendangregg.com/perf.html]][:performance]
[27:01][Switch back to using xmalloc() and enable the full :"code generation" code][:memory]
[27:29][:Run it to see that it's way slower, and investigate why][:performance]
[36:04][@praet_a51][processExplorer has a search feature that can locate file locks]
[36:36][Continue to investigate this slowdown][:performance :run]
[37:38][Revert buf__grow() to the old committed code][:memory]
[38:35][:Run it to find that it's still very slow][:performance]
[39:24][@badflydog][If you go to 'Users' in Task Manager you can see if the CPU usage comes from other places than your account]
[40:49][@elventhief][Any VTune service process running?]
[40:53][Reboot][:admin]
[/video]

View File

@ -0,0 +1,2 @@
[video member=pervognsen stream_platform=twitch project=bitwise title="More Code Generation, Part 2" vod_platform=youtube id=PVnRVadB7g0 annotator=]
[/video]