Algorithms/go-perfbook
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

explicitly call out TODOs

Browse Source
This commit is contained in:
Damian Gryski 2018-03-04 10:34:44 -08:00
parent 48db8aa7af
commit 99e82b8791
1 changed files with 3 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
performance.md
View File

@ -152,6 +152,8 @@ Dan Luu's previously mentioned talk on [BitFunnel performance
estimation](http://bitfunnel.org/strangeloop) shows an example of using rough estimation](http://bitfunnel.org/strangeloop) shows an example of using rough
calculations to determine if your target performance figures are reasonable. calculations to determine if your target performance figures are reasonable.
TODO: Programming Pearls has "Fermi Problems". Knowing Jeff Dean's slide helps.
For greenfield development, you shouldn't leave all benchmarking and For greenfield development, you shouldn't leave all benchmarking and
performance numbers until the end. It's easy to say "we'll fix it later", but performance numbers until the end. It's easy to say "we'll fix it later", but
if performance is really important it will be a design consideration from the if performance is really important it will be a design consideration from the
@ -163,7 +165,7 @@ in the development cycle when a more complete view of the system performance
is available. Any full-system profiles you do while the system is incomplete is available. Any full-system profiles you do while the system is incomplete
will give a skewed view of where the bottlenecks will be in the finished system. will give a skewed view of where the bottlenecks will be in the finished system.
Death by 1000 cuts. TODO: How to avoid/detect "Death by 1000 cuts" from poorly written software.
Write code that you can benchmark. Profiling you can do on larger systems. Write code that you can benchmark. Profiling you can do on larger systems.
Benchmarking you want to test isolated pieces. You need to be able to extract Benchmarking you want to test isolated pieces. You need to be able to extract